Insights into new targets, modalities, and technologies for improving human health, curated for Peter Schultz and the Scripps Research community. https://github.com/andrewsu/ai-nuggets en-us AI Nuggets by the Su Lab Peter Schultz false Your daily briefing has been paused because no downloads have been detected in about two weeks. As soon as you download an episode, generation will automatically resume. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-18-goodbye Sat, 18 Jul 2026 09:00:11 +0000 7 Your daily briefing has been paused because no downloads have been detected in about two weeks. As soon as you download an episode, generation will automatically resume. Four bioRxiv reports spanning a Wnt-pathway degrader that finally explains why enzymatic tankyrase inhibitors underperformed for a decade, a David Baker generative framework for de novo transmembrane β-barrel nanopores with custom lumens, a paradoxical GPCR agonist strategy that extends survival in an ALS mouse model by driving receptor internalization to release an oligodendrocyte differentiation brake, and a digital twin of in vitro pancreatic islet differentiation that surfaces two new causal regulators of endocrine versus exocrine fate. (1) Q. Wang, L. Li, L. You, S. Wang, L. Han, B. Wang, L. Yao, Y. Addepalli, Y. Lu, I. Mender, A. M. Flusche, C. Kim, N. Yarravarapu, A. Lemoff, L. Lum, J. W. Shay, Y. Yu, C. Chen (UT Southwestern) reframe a longstanding failure of Wnt drug discovery. Enzymatic TNKS inhibitors like IWR1 and XAV939 stabilize the β-catenin destruction complex on paper but have consistently disappointed. Catalytic inhibition traps TNKS inside the destruction complex, and trapped TNKS uses its scaffolding function to nucleate large AXIN puncta that visibly rigidify the DC — FRAP shows β-catenin turnover slows to a crawl inside those puncta, and catalytically dead TNKS mutants still form puncta, so this is scaffolding, not enzymology. IWR1-POMA (IWR1 warhead + pomalidomide-cereblon binder) degrades TNKS with DC50 ~60 nM, dissolves the puncta rather than freezing them, shows essentially no off-target hits across the PARP family or 79 other NAD-dependent enzymes by TMT proteomics, and delivers much deeper β-catenin suppression than IWR1. In PDM-7 patient-derived colorectal cancer organoids carrying truncating APC mutations, the degrader kills where the inhibitor does nothing. General lesson: when an enzymatic inhibitor plateaus at partial pathway suppression, the target's scaffolding function may be doing the residual signaling — and only degradation removes both. (2) A. Philomin, R. Sonigra, S. Majumder, H.-J. Lin, Y. Li, F. Xue, R. D. Kibler, C. Baldus, E. Trapido, A. Medeiros, B. Coventry, A. Bera, A. Kang, J. Mendoza, M. Kumar, Y. Yang, and D. Baker present a generative AI framework for de novo transmembrane β-barrel nanopores. TMB_RFD2 (RFdiffusion2 finetuned on a curated 117,447-member TMB distillation set, orders of magnitude larger than the ~699 β-barrel structures in the PDB) generates asymmetric barrel backbones with tunable strand count, glycine-twist and β-bulge placement. TMB_MPNN (ProteinMPNN with a 96-neighbor receptive field, finetuned on the same set) designs sequences that actually express and fold. Forty-eight designs from 10 to 16 strands express in E. coli, refold in detergent, and conduct ions with clean single-channel traces; two crystal structures agree with the designed models within 1 Å Cα RMSD including deliberately placed glycine twists. Motif scaffolding drops a three-histidine copper coordination site into the pore lumen; single-channel recordings show copper-specific gating that goes away when the histidines are mutated. A larger 14-strand design supports helicase-driven single-strand DNA threading — the first published enzymatic DNA threading trace through an asymmetric de novo TMB with no soluble domain. A hydrophobic-thickness-aware sequence design mode produces longer barrels that insert into synthetic PEO-b-PBO/DPhPC hybrid membranes at 200 mV where standard barrels fail. Nanopores as programmable devices with custom lumen chemistry, tunable hydrophobic thickness, and scaffold-ready binding sites. (3) S. Raffaele, T. Bonifacino, F. C. Mannella, N. Nguyen, C. Torazza, D. Marangon, E. M. Chinosi, H. D. Schroder, E. K. Hejbol, K. Madsen, L. Marchetti, M. L. Trincavelli, M. Milanese, M. P. Lecca and colleagues target the oligodendroglial GPCR GPR17 in ALS. GPR17 is transiently expressed as OPCs mature and must be downregulated for full myelination; persistent GPR17 locks cells in an immature state. Post-mortem human ALS spinal cord shows exactly this signature — bulk RNA-seq elevation of GPR17 across cervical, thoracic, and lumbar levels, single-nucleus data showing an expansion of GPR17-positive immature oligodendrocytes, and IHC confirming spinal-cord-restricted accumulation not seen in cortex. The counterintuitive move is to hit the receptor with an agonist rather than an antagonist, because sustained agonism drives GPR17 internalization and functional downregulation, releasing the differentiation block. Brain-penetrant second-generation agonist galinex (sub-nanomolar potency in GTPγS assay), delivered by osmotic minipump at 10 mg/kg/day from the early symptomatic P90 stage, extends survival, delays weight loss, and improves rotarod, beam-balance, extension-reflex, gait, and grip-strength in female SOD1G93A mice, with restored mature oligodendrocyte density, preserved myelin, and better motor-neuron survival. Male mice show almost none of this, so the sex dependence is real and warrants mechanistic work before translation. (4) E. E. Sanchez-Castro, M. Ishahak, T. Le, M. M. Maestas, D. C. Hernandez-Rincon, N. Mukherjee, K. Bradley, J. Lu, S. E. Gale, J. R. Millman (WashU) build an integrated digital twin of in vitro SC-islet differentiation from 400,603 cells across 61 datasets, 4 cell lines, and 7 differentiation protocols, anchored on a new snMulti-seq map (69,535 nuclei, 9 timepoints, 33 days, 13 cell states). Over 1,000 in silico transcription-factor knockouts prospectively validate two regulators. STAT1 — never previously implicated in pancreatic development — is predicted to push cells toward exocrine fate; CRISPRi knockdown and ruxolitinib both selectively suppress exocrine markers (CAII, CFTR) on days 9-13 while leaving β-cell identity (INS, NKX6-1, MAFA) and glucose-stimulated insulin secretion intact. ZEB1 plays a dynamic role — early on it is required for endocrine specification, but persistent activity later diverts endocrine progenitors toward an off-target serotonergic islet-cell lineage rather than β cells. A window-specific shRNA on days 15-17 collapses serotonin content (TPH1, SLC18A1) and boosts the NKX6-1+/C-peptide+ β-cell fraction without touching insulin content. For an SC-islet manufacturing program this is directly actionable — model-nominated regulators with defined intervention windows, translated into small-molecule or CRISPRi handles that steer fate at branch points. All four preprints have full-body renders available via r.jina.ai (v2/v3 revisions on bioRxiv); summaries grounded in the paper body. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-17-tnks-protac-denovo-tmb-nanopore-gpr17-als-sc-islet-digital-twin Fri, 17 Jul 2026 12:00:00 +0000 360 Four bioRxiv reports: (1) Wang, Lu, Mender, Lum, Shay, Yu, Chen (UT Southwestern) — TNKS PROTAC IWR1-POMA (DC50 ~60 nM) reframes why enzymatic tankyrase inhibitors underperformed in WNT-driven cancer. Catalytic inhibition traps TNKS in the β-catenin destruction complex, where its scaffolding function nucleates AXIN puncta that rigidify the DC (FRAP-confirmed) and impede β-catenin turnover; catalytically dead TNKS still forms puncta. IWR1-POMA degrades with essentially no PARP/NAD-enzyme off-targets, dissolves the puncta, and kills PDM-7 patient-derived APC-truncating CRC organoids where the enzymatic inhibitor is inert. Lesson: plateaued enzymatic efficacy may signal a scaffolding function that only degradation removes. (2) Philomin/Sonigra/Majumder/Baker — TMB_RFD2 (RFdiffusion2 finetuned on 117,447-member TMB distillation set) plus TMB_MPNN (96-NN ProteinMPNN finetuned on same set) enables generative design of asymmetric transmembrane β-barrel nanopores. 48 designs (10-16 strands, 0.7-1.5 nm pores) functional in electrophysiology; two crystal structures within 1 Å Cα RMSD confirm backbone including specified glycine twists. His3-Cu coordination motif scaffolded into pore lumen gives Cu2+-specific gating; larger 14-strand design supports helicase-driven ssDNA threading (first for an asymmetric de novo TMB with no soluble domain); hydrophobic-thickness-aware design mode enables function in synthetic PEO-b-PBO/DPhPC hybrid membranes at 200 mV. (3) Raffaele/Bonifacino/Fumagalli/Lecca — GPR17 as druggable target in ALS. Post-mortem human ALS spinal cord shows bulk-RNA-seq upregulation of GPR17 and single-nucleus expansion of GPR17+ immature oligodendrocytes; IHC restricts to spinal cord. Counterintuitive strategy: sustained agonism drives receptor internalization and functional downregulation, releasing the differentiation brake. Brain-penetrant second-gen agonist galinex (sub-nM GTPγS EC50), 10 mg/kg/day osmotic pump from P90, extends survival and improves rotarod/beam-balance/gait/grip-strength in female SOD1G93A mice with restored mature oligodendrocytes and preserved myelin and motor neurons; male mice show no survival benefit. (4) Sanchez-Castro/Ishahak/Millman (WashU) — integrated digital twin of in vitro SC-islet differentiation (400,603 cells, 61 datasets, 4 cell lines, 7 protocols) anchored on new snMulti-seq map (69,535 nuclei, 33 days, 13 states). Over 1,000 in silico TF knockouts validate STAT1 as an exocrine-fate driver (CRISPRi + ruxolitinib on days 9-13 suppress CAII/CFTR without touching INS/NKX6-1/MAFA or glucose-stimulated insulin secretion) and ZEB1 as a dynamic endocrine regulator that later diverts progenitors toward off-target serotonergic islet cells (shRNA on days 15-17 collapses TPH1/SLC18A1 and boosts NKX6-1+/C-peptide+ β cells without touching insulin). Model-nominated regulators with defined intervention windows for SC-islet manufacturing. false Four bioRxiv preprints from the last 24-48 hours spanning covalent remodeling of cereblon that reframes molecular glue chemistry beyond the G-loop degron, a covalent inhibitor of the cancer stem cell target ALDH1A3 caught in two mutually exclusive conformations inside a single active site, a resident alveolar macrophage IL-34/IGF-1 axis that fuels KRAS-mutant lung cancer and protects it from ferroptosis, and a structure-guided peptide design pipeline that finds remote functional sequence space from tiny training sets. (1) A. H. de la Peña, J. T. Cruite, J. Che, M. E. Matyskiela, P. P. Chamberlain, E. S. Fischer (Dana-Farber), and L. H. Jones (Dana-Farber) solve the cryo-EM structure of a ternary complex in which cereblon and CRL4-DDB1 recruit the neosubstrate NTAQ1, driven not by a conventional imide but by the covalent molecular glue EM12-FS. EM12-FS does not bind CRBN reversibly the way IMiDs do — it imidazylates CRBN His353, repositions the sensor loop, and eliminates a steric clash that had blocked NTAQ1 engagement by reversible CRBN binders. The engineered interface is stabilized by a distinctive T-shaped C-H/π interaction between sulfated His353 and NTAQ1 Phe126. Biochemistry and mutations map ternary complex determinants. The reframe is that induced-proximity degrader pharmacology does not have to accept CRBN as it comes off the shelf: site-specific synthetic modification of the E3 ligase itself expands neosubstrate specificity beyond the G-loop degron, and covalent engineering of the ligase is a legitimate new axis for degrader design. (2) D. Covaleda, D. Vizarraga, T. Upadhyay, J. Zhu, D. Abegg, R. Pequerul, M. Hugo, A. Adibekian, I. Fita, X. Parés, F. X. Avilés, M. Bogyo, and J. Farrés target aldehyde dehydrogenase 1A3, an isoform overexpressed in tumors and long associated with cancer stem cell traits and therapy resistance. They exploit ALDH's secondary esterase activity to build an in vitro assay and screen a 3,200-compound cysteine-focused covalent fragment library against the highly conserved catalytic cysteine at the intersection of substrate and cofactor channels. The winner, Z34, is an acrylamide fragment with sub-micromolar potency against both ALDH1A1 and ALDH1A3; biophysics confirms time-dependent, covalent, irreversible binding to the active-site cysteine. The 2.26 Å cryo-EM structure of the ALDH1A3-Z34 complex shows the same fragment, tethered by the same cysteine covalent bond, occupies two mutually exclusive conformations — one fills the substrate-binding channel, the other the cofactor-binding channel. Two productive elaboration vectors are available off a single warhead, a rare gift in covalent medicinal chemistry: the fragment can be grown in either direction for isoform selectivity or to reach beyond the immediate active site. (3) J. Zak, H. Chen, E. Wang, P. Ozark, G. Mognol, M. D.-Y. Park, N. Fournier, P. Chaudary, J. Hu, R. Shepard, A. Ghebremedin, M. Paradise, J. Rivera, W. J. Harris, Z. Xu, A. Ramadan, B. Lim, M. Colonna, M. Merad, M. De Palma, M. Onaitis, and J. A. Varner (UCSD) show that KRAS-G12D lung epithelial tumors do not just tolerate the tissue-resident, embryonically-derived alveolar macrophage compartment — they expand it, using tumor-cell-secreted IL-34 to drive alveolar macrophage proliferation, recapitulating a signal from late embryonic lung development. The macrophages return the favor by secreting IGF-1, which drives tumor cell proliferation and protects tumor cells from ferroptosis. Neutralizing or genetically deleting IL-34 in the mouse model collapses the loop: IGF-1 goes down, macrophage and tumor proliferation drop, tumor progression stops. Human data lines up — high IL34 + high IGF1 correlate with poor survival in KRAS-G12D/V lung adenocarcinomas and other solid tumors. Two independently druggable clinical arms — IL-34 blockade cutting off macrophage recruitment or IGF-1 blockade cutting off the mitogenic and ferroptosis-protective return signal — into a genotype that has been very hard to attack head-on. (4) P. Bayat, S. J. Perkins, S. Clancy, S. S. Patel, R. F. Yin, K. Bozovicar, S. Singh, S. Shrestha, Z. Moustafa, R. Zayani, I. Iwe, S. Bayat, P. Kelly, J. R. J. Vigar, V. Y. White, M. Xie, M. Simchi, S. Palter, J. Nguyen, I. Y. Zeisler, B. Wu, and K. Pardee build MDMI (Minimal Data, Maximal Insight), a two-stage pipeline for peptide-protein interface design that leans on structure rather than sequence. Instead of feeding a machine learning model with peptide sequences alone, they extract three-dimensional interface geometry and binding energetics from predicted peptide-protein complexes and pair the structure-aware predictor with a genetic algorithm for sequence exploration. On an all-negative benchmark panel the sequence-only baseline hallucinates binders 70% of the time; MDMI collapses that to near zero and produces ~4× more high-confidence in silico binders than state-of-the-art baselines. The split-GFP functional test finds active peptides up to 38% diverged from wild-type in Stage 1, and motif-guided recombination pushes past 50% divergence in Stage 2, revealing two distinct functional clusters in sequence space. A top candidate expressed as a full-length GFP fusion retains a GFP-like emission profile — the redesigned peptide is refolding into a fluorescent scaffold. Structure-informed pipelines can reach genuinely distant functional sequence space from small annotated datasets, which is the regime therapeutic peptide analog discovery lives in. All four preprints are v1 posted 2026-07-14/15 on bioRxiv; bodies not yet rendered — summaries are grounded in the full abstracts per PIPELINE.md guidance. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-16-crbn-covalent-glue-aldh1a3-il34-kras-lung-mdmi-peptide Thu, 16 Jul 2026 12:00:00 +0000 387 Four bioRxiv preprints from the last 24-48 hours: (1) de la Peña, Cruite, Che, Matyskiela, Chamberlain, Fischer, Jones (Dana-Farber + Neomorph) — covalent molecular glue EM12-FS imidazylates CRBN His353 to reprogram induced-proximity pharmacology beyond the G-loop degron. Cryo-EM structure of the NTAQ1-EM12-FS-CRBN-DDB1 ternary complex reveals a non-canonical neosubstrate interface stabilized by a T-shaped C-H/π interaction between sulfated His353 and NTAQ1 Phe126; the covalent CRBN modification repositions the sensor loop to eliminate the steric clash that prevents NTAQ1 engagement by reversible IMiDs. Site-specific synthetic modification of CRBN itself as a legitimate new axis for degrader design. (2) Covaleda, Vizarraga, Upadhyay, Zhu, Abegg, Pequerul, Hugo, Adibekian, Fita, Parés, Avilés, Bogyo, Farrés — cysteine-focused covalent fragment screen (3,200 fragments) against the catalytic cysteine of ALDH1A3 via ALDH esterase activity identifies acrylamide fragment Z34 with sub-micromolar potency against ALDH1A1 + ALDH1A3. 2.26 Å cryo-EM structure of the ALDH1A3-Z34 complex shows the same warhead in two mutually exclusive covalent conformations — one occupying the substrate channel + one occupying the cofactor channel — two productive elaboration vectors off a single covalent handle for isoform selectivity or extended active-site engagement. (3) Zak, Chen, Wang, Ozark, Mognol, Park, Fournier, Chaudary, Hu, Shepard, Ghebremedin, Paradise, Rivera, Harris, Xu, Ramadan, Lim, Colonna, Merad, De Palma, Onaitis, Varner (UCSD) — KRAS-G12D lung epithelial tumors expand tissue-resident, embryonically-derived alveolar macrophages by secreting IL-34 (recapitulating a late embryonic lung development signal); alveolar macrophages return IGF-1, which drives tumor cell proliferation and protects tumor cells from ferroptosis. Genetic deletion or neutralization of IL-34 collapses the loop — IGF-1 goes down, macrophage + tumor proliferation drop, tumor progression stops. High IL34 + high IGF1 correlate with poor survival in KRAS-G12D/V lung adenocarcinomas and other solid tumors. Two independently druggable clinical arms into a hard-to-attack genotype. (4) Bayat, Perkins, Clancy, Patel, Yin, Bozovicar, Singh, Shrestha, Moustafa, Zayani, Iwe, Bayat, Kelly, Vigar, White, Xie, Simchi, Palter, Nguyen, Zeisler, Wu, Pardee — Minimal Data Maximal Insight (MDMI) is a two-stage structure-guided pipeline for peptide-protein interface design using only tiny annotated datasets: 3D interface geometry + binding energetics from predicted peptide-protein complexes feed a structure-aware predictor paired with a genetic algorithm for sequence exploration. False-positive rate on all-negative benchmark collapses from 70% (sequence-only baseline) to near zero and produces ~4× more high-confidence in silico binders than SOTA. Split-GFP functional test finds active peptides up to 38% diverged from wild-type in Stage 1 and >50% divergence via motif-guided recombination in Stage 2, revealing two distinct functional clusters in sequence space; top candidate expressed as full-length GFP fusion retains GFP-like emission. Structure-informed pipelines can reach remote functional sequence space from small annotated data — the regime therapeutic peptide analog discovery lives in. false Four bioRxiv preprints from the last 24-48 hours spanning a structural rescue of a discarded first-generation rifamycin against rifampicin-resistant RNA polymerase, a Sanford Burnham Prebys degrader campaign that shows collapsing the SAGA histone acetyltransferase complex with a PROTAC dissolves the ENL-anchored transcriptional condensates holding AML identity in place, an NMDA-receptor axis that lets therapy-tolerant cancer persister cells survive ferroptosis, and a lineage-tracing MSKCC study that pins pancreatic tumor maintenance on a single plastic cell state through a single cytokine. (1) H. Mosaei, Y. Shin, V. N. Kozhevnikov, P. G. Waddell, M. J. Hall, K. S. Murakami (Penn State), and N. Zenkin (Newcastle) revisit rifamycin B, the earliest member of the rifamycin family and precursor of the clinically used derivatives, previously sidelined for chemical instability and weak antibacterial activity. Under contemporary assay conditions, Rif B is stable enough to characterize and retains inhibitory activity against RNAP variants carrying clinically relevant rifampicin-resistance mutations. They report the first crystal structure of Rif B on its own and a co-structure of Rif B bound to bacterial RNAP. The distinctive C-4 O-carboxymethyl substituent forms an intramolecular contact in the free molecule and re-partners as a salt bridge with fork loop 2 of the RNAP β-subunit upon binding; that additional interaction — absent from rifampicin — explains the reduced sensitivity of Rif B to resistance-associated substitutions and identifies the C-4 position as an underexplored handle for rational rifamycin modification. Rif B is redefined as a mechanistically distinct scaffold and reopens the C-4 chemistry route into RNAP for resistant TB and Gram-positive infections. (2) A. Deshpande, C.-Y. Chiang, M. Perales Garcia, N. Niranjan, N. Sinha, D. Finlay, A. M. Stevens, E. Zahn, B. A. Garcia, I. Jeremias, M. Wunderlich, K. Jensen-Pergakes, A. Udyavar, A. Carr, A. R. Nager, Y. Yang, R. Murad, C. Jones, S. O'Connell, T. Paul, K. Vuori, and A. J. Deshpande (Sanford Burnham Prebys) use genome-scale dependency data across >1,000 cancer cell lines to identify the SAGA histone acetyltransferase complex as a selective chromatin dependency in AML and hematologic malignancies. Pharmacological degradation of the SAGA catalytic subunits KAT2A and KAT2B using a cereblon-recruiting PROTAC (GSK983/GSK699) delivers potent broad-spectrum antileukemic activity across genetically diverse AML cell lines, primary patient samples, and an isogenic KMT2A-rearranged model with cooperating oncogenic mutations. Mechanistically, KAT2A/B degradation drives a genome-wide reduction in H3K9ac with the largest asymmetric losses concentrated at core AML oncogene loci — MYC, MYB, the HOXA cluster — displaces the chromatin reader ENL from those loci, dissolves ENL-anchored transcriptional condensates, and disrupts the Super Elongation Complex. KAT2A/B also acetylates the SEC itself (ENL, AFF1, AFF3), linking the enzyme to both histone and non-histone regulation of transcriptional elongation. A chromatin-to-condensate axis with a molecule attached and a mechanism-based pan-AML degrader strategy. (3) S. Punzi, I. Villanti, G. Gatti, D. Cittaro, G. Crupi, M. Prunella, N. Altini, G. Casaroli, E. Guerrera, G. F. M. Gallo, C. Felici, O. A. Botrugno, E. Tanzi, V. Bevilacqua, A. Nai, L. Silvestri, and G. Tonon (IRCCS San Raffaele, Milan) examine drug-tolerant persister cells left behind after standard chemotherapy and find ferroptosis is the programmed cell death pathway most deregulated. Platinum-based persisters become highly ferroptosis-sensitive, so pairing platinum with ferroptosis inducers ablates persister CRC cells. Topoisomerase-inhibitor persisters go the opposite direction, raising intracellular iron and becoming ferroptosis-resistant. Mechanistically, topoisomerase inhibitors turn on the Xc− antiporter axis (SLC7A11 + CD44), which pulls in cystine to load GPX4 and dumps glutamate extracellularly. That extracellular glutamate binds NMDA receptors on the same cancer cells, driving Ca²⁺ entry and an AKT/NFE2L2 cytoprotective program; NFE2L2 also increases ER-to-mitochondria distance and reduces mitochondrial ROS. Combined inhibition of the Xc− antiporter and the NMDAR/NFE2L2 axis re-sensitizes persisters to ferroptosis. A mechanistic argument for repurposing existing clinical NMDAR blockers — memantine, ketamine, ifenprodil — into oncology combinations. (4) A. Singhal, K. Ryan, S. Rose, H. Styers, J. Kim, N. Pasnuri, A. Moore, J. Llamosas, E. Chen, J. Adams, A. Nandula, R. Sharma, Z. Li, T. Nawy, Y. Yan, N. Tezcan, O. Basturk, M. H. Sherman, D. Pe'er, and T. Tammela (MSKCC) use lineage tracing and genetic ablation in autochthonous pancreatic ductal adenocarcinoma to interrogate the two dominant cancer cell states — classical and basal. Plasticity is asymmetric: basal cells are highly plastic and can regenerate classical cells, whereas classical cells exhibit limited plasticity. Ablation of basal but not classical cells triggers rapid, durable tumor collapse driven by loss of immunosuppressive cancer-associated fibroblasts, macrophage repolarization, and reprogrammed tumor cytokine milieu, culminating in tumor destruction by cytotoxic lymphocytes. Knockout of a single cytokine — GM-CSF — specifically in basal cells recapitulates macrophage repolarization, lymphocyte recruitment, and tumor shrinkage. A cell-state-defined immunosuppressive circuit maintains PDAC, and GM-CSF is a defined therapeutic node into a tumor type that has resisted essentially every immunotherapy tried against it. All four preprints are v1 posted 2026-07-13/14 on bioRxiv; bodies not yet rendered — summaries are grounded in the full abstracts per PIPELINE.md guidance. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-15-rifamycin-b-rnap-saga-kat2a-protac-nmdar-persister-basal-pdac Wed, 15 Jul 2026 12:00:00 +0000 414 Four bioRxiv preprints from the last 24-48 hours: (1) Mosaei, Shin, Kozhevnikov, Waddell, Hall, Murakami (Penn State), Zenkin (Newcastle) — rifamycin B, the earliest member of the rifamycin family and precursor of clinically used derivatives, previously sidelined for chemical instability and weak antibacterial activity, retains inhibitory activity against RNAP variants carrying clinically relevant rifampicin-resistance mutations. First crystal structure of Rif B on its own and co-structure of Rif B bound to bacterial RNAP show the distinctive C-4 O-carboxymethyl substituent forms an intramolecular contact in the free molecule and re-partners as a salt bridge with fork loop 2 of the RNAP β-subunit upon binding — the additional interaction absent from rifampicin that tolerates the resistance mutations. C-4 is identified as an underexplored handle for rational rifamycin modification. Rif B is redefined as a mechanistically distinct scaffold and a chemistry route back into RNAP for resistant TB and Gram-positive infections. (2) A. Deshpande, Chiang, Perales Garcia, Niranjan, Sinha, Finlay, Stevens, Zahn, Garcia, Jeremias, Wunderlich, Jensen-Pergakes, Udyavar, Carr, Nager, Yang, Murad, Jones, O'Connell, Paul, Vuori, A. J. Deshpande (Sanford Burnham Prebys) — genome-scale dependency data across >1,000 cancer cell lines identifies SAGA histone acetyltransferase as a selective chromatin dependency in AML and hematologic malignancies. Cereblon-recruiting PROTAC (GSK983/GSK699) degrades SAGA catalytic subunits KAT2A + KAT2B; delivers potent broad-spectrum antileukemic activity across genetically diverse AML cell lines, primary patient samples, and an isogenic KMT2A-rearranged model with cooperating oncogenic mutations. Mechanism: KAT2A/B degradation drives a genome-wide reduction in H3K9ac with the largest asymmetric losses at core AML oncogene loci (MYC, MYB, HOXA cluster), displaces ENL from those loci, dissolves ENL-anchored transcriptional condensates, and disrupts the Super Elongation Complex. KAT2A/B also acetylates the SEC itself (ENL, AFF1, AFF3), tying the enzyme to both histone and non-histone regulation of transcriptional elongation. Chromatin-to-condensate axis with a molecule attached, and a mechanism-based pan-AML degrader strategy. (3) Punzi, Villanti, Gatti, Cittaro, Crupi, Prunella, Altini, Casaroli, Guerrera, Gallo, Felici, Botrugno, Tanzi, Bevilacqua, Nai, Silvestri, Tonon (IRCCS San Raffaele, Milan) — cancer persister cells left behind after chemotherapy show ferroptosis as the most deregulated cell death pathway. Platinum persisters are ferroptosis-sensitive (platinum + ferroptosis inducers ablate persister CRC cells). Topoisomerase-inhibitor persisters accumulate intracellular iron and become ferroptosis-resistant. Mechanism: topo inhibitors turn on the Xc− antiporter (SLC7A11 + CD44), loading GPX4 via cystine import and dumping glutamate outside the cell; that glutamate binds NMDA receptors on the same cancer cells, driving Ca²⁺ entry and an AKT/NFE2L2 cytoprotective program; NFE2L2 also increases ER-to-mitochondria distance and reduces mitochondrial ROS. Combined inhibition of Xc− + NMDAR/NFE2L2 re-sensitizes persisters to ferroptosis — mechanistic argument for repurposing existing clinical NMDAR blockers (memantine, ketamine, ifenprodil) into oncology combinations. (4) Singhal, Ryan, Rose, Styers, Kim, Pasnuri, Moore, Llamosas, Chen, Adams, Nandula, Sharma, Li, Nawy, Yan, Tezcan, Basturk, Sherman, Pe'er, Tammela (MSKCC) — lineage tracing + genetic ablation in autochthonous PDAC show basal cancer cells are highly plastic (regenerate classical cells) while classical cells exhibit limited plasticity. Ablation of basal but not classical cells triggers rapid + durable tumor collapse driven by loss of immunosuppressive CAFs, macrophage repolarization, and reprogrammed tumor cytokine milieu, culminating in tumor destruction by cytotoxic lymphocytes. Knockout of a single cytokine — GM-CSF — specifically in basal cells recapitulates macrophage repolarization + lymphocyte recruitment + tumor shrinkage. A cell-state-defined immunosuppressive circuit maintains PDAC, and GM-CSF is a defined therapeutic node into a tumor type that has resisted essentially every immunotherapy tried against it. false Four bioRxiv preprints from the last 24-48 hours spanning a de novo miniprotein platform for kinase modulation outside the ATP site, cryo-EM resolution of a decade-old chemical biology mystery, a cross-indication proteostasis therapeutic node with an in-vivo-active chemical probe, and structure-guided repurposing of a bacterial retron as a designer antimicrobial. (1) M. S. Bauer, G. R. Lee, B. Coventry, K. A. Klupt, A. M. Fernández-Escamilla, S. Kumar, M. S. Donald Paladino, D. Li, M. Glögl, D. Lietha, E. Muratspahić, T. Schlichthärle, X. Wang, L. Schmiderer, S. Kenny, B. Faezov, W. Chen, A. F. Shida, Y. Hsia, R. D. Kibler, M. B. Elowitz (Caltech + HHMI), B. Nabet (Fred Hutchinson), and D. Baker (Institute for Protein Design, UW + HHMI) attack kinase selectivity by targeting the kinase domain itself with de novo designed genetically encoded miniproteins. From 96 binders designed to stabilize distinct conformational states of focal adhesion kinase (FAK), 33 modulate activity; biochemical characterization of the four most potent shows two low-nanomolar FAK inhibitors and two designs that potentiate FAK activity by more than two-fold — activation is a design objective, not a byproduct. When expressed inside cells, the same designs recapitulate both inhibition and activation, showing conformational-state binders can directly tune kinase signaling in living cells. Kinase-domain conservation is reframed as leverage rather than obstacle: the FAK inhibitor scaffolds are redesigned against Src, demonstrating retargetability. The platform delivers three things that ATP-competitive small-molecule discovery does not — an ATP-site-orthogonal way to target kinases, on-demand activator design, and a discovery route to novel modulatory sites on the kinase domain. (2) J. A. Lee, X. Gu, C. Chan, V. S. Robertson, V. Garcia-Ruiz, Y. E. Li, A. H. Ngo, P. Alabi, V. Denic (Harvard), J. K. Sello, and W. Clemons (Caltech) solve the long-standing mystery of Retro-1 and Retro-2, structurally distinct small molecules that have served for a decade as chemical biology probes for retrograde trafficking blockade without a known target. Both compounds target Get3, the ATPase chaperone of the guided entry of tail-anchored proteins (GET) pathway that inserts tail-anchored SNAREs required for retrograde transport into the ER membrane. Cryo-EM shows both compounds occupy a cryptic pocket in Get3 and allosterically trap it in a stalled complex with upstream pathway components. Two implications matter. First, the GET pathway is now a validated therapeutic vulnerability for pathogen and toxin entry — a druggable node no program has been aimed at deliberately. Second, this is a clean example of complex-stabilization pharmacology, sibling to molecular glues and allosteric splice modulators: the drug does not block a site, it glues a productive machine into a nonproductive intermediate. A cryptic pocket, a co-structure, and two structurally distinct starting hits make the medicinal-chemistry path unusually well-lit. (3) M. Riedl Khursigara, A. C. Goss, M. Kost-Alimova (Broad Institute), K. Keller, C. D. De Mata, R. Muraleedharan, E. R. Collantes, M. Brown, E. Grinkevich, F. M. Arines, J. Lin, P. Byrne, S. Bazua Valenti, et al. with senior author A. Greka (Broad) and Massachusetts Eye and Ear collaborators identify a hetero-oligomeric cargo receptor complex — TMED7 paired with TMED2, 9, and 10 — that acts as an entrapment node holding structurally and functionally diverse mutant clients within the post-ER secretory pathway via TMED7 binding to the integral Golgi protein GRASP55. The reframe: proteinopathies as different as autosomal-dominant tubulointerstitial kidney disease and rhodopsin-driven retinitis pigmentosa are not usually thought of as sharing a druggable mechanism, but this work argues they do, because the same TMED entrapment complex is what holds toxic mutant cargoes in place long enough to damage the cell. Genetic ablation of the entrapment node clears accumulated cargo. The small molecule BRD7635 does the same pharmacologically, reversing histopathological hallmarks and rescuing function in vivo in both a kidney proteinopathy mouse model and a retinitis pigmentosa mouse model, including measurable vision preservation. An actionable proteostasis target with an in-vivo-active chemical probe operating across two clinically distinct indications. (4) G. N. Hibshman, L. Wang, N. MacRae, K. Zhang, A. Florez, S. Shipman (Gladstone Institutes + UCSF), and E. Nogales (UC Berkeley + HHMI) determine the cryo-EM structure and activation mechanism of retron-Kva2, a type IX retron from the human pathogen Klebsiella variicola. Retron-Kva2 assembles into an asymmetric higher-order ribonucleoprotein complex that sequesters a toxic dimeric HEPN RNase at its core. They identify the natural phage trigger as T5 protein D5, which activates the retron through structural mimicry — its helix-turn-helix fold mimics the retron's own winged-helix protein, binding the msDNA sensor and conformationally remodeling the assembly to release HEPN-mediated tRNA cleavage and growth arrest. The critical insight is that retron-Kva2 surveils a structural fold (not a primary sequence) via msDNA binding, providing a broadly exploitable pathway for programmable activation. They computationally design de novo synthetic triggers — unrelated in sequence to D5 — that activate retron-Kva2-mediated bacterial growth arrest in vivo. Bacterial defense is repurposed into a structure-guided precision antimicrobial platform, a class where new mechanisms are badly needed. All four preprints are v1 posted 2026-07-13 on bioRxiv; bodies not yet rendered — summaries are grounded in the full abstracts per PIPELINE.md guidance. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-14-baker-denovo-kinase-get3-retro-tmed-proteinopathy-retron-antimicrobial Tue, 14 Jul 2026 12:00:00 +0000 342 Four bioRxiv preprints from the last 24-48 hours: (1) Bauer, Lee, Coventry, Klupt, Fernández-Escamilla, Kumar, Donald Paladino, Li, Glögl, Lietha, Muratspahić, Schlichthärle, Wang, Schmiderer, Kenny, Faezov, Chen, Shida, Hsia, Kibler, Elowitz (Caltech), Nabet (Fred Hutchinson), Baker (IPD/UW) — de novo designed genetically encoded miniproteins that bind the FAK kinase domain and toggle between low-nanomolar inhibition and >2× activation of the same enzyme. 96 designs stabilizing distinct FAK conformational states → 33 modulate activity; 4 most potent characterized biochemically show 2 low-nanomolar inhibitors + 2 >2-fold activators; cellular expression preserves both effects in situ. Kinase-domain conservation is leveraged rather than fought — FAK inhibitor scaffolds redesigned against Src as demonstration. Establishes a genetically encoded, ATP-site-orthogonal platform for kinase control + on-demand activator design + a discovery route to novel modulatory surfaces on the kinase domain. (2) Lee, Gu, Chan, Robertson, Garcia-Ruiz, Li, Ngo, Alabi, Denic (Harvard), Sello, Clemons (Caltech) — cryo-EM mechanism of Retro-1 and Retro-2, structurally distinct decade-old chemical-biology probes for retrograde trafficking with an unresolved target. Both compounds bind a cryptic pocket in Get3, the ATPase chaperone of the GET pathway for tail-anchored protein insertion (including retrograde-required SNAREs), allosterically trapping Get3 in a stalled complex with upstream pathway components. Establishes the GET pathway as an unsuspected pharmacological vulnerability for pathogen/toxin entry, validates complex-stabilization as a therapeutic strategy (molecular-glue parallel), and delivers cryptic pocket + co-structure + 2 structurally distinct starting hits for medicinal chemistry. (3) Riedl Khursigara, Goss, Kost-Alimova (Broad), et al., senior author Anna Greka (Broad), with Massachusetts Eye and Ear collaborators — TMED7-TMED2/9/10 hetero-oligomeric cargo receptor entrapment complex is a shared therapeutic node across genetically and clinically distinct proteinopathies. TMED7 tethers structurally/functionally diverse mutant secretory clients to Golgi protein GRASP55, holding them in the post-ER secretory pathway long enough to cause disease. Genetic disruption or the small molecule BRD7635 clear accumulated cargo, reverse histopathology, and rescue function in vivo in both a kidney tubulointerstitial proteinopathy mouse model and a retinitis pigmentosa mouse model (with vision preservation). Actionable proteostasis target with in-vivo-active chemical probe operating across two clinically distinct indications. (4) Hibshman, Wang, MacRae, Zhang, Florez, Shipman (Gladstone/UCSF), Nogales (UC Berkeley + HHMI) — cryo-EM structure + activation mechanism of retron-Kva2, a type IX retron from the human pathogen Klebsiella variicola. Retron assembles into asymmetric higher-order RNP complex sequestering toxic dimeric HEPN RNase at core; natural phage trigger identified as T5 protein D5, which activates the retron through structural mimicry — its helix-turn-helix fold mimics the retron's own winged-helix protein to bind the msDNA sensor and conformationally remodel the assembly to release HEPN-mediated tRNA cleavage and growth arrest. Recognition is of structural fold (not primary sequence) via msDNA binding — providing a broadly exploitable pathway for programmable activation. Computationally designed de novo synthetic triggers (unrelated in sequence to D5) activate retron-Kva2-mediated bacterial growth arrest in vivo — bacterial defense repurposed into a structure-guided precision antimicrobial platform. false Four bioRxiv preprints from the last 24-48 hours spanning a T-cell-directed CAR-T strategy for rheumatoid arthritis, structural biology of a new-old druggable metabolic target, a crisp negative result on multi-state AI structure prediction, and a mechanism-grounded organ-on-chip comparison of two calcineurin inhibitors. (1) C. Gur, L. Ravkaie, R. Sharet-Eshed, R. Shalita, R. Avellino, E. Rauchbach, K. Xie, E. David, G. Yagel, M. Zada, M. Ben Yehuda, K. Mazuz, M. N. Von Locquenghien, H. Peleg, Y. Naparstek, K. Atlan, S. Kfir-Erenfeld, Y. Kuznetsov, R. Tzemach, M. Lidar, A. Balbir-Gurman, T. S. Phan, K. Freitag, and I. Amit (Weizmann Institute) apply single-cell multi-omics to human RA synovium and identify PDCD1 (the gene for PD-1) as a selective marker of the pathogenic disease-associated T-cell state — the T-cell arm that B-cell-directed CD19 CAR-Ts leave untouched. They build a PD-1-directed CAR that potently eliminates these cells in vitro and in vivo with marked attenuation of synovitis in RA models, and layer on two engineering choices that carry the safety story. First, CAR activity is gated by an NR4A2-driven biosensor that switches the cells on only inside inflamed synovium — a tissue-restriction circuit rather than a target-restriction circuit, which is the right frame because PD-1 is also expressed on healthy antigen-experienced T cells throughout the body. Second, the cells are armored to secrete soluble TNF receptor II locally, absorbing the TNF-α that drives baseline synovial inflammation, dampening the interferon response that CAR engagement itself triggers, and pushing the tissue myeloid compartment toward a reparative state. This is the plausible T-cell CAR-T complement to B-cell-directed autoimmune CAR-Ts, and the biosensor-gated tissue restriction is a genuinely different concept from just picking a synovium-specific antigen. (2) D. B. Sauer, J. Song, J. J. Marden, B. Wang, K. Sowerby, J. C. Sudar, W. J. Rice, and D.-N. Wang (NYU School of Medicine) determine the human NaCT (SLC13A5) sodium-citrate cotransporter in three states by single-particle cryo-EM: sodium-free, sodium-bound, and sodium-plus-substrate-bound. NaCT imports tricarboxylate and dicarboxylate TCA-cycle intermediates driven by the inward sodium gradient, and it is a genetically validated target — loss-of-function mutations cause an autosomal-recessive epileptic encephalopathy, and pharmacologic inhibition has been pursued for metabolic disease and aging. The mechanistic result is that in human NaCT, sodium and substrate bind simultaneously, distinct from the sequential-binding, conformational-selection mechanism established for the bacterial DASS homolog VcINDY. For orthosteric inhibitor design that human-vs-bacterial mechanistic split matters — the pharmacophore inherited from VcINDY crystal structures does not necessarily transfer, and three cryo-EM states of the human protein with and without cargo is what a serious NaCT medicinal-chemistry campaign has been waiting for. (3) M. Ye, Y.-H. Wang, M. Brogi, J. M. Parks (Oak Ridge), K. M. Kuo, and J. C. Gumbart (Georgia Tech) run a careful head-to-head of AlphaFold-3, Boltz-2, Chai-1, and BioEmu on four textbook multi-state proteins (Pf-MATE transporter, LAO binding protein, SecA translocation ATPase, β2 adrenergic receptor), scoring state-bias and sampling breadth against experimental references. Two findings hit hard: all four models collapse toward whichever conformer is best-represented in the PDB, and small-molecule ligands have weak and inconsistent effects on which state gets predicted, while large protein binding partners do drive clear conformational switching. MSA-clustering workarounds (AF-Cluster and random subsampling) inherit the same bias, so this is not a fix one architecture away. For AI-guided structure-based drug discovery this is a crisp negative result: current predictors are still not the right primary tool for choosing which conformational state to dock into, and the field cannot pretend a ligand modulates conformer selection when the models don't act like it does. (4) K. S. Aryeh, Y. P. Tsang, E. W. Hsu, C. K. Yeung, J. MacDonald, T. K. Bammler, J. Himmelfarb, L. M. Rehaume, and E. J. Kelly (University of Washington, with Aurinia) run cyclosporine A against voclosporin — the calcineurin inhibitor FDA-approved for lupus nephritis — in both 2D human proximal-tubule epithelial monolayers and a perfused 3D kidney microphysiological system. The 2D readout is silent (neither drug kills the cells outright). The perfused 3D chip is where they separate. Confocal tomography shows cyclosporine fragments mitochondria while voclosporin preserves reticular architecture; TMRM flow cytometry confirms voclosporin preserves mitochondrial membrane potential relative to cyclosporine. RNA-seq finds cyclosporine enriches for unfolded-protein response and ER stress, p21-driven G2/M checkpoint arrest, and — the striking piece — a transcriptional signature consistent with ferroptosis priming; voclosporin instead induces adaptive ER chaperone and ERAD programs without tripping the canonical UPR sensors and does not suppress the cell cycle. Two things to take home: an MPS with mechanistic readouts catches sublethal tubular injury that 2D culture and even a KIM-1 biomarker miss, and calcineurin inhibitors are not interchangeable at the mitochondrial-and-proteostatic-stress layer. All four preprints are v1 posted 2026-07-11 through 2026-07-12 on bioRxiv; bodies not yet rendered — summaries are grounded in the full abstracts per PIPELINE.md guidance. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-13-pd1-car-t-arthritis-nact-cotransporter-af3-multistate-voclosporin-mps Mon, 13 Jul 2026 12:00:00 +0000 357 Four bioRxiv preprints from the last 24-48 hours: (1) Gur, Ravkaie, Sharet-Eshed, Shalita, Avellino, Rauchbach, Xie, David, Yagel, Zada, Ben Yehuda, Mazuz, Von Locquenghien, Peleg, Naparstek, Atlan, Kfir-Erenfeld, Kuznetsov, Tzemach, Lidar, Balbir-Gurman, Phan, Freitag, Amit (Weizmann) — synovium-restricted armored PD-1-targeted CAR-T cells for rheumatoid arthritis. Single-cell multi-omics of human RA identifies PDCD1 (PD-1) as a selective marker of pathogenic synovial disease-associated T cells — the T-cell arm that CD19 CAR-Ts leave untouched. PD-1-directed CAR eliminates these cells in vitro and in vivo with marked attenuation of synovitis in RA models. Two engineering choices carry the safety story: NR4A2-driven biosensor gates CAR activity to inflamed synovium (tissue-restriction rather than target-restriction, correct because PD-1 is also on healthy antigen-experienced T cells throughout the body), and cells are armored to secrete soluble TNF receptor II locally to absorb baseline TNF-α, dampen the interferon response CAR engagement triggers, and push the myeloid compartment toward a reparative state. Plausible T-cell CAR-T complement to B-cell-directed autoimmune CAR-Ts. (2) Sauer, Song, Marden, Wang, Sowerby, Sudar, Rice, Wang (D.-N. Wang, NYU School of Medicine) — cryo-EM structures of human sodium-citrate cotransporter NaCT (SLC13A5) in three states (sodium-free, sodium-bound, sodium+substrate-bound) showing sodium and substrate bind simultaneously rather than sequentially. Distinct from the sequential-binding, conformational-selection mechanism established for the bacterial DASS homolog VcINDY. Matters because NaCT is genetically validated (loss-of-function causes autosomal-recessive epileptic encephalopathy; pharmacologic inhibition pursued for metabolic disease and aging) and human-vs-bacterial mechanistic split means the VcINDY pharmacophore does not necessarily transfer to human NaCT orthosteric design. (3) Ye, Wang, Brogi, Parks (Oak Ridge), Kuo, Gumbart (Georgia Tech) — head-to-head benchmark of AlphaFold3, Boltz-2, Chai-1, and BioEmu on four canonical multi-state proteins (Pf-MATE, LAO, SecA, β2AR). All four models collapse toward whichever conformer is best-represented in the PDB; small-molecule ligands have weak and inconsistent effects on state selection while large protein binding partners drive clear conformational switching. MSA-clustering workarounds (AF-Cluster and random subsampling) inherit the same bias — not a fix one architecture away. Crisp negative result: current AI predictors are still not the right primary tool for choosing which conformational state to dock into. (4) Aryeh, Tsang, Hsu, Yeung, MacDonald, Bammler, Himmelfarb, Rehaume, Kelly (University of Washington, with Aurinia) — voclosporin vs cyclosporine A in perfused human proximal-tubule microphysiological system. 2D readout silent (neither drug kills cells outright). 3D perfused chip separates them at the mechanism level — confocal tomography shows CsA fragments mitochondria while VCS preserves reticular architecture, TMRM flow cytometry confirms VCS preserves ΔΨm relative to CsA, RNA-seq finds CsA enriched for UPR/ER stress + p21-driven G2/M arrest + ferroptosis-priming transcriptional signature while VCS induces adaptive ER chaperone + ERAD programs without tripping canonical UPR sensors. MPS with mechanistic readouts catches sublethal tubular injury 2D and KIM-1 miss; calcineurin inhibitors not interchangeable at the mitochondrial-and-proteostatic-stress layer. false Four bioRxiv preprints from the last 24-48 hours spanning a new GPCR target for type-1 diabetes, a modular base-editor precision engineering strategy, a chem-biology-through-CRISPR differentiation-therapy target for AML, and the first wet-lab-validated quantum-classical generative peptide design. (1) N. A. Haq, K. W. Toczyska, A. Islam, O. E. Olaniru, Y. Lei, M. Hu, M. Zhao, R. Müller, M. K. M. Mirza, N. H. Fine, D. J. Hodson, S. J. Persaud (King's College London), A. G. Beck-Sickinger (Leipzig), J. Pearson, and G. A. Bewick address the central limitation of current type-1 diabetes disease-modifying therapy — that teplizumab and JAK-1/2 inhibitors like baricitinib buy time by dampening the immune attack without making β-cells themselves harder to kill. Using RNAscope, cell sorting, and a TAMRA-labeled derivative of the long-acting selective Y4R agonist K22, they establish that among the four neuropeptide Y receptor subtypes, Y4R is enriched at the β-cell surface in both mouse and human islets. K22 has no effect on baseline glucose-stimulated insulin secretion or systemic glucose tolerance — the receptor is silent under healthy conditions and reveals itself only under stress. Under cytokine, streptozotocin, lipotoxic, and ER stress insult, Y4R agonism reduces caspase-3/7 activation, sustains insulin release, and promotes proliferation, while switching on a coordinated β-cell resilience program: KEAP1-NRF2-driven antioxidative and proteostatic activation, suppression of EIF2 signalling and downstream biosynthetic/ER-stress pathways, and reinforced β-cell identity and insulin processing. In parallel, Y4R activation damps pathogenic chemokines (CXCL10, CCL3/4/7) and IL-6 in stressed islets while preserving IL-2 and Foxp3 signals, and blocks activated human PBMC migration/invasion toward cytokine-stressed human islets. In the stringent NY8.3 CD8+ T-cell adoptive-transfer model of autoimmune diabetes, systemic K22 significantly delays disease onset. Y4R emerges as a β-cell-centric GPCR target that couples intrinsic resilience with local immune modulation — the complementary strategy the field has wanted alongside immune-directed drugs, and directly relevant to islet-replacement therapies as well. (2) M. M. Müller, N. T. Southern, and D. Niopek (Heidelberg) attack the persistent bystander-editing problem of the high-activity TadA8e adenine base editor with a design that is orthogonal to the prevailing deaminase-mutation strategy. The prevailing fix — weakening the deaminase with point mutations — trades on-target activity for precision. This work instead asks whether physically constraining the deaminase's spatial reach, by inserting a bulky autonomously folding protein domain inside TadA, can preserve on-target catalysis while sterically blocking access to bystander adenines in the exposed R-loop ssDNA. ProDomino computational prediction plus structure-guided site selection identifies multiple accepting sites; L68 is standout, accepting LOV2, a PDZ domain, or superfolder-GFP with high retained editing activity — and critically, the insert identity barely matters, meaning window narrowing is driven by position not domain, i.e., a purely structural constraint. Across 5 endogenous loci, the L68-sfGFP variant preserves near-wild-type editing at the central target adenine while collapsing the effective editing window from positions A2-A12 down to A4-A8, and in an orthogonal R-loop assay nearly eliminates Cas-independent off-target deamination in trans (a class of off-target editing that has resisted deaminase-only optimization). Because sfGFP folds correctly inside the deaminase, one variant delivers narrowed on-target window, suppressed Cas-independent off-targets, and a live-cell fluorescent tag simultaneously. Internal domain-insertion becomes a modular precision knob for CRISPR base editors that is cleanly separable from — and stackable with — deaminase-tuning approaches, and the strategy transfers directly to other CRISPR effectors. (3) S. M. Amos, C.-C. Chen, Y. Xiang, K. Motoyama, T. Gonzalez-Robles, V. Narendra, G. Johnson, H. T. Lee, Y.-J. Ho, I. Celikoyar, Z. Ye, S. Guo, C. Glickman, N. O'Hearn, O. Sarkar, A. Arroyo-Ortega, T. Devine, M. J. Pagano, K. Ruggles, F. J. Sanchez-Rivera, A. N. Koehler (Broad Ludwig Center), S. W. Lowe (Memorial Sloan Kettering + HHMI), and Y. M. Soto-Feliciano (MIT) run a focused CRISPR screen in acute leukemia and land on TRIM28, a multi-domain chromatin adaptor best known as a KRAB-zinc-finger co-repressor. Depletion impairs leukemia proliferation in vitro and in vivo, and — more strikingly — activates neutrophil differentiation programs, driving leukemia cells toward mature, functionally neutrophil-like states with reduced leukemic potential. Integrative transcriptomic and chromatin profiling place TRIM28 as a co-repressor of neutrophil-associated loci independently of H3K9 methylation, a departure from its canonical KRAB-ZNF/SETDB1 axis. The chem-biology payoff, and what makes this a Scripps-relevant read, is that the group developed a selective small-molecule inhibitor that binds the TRIM28 PHD-bromodomain (US patent application filed), phenocopies genetic knockdown across biochemical and cellular assays, has low-micromolar anti-leukemia activity, induces neutrophil differentiation, and synergizes with menin inhibitors already in the clinic. Target discovery, mechanism of action, and a first-generation chemical probe delivered in one preprint — a differentiation-therapy hypothesis with a molecule attached, ready for medicinal-chemistry optimization. (4) E. S. Engdal, J. Funk, O. Bacarreza, L. Machado, K. H. Johansen, J. Kemming, T. Farnsworth, V. Brasas, R. Y. L. Lefevre-Morand, M. Slysz, O. L. Noerregaard, O. A. D. A. Sandberg, A. Makarovskiy, P. Lodahl, C. G. Acevedo-Rocha (HERVolution Therapeutics), K. Kurowski, S. R. Hadrup (Technical University of Denmark), W. R. Clements (ORCA Computing), and T. Jenkins (Sparrow Quantum) present the first end-to-end hybrid quantum-classical pipeline for de novo design of MHC class I-binding peptides: a generative adversarial network coupled to latent-space priors sampled from a real photonic quantum processor. The premise is that the default factorized-Gaussian prior in classical generative models under-explores biomolecular sequence space, and structured non-classical distributions from quantum hardware can serve as an alternative inductive bias — one that concentrates gains where classical training data is thinnest. Across 131 HLA class I alleles evaluated in silico, quantum-derived priors increased the yield of predicted strong binders, with the largest relative gains on the understudied alleles where classical baselines perform worst — the regime that matters most for personalized immunotherapy and vaccine design in patients whose HLA types are under-served by existing training data. Mechanistically, gains coincide with broader sequence exploration at non-anchor positions while anchor specificity is preserved, consistent with a targeted expansion of accessible sequence diversity. On three of those understudied alleles, the team took computational designs into peptide-MHC stability ELISAs and confirmed that quantum-designed peptides are potent stabilisers of the pMHC class I complex — the first wet-lab validation of quantum-primed generative design on a real therapeutic-adjacent task, and the effect concentrates precisely where new inductive biases should matter, i.e., the small-data corner where deep generative models tend to memorize rather than explore. All four preprints are v1 posted 2026-07-10 on bioRxiv; the Y4R T1D and ABE-domain-insertion bodies are fully rendered, while the TRIM28 and quantum-classical MHC-peptide bodies remain abstract-only per bioRxiv's rendering lag — summaries labeled accordingly per PIPELINE.md. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-12-y4r-t1d-abe-domain-insert-trim28-aml-quantum-mhc-peptide Sun, 12 Jul 2026 12:00:00 +0000 419 Four bioRxiv preprints from the last 24-48 hours: (1) Haq, Toczyska, Islam, Olaniru, Lei, Hu, Zhao, Müller, Mirza, Fine, Hodson, Persaud (King's College London), Beck-Sickinger (Leipzig), Pearson, Bewick — selective neuropeptide Y4 receptor (Y4R) agonism as new disease-modifying target for type-1 diabetes. RNAscope, cell sorting, and TAMRA-K22 fluorescent-ligand competition localize Y4R to β-cell surface in mouse and human islets; the receptor is silent in health (no effect on baseline GSIS or systemic glucose tolerance). Under cytokine, streptozotocin, lipotoxic, and ER stress, long-acting selective Y4R agonist K22 activates a coordinated β-cell resilience program (KEAP1-NRF2 antioxidative activation, EIF2-arm suppression, reinforced β-cell identity and insulin processing), damps CXCL10/CCL3/4/7/IL-6 while preserving IL-2 and Foxp3, blocks activated human PBMC migration/invasion toward stressed human islets, and significantly delays disease onset in the NY8.3 CD8+ T-cell adoptive-transfer model. First in-vivo evidence for a β-cell-centric GPCR that pairs intrinsic cytoprotection with local immunomodulation — a complement to teplizumab that acts on the target cell rather than the immune system alone. (2) Müller, Southern, Niopek (Heidelberg) — internal domain-insertion base editor engineering. The high-activity TadA8e adenine base editor tolerates internal insertion of bulky autonomously folding protein domains at multiple ProDomino-predicted and structure-guided sites, with position (not domain identity) governing outcome. Insertion at L68 with LOV2, PDZ, or superfolder-GFP preserves near-wild-type editing at the central target adenine across 5 endogenous loci, collapses the effective editing window from A2-A12 down to A4-A8, and in an orthogonal R-loop assay nearly eliminates Cas-independent trans off-target deamination — an off-target class that has resisted deaminase-only optimization. The sfGFP variant simultaneously delivers a fluorescent, trackable editor. Internal domain-insertion is a modular precision knob orthogonal to and stackable with deaminase-mutation approaches, immediately transferable to other CRISPR effectors. (3) Amos, Chen, Xiang (Soto-Feliciano lab MIT) with Scott Lowe (MSK + HHMI) and Angela Koehler (Broad Ludwig) — TRIM28 as pro-differentiation therapeutic target in acute myeloid leukemia. Focused CRISPR screen identifies TRIM28 dependency in acute leukemia; depletion impairs proliferation in vitro and in vivo, activates neutrophil differentiation programs, and drives leukemia cells toward mature, functionally neutrophil-like states with reduced leukemic potential. Integrative transcriptomic and chromatin profiling place TRIM28 as a co-repressor of neutrophil-associated loci independently of H3K9 methylation — a departure from its canonical KRAB-ZNF axis. The group develops a selective small-molecule inhibitor of the TRIM28 PHD-bromodomain (US patent application filed) that phenocopies the genetic knockdown across biochemical and cellular assays, has low-micromolar anti-leukemia activity, induces neutrophil differentiation, and synergizes with menin inhibitors already in the clinic. Target discovery, mechanism, and first-generation chemical probe in one preprint — a differentiation-therapy hypothesis with a molecule attached. (4) Engdal, Funk, Bacarreza, Machado, Johansen, Kemming, Farnsworth, Brasas, Lefevre-Morand, Slysz, Noerregaard, Sandberg, Makarovskiy, Lodahl, Acevedo-Rocha (HERVolution Therapeutics), Kurowski, Hadrup (Technical University of Denmark), Clements (ORCA Computing), Jenkins (Sparrow Quantum) — first end-to-end hybrid quantum-classical generative pipeline for de novo MHC class I-binding peptide design. A generative adversarial network is coupled to latent-space priors sampled from a real photonic quantum processor, providing a structured non-classical distribution as an alternative inductive bias to the default factorized-Gaussian prior in classical generative models. Across 131 HLA class I alleles evaluated in silico, quantum-derived priors increased the yield of predicted strong binders, with the largest relative gains on understudied alleles where classical baselines perform worst; gains coincide with broader sequence exploration at non-anchor positions while anchor specificity is preserved. On three of those understudied alleles, quantum-designed peptides were validated in vitro via peptide-MHC stability ELISAs as potent stabilizers of the pMHC class I complex — the first wet-lab validation of quantum-primed generative design on a real therapeutic-adjacent task, with the effect concentrated precisely in the small-data regime that matters most for personalized immunotherapy and vaccine design. false Four bioRxiv preprints from the last 24-48 hours spanning a new gene-therapy-delivered targeted protein degradation modality, a modern biomarker-guided repositioning of an abandoned DNA-damaging prodrug, drug-discovery methods against a pediatric neurotropic virus, and a target-biology reframe of a hot epitranscriptomic writer. (1) J. L. Yang, K. Y. Loh, C. R. Sandoval Espinoza, D. Schuster, K. Deisseroth, and C. R. Bertozzi (Stanford) rebuild the LYTAC concept as a genetically encoded protein — murine IGF2 fused via a 20-aa glycine-serine linker to an IgG-binding nanobody, engaging the lysosomal-trafficking CI-MPR/IGF2R receptor to route IgG for lysosomal degradation. Packaged in AAV8, a single dose gives ~77-day sustained expression and ~85-89% depletion of endogenous IgG1/IgG2b/IgG2c by mass spectrometry, with disease benefit in a passive-transfer rheumatoid-arthritis model and no detectable anti-GELYTAC antibodies. Two engineering flourishes matter for the platform: (i) a Granzyme-B-cleavable conditional variant restricts activity to sites of cytotoxic-lymphocyte activation, and (ii) an LMI070/branaplam-responsive Xon splice-switch cassette provides on-demand small-molecule control over expression. Contrasted explicitly with Biohaven's weekly-dosed GalNAc-peptide BHV-1300 and Lycia's catalytic LYTACs (IgE rebounds by ~21 days), this is a serious argument that the durability problem in extracellular targeted protein degradation is solvable genetically rather than pharmacologically. (2) J. L. Elia, J. Hill, C. D. Heer, S. Smolev, A. M. Sykes, S. R. Arbelaez, K. N. Lucas, S. S. Johnson, R. K. Sundaram, S. B. Herzon, and R. S. Bindra (Yale) run a focused screen of DNA-damaging agents through isogenic HR+/HR- cell pairs and find that CB1954, a nitrobenzamide-aziridine prodrug abandoned decades ago, is highly selective for BRCA2-deficient tumor cells. It forms DNA interstrand crosslinks regardless of HR status, but selectively drives DNA-damage signaling, apoptosis, and clonogenic-survival loss in HR-deficient cells. A targeted DDR CRISPR screen and isogenic validation localize the operative dependency to the Fanconi anemia and HR pathways, not MMR/NER. A bifurcated mechanism emerges: NQO2-dependent bioactivation drives the HRD-selective cytotoxicity, while intrinsic aziridine reactivity accounts for residual non-selective activity — meaning selectivity can be dialed in medicinal-chemically. Favorable preclinical PK and genotype-dependent antitumor activity in BRCA2-deficient xenografts. The broader lesson is that the pre-biomarker DNA-damaging-agent back catalogue is worth re-screening against modern isogenic models. (3) I. Biswas, Q. Wang, J. T. McCann, E. P. Tchesnokov, L. Nguyen, M. Saini, J. Cantero, J. L. Revalde, M. Gotte, A. Renslo, R. J. Neitz, M. R. Arkin (UCSF), E. Arnold, and F. X. Ruiz (Rutgers) run parallel crystallographic-fragment and biochemical HTS campaigns on the EV-D68 3D polymerase — a target for a pediatric neurotropic virus with no approved antivirals or vaccine. The 650-fragment CFS returns ~10% hits distributed across the RNA template channel, active site, and RNA primer channel, plus two previously undescribed ligandable pockets — Thumb Site II and the Index-Middle Finger Pocket — with inhibition confirmed for compounds binding each site. A PicoGreen fluorescence HTS of 50,000 ChemBridge compounds (0.77% hit rate) surfaces a 5-aminoindazole scaffold, with hit-to-lead compound 727590 at IC50 25 µM and defined SAR. A dual-modality ligand map plus a screen-derived scaffold is a textbook discovery-program opening on a target that had none. (4) R. J. Kostlan, J. T. Phoenix, A. Budreika, M. G. Ferrari, C. F. Deegan, E. T. Warren, P. S. Bawa, C. S. Rogers, D. Dureja, M. Ali, G. R. Hancock, K. S. Young, G. Gupta, A. Solanki, D. J. Vander Griend, S. W. Fanning, and S. Kregel (Loyola Chicago) show that METTL3 sustains AR-driven prostate cancer through a non-enzymatic scaffold function. Genetic depletion of METTL3 blocks proliferation of AR+ lines while sparing untransformed prostate epithelial cells, yet pharmacologic METTL3 catalytic inhibitors reduce global mRNA m6A without affecting proliferation. A catalytically dead METTL3 mutant rescues knockdown as well as wild-type. Mechanistically, METTL3 directly interacts with AR, co-localizes with it on chromatin, and knockdown reduces AR cistromic occupancy. The clinical implication is clean: the correct modality for METTL3 in this setting is a scaffold-degrader (PROTAC or molecular glue) — not the STC-15-class catalytic inhibitors currently in trials. All four preprints are v1 posted 2026-07-09 through 2026-07-10 on bioRxiv; the Bertozzi GELYTAC paper has full body text, the other three summaries are abstract-grounded per PIPELINE.md. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-11-aav-gelytac-cb1954-hrd-ev-d68-3dpol-mettl3-ar-scaffold Sat, 11 Jul 2026 12:00:00 +0000 350 Four bioRxiv preprints from the last 24-48 hours: (1) Yang, Loh, Sandoval Espinoza, Schuster, Deisseroth, Bertozzi (Stanford) — genetically encoded AAV-delivered LYTACs (GELYTACs). Fusion of murine IGF2 (engages CI-MPR/IGF2R lysosomal-trafficking receptor) and an IgG-binding nanobody, packaged in AAV8 for continuous in-vivo expression. Single dose gives ~85-89% depletion of endogenous IgG1/IgG2b/IgG2c by mass spec ~77 days post treatment, disease benefit in passive-transfer rheumatoid arthritis, and no detectable anti-GELYTAC antibodies. Two engineering flourishes: Granzyme-B-cleavable variant restricts activity to sites of cytotoxic-lymphocyte activity, and an LMI070/branaplam-responsive Xon splice-switch cassette gives on-demand small-molecule control over expression. Contrasts with Biohaven's weekly-dosed BHV-1300 GalNAc-peptide degrader and Lycia's catalytic LYTACs (IgE rebounds ~21 days) — argues durability in extracellular TPD is solvable genetically rather than pharmacologically. (2) Elia, Hill, Heer, Smolev, Sykes, Arbelaez, Lucas, Johnson, Sundaram, Herzon, Bindra (Yale) — CB1954, a nitrobenzamide-aziridine prodrug abandoned decades ago, is a BRCA2-selective agent identified by biomarker-guided isogenic HR+/HR- screen. Forms DNA interstrand crosslinks independent of HR status but selectively drives DDR signaling + apoptosis + clonogenic-survival loss in HR-deficient cells; targeted DDR CRISPR screen localizes dependency to FA/HR pathways not MMR/NER; bifurcated mechanism — NQO2-dependent bioactivation drives HRD-selective cytotoxicity, aziridine reactivity drives residual non-selective activity — medicinally tunable. Genotype-dependent activity in BRCA2-deficient xenografts. Vindicates re-screening the pre-biomarker DNA-damaging agent back catalogue. (3) Biswas, Wang, McCann, Tchesnokov, Nguyen, Saini, Cantero, Revalde, Gotte, Renslo, Neitz, Arkin (UCSF), Arnold, Ruiz (Rutgers) — parallel CFS + HTS on EV-D68 3Dpol. 650-fragment crystallographic screen returns ~10% hits distributed across RNA template channel + active site + RNA primer channel + two previously undescribed ligandable pockets (Thumb Site II and Index-Middle Finger Pocket) with inhibition confirmed for compounds binding each. PicoGreen HTS of 50K ChemBridge compounds (0.77% hit rate) surfaces a 5-aminoindazole scaffold; lead 727590 IC50 25 µM with defined SAR. Textbook discovery-program opening on a pediatric-antiviral target that had none. (4) Kostlan, Phoenix, Budreika, Ferrari, Deegan, Warren, Bawa, Rogers, Dureja, Ali, Hancock, Young, Gupta, Solanki, Vander Griend, Fanning, Kregel (Loyola Chicago) — METTL3 as a non-enzymatic AR co-regulator in prostate cancer. Genetic METTL3 depletion blocks AR+ line proliferation but not untransformed epithelia, while catalytic inhibitors reduce global mRNA m6A without affecting proliferation. Catalytically dead METTL3 mutant rescues knockdown as well as wild-type. Direct AR interaction + chromatin co-localization + reduced AR cistromic occupancy on knockdown. Reframes clinical strategy: correct modality is a scaffold-degrader (PROTAC/molecular glue), not the STC-15-class catalytic inhibitors currently in trials. false Four bioRxiv preprints from the last 24-48 hours spanning gene-therapy conditioning, synthetic biology, cancer metabolism, and mechano-immunology. (1) Romina Marone, Rosalba Lepore, Kiriaki Paschoudi, Jessica Zuin, Alessandro Sinopoli, Anna Camus, Thomas Burgold, Ewelina Bartoszek, Diego Calabrese, Mylene Toranelli, Julia Wittwer, Manuel Rhiel, Geoffroy Andrieux, Chang Li, Alvin Hsu, Amelie Wiederkehr, Lisa C. Wellinger, Eva-Maria Grossjohann, Emiel Ten Buren, Julie Brault, Laura Garcia Prat, Frank Lehmann, Valentin Do Sacramento, Christopher Divsalar, Saniye Yumlu, David R. Liu (Broad + HHMI), Andre Lieber (Seattle), Toni Cathomen (Freiburg), Tatjana I. Cornu, Evangelia Yannaki (Thessaloniki), Stefanie Urlinger, and Lukas T. Jeker (Basel + Cimeio Therapeutics) attack the biggest remaining obstacle in curative ex-vivo gene therapy for hemoglobinopathies — the myeloablative conditioning that clears the bone marrow niche. They generate a new CD117-blocking monoclonal antibody, CIM058, that potently disables wild-type HSPCs, and use prime editing to install a single-nucleotide epitope shift in human CD34+ cells so the antibody no longer touches transplanted cells. The two components are complementary: the antibody clears the endogenous niche without alkylator chemotherapy, and the edited donor cells are invisible to it, enabling in-vivo selection. Combined, they ameliorate disease phenotype in a β-thalassemia model. If confirmed in primates and eventually in humans, this attacks the busulfan toxicity that has kept ex-vivo sickle cell and thalassemia gene therapy from broad rollout, putting two orthogonal engineered modalities — a shielding-epitope prime edit and a receptor-blocking antibody — into the same protocol. (2) Nathaniel J. Gaut, Christopher Deich, Brock Cash, Tanner Hoog, Aaron E. Engelhart, and Katarzyna P. Adamala (Minnesota) report a synthetic cell built from purified components that runs a complete cell cycle end to end. The genome is a 90kb multi-plasmid design encoding PURE-style translation, Phi29 rolling-circle genome replication, and a genetically encoded feeding mechanism in which an internally expressed α-hemolysin variant binds Ni-NTA lipids on incoming feeder liposomes, driving fusion that supplies fresh membrane and translation machinery. Because feeding requires an internally translated protein, a mutation that raises expression of that protein directly increases growth, offspring number, and fitness — coupling genotype to heritable selection inside the vesicle in a way earlier minimal cells did not. They demonstrate five generations of cell cycle, spread of a beneficial mutation through a population under resource competition, and genetically encoded rather than mechanical division. All the hallmarks — encoded metabolism, replication, growth, division, heritable selection — are running in one operational, chemically defined system. That is the object synthetic biology has been building toward, and the platform it opens for programmable therapeutic living materials and directed evolution. (3) Emily M. Clay, Xiaohai Shi, Elizabeth A. Kolar, Yanqiu Liu, Bachchu Lal, and Paul A. Watkins (Kennedy Krieger + Johns Hopkins Neurology) nominate ACSVL3 — an acyl-CoA synthetase that activates very-long-chain fatty acids — as a metabolic dependency of glioma largely absent in normal glia. Genetic depletion of ACSVL3 slows U87MG and Mayo-22 glioma cells and their xenografts; pharmacologically, the natural-product inhibitor grassofermata (CB5) reproduces the knockout phenotype at low-micromolar concentrations, blocking β-oxidation across C16-C24 fatty acids, inducing GFAP+ astrocytic differentiation, and reducing invasion. Normal human fibroblasts are unaffected at doses that stop glioma cells. In NOD/SCID mice, up to 32 mg/kg/day of the drug is grossly well tolerated intraperitoneally, and a much lower 2 mg/kg/day slows U87MG xenografts. Glioblastoma has famously few actionable dependencies; identifying a lipid-activation enzyme that is glioma-enriched and druggable in vivo with an off-the-shelf small molecule reopens a class of chemistry-forward starting points for a disease that badly needs them. (4) Yunus Alapan, Jaehoon Kim, Kiyoon Min, Alexander Shih, Samuel N. Lucas, Taehee Yoon, Paul A. Archer, Heather K. Lin, Ruby Freeman, Megen C. Wittling, Megan Wyatt, Chrystal M. Paulos, Sarwish Rafiq, and Susan Napier Thomas (Georgia Tech + Emory Winship Cancer Institute) show, using microphysiological transmigration devices and in vivo models, that CD8+ T cells squeezing through confined pores lose surface PD-1 within minutes via ubiquitin-mediated proteasomal degradation, and gain function and fitness in parallel. The mechanism is conserved across species and applies to human tumor-infiltrating lymphocytes and CAR-T cells; the imprint correlates with disease outcome in human melanoma. This flips the framing on checkpoint expression, which has been treated as a purely molecular readout of exhaustion; here, the mechanical history of the cell — where it has been and what it squeezed through — is a co-equal determinant. Immediately practical: a mechano-priming step in the manufacturing of CAR-T or TIL products could deliver checkpoint-blockade-like fitness without engaging the PD-1 pathway pharmacologically. All four preprints are v1 posted 2026-07-08 through 2026-07-09 on bioRxiv; bodies not yet rendered — summaries are grounded in the full abstracts per PIPELINE.md guidance. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-10-cd117-shielded-hspc-adamala-synthcell-acsvl3-glioma-pd1-mechano Fri, 10 Jul 2026 12:00:00 +0000 339 Four bioRxiv preprints from the last 24-48 hours: (1) Marone, Lepore, Paschoudi, Zuin, Sinopoli, Camus, Burgold, Bartoszek, Calabrese, Toranelli, Wittwer, Rhiel, Andrieux, Li, Hsu, Wiederkehr, Wellinger, Grossjohann, Ten Buren, Brault, Garcia Prat, Lehmann, Do Sacramento, Divsalar, Yumlu, Liu (Broad + HHMI), Lieber (Seattle), Cathomen (Freiburg), Cornu, Yannaki (Thessaloniki), Urlinger, Jeker (Basel + Cimeio Therapeutics) — CD117 epitope-shielded HSPC transplantation. New anti-CD117 monoclonal antibody CIM058 blocks wild-type HSPCs. Prime editing of human CD34+ HSPCs installs a single-nucleotide epitope shift that renders the transplanted cells invisible to CIM058. Combined, the two components ameliorate disease phenotype in a β-thalassemia model — the first plausible route out of busulfan/myeloablative alkylator chemotherapy conditioning for ex-vivo hemoglobinopathy gene therapy, coupling receptor-blocking antibody with orthogonal prime-edited epitope shield in one protocol. (2) Gaut, Deich, Cash, Hoog, Engelhart, Adamala (Minnesota) — first fully chemically defined synthetic cell running a complete cell cycle. 90kb multi-plasmid genome, PURE translation, Phi29 rolling-circle genome replication, and a genetically encoded feeding mechanism in which internally translated α-hemolysin binds Ni-NTA lipids on incoming feeder liposomes to drive fusion supplying fresh membrane and translation machinery. Because feeding requires an internally expressed protein, genotype (αHL-expression variant) directly couples to growth, offspring number, fitness, and heritable selection under resource competition. Demonstrates 5 generations, spread of a beneficial αHL-expression mutation through a population, and genetically encoded (not mechanical) division. All the hallmarks of life — encoded metabolism, replication, growth, division, heritable selection — running in one operational, chemically defined system. (3) Clay, Shi, Kolar, Liu, Lal, Watkins (Kennedy Krieger + Johns Hopkins Neurology) — first-in-class metabolic target for glioblastoma. ACSVL3 (very-long-chain acyl-CoA synthetase 3) is glioma-enriched vs normal glia, genetic depletion slows U87MG + Mayo-22 xenografts, and natural-product inhibitor grassofermata (CB5) reproduces the knockout phenotype at 3 μM: blocks β-oxidation of C16-C24 fatty acids, induces GFAP+ astrocytic differentiation, reduces invasion. Normal human fibroblasts unaffected at 10 μM. NOD/SCID mice tolerate up to 32 mg/kg/day intraperitoneally with no obvious side effects; 2 mg/kg/day slows U87MG xenografts. Reopens chemistry-forward starting points for a disease with famously few actionable dependencies. (4) Alapan, Kim, Min, Shih, Lucas, Yoon, Archer, Lin, Freeman, Wittling, Wyatt, Paulos, Rafiq, Thomas (Georgia Tech + Emory Winship) — confined T cell migration mechanically controls PD-1. CD8+ T cells squeezing through confined pores lose surface PD-1 within minutes via ubiquitin-mediated proteasomal degradation, with concurrent gain in function and fitness. Mechanism conserved across species; applies to human TIL and CAR-T; imprint correlates with disease outcomes in human melanoma. Reframes checkpoint expression as co-determined by mechanical history of tissue traversal, not just molecular exhaustion cues, and suggests a mechano-priming manufacturing step for adoptive cell therapy that could deliver checkpoint-blockade-like fitness without engaging PD-1 pharmacologically. All four preprints are v1 posted 2026-07-08 through 2026-07-09 on bioRxiv; bodies not yet rendered — summaries are grounded in the full abstracts per PIPELINE.md. false Four bioRxiv preprints from the last 24-48 hours spanning computational protein design for targeted protein degradation, chemical biology of intralysosomal amyloid formation, orphan-transporter deorphanization, and TP53-loss synthetic-lethal metabolism. (1) Bram Mylemans, Boyana Korona, Amanda M. Acevedo-Jake, Ailsa MacRae, Thomas A. Edwards, Danny T. Huang (CRUK Beatson), Andrew J. Wilson, Laura S. Itzhaki (Cambridge), and Dek N. Woolfson (Bristol) demonstrate hetero-bifunctional de novo–designed proteins as alternatives to small-molecule PROTACs for targeted protein degradation. Small-molecule PROTACs are limited by (i) fussy dependence on linker length/orientation/lysine geometry around the target and (ii) the cereblon/VHL duopoly that dominates the clinical E3 repertoire. The authors engineer a small, highly stable helix-turn-helix scaffold, computationally graft in a hot-spot–driven binding site for anti-apoptotic cancer target BCL-xL on one face, and place a short linear motif (SLiM) that recruits the E3 ubiquitin ligase KLHL20 on another. The resulting bifunctional protein binds both partners in vitro, degrades BCL-xL in cells, and drives apoptosis in a cancer cell line. Because binding-site geometry is dialed in by the designed scaffold rather than by an empirical linker, and because SLiMs can in principle access a much wider slice of the human E3 repertoire, this is a proof of concept that computational protein design can open ternary-complex modalities that were structurally out of reach for small molecules. (2) Ryan D. Elias, Sarah Allen, Ismail Ismet Demiralp, R. Tyler O'Neill, Jan-Hannes Shäfer, Hana Siems, Elizabeth A. Montabana, Utz H. Ermel, Cassidy Ash, Fatima Abdurrob, David A. Yacoubian, Ori L. Lederberg, Daniel Serwas, David A. Agard, Benjamin F. Cravatt (Scripps), and Jeffery W. Kelly (Scripps) finally nail the mechanism of the widely used lysosomal-damage reagent L-leucyl-L-leucyl methyl ester (LLOMe). Cathepsin C (CTSC), a lysosomal dipeptidase, uses a subtiligase-like thioester intermediate to ligate the dipeptide methyl ester into (Leu-Leu)n oligopeptides that exceed critical concentration and self-assemble into cross-β-sheet amyloid fibrils inside the lysosome (X-ray diffraction rings at 11.5 and 4.8 Å; AmyTracker/thioflavin T/Proteostat/Hoechst all confirm). Cryo-electron tomography of TMEM192-GFP–affinity-isolated lysosomes shows lumenal fibrils in physical contact with broken membranes; membranolytic activity is coupled specifically to active fibril growth (not deposition), and remnant fibrils after LLOMe washout are membranolytically inert but persist for hours. Sequence variation — substituting phenylalanine to give LFOMe, FLOMe, or FFOMe — retunes the aggregate structure, the ESCRT-mediated (CHMP4B) lysosomal-repair response, and the fibril's ability to cross-seed cytosolic Tau(P301S) at lysosomal sites; α-synuclein(A53T) is not cross-seeded, arguing a specific structural match. LC-MS with LOMe, LLLOMe, and GLLOMe substrates places the ligation mechanism as CTSC aminopeptidase → CTSC-dipeptide thioester → either hydrolysis or nucleophilic attack by an incoming peptide N-terminus, in analogy with subtiligase and native chemical ligation. Two implications: numerous prior lysosomal-damage experiments in the neurodegeneration literature need to be reinterpreted through a hidden intralysosomal amyloid load, and the CTSC-dipeptide ligation platform now becomes a modular chemical-biology tool for generating defined amyloid structures inside living cells — a rare experimental handle on the endolysosomal-breakout step of prion-like tau spreading. (3) Corey N. Cunningham, Alex J. Bott, Charles H. Adelmann, Matthew Shields, Katarina E. Heyden, Jonathan G. Van Vranken, Alvaro J. Narbona-Perez, Juan A. Cantres-Velez, Guillaume Adelmant, Nathan M. Krah, Steven H. Gygi, David M. Sabatini, and Jared Rutter (HHMI Utah + Broad) identify SLC16A6 as the long-sought orphan melanosomal tyrosine transporter. Tyrosine is the mandatory substrate for melanin synthesis and must cross the melanosomal membrane, but the responsible transporter had eluded biochemistry for decades. Genetic screens place SLC16A6 expression downstream of the SOX10–MITF master regulatory axis that governs the entire melanosomal program. Redirecting SLC16A6 to the plasma membrane with a trafficking-disrupting S240A mutation lets the authors prove direct tyrosine transport, competitively inhibited by other bulky amino acids, and to demonstrate SLC16A6-sufficiency for in vitro melanosomal tyrosine uptake. Depleting SLC16A6 does not merely reduce pigmentation but collapses melanosome biogenesis and depletes most melanosomal proteins, arguing that substrate loading and organelle identity are causally coupled. A clean template for pairing directed genetic screens with a trafficking-mutant handle to force a reluctant intracellular transporter into a measurable regime — the strategy that has stranded most orphan SLC transporters. (4) Julia Urgel-Solas, Anna Benito, Laura Ortet, Savvas Kourtis, Etna Abad, Albert Coll Manzano, Anna Shevzov-Zebrun, Camilla Reiter Elbaek, Carlos Martinez, Matthew Vander Heiden (MIT), Sara Sdelci, and Ana Janic (CRG Barcelona) show that TP53 loss creates a nuclear-compartment metabolic vulnerability. TP53-deficient cells acquire a requirement for de novo purine biosynthesis, with a pronounced dependence on adenosine-related metabolism; perturbing the pathway increases DNA damage and specifically depletes nuclear ATP. Histone methylation rises within minutes of purine-synthesis perturbation or acute DNA damage — fast enough that the authors argue methylation is an intrinsic, early feature of the nuclear stress response rather than downstream epigenetic remodeling; interfering with methylation itself reduces nuclear ATP and impairs DNA damage resolution. Genetic disruption of the purine-pathway enzyme PFAS (phosphoribosylformylglycinamidine synthase) selectively impairs the growth of TP53-deficient tumors in vivo, establishing that the vulnerability is druggable rather than a cell-culture artifact. Nuclear adenosine metabolism reframes as a compartmentalized adaptive response to genotoxic stress, and chromatin methylation as a fast nuclear metabolic-signal integrator — a synthetic-lethal window against the single most common oncogenic event in a pathway that already has clinical inhibitors. All four preprints are v1 or freshly reposted on bioRxiv this week; bodies are variably rendered — the Kelly/Cravatt and de novo TPD papers have full body text, SLC16A6 and TP53/purine bodies are still abstract-only per bioRxiv's rendering lag, and summaries are labeled accordingly per PIPELINE.md. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-08-denovo-protein-protac-ctsc-amyloid-slc16a6-tp53-purine Wed, 08 Jul 2026 12:00:00 +0000 378 Four bioRxiv preprints from the last 24-48 hours: (1) Mylemans, Korona, Huang (CRUK Beatson) + Itzhaki (Cambridge) + Woolfson (Bristol) — de novo–designed protein PROTACs. Computationally designed helix-turn-helix scaffold with hot-spot binder for BCL-xL on one face + short linear motif (SLiM) for KLHL20 E3 ligase recruitment on another face. Bifunctional design binds both partners in vitro, degrades BCL-xL in cells, drives apoptosis in cancer line. Sidesteps two central bottlenecks of small-molecule PROTACs: fussy dependence on empirical linker geometry and the cereblon/VHL duopoly on E3 ligases. Proof of concept that computational protein design can open ternary-complex modalities out of reach for small molecules. (2) Elias, Serwas, Agard, Cravatt (Scripps), Kelly (Scripps) — intralysosomal amyloidogenesis by Cathepsin C. LLOMe, the decades-old lysosomal-damage reagent, is a prodrug that CTSC ligates via a subtiligase-like thioester intermediate into (Leu-Leu)n oligopeptides that self-assemble into cross-β-sheet amyloid fibrils inside the lysosome. Cryo-electron tomography of affinity-isolated lysosomes shows lumenal fibrils in contact with broken membranes; membranolytic activity is coupled specifically to active fibril growth, not deposition. Dipeptide-sequence variation (LFOMe/FLOMe/FFOMe) retunes aggregate structure, ESCRT-repair response, and cross-seeding of cytosolic Tau(P301S) — but not α-synuclein(A53T). Recontextualizes prior lysosomal-damage neurodegeneration literature and delivers a modular chemical-biology tool for generating defined amyloid structures inside living cells. (3) Cunningham, Bott, Adelmann + Rutter + Sabatini (HHMI Utah + Broad) — SLC16A6 is the missing melanosomal tyrosine transporter. Genetic screens place SLC16A6 downstream of the SOX10–MITF master melanogenesis axis. S240A trafficking mutant redirects SLC16A6 to plasma membrane and enables direct proof of tyrosine transport, competitively inhibited by other bulky amino acids. SLC16A6 depletion collapses melanosome biogenesis and depletes most melanosomal proteins — substrate loading and organelle identity are causally coupled. Template for orphan SLC transporter deorphanization. (4) Urgel-Solas, Sdelci + Janic (CRG Barcelona) + Vander Heiden (MIT) — TP53 loss creates a nuclear adenosine dependency. TP53-deficient cells require de novo purine biosynthesis with pronounced adenosine-metabolism dependence; perturbation raises DNA damage and specifically drops nuclear ATP. Histone methylation rises within minutes of perturbing purine synthesis or acute DNA damage — argued to be an intrinsic feature of the nuclear stress response. Genetic knockout of the purine-pathway enzyme PFAS selectively impairs TP53-deficient tumor growth in vivo — a synthetic-lethal window against the single most common oncogenic event in a pathway with existing clinical inhibitors. false Four bioRxiv preprints from the last 24-48 hours spanning AI-native virtual screening, covalent chemical biology, spatial metabolomics, and structure-guided ion-channel drug discovery. (1) Noah Getz, Geoffrey Smith, Avene Colgan, Vincent Fan, Luca Cavalleri, Francesco Capponi, Jeremy Wohlwend, Anthony Gitter, Joshua Kritzer, Madison Maiorano, Nathan Wlodarchak, Gabriele Corso, and Saro Passaro (Boltz PBC + academic collaborators) present BoltzMol-1, a small-molecule hit discovery pipeline centered on an optimized version of Boltz-2 explicitly adapted for prospective discovery. Model-driven ranking is coupled to procurement from commercial catalogs; an ADMET triage layer for kinetic solubility (logS), lipophilicity (logD), and Caco-2 permeability filters candidates before synthesis/purchase. On a ten-target panel most of which had no representation in the underlying affinity training data, the pipeline identified functional actives or binders on 6 of 10 targets with experimental budgets of 28-96 compounds per target, including receptors and enzymes traditionally considered difficult for structure- or ligand-based approaches. Concrete data point on where AI-native virtual screening sits vs campaign-scale HTS. (2) Chao Wang, Polina E. Barzova, Jessica Robles, Ethan S. Toriki, Francisco J. Garcia, Jeffrey M. McKenna, Markus Schirle, and Ziyang Zhang (UC Berkeley + Novartis) introduce chiral β-lactone electrophiles as a stereo-encoded covalent warhead class. The C481S mutation at BTK's active-site cysteine is the dominant clinical resistance mechanism against ibrutinib in mantle cell lymphoma and CLL because it eliminates the anchor cysteine every clinical BTK covalent inhibitor requires. The asymmetric β-lactone warhead achieves stereoselective covalent modification of both wild-type BTK and the C481S mutant, but at distinct sites of reactivity — meaning the same warhead class covers wild-type and mutant enzymes through orthogonal chemistry. Individual enantiomers of kinase-directed β-lactone probes preferentially engage different subsets of the kinome, establishing stereochemistry as an intrinsic selectivity filter in covalent drug discovery. (3) Chan-Mi Yoo, Ji-Young Jo, Chan-Ru Choi, Yun Seok Park, Yun Ju Cha, Sun Jung, Jangwon Kang, Jinsu Kim, Yeon Pyo Kang, Tae Hyeon Yoo, Ji-Sook Kim, and Hyun-Woo Rhee (Seoul National University) introduce DESTNI, an engineered desthiobiotin ligase derived from TurboID via yeast-display directed evolution, and establish it as a proximity-labeling platform for spatially resolved profiling of amine-containing metabolites in live cells. An integrated annotation framework combines synthetic desthiobiotin-metabolite standards, in vitro DESTNI-reactive metabolite discovery, and machine-learning tandem-MS prediction of DTB-derivatized metabolites and oligopeptides. Organelle-targeted DESTNI recovers compartment-enriched amine-metabolite signatures: mitochondrial-matrix-enriched glycine, 5-aminolevulinic acid, ornithine, and spermidine adducts, plus nuclear-enriched γ-aminobutyric acid and 5-aminovaleric acid adducts. Puts spatial metabolomics on the same platform footing as spatial proteomics for the first time. (4) Guoyi Luo, Xi Zhang, Hongfei Xia, Zhijie Wei, Zhiyong Zhang, Jiawei Sun, Zhen Zhang, Yiyu Peng, Hui Liu, Xiaolin Huang, Peng Cao, Mingqiang Rong, Yonghua Yu, and Cheng Tang (Hunan Normal University) report MPiN, a highly selective state-dependent KV1.3 inhibitor with in vivo efficacy in a mouse psoriasis model, and the structural pocket it engages. KV1.3 is a validated autoimmune drug target, but every previously disclosed inhibitor has converged on the same conserved sites with recurring selectivity and clinical failure problems. MPiN engages a previously uncharacterized extracellular allosteric pocket framed by the PP1/PP2 turret loops, the pore helix, and the outer S5/S6 helices, acting through a bidirectional pore-to-sensor coupling that simultaneously constricts the selectivity filter and facilitates voltage-sensor activation. Despite the pocket being highly conserved across KV1 paralogs, subtype selectivity is dictated by peripheral residues G427 and H451, which define pocket geometry without directly contacting the ligand — a geometry-driven non-contact selectivity mechanism. Opens a structurally distinct chemical space on a validated but historically frustrating target. All four preprints are v1 posted in the last 24-72h; bodies are not yet rendered on bioRxiv — summaries are abstract-grounded per PIPELINE.md. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-07-boltzmol-btk-c481s-blactone-destni-kv13-mpin Tue, 07 Jul 2026 12:00:00 +0000 359 Four bioRxiv preprints from the last 24-48 hours: (1) Getz, Smith, Colgan, Fan, Cavalleri, Capponi, Wohlwend, Gitter, Kritzer, Maiorano, Wlodarchak, Corso, Passaro (Boltz PBC + academic collaborators) — BoltzMol-1, a small-molecule hit discovery pipeline centered on an optimized version of Boltz-2 for prospective discovery. Model-driven ranking + procurement from commercial catalogs + ADMET triage layer (kinetic solubility, lipophilicity, Caco-2 permeability) at ranking time. On a ten-target panel with most targets having no representation in underlying affinity training data, identified functional actives or binders on 6 of 10 targets with experimental budgets of 28-96 compounds per target, including receptors and enzymes traditionally hard for structure- or ligand-based methods. Concrete data point on where AI-native virtual screening sits vs campaign-scale HTS. (2) Wang, Barzova, Robles, Toriki, Garcia, McKenna, Schirle, Zhang (UC Berkeley + Novartis) — chiral β-lactone electrophiles as stereo-encoded covalent warheads. Asymmetric β-lactone achieves stereoselective covalent modification of both wild-type BTK and ibrutinib-resistant BTK(C481S) at distinct sites of reactivity — same warhead class covers wild-type and mutant via orthogonal chemistry despite the loss of anchor Cys481. Individual enantiomers of kinase-directed β-lactone probes preferentially engage different kinome subsets. Establishes stereochemistry as an intrinsic selectivity filter in covalent drug discovery. (3) Yoo, Jo, Choi, Park, Cha, Jung, Kang, Kim, Kang, Yoo, Kim, Rhee (Seoul National University) — DESTNI, a desthiobiotin ligase engineered from TurboID by yeast-display directed evolution, plus synthetic DTB-metabolite standards + in vitro DESTNI profiling + ML tandem-MS predictor for annotation. Extends proximity labeling from spatial proteomics to spatial metabolomics in live cells, with organelle-resolved amine-metabolite maps (mitochondrial matrix: glycine + 5-aminolevulinic acid + ornithine + spermidine; nuclear: GABA + 5-aminovaleric acid). (4) Luo, Zhang, Xia, Wei, Zhang, Sun, Zhang, Peng, Liu, Huang, Cao, Rong, Yu, Tang (Hunan Normal University) — MPiN, a highly selective state-dependent KV1.3 inhibitor with in vivo psoriasis-model efficacy engages a previously uncharacterized extracellular allosteric pocket (PP1/PP2 turret loops + pore helix + outer S5/S6 helices) via bidirectional pore-to-sensor coupling that constricts the selectivity filter and facilitates voltage-sensor activation simultaneously. Subtype selectivity across the conserved KV1 paralog pocket is dictated by peripheral residues G427 and H451 that never directly contact the ligand — non-contact selectivity opens a distinct chemical space on a validated but historically frustrating autoimmune target. All four preprints are v1 posted in the last 24-72h; bodies are not yet rendered on bioRxiv — summaries are abstract-grounded per PIPELINE.md. false Four bioRxiv preprints from the last 24-48 hours spanning AI-native enzyme engineering, chemical biology of amyloids, natural-product target ID, and ML-guided phenotypic antibiotic discovery. (1) Zhenjia Gan, Yuzhi Xu, Junde Xu, Zhihao Wu, Juping Huang, Jiabin Yin, Guangyong Chen, and John Z. H. Zhang (Shenzhen University of Advanced Technology) introduce CatESO, a differentiable enzyme sequence optimizer that backpropagates through a cross-modal K-cat predictor under continuous sequence relaxation, co-optimizing substrate-specific catalytic turnover, evolutionary plausibility (ESM-2), and foldability (ESMFold pLDDT) in one end-to-end framework. Across seven OOD enzymes spanning EC classes 1-7, CatESO raised predicted K-cat for the vast majority of designs with median 1.52× fold change and every variant retaining pLDDT >70. Head-to-head against RFdiffusion3-LigandMPNN and ZymCtrl, CatESO struck a better balance between predicted activity and structural confidence. Carries differentiable protein design beyond structure- and binding-centred goals to enzyme catalytic function — the missing objective for biocatalysis and greener process chemistry. (2) Anthony Balistreri, Mark Gomulinski, Matthew R. Chapman (Michigan), and Jeffery W. Kelly (Scripps) show that soluble TREM2 — the shed microglial extracellular domain previously shown to inhibit amyloid-β — is a sub-stoichiometric inhibitor of the bacterial functional amyloid CsgA, the major protein component of curli that scaffolds biofilms in uropathogenic E. coli and other proteobacteria. Kinetic modeling localizes the inhibition to primary and secondary nucleation, not fiber elongation — the same mechanistic signature reported for its human-amyloid activity. Exogenously added sTREM2 suppresses curli-dependent pellicle biofilm formation without affecting bacterial growth. Reframes sTREM2 as a broad amyloid inhibitor operating across the human-microbe interface and strengthens the case for cross-kingdom amyloid biology as a therapeutic axis, given emerging evidence that curli-like bacterial amyloids act as peripheral triggers for synucleinopathy and gut-brain neurodegeneration. (3) James Martin II (Rowan-Virtua School of Osteopathic Medicine) builds a control-validated pan-proteome deep-learning drug-target-interaction pipeline — GNN ligand encoder + ESM-2 protein language-model encoder + bidirectional cross-attention — with bias-corrected ranking and control-anchored docking. Uses losartan as a known-target control to identify and correct a frequent-hitter bias in raw DTI predictions. Applied to the orphan 2022 Streptomyces/Achromobacter co-culture natural product ligiamycin A, the corrected predictions cluster on class-A GPCRs and put orphan receptor GPR35 in front. Matched positive/negative-control docking scores ligiamycin A at -8.1 kcal/mol on GPR35, essentially tied with zaprinast (-8.3) against non-binder floor -5.5; FFAR1 is excluded and histamine H2 is inconclusive. The methodological deliverable is a reusable, control-validated recipe for deorphanizing bioactive microbial natural products at pan-proteome scale — the bottleneck that has stranded most secondary metabolites without a target. (4) Lia R. Serrano, Andrew Zhou, Ziming Wei, Kee-Lee K. Stocks, Yasha Ektefaie, Peter J. Gwynne, Eric Chen, Inna Krieger, James Sacchettini (Texas A&M), Bree Aldridge (Tufts), Linden T. Hu (Tufts), and Maha R. Farhat (Harvard Medical School) systematically evaluate three active-learning strategies for whole-cell bacterial bioactivity prediction on retrospective M. tuberculosis HTS data, identify a novelty × hit-rate-balanced optimum, and deploy it in a closed-loop live campaign against Borrelia burgdorferi. AL raised the experimental hit rate 5× (0.2% → 1.0%). Prospective in silico selection of chemically diverse compounds across bacterial species delivered 53× enrichment over investigator-directed screening, and 100% of experimentally validated predicted hits showed intended narrow-spectrum B. burgdorferi activity. Concrete quantification of how much a calibrated AL loop can extend a whole-cell screen out of distribution without additional experimental cycles — inside the campaign infrastructure that antibiotic HTS groups already run. All four preprints are v1 posted in the last 24-72h; bodies are not yet rendered on bioRxiv — summaries are abstract-grounded per PIPELINE.md. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-06-cateso-strem2-csga-ligiamycin-gpr35-al-antibiotic Mon, 06 Jul 2026 12:00:00 +0000 361 Four bioRxiv preprints from the last 24-48 hours: (1) Gan, Xu, Xu, Wu, Huang, Yin, Chen, Zhang (Shenzhen U. Advanced Technology) — CatESO, differentiable enzyme sequence optimizer that backpropagates through a cross-modal K-cat predictor under continuous sequence relaxation, co-optimizing substrate-specific catalytic turnover, evolutionary plausibility (ESM-2), and foldability (ESMFold pLDDT) end-to-end. Median 1.52× predicted K-cat fold change across seven OOD enzymes spanning EC classes 1-7 with every variant retaining pLDDT >70. Beats RFdiffusion3-LigandMPNN and ZymCtrl on predicted activity vs structural confidence. Carries differentiable protein design beyond structure/binding to enzyme catalytic function — the missing objective for biocatalysis and greener process chemistry. (2) Balistreri, Gomulinski, Chapman (Michigan), Kelly (Scripps) — soluble TREM2 is a cross-kingdom amyloid inhibitor. Sub-stoichiometric inhibition of the bacterial functional amyloid CsgA (curli). Kinetic modeling localizes the action to primary + secondary nucleation, not fiber elongation — same signature as sTREM2's Aβ activity. Exogenous sTREM2 suppresses curli-dependent pellicle biofilm formation without affecting bacterial growth. Reframes sTREM2 as a broad amyloid inhibitor across the human-microbe interface, strengthening the case for curli-like bacterial amyloids in gut-brain neurodegeneration. (3) Martin (Rowan-Virtua) — control-validated pan-proteome DTI pipeline (GNN ligand + ESM-2 protein + bidirectional cross-attention + losartan-anchored frequent-hitter bias correction + matched positive/negative control docking) nominates orphan class-A GPCR GPR35 as the target of the 2022 Streptomyces/Achromobacter co-culture natural product ligiamycin A. Docking: -8.1 kcal/mol at GPR35 agonist pocket, essentially tied with zaprinast (-8.3) against non-binder floor -5.5; FFAR1 excluded, histamine H2 inconclusive. Reusable recipe for deorphanizing bioactive natural products at pan-proteome scale. (4) Serrano, Sacchettini (Texas A&M), Aldridge + Hu (Tufts), Farhat (Harvard Medical School) — active-learning strategy for whole-cell phenotypic antibiotic discovery. Novelty × hit-rate-balanced AL identified on retrospective M. tuberculosis HTS; closed-loop deployment against Borrelia burgdorferi raises experimental hit rate 5× (0.2% → 1.0%). Prospective in silico selection of chemically diverse compounds delivers 53× enrichment vs investigator-directed screening with 100% narrow-spectrum B. burgdorferi activity in validated hits. Concrete quantification of how much a calibrated AL loop can extend whole-cell screens OOD without more experimental cycles. All four preprints are v1 posted in the last 24-72h; bodies are not yet rendered on bioRxiv — summaries are abstract-grounded per PIPELINE.md. false Four bioRxiv preprints from the last 24-48 hours spanning immunotherapy-toxicity genetics, tumor-cell antigen presentation, synthetic-cell design, and chemical-biology deorphanization. (1) J. Richard Brewer, Ailin Han, Amin H. Nassar, Elias Bou Farhat, Holly N. Blackburn, Tianli Xiao, Haris Mirza, Walter K. Mowel, Esen Sefik, Saskia Hartner, Michael Chiorazzi, Takeshi Itoh, Min-Hee Oh, Matthew Z. Madden, Athreya Rangavajhula, Elio Adib, Mustafa J. Saleh, Marc Machaalani, Mehrdad Rakaee, Masoud Tafavvoghi, Claire Quattropani, Nicole Gazetos, David Gerber, Farjana Fattah, Jeffrey A. SoRelle, Dominic Choo, Mitchell S. von Itzstein, Shannon Bevans-Fonti, Mohammad Ghanbar, Karthik Suresh, Thomas Mazumder, Chun J. Ye, Toni K. Choueiri, Alexander Gusev, and Richard A. Flavell (Yale + Dana-Farber + Broad) identify DPP9 as the first genetic susceptibility gene for checkpoint-inhibitor pneumonitis (CIP) in a clinico-genomics cohort of ~4,400 immune checkpoint inhibitor patients. CIP is the leading cause of immunotherapy-related mortality with case-fatality approaching 10%. DPP9 suppresses CARD8 inflammasome activation and IL-18 secretion in human monocytes; plasma IL-18 is selectively elevated in patients before CIP onset; myeloid-restricted deletion of Dpp8 and Dpp9 in mice recapitulates pulmonary hallmarks — granuloma formation, IFNγ T-cell accumulation, and monocyte-derived macrophage infiltration — all driven by excess IL-18. Nominates IL-18 blockade as the mechanistically-rational therapeutic intervention. (2) Charlie Chung, Elif Ozcelik, Jialin Zhang, Zeli Shen, Onur Eskiocak, Yihan Qin, Jill Habel, Kadir Ozler, Santhilal Subhash, Zakaria Aminzada, Gagan Dev, Libia Garcia, Scott K. Lyons, Timothy Hand, David E. Rivadeneira, James G. Fox, Peter M. K. Westcott, Meri Rogava, and Semir Beyaz (Cold Spring Harbor Laboratory) show that a Helicobacter-containing microbiome suppresses primary tumor growth and metastasis in orthotopic mouse colorectal-cancer models via induction of tumor-cell-intrinsic MHC-II through CIITA. Tumor CIITA knockout abrogates the microbial protection; enforced CIITA expression sensitizes microsatellite-stable tumors to combined PD-1 and CTLA-4 blockade — the ~90% of metastatic-colon-cancer patients who get almost no benefit from checkpoint inhibitors. Human MSS patient-derived organoids replicate the phenotype: elevating cancer-cell MHC-II drives tumor-cell apoptosis on co-culture with autologous immune cells. Reframes resensitization as turning the tumor cell into a competent antigen-presenting cell rather than modulating T-cell exhaustion or cold-tumor infiltration. (3) Basak Velioglu Ulubas, Orion Venero, David Garenne, Vincent Noireaux, Aaron E. Engelhart, and Katarzyna P. Adamala (Minnesota) build a programmable cell-free synthetic cell in which translation depends on tyrosine and impose tyrosine scarcity as a tunable metabolic bottleneck. Adding phenylalanine hydroxylase (PAH) — the enzyme that fails in phenylketonuria — as a minimal biosynthetic module reconstitutes phenylalanine → tyrosine → translation coupling. Encapsulated in liposome compartments the system produces three distinct synthetic-cell populations, with the metabolically-active population showing significantly higher protein-production fitness. Foundational step toward autonomous, self-regulating minimal cells and an experimentally tractable origins-of-life testbed for how primitive cells bootstrapped internal metabolism from environmental scarcity. (4) Aakash Chandramouli and Siddhesh Kamat (IISER Pune) identify lipoprotein lipase (LPL) as a previously unrecognized mammalian neutral cholesteryl-ester hydrolase using systematic activity-based protein profiling + LC-MS across mouse tissues and macrophages. Activity is enriched in membrane fractions and sensitive to broad-spectrum metabolic-serine-hydrolase inhibitors. Pharmacological screening excludes NCEH1 and hormone-sensitive lipase (LIPE) and instead points to tetrahydrolipstatin-sensitive LPL. Recombinant wild-type LPL, but not the S159A catalytic-dead variant, efficiently hydrolyzes cholesteryl esters in vitro, dependent on LMF1 co-expression for enzyme maturation. Expands LPL's functional footprint beyond triglyceride hydrolysis and reopens medicinal-chemistry questions around LPL pharmacology, familial-chylomicronemia-syndrome activator programs, and isoform-selective serine-hydrolase inhibitors. All four preprints are v1; bodies are not yet rendered on bioRxiv — summaries are abstract-grounded per PIPELINE.md. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-05-dpp9-cip-il18-mhcii-microbiome-crc-pah-synthcell-lpl-cholesterylesterase Sun, 05 Jul 2026 12:00:00 +0000 318 Four bioRxiv preprints from the last 24-48 hours: (1) Brewer, Nassar, Choueiri, Gusev, Flavell (Yale + Dana-Farber + Broad) — DPP9 loss-of-function is the first genetic susceptibility gene for checkpoint-inhibitor pneumonitis (CIP) in a clinico-genomics cohort of ~4,400 immune checkpoint inhibitor patients. DPP9 suppresses CARD8 inflammasome + IL-18 secretion in human monocytes; plasma IL-18 is selectively elevated pre-CIP; myeloid Dpp8/Dpp9 KO mice recapitulate granuloma formation, IFNγ T cells, and monocyte-macrophage infiltration driven by excess IL-18. Nominates IL-18 blockade as mechanistically-rational therapy for the #1 cause of immunotherapy mortality. (2) Chung, Beyaz, Westcott, Rogava (Cold Spring Harbor Laboratory) — Helicobacter-containing microbiome suppresses primary CRC growth + metastasis in orthotopic mouse models via induction of tumor-intrinsic MHC-II through CIITA. Tumor CIITA KO abrogates protection; enforced CIITA sensitizes microsatellite-stable tumors to combined PD-1 + CTLA-4 blockade — the ~90% of metastatic-CRC patients who currently get almost no checkpoint-inhibitor benefit. Human MSS patient-derived organoids replicate the phenotype. Reframes resensitization as turning the tumor cell into a competent antigen-presenting cell. (3) Velioglu Ulubas, Noireaux, Engelhart, Adamala (Minnesota) — programmable cell-free synthetic cell coupling metabolism to translation via a phenylalanine-hydroxylase (PAH) module in liposome compartments. Tyrosine-dependent translation + tunable tyrosine bottleneck + PAH biosynthetic module → three distinct synthetic-cell populations; metabolically-active population shows significantly higher protein-production fitness. Foundational step toward autonomous minimal cells + experimentally tractable origins-of-life testbed. (4) Chandramouli, Kamat (IISER Pune) — identifies lipoprotein lipase (LPL) as a previously unrecognized mammalian neutral cholesteryl-ester hydrolase via activity-based protein profiling + LC-MS. Excludes NCEH1 + hormone-sensitive lipase, points to tetrahydrolipstatin-sensitive LPL. Recombinant wild-type LPL, but not S159A catalytic-dead variant, hydrolyzes CEs in vitro dependent on LMF1 maturation. Expands LPL functional footprint beyond triglyceride hydrolysis, reopens medicinal-chemistry questions around LPL pharmacology, FCS-activator programs, and isoform-selective serine-hydrolase inhibitors. All four preprints are v1; bodies not yet rendered on bioRxiv — summaries are abstract-grounded per PIPELINE.md. false Four bioRxiv preprints from the last 48 hours spanning drug repurposing, AI-native protein design, RNA druggability, and precision immunotherapy. (1) Giselle Sanchez-Guerrero, David Umbaugh, Nga Nguyen, Hartmut Jaeschke, and Anup Ramachandran at the University of Kansas Medical Center make a therapeutic case for delayed activation of the adenosine A2B receptor in acetaminophen overdose. Acetaminophen remains the leading cause of drug-induced acute liver failure in the United States and N-acetylcysteine works only in a narrow early window that most overdose patients miss. In fasted mice given 300 mg/kg acetaminophen, activating the A2B receptor 6 or 9 hours later — after injury is established — accelerated liver recovery: faster Kupffer-cell repopulation, macrophage migration to and resolution of necrotic areas, hepatocyte cell-proliferation and metabolism programs, and upregulation of lipid-droplet genes such as Cidec and Plin2 to fuel regeneration. The mechanism is not blocking the initial toxic insult but reprogramming the innate-immune and metabolic response after the fact, opening a late-window therapeutic slot in an indication where the standard of care has essentially none. (2) Ho Yeung Chim, Mubarak Olanrewaju Idris, Dominic Rieger, Phillip Schlegel, Nicolas Manuel Goldbach, Miguel Atienza Juanatey, Bhoomika Basu Mallik, Stephen Buckley, Samrat Basak, Sandrine Georgeon, Kelvin Lau, Florence Pojer, Leonard Kaysser, Philip Tinnefeld, Clara T. Schoeder, Bruno E. Correia (EPFL), and Alena Khmelinskaia (LMU Munich / TU Dresden) present a generative framework that starts from an isolated interaction module and grows a homo-oligomer around it. Prior de novo oligomer design has largely produced fixed assemblies; the new twist is designs that assemble responsively. Experimentally validated designs are accurate and explore new-to-nature topologies. Two functional demonstrations: chemical-trigger designs oligomerize conditionally when their ligand is present, and phosphorylation-driven designs oligomerize reversibly under a kinase-phosphatase switch. The team then bolts these assemblies to real biological outputs — ligand-dependent membrane binding and phosphorylation-controlled gene-regulatory switches — sketching a route to designer protein circuits whose stoichiometry is set by an external signal, exactly what is missing for programmable sensors, conditional biotherapeutic assembly, and layered synthetic-biology regulatory logic. (3) Lukasz T. Olenginski and Robert T. Batey at CU Boulder go after a foundational question for RNA-targeted drug discovery — whether the transient cryptic pockets that expand druggability in proteins are also accessible in RNA. Using an env8 cobalamin riboswitch, they show that beta-axial substituted cobalamin ligands bind by displacing a conserved adenosine (A20) from the RNA core, exposing a previously hidden site. Selective abasic substitutions plus ITC and fluorescence measurements quantify the energetic penalty of that base displacement at approximately 1.4 kcal/mol, a modest cost that keeps the cryptic conformer thermally accessible. Computational conformational sampling then finds analogous low-energy cryptic sites in structurally unrelated HIV-1 and HCV RNAs, and a ligand originally raised against the riboswitch cryptic site docks against the newly exposed surfaces of both. For RNA-targeted small-molecule programs, the pitch is that RNA is more druggable than its native structures suggest — cryptic base-displacement pockets are a systematically searchable class of binding surface, and both the energetic ceiling and the computational recipe are now in hand. (4) Sanjay Pathak, Ronia Ahmed, Naomi Nagy, Seungjin Lee, Carly Bader, Sushmita Regmi, Betty Iliopoulou, Pei Chen, Bhavya Gupta, Alonso Villar-Prados, Yong Bin Kim, Noha Hussein, Erica Soohoo, Alexander Twoy, Avnesh Thakor, Kristina Jensen, Paul J. Utz, Mark M. Davis, Justin Annes, and Everett Meyer at Stanford identify CD38 as a marker of the autoreactive T-cell population driving type 1 diabetes. Existing T-cell-depleting therapies delay progression of stage 2 and 3 disease but hit T regulatory cells as collateral damage — the wrong direction for tolerance restoration. In non-obese diabetic mice and recently diagnosed patients, CD38-positive T cells drive disease: adoptive transfer of CD38-positive but not CD38-negative cells into humanized diabetic-mouse models induced diabetes. Depleting CD38-positive cells with an anti-CD38 monoclonal antibody prevented insulitis and diabetes onset while sparing CD4+CD25+ Tregs and preserving islet function, with a bonus signal that senescent immunogenic beta cells were also cleared. Since anti-CD38 antibodies are clinically established in multiple myeloma, this is a fast-moving repositioning candidate as a transient tolerance-restoring intervention in early T1D — precision immunotherapy with an off-the-shelf agent. Papers 1, 2, and 3 are v1 with bodies not yet rendered on bioRxiv — summaries are abstract-grounded; paper 4 is v2 posted 2026-07-03. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-04-a2b-adenosine-liver-recovery-interface-oligomer-cryptic-rna-cd38-t1d Sat, 04 Jul 2026 12:00:00 +0000 300 Four bioRxiv preprints from the last 48 hours: (1) Sanchez-Guerrero, Umbaugh, Nguyen, Jaeschke, Ramachandran (KU Medical Center) — delayed activation of the adenosine A2B receptor at 6 or 9 hours post-APAP accelerates liver recovery in mice via faster Kupffer-cell repopulation, macrophage resolution of necrotic areas, hepatocyte proliferation, and lipid-metabolism reprogramming (Cidec, Plin2), opening a late-window therapeutic slot beyond the NAC cutoff for the #1 cause of drug-induced acute liver failure in the US. (2) Chim, Idris, Rieger, Schlegel, Goldbach, Juanatey, Mallik, Buckley, Basak, Georgeon, Lau, Pojer, Kaysser, Tinnefeld, Schoeder, Correia (EPFL), Khmelinskaia (LMU Munich / TU Dresden) — interface-seeded generative framework for de novo design of responsive homo-oligomers from isolated interaction modules. Experimentally validated new-to-nature topologies, chemical-trigger conditional oligomerization, phosphorylation-driven reversible oligomerization, ligand-dependent membrane binding, and phosphorylation-controlled gene-regulatory switches. Route to designer protein circuits with externally programmable stoichiometry for sensors, conditional biologics, and synthetic-biology regulatory logic. (3) Olenginski, Batey (CU Boulder) — cryptic RNA binding sites are energetically accessible and chemically addressable. Using an env8 cobalamin riboswitch, beta-axial cobalamin derivatives displace a conserved adenosine (A20) to expose a hidden pocket; ITC and fluorescence measurements quantify the energetic penalty at ~1.4 kcal/mol. Computational sampling recapitulates analogous low-energy cryptic sites in HIV-1 and HCV RNAs with docking-consistent ligand engagement. Establishes cryptic base-displacement pockets as a systematically searchable class for RNA-targeted drug discovery. (4) Pathak, Ahmed, Nagy, Lee, Bader, Regmi, Iliopoulou, Chen, Gupta, Villar-Prados, Kim, Hussein, Soohoo, Twoy, Thakor, Jensen, Utz, Davis, Annes, Meyer (Stanford) — CD38 marks the pathogenic autoreactive T-cell population in type 1 diabetes. Adoptive transfer of CD38-positive but not CD38-negative cells drives disease in humanized diabetic mice. Anti-CD38 monoclonal antibody selectively depletes diabetogenic T cells while sparing CD4+CD25+ Tregs and preserving islet function; clears senescent immunogenic beta cells. Since anti-CD38 antibodies are clinically established in myeloma, this is a fast-moving repositioning candidate as transient tolerance-restoring precision immunotherapy for early T1D. Papers 1-3 are v1 with bodies not yet rendered on bioRxiv — summaries are abstract-grounded; paper 4 is v2 posted 2026-07-03. false Your daily briefing has been paused because no downloads have been detected in about two weeks. As soon as you download an episode, generation will automatically resume. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/scripps-biomed-brief 2026-07-03-goodbye Fri, 03 Jul 2026 09:00:16 +0000 8 Your daily briefing has been paused because no downloads have been detected in about two weeks. As soon as you download an episode, generation will automatically resume. Four bioRxiv preprints spanning cell therapy, enzyme therapy, genome engineering, and chemical biology. (1) Ruquaiya Alam, Shantanu Kumar, Rohit Shukla, Nisha Chaudhary, Juli Gupta, Akshita Sinha, Rituparna Chaudhuri, Mohan Ranganathan, Kashif Husain, Nisar Rahim Shaikh, Deeksha Joshi, Jahnvi Hora, Syed Ahmad Ali, Prasad Iyer, Irfan Ahmad Mir, Mohammad Husian, Vishu Hari, Amit Kumar Srivastava, Ulaganathan Mabalirajan, Gaurav Kharya, Sivaprakash Ramalingam, Asimul Islam, and Tanveer Ahmad describe viroVbot, a next-generation in vivo CAR-T platform. A computational immunogenicity predictor (CIMMEX) plus a 22,562-glycoprotein-sequence envelope screen yielded 641 vesiculovirus homologs, and Piry virus glycoprotein (PIRYV) emerged as the lead: lower predicted MHC-epitope density, reduced human seroprevalence, and lower T-cell activation than VSV-G. Native receptor binding is deleted (ePIRYVᴿᴿᴰ) and CD3/CD7 nanobodies are displayed for T-cell-selective transduction; CAR-TRAP producer cells reject B-cell transduction; machine-learning-optimized T-cell-specific promoters restrict CAR expression to lymphocytes; hepatocyte expression is suppressed and phagocytic uptake is blocked. In humanized xenografts, viroVbot3 generates potent BCMA/CD19 responses against multiple myeloma and Claudin-18.2 responses against gastric cancer, with sequential redosing enabled by envelope rotation and minimal off-target organ biodistribution. Frames in vivo CAR-T as a redosable modality rather than a single-shot procedure. (2) Keyi Zhang, Wenhao Ma, Zhijie Wu, Zipei Ren, Congcong Chen, Yan Xia, Dan He, Ziting Yu, Hongze Niu, Jing Qin, Pan Gao, Wenyuan Yang, Yangguang Dai, Xiu Li, Zheyue Dong, Yue Wang, Xiaoyan Dong, Cheng Chen, and Xiaobing N. Wu report IdeSM33, a structure-guided K167R/D226E double mutant of the Streptococcus pyogenes IgG-specific protease IdeS with a fourfold increase in IgG Fc-binding affinity by biolayer interferometry, likely attributable to strengthened hydrogen bonding at the Fc interface. IdeSM33 cleaves serum IgG faster in vitro and depletes circulating IgG in rabbits at doses as low as 0.005 mg/kg — 40× below wild-type. A single 0.2 mg/kg dose degrades binding and neutralizing antibodies against AAV9 within 1–2 days and restores hepatic AAV9 transduction in pre-immunized animals. Doses above 0.2 mg/kg yield higher plasma exposures without proportional pharmacodynamic gains, arguing for a submilligram therapeutic ceiling. Reopens both the seropositive-treatment-naive population and the redosing question for AAV gene therapies, in addition to indications in autoantibody disease and transplant desensitization. (3) Shirin Fatma, Shubham Dubey, Vinh Truong, Seong Guk Park, Hailey Wallace, Alfredo Jose Florez Ariza, Isabella King, and Elizabeth H. Kellogg report cryo-EM structures of the Vibrio cholerae type I-F3 CRISPR-associated transposase (VchCAST) holo integration complex. The structures reveal sequential conformational licensing across Cascade-TniQ, TnsC, and TnsA/B, with the catalytically competent state stabilized only in the fully assembled on-target complex. Comparison with earlier partial-assembly CAST structures identifies conserved architectural features across type I-F3 diversity, arguing for a common licensing logic. The mechanistic significance is a rational handle for improving fidelity and yield of RNA-guided kilobase-cargo genomic insertion without double-strand breaks — the modality-defining advantage of CASTs over Cas9-based editing. (4) Emily Dixon, Farid Azimian, Alicia Joby-Chacko, Ryan Tatum, Chad Boykin, Yi-Hsuan Chen, and Qun Lu (East Carolina) identify a first-in-class small-molecule modulator of the ARF1-C9orf72:SMCR8:WDR41 (CSW) interface. Phospho-ASAP1 (Tyr782) — a marker of suppressed ARF-GAP activity leaving ARF1 in a sustained GTP-bound form — is elevated in the posterior frontal motor cortex of both sporadic ALS and C9orf72-mutant ALS patients, and C9orf72 overexpression fragments the Golgi apparatus similarly to the ARF inhibitor brefeldin A. Hierarchical virtual screening (HTVS → SP → XP), MM/GBSA free-energy filtering, and in-silico ADME/T on the 40,183,714-compound MCULE library, focused on the pocket around SMCR8 Arg147, converge on MCULE-5095997944: 167 nM steady-state dissociation constant on purified ARF1 by SPR (vs 1.41 mM for non-binder ZCL278), reduces GTP-loaded ARF1 after activation, and rescues ARF1-dependent Golgi morphology. The first chemical handle on a genetically anchored protein-protein interaction implicated in ALS/FTD, unifying sporadic and C9-mutant disease under an ARF-GAP-deficiency pharmacology and providing a starting point for medicinal chemistry in an indication with essentially no disease-modifying therapy. Papers 1–3 are v1 with bodies not yet rendered on bioRxiv — summaries are abstract-grounded; paper 4 is v2 posted 2026-07-01 with full text rendered. 2026-07-02-virovbot-in-vivo-car-t-idesm33-vchcast-arf1-csw-c9orf72 Thu, 02 Jul 2026 12:00:00 +0000 429 Four bioRxiv preprints spanning cell therapy, enzyme therapy, genome engineering, and chemical biology: (1) Alam, Kumar, Shukla et al. with Ahmad (Jamia Millia Islamia) — viroVbot, a next-generation in vivo CAR-T platform. Computational immunogenicity prediction (CIMMEX) plus 22,562-sequence envelope screen yields Piry virus glycoprotein (PIRYV) as the lead: lower MHC-epitope density, seroprevalence, and T-cell activation than VSV-G. Native RBD deleted; CD3/CD7 nanobodies displayed; CAR-TRAP producer cells reject B-cell transduction; ML-optimized T-cell-specific promoters restrict CAR expression to lymphocytes. Humanized xenografts show potent BCMA/CD19 (multiple myeloma) and Claudin-18.2 (gastric) responses with sequential redosing via envelope rotation. Frames in vivo CAR-T as a redosable modality. (2) Zhang, Ma, Wu et al. with Wu (Beijing E-Town) — IdeSM33, structure-guided K167R/D226E double mutant of S. pyogenes IdeS with fourfold higher IgG Fc affinity. Depletes circulating IgG in rabbits at 0.005 mg/kg (40× below WT); single 0.2 mg/kg dose degrades anti-AAV9 binding and neutralizing antibodies within 1–2 days and restores hepatic AAV9 transduction in pre-immunized animals. Reopens seropositive-treatment-naive population and AAV redosing question for gene therapy. (3) Fatma, Dubey, Truong et al. with Kellogg (St Jude) — cryo-EM structures of the V. cholerae type I-F3 CAST holo integration complex reveal sequential conformational licensing across Cascade-TniQ, TnsC, and TnsA/B; catalytically competent state stabilized only in the fully assembled complex; conserved architecture across type I-F3 CAST diversity. Rational handle for raising fidelity and yield in kilobase-cargo RNA-guided genomic insertion without double-strand breaks. (4) Dixon, Azimian, Joby-Chacko et al. with Lu (East Carolina) — MCULE-5095997944, first-in-class small molecule from 40M-compound MCULE virtual screen targeting the ARF1-C9orf72:SMCR8:WDR41 interface around SMCR8 Arg147. 167 nM SPR dissociation constant on ARF1; reduces GTP-loaded ARF1 after activation; rescues Golgi morphology. Elevated phospho-ASAP1 (Tyr782) in motor cortex unifies sporadic and C9-mutant ALS under a common ARF-GAP-deficiency pharmacology and provides a starting point for medicinal chemistry in ALS/FTD. Papers 1–3 are v1 with bodies not yet rendered — summaries are abstract-grounded; paper 4 is v2 posted 2026-07-01 with full text rendered. false Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Clyde Smith, Ailiena O. Maggiolo, Chasity Jonosko, Maura E. Charette, Nilakshi Paul, Marta Toth, Guillermo Calero, Sierra M. Carr, Silvia Russi, and colleagues with Sergei B. Vakulenko (Notre Dame), Alexander Deiters (Pittsburgh), and Aina E. Cohen (Stanford / SLAC) present CAGE-TRX, a genetically encoded active-site photocaging strategy for time-resolved crystallography. Time-resolved crystallography synchronizes catalysis inside the crystal and captures transient intermediates, but existing pump-probe schemes lean on caged substrates that must be redesigned for each enzyme class. CAGE-TRX flips the strategy by installing a photocaged non-canonical amino acid via genetic code expansion at a catalytic residue. Light removes the cage, restores the active-site side chain, and initiates the reaction in unison across the crystal, enabling both pump-release-quench-probe cryocrystallography and pump-release-probe room-temperature serial crystallography. Using β-lactamases as a model, the team demonstrates efficient decaging, activity recovery, and direct structural visualization of reaction intermediates. The significance for drug discovery: mechanistic snapshots of catalysis have historically depended on custom substrate chemistry per enzyme class; CAGE-TRX decouples reaction initiation from substrate design and makes serial time-resolved crystallography accessible for any enzyme where a catalytic residue can host a photocage — including the many drug-discovery-relevant enzyme families that have resisted mechanistic structural characterization. (2) Chuanxu Cheng, Quanxin Ning, Juan Du, Jianati Dawulieti, Chenyang Guo, Madi Sun, Kunbao Zhang, Haijun Li, Qunjie Bi, Jian Li, Zhen Wu, Hanyao Huang, Zhi-Liang Ji, Jin-Zhi Du, Chao Yang, Dan Shao and colleagues with Kam Leong (Columbia and South China University of Technology) report an 8-arm PEG-conjugated netilmicin as a multivalent NET-associated chromatin neutralizer for severe sepsis. Neutrophil extracellular traps are protective locally but disastrous when they spill into circulation — the exposed DNA-histone chromatin lattice engages TLR4 and TLR9 on macrophages and drives systemic cytokine release. From a library of multivalent aminoglycoside-displayed materials, 8-arm PEG-netilmicin emerges as the lead: compared to 2- and 4-arm counterparts, it binds DNA-histone chromatin complexes with substantially higher avidity through stable multivalent noncovalent contacts, retains full aminoglycoside antibacterial activity, and suppresses NET-induced TLR4/TLR9 signaling in macrophages. In severe septic mice, intravenously administered 8-arm netilmicin accumulates preferentially in inflamed tissues, reduces bacterial dissemination and systemic cytokine output, protects multiple organs, and improves survival. The conceptual advance is naming the NET-associated DNA-histone chromatin complex itself — not the neutrophil, not free DNA — as the actionable extracellular target, and demonstrating that a rationally designed multivalent material can hit that target selectively while sparing the antibacterial pharmacology of an approved drug class. (3) Christopher Mullally, Bojana Stefanovska, Yanjun Chen, Harshita Gupta, Michael Carpenter, and Reuben Harris introduce the collapsing retrovirus (CRV), a strategy for delivering viro-toxic cargoes that cannot be expressed in producer cells. The motivating example is the DNA cytosine deaminase APOBEC3B, which mutates viral cDNA during reverse transcription and collapses vector titer. The APOBEC3B minigene is interrupted by a translation stop cassette flanked by direct repeats of the surrounding APOBEC3B sequence — the intact minigene is present in producer cells only as a disrupted, non-functional form. During transduction, the retroviral reverse transcriptase, which naturally templates high-frequency recombination across direct repeats, excises the stop cassette and reconstitutes the full-length cargo in the transduced cell. Reconstitution reaches ~90% on its own and approaches ~100% when the minigene is coupled in-frame to a selectable marker. Beyond APOBEC3B delivery for functional deamination studies, the broader value is a generic delivery paradigm for any cargo whose expression during vector production would otherwise self-defeat — restriction factors, toxins, cell-autonomous cytokines, gain-of-function CRISPR effectors — extending retroviral gene delivery into a class of payloads that has been effectively off-limits. (4) Tianli Xiao, James R. Brewer, Maximillian J. Carlino, Ailin Han, Yamato Takabe, Chia-Yi Lee, Fengrui Zhang, Mi Chen, Holly N. Blackburn, Amin H. Nassar, Qiankun Wang, Kristen Brennand, Liang Shan, Essen Sefik, Diane S. Krause and colleagues with Richard Flavell at Yale resolve the puzzle of why loss-of-function DPP9 mutations cause severe pancytopenia in Hatipoglu syndrome patients (necessitating bone marrow transplantation) while Dpp9 mutant mice have normal hematopoiesis. They combine CRISPR editing of primary human CD34⁺ hematopoietic stem and progenitor cells with engraftment into MISTRG6 humanized mice to run reverse genetics directly on the human system in vivo. Human DPP9 deletion is efficient and durable and faithfully reproduces peripheral and bone-marrow cytopenia through a cell-intrinsic mechanism — with almost no transcriptional response, implicating post-transcriptional regulation. Mechanistically, DPP9 loss activates the CARD8 inflammasome and drives pyroptosis of human HSPCs, while the closely related NLRP1 inflammasome is dispensable. The therapeutic lever this identifies is CARD8 — a target with small-molecule inhibitors now moving toward the clinic in other inflammasomopathy indications — and the broader methodological point is that MISTRG6 humanized-mouse reverse genetics on primary human HSPCs is now a tractable route to interrogate human-specific hematopoietic disease mechanisms that mouse knockouts miss. 2026-07-01-cage-trx-net-chromatin-8arm-crv-apobec-card8-dpp9 Wed, 01 Jul 2026 12:00:00 +0000 417 Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Smith, Maggiolo, Jonosko et al. with Vakulenko (Notre Dame), Deiters (Pittsburgh), and Cohen (Stanford/SLAC) — CAGE-TRX, a genetically encoded active-site photocaging strategy for time-resolved crystallography. A photocaged non-canonical amino acid is installed via genetic code expansion at a catalytic residue; light removes the cage, restores the side chain, and starts the reaction in unison across the crystal, enabling pump-release-quench-probe cryocrystallography and pump-release-probe room-temperature serial crystallography without redesigning the substrate. Demonstrated with β-lactamases: efficient decaging, activity recovery, and direct visualization of reaction intermediates. Decouples reaction initiation from substrate design and opens serial time-resolved crystallography to any enzyme class with a photocageable catalytic residue — including drug-discovery-relevant enzyme families that have resisted mechanistic structural characterization. (2) Cheng, Ning, Du et al. with Leong (Columbia / SCUT) — 8-arm PEG-conjugated netilmicin as a multivalent NET-associated chromatin neutralizer. 8-arm construct binds DNA-histone chromatin complexes with substantially higher avidity than 2- and 4-arm counterparts through stable multivalent noncovalent contacts, retains full aminoglycoside antibacterial activity, and suppresses NET-driven TLR4/TLR9 macrophage inflammation. In severe septic mice, intravenous 8-arm netilmicin accumulates in inflamed tissue, reduces bacterial dissemination and cytokine output, protects multiple organs, and improves survival. Names NET-associated chromatin as an actionable extracellular target and demonstrates multivalent chromatin targeting as a rational modality for sepsis. (3) Mullally, Stefanovska, Chen, Gupta, Carpenter, Harris — the collapsing retrovirus (CRV). APOBEC3B minigene interrupted by a translation stop cassette flanked by direct repeats of surrounding sequence is reconstituted during transduction by retroviral RT-catalyzed recombination across the repeats. Full-length cargo restored at ~90% efficiency (approaching 100% when coupled in-frame to a downstream selectable marker). Generic delivery paradigm for any viro-toxic payload that cannot be expressed in producer cells — restriction factors, toxins, cell-autonomous cytokines, gain-of-function CRISPR effectors. (4) Xiao, Brewer, Carlino et al. with Flavell (Yale) — MISTRG6 humanized-mouse reverse genetics combining CRISPR editing of primary human CD34⁺ HSPCs with in vivo engraftment reproduces the pancytopenia of human DPP9 deficiency (Hatipoglu syndrome). Cell-intrinsic, post-transcriptional mechanism (minimal transcriptional response) driven by CARD8 inflammasome-mediated HSPC pyroptosis; NLRP1 is dispensable. Explains why Dpp9 mutant mice have normal hematopoiesis and identifies CARD8 — with small-molecule inhibitors advancing clinically in other inflammasomopathies — as the actionable drug target. false Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Petar Hristov, Pooria Daneshvar Kakhaki, Talia Tzadikario, and colleagues with Ryan A. Flynn at Boston Children's, Jeffrey D. Esko at UC San Diego, Jian Liu, and Miten Jain introduce HS-nano-seq, the first nanopore platform for intact, single-molecule analysis of heparan sulfate. Heparan sulfate is one of the most consequential extracellular biopolymers — it sequesters and presents growth factors, organizes FGFR and Wnt signaling, gates host-pathogen entry, and shapes coagulation — but it cannot be sequenced; existing methods cleave the chain to fragments and read aggregate composition. The conceptual move is to treat heparan sulfate as biochemically analogous to a nucleic acid: rapid nucleic-acid purification, DNA-adapter conjugation to form HS-DNA chimeras resolved as discrete species by capillary electrophoresis, and single-molecule nanopore reads whose ionic-current fingerprints encode sulfation patterns. Synthetic standards classify by chain length and sulfation, and cell-derived heparan sulfate reveals heterogeneity in binding motifs across cell types that bulk methods cannot see. The broader claim is that the framework extends to other extracellular matrix glycopolymers, bringing a molecular class that has resisted systematic interrogation onto the same sequencing workflow as nucleic acids. (2) Giulia Saronio, Gaia Antonini, Ankit Jain, and colleagues with Roger Geiger in Bellinzona, plus Andrew deMello and Stavros Stavrakis at ETH Zurich, built a droplet single-cell CRISPR screen that co-encapsulates a single primary human CD8⁺ T cell with a cancer cell, measures rapid target-cell death within the droplet, and recovers the sgRNA from phenotype-sorted droplets. The screen recovered canonical nodes — TCR signaling, synapse formation, granule exocytosis, PTEN, RASA2, FOXO1, AFAP1L2 — and validated hits across bispecific engager and TCR-engineered settings. The unexpected finding is that knocking out RPTOR or RHEB — components that should support T cell function through mTORC1-driven anabolism — instead enhanced rapid cytotoxic execution while reducing mTORC1 output, increasing AKT phosphorylation, and attenuating anabolic programs. Transient pharmacologic mTORC1 inhibition reproduced this rapid-killing state and improved antitumor activity after adoptive transfer. The therapeutic implication is concrete: a brief mTORC1 inhibitor pulse around the point of CAR-T or TCR-T infusion may shift transferred cells from growth toward execution. (3) Adini Q. Arifah, Constantinos Patinios, Rachael Larose, and Chase L. Beisel show that Cas9, when its tracrRNA is reprogrammed to hybridize a cellular RNA, can be repurposed for both A-to-I RNA editing and trans-splicing. Fusing the ADAR2 deaminase domain to dead SpyCas9 and systematically engineering the reprogrammed tracrRNA (Rptr) achieved efficient and tunable A-to-I editing with on- and off-target profiles comparable to dCas13. The platform extended to the compact CjeCas9 after HNH deletion and PAM-interacting domain mutations, and Rptrs that block splicing enabled both 3′ and 5′ trans-splicing. Two consequences: a second, mechanistically independent RNA-editing chassis built on the most heavily engineered nuclease in biotechnology, and a direct programmable handle on alternative splicing — which dominates therapeutic axes in neurodegeneration, muscular dystrophy, and cancer. (4) Josephine L. Ramirez, Qixiang He, Tanner Wiegand, and colleagues with Samuel H. Sternberg and Israel S. Fernandez at Columbia present cryo-EM reconstructions of two evolutionarily diverse DRT10 reverse-transcriptase-noncoding-RNA systems. DRT10 is a bacterial defense-associated reverse transcriptase that catalyzes protein-primed tandem-repeat DNA synthesis in a mechanism strikingly analogous to eukaryotic telomerase; the structural basis was unknown. The structures reveal an unanticipated 2:1 architecture, with two RT monomers binding opposite sides of a single pseudo-symmetric ncRNA. Biochemistry shows each monomer reverse-transcribes the template encoded on its respective side, but only one generates the long repetitive product, with the template sequence defined by the distance between two flanking stem-loop anchors. Combined with prior work on DRT2, DRT3, and DRT9, the result establishes a conserved mechanistic logic for ncRNA-templated tandem-repeat synthesis across Class 2 UG antiviral systems. The horizon application is a programmable in-cell DNA writer with a tunable repeat length, complementary to prime editing for inserting repeat-rich sequences. 2026-06-30-hs-nano-seq-droplet-tcell-crispr-rptrs-cas9-rna-edit-drt10 Tue, 30 Jun 2026 12:00:00 +0000 377 Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Hristov, Daneshvar Kakhaki, Tzadikario et al. with Flynn (Boston Children's), Esko (UCSD), Liu, and Jain — HS-nano-seq, the first single-molecule nanopore platform for intact heparan sulfate. Treats HS as biochemically analogous to a nucleic acid: rapid nucleic-acid purification + DNA-adapter conjugation + capillary-electrophoresis resolution + nanopore ionic-current fingerprints that encode sulfation patterns. Synthetic standards classify by length and sulfation; cell-derived HS reveals chain-level heterogeneity in binding motifs across cell types. First sequencing-style readout for an ECM glycopolymer; framework extends to other glycan classes. (2) Saronio, Antonini, Jain et al. with Geiger (Bellinzona) + deMello + Stavrakis (ETH) — droplet single-cell CRISPR screen co-encapsulates one primary human CD8⁺ T cell with one cancer cell; measures rapid target-cell death and recovers sgRNAs from phenotype-sorted droplets. Recovered TCR signaling, synapse formation, granule exocytosis, PTEN, RASA2, FOXO1, AFAP1L2 — validated across bispecific engager and TCR-engineered settings. Counterintuitive RPTOR or RHEB knockout enhances rapid cytotoxic execution while reducing mTORC1 output, raising AKT phosphorylation, and attenuating anabolic programs; transient pharmacologic mTORC1 inhibition reproduces rapid-killing state and improves antitumor activity after adoptive transfer. Brief mTORC1 brake as a concrete CAR-T/TCR-T manufacturing lever. (3) Arifah, Patinios, Larose, Beisel — Rptrs: tracrRNAs reprogrammed to hybridize cellular RNAs recruit dSpyCas9-ADAR2dd to drive A-to-I RNA editing with on/off-target profiles comparable to dCas13; compact CjeCas9 tailored for RNA targeting via HNH deletion + PAM-interacting domain mutation; splice-blocking Rptrs enable 3′ and 5′ trans-splicing. Mechanistically independent second RNA-editing chassis on the most heavily engineered nuclease in biotechnology + direct handle on alternative splicing. (4) Ramirez, He, Wiegand et al. with Sternberg + Fernandez (Columbia) — cryo-EM of two evolutionarily diverse DRT10 antiviral RT systems reveals 2:1 RT-ncRNA architecture; two RT monomers bind opposite sides of pseudo-symmetric ncRNA, each templated from its own face, but only one generates the long repetitive product; repeat length defined by distance between flanking stem-loop anchors. Conserved mechanistic logic for ncRNA-templated tandem-repeat DNA synthesis across Class 2 UG antiviral systems; horizon application as programmable in-cell DNA writer with tunable repeat length, complementary to prime editing for repeat-rich insertions. false Four bioRxiv preprints, all v1 posted in the last 48 hours, full text retrieved via Jina reader proxy for all four. (1) Lorenzo Serra, Claire O'Brien, Molly Patterson, and colleagues with Alexander Price at the University of Pennsylvania show that pharmacological inhibition of the host spliceosome selectively suppresses nuclear-replicating viruses while leaving cytoplasmic RNA viruses unaffected. Low-dose SF3B1 inhibitors — two mechanistically distinct spliceosome-targeting compounds already in clinical oncology development — or genetic depletion of SF3B1 disrupted splicing of complex adenoviral late transcripts while largely sparing simple early transcripts. The consequence was impaired adenoviral DNA replication, reduced late protein accumulation, and multi-log decreases in infectious progeny output. The same inhibitors blocked RNA splicing and replication of herpes simplex virus 1 (a DNA virus) and influenza A (an RNA virus that requires host nuclear processing), while cytoplasmic RNA viruses lacking spliceosome dependence were unaffected — confirming on-target mechanism. The frame is that spliceosome complexity — particularly the multi-exon, interdependent transcriptional programs that nuclear viruses evolved to exploit — creates a selective vulnerability: the virus depends on correct splicing more acutely than the host does under low-dose inhibition. Broad-spectrum antiviral activity across adenovirus, HSV-1, and influenza A with compounds already proven safe in cancer patients represents a near-term path to a new antiviral class orthogonal to all existing approved antivirals. (2) Shahama Taifour, Christopher Wallis, Edina Wang, and colleagues with Pilar Blancafort at the University of Western Australia describe the first non-viral, in vivo delivery platform for dCas9-KRAB ribonucleoproteins (RNPs) using guanidino-functionalized PAMAM dendrimers. The target is EWSR1-FLI1, the chimeric transcription factor that drives approximately 85% of Ewing sarcoma cases — an aggressive pediatric malignancy with no small-molecule strategy for directly targeting the oncogenic driver. EWSR1-FLI1 lacks the binding pockets required for conventional drugs. Intracellular delivery of dCas9-KRAB/RNPs loaded in PAMAM polymers achieves robust EWSR1-FLI1 repression both in established cell line xenografts and in patient-derived xenografts (PDXs), accompanied by potent antitumor effects. The PAMAM vehicle is non-immunogenic compared to viral delivery, compatible with clinical manufacture, and the guide RNA component is swappable, making the platform broadly applicable to any hard-to-drug oncogenic transcription factor or fusion protein. First non-viral in vivo CRISPRi RNP delivery demonstration. (3) Leigh-Ana Rossitto, Tsanwen Lu, Yuliang Ma, Pallavi Kaila Sharma, and colleagues with Susan Taylor, David Gonzalez, Xu Chen, and J. Silvio Gutkind at UC San Diego perform multiplexed mass spectrometry-based proteomic and phosphoproteomic profiling of cell lines with selective double knockout of type I (RIα/RIβ) or type II (RIIα/RIIβ) PKA regulatory subunits under basal conditions and glycolytic stress. RI and RII loss have strikingly opposite consequences: RI KO increases both the abundance and kinase activity of the PKA catalytic Cα isoform and broadly elevates substrate phosphorylation, while RII KO decreases the Cβ isoform's abundance, activity, and phosphorylation output. Both genotypes impair metabolic flexibility to glycolytic stressors, but in opposing directions. The most consequential differentially phosphorylated site maps to Tau, the microtubule-associated protein whose hyperphosphorylation is a hallmark of Alzheimer's disease. RI KO increases Tau phosphorylation through two mechanisms: elevated Cα kinase activity and a higher binding affinity of Tau for the negatively charged flexible linker of RII subunits — a newly described regulatory interaction that had not been characterized previously. The work establishes PKA subunit identity as a regulatory axis for Tau phosphorylation, with direct implications for how kinase-targeting strategies in Alzheimer's disease should account for PKA isoform specificity. From Scripps Research's perspective, the Taylor lab at UCSD is a longstanding center of PKA structural biology and this work opens a new target surface — the PKA RI-Tau axis — for Alzheimer's drug discovery. (4) Priyanda Giri, Venkatesh Yarra, Moriah Mathis, and colleagues with Rahul Kohli, Ryan Mehl, and E. James Petersson describe a genetic code expansion strategy that installs a switchable fluorophore-quencher pair — acridonylalanine (Acd, fluorophore) and methyltetrazinyl phenylalanine (Tet, quencher) — into proteins expressed in E. coli using amber suppression at two sites simultaneously. No post-translational labeling step is required. The methyltetrazine quencher can be toggled off by bioorthogonal tetrazine ligation or photochemically, providing an internal control experiment within the same labeled protein. Mechanistic studies using Stern-Volmer quenching, fluorescence lifetime measurements, and proline-ruler peptides establish the distance dependence of quenching and validate the physical model. Applied to calmodulin (a calcium-sensing conformational switch), RecA (a DNA damage sensor and driver of SOS activation), and LexA (the transcriptional repressor whose RecA-catalyzed cleavage unlocks the SOS regulon and promotes acquired antibiotic resistance), the constructs report molecular-level conformational changes without external perturbation. The system also enables high-throughput drug screening directly against protein dynamics — allosteric site formation, disordered region compaction, protein-protein interface remodeling — using a standard fluorescence readout and without post-translational labeling, removing a bottleneck in screening for modulators of intrinsically dynamic targets. 2026-06-29-spliceosome-antiviral-crispri-ewsr1-pka-tau-gce-fret Mon, 29 Jun 2026 12:00:00 +0000 454 Four bioRxiv preprints, all v1 posted in the last 48 hours, full text retrieved for all four: (1) Serra, O'Brien, Patterson et al. with Price (UPenn) — low-dose SF3B1 inhibitors selectively disrupt splicing of complex adenoviral late transcripts, impairing DNA replication and producing multi-log drops in infectious output; same mechanism blocks HSV-1 and influenza A but not cytoplasmic RNA viruses. Shared spliceosome dependence across nuclear-replicating pathogens as a broad-spectrum antiviral vulnerability exploitable with compounds already in clinical oncology. (2) Taifour, Wallis, Wang et al. with Blancafort (UWA) — first non-viral in vivo delivery of dCas9-KRAB RNPs via guanidino-PAMAM dendrimers silences EWSR1-FLI1, the hard-to-drug fusion oncogene driving ~85% of Ewing sarcoma, in established and patient-derived xenografts with potent antitumor effects. Guide-RNA-swappable platform extensible to any hard-to-drug oncogenic transcription factor. (3) Rossitto, Lu, Ma et al. with Taylor/Gonzalez/Chen/Gutkind (UC San Diego) — multiplexed proteomics/phosphoproteomics of RIα/RIβ vs. RIIα/RIIβ double knockouts: RI KO elevates Cα activity and substrate phosphorylation; RII KO reduces Cβ activity; both impair metabolic flexibility. Most differentially regulated site is Tau — RI KO increases Tau phosphorylation through elevated Cα activity and newly discovered higher Tau affinity for the RII flexible linker. PKA subunit identity as a regulatory axis for Tau hyperphosphorylation in Alzheimer's disease. (4) Giri, Yarra et al. with Kohli/Mehl/Petersson — genetic code expansion dual-encodes acridonylalanine (Acd fluorophore) + methyltetrazinyl phenylalanine (Tet quencher) into proteins in E. coli; Tet toggle by bioorthogonal or photochemical reaction provides internal control. Applied to calmodulin, RecA, and LexA (whose cleavage by RecA drives acquired antibiotic resistance); reports conformational dynamics at molecular resolution and enables HTS drug screening without post-translational labeling. false Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Bo Liu, Benjamin Jung Fair, Zhiling Kuang, and colleagues with Jingxin Wang at the University of Chicago report CP3, a risdiplam-derived compound that depletes cellular prion protein (PrP) by activating a cryptic exon in the PRNP gene, routing the resulting mRNA for nonsense-mediated decay. Risdiplam corrects a splicing defect in SMA; CP3 exploits the same chemistry in the opposite direction — introducing a new exon rather than restoring a skipped one — to redirect PRNP mRNA away from functional protein. Cellular PrP falls by roughly 70%. Mechanistically, CP3 activity is strictly dependent on the alternative splicing factor Luc7L, a previously unrecognized requirement for small-molecule cryptic exon activation. Co-administration of CP3 with PTC258, a Luc7L activator, lowers PrP levels in the brains of transgenic mice. Prion diseases (CJD, fatal familial insomnia, GSS, kuru) are uniformly fatal with no approved disease-modifying therapy; lowering PrP before misfolding propagates is the most validated preventive strategy, and CP3 provides the first orally deployable small-molecule framework for doing so. (2) Sizhun Li, Chelsea May Brown, Ruby Hao Sun, Phillip Stansfeld, and Shu-Sin Chng at the National University of Singapore reconstitute MmpL3 — the essential mycobacterial inner membrane flippase for trehalose monomycolate (TMM), a mycolic acid precursor — in artificial lipid bilayers and quantify its proton translocation activity. MmpL3 is among the highest-confidence drug targets in tuberculosis: without it, bacteria cannot build their outer membrane and die. The team finds that known MmpL3 inhibitors binding in the central channel do not suppress proton pumping in this assay. What does suppress it completely is cardiolipin (CL), a phospholipid abundant in the mycobacterial inner membrane. They map the CL-binding site computationally and show that an MmpL3 variant with substitutions at that site is insensitive to CL in vitro and unable to support growth of Mycobacterium smegmatis, confirming the CL-binding interaction is essential. The CL-binding surface is structurally distinct from the channel sites targeted by all existing inhibitor classes, opening a new druggable address on a target that has already undergone extensive medicinal chemistry. (3) Ashleigh Kropp, Jamie Blaza, Alison Parkin, Rhys Grinter, and colleagues across Monash University, the University of Edinburgh, York, and collaborating institutions report that Bacillus subtilis forms a quinone-transporting pseudomembrane composed of filaments of the NADH dehydrogenase Ndh and the quinone-transporting protein Ncp. Cellular respiration is surface-area limited because hydrophobic quinone carriers must diffuse through lipid bilayers; eukaryotes solved this with mitochondria, Gram-negative bacteria with internal membrane invaginations. Gram-positive bacteria (Bacillota) were thought to lack an equivalent solution. Ndh and Ncp co-assemble with phospholipids into complexes with a solvent-excluded hydrophobic lumen that sequesters quinones, and these complexes polymerize into filaments that link chambers into a continuous quinone conduit, amplifying electron transport capacity without consuming additional membrane. Cryo-EM, lipidomics, and molecular dynamics characterize the structure. Phylogenetic analysis shows the system is widespread across Bacillota, which includes Staphylococcus, Streptococcus, Clostridium, and Bacillus. The Ndh-Ncp filament is unique to Gram-positive bacteria and essential for their respiratory metabolism, with no structural equivalent in humans. (4) Ashwani Kumar Gupta, Ekta Minocha, Jiao-Jing Wang, Zhenxiao Tu, Zheng J. Zhang, and Jason Wertheim at Northwestern and the University of Arizona describe reconstituted kidney scaffolds seeded with both iPSC-derived nephron progenitors and ureteric bud progenitors simultaneously. Using both progenitor types produces more advanced nephron structures than either alone, with interconnected nephron tubules and collecting ducts evident by structural analysis. Engraftment of recellularized scaffolds into immunocompromised mice yields vascularization and maturation; vascular anastomosis with the host is confirmed by erythrocytes present inside the graft vasculature. Nephron-secreted proteins are detected in the host mouse urine, demonstrating functional secretion and drainage. The platform is also presented for drug screening, where a three-dimensional, vascularized, nephron-bearing construct allows pharmacological assessment of renal function in vitro and in vivo. 2026-06-28-prion-splice-cp3-mmpl3-cardiolipin-quinone-filaments-kidney-scaffold Sun, 28 Jun 2026 12:00:00 +0000 312 Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Liu, Fair, Kuang et al. with Wang (U Chicago) — CP3, a risdiplam-derived compound, activates a cryptic exon in PRNP via Luc7L-dependent splice switching, routing PrP mRNA for degradation and achieving ~70% cellular PrP depletion; co-administration with Luc7L activator PTC258 lowers PrP in transgenic mouse brains. First small-molecule framework for PrP depletion in prion disease; demonstrates cooperative small-molecule splicing modulation as a PrP-reduction strategy. (2) Li, Brown, Sun et al. with Chng (NUS) — MmpL3 reconstituted in lipid bilayers: cardiolipin (CL) abolishes proton translocation; channel-binding TB inhibitors do not. CL-binding site mapped computationally; site mutant loses CL regulation in vitro and fails to support M. smegmatis growth — confirming functional essentiality. New druggable surface on TB's most validated target, orthogonal to all existing inhibitor classes. (3) Kropp, Zdorevskyi et al. with Grinter/Blaza (Monash/Edinburgh) — Ndh + Ncp co-assemble with phospholipids into filaments containing a continuous hydrophobic quinone-sequestering lumen, bypassing membrane surface-area limits for electron transport in Gram-positive bacteria. Cryo-EM confirmed; widespread across Bacillota (Staph/Strep/Clostridium/Bacillus); no human structural counterpart. (4) Gupta, Minocha, Wertheim et al. (Northwestern / U Arizona) — iPSC nephron + ureteric bud progenitor co-seeding of decellularized kidney scaffolds produces interconnected collecting ducts; in vivo engraftment achieves host vascular anastomosis (erythrocytes in graft vasculature) and nephron-secreted proteins in mouse urine. Functional integration proof-of-concept for transplantable bioengineered kidney tissue and drug-screening platform. false Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Sydney Victoria Oraskovich, Kazuomori Lewis, and colleagues with David Schaffer at UC Berkeley introduce ELiPS (Expression-Linked Promoter Selection), a high-throughput platform for engineering short synthetic promoters for AAV-based gene therapy. The 5 kb AAV packaging limit constrains therapeutic transgene size, and conventional promoters — CMV, CAG — consume 600–900 bp before the transgene begins. ELiPS assembles libraries of more than one million promoter variants by Golden Gate cloning of random combinations of transcription-factor binding sites upstream of a minimal promoter, GFP reporter, and a unique 3' barcode; expression level is read by sequencing rather than imaging across millions of AAV-transduced cells. The screen identified synthetic ~250 bp promoters surpassing CMV and CAG in vitro. One selected variant drove therapeutic expression of B-domain-deleted Factor VIII in vivo at levels comparable to a liver-specific benchmark, providing proof of concept for a full-length hemophilia A transgene. The platform is substrate-agnostic — swapping the transcription-factor binding-site alphabet repeats the workflow for any target cell type. (2) Kewen Peng, Suryadeep Chakraborty, and colleagues with Hening Lin at Cornell and HHMI report a direct-to-biology campaign against SARS-CoV-2 and MERS-CoV macrodomain 1 (Mac1). Mac1 is a conserved viral ADP-ribosylhydrolase that strips the ADP-ribose mark the host cell places on viral proteins to suppress replication, so its inhibition should restore innate immune signaling. The starting compound, GS-441524 (parent nucleoside of remdesivir), binds Mac1 at ~10 µM. Direct-to-biology iterates without purification: amide-coupling reactions diversify the scaffold and crude products feed directly into a mix-and-read fluorescence polarization assay, enabling hundreds of analogues per day. After iterative rounds, KP-S54 reached 44 nM against SARS-CoV-2 Mac1 and 91 nM against MERS-CoV Mac1. A co-crystal structure of a related derivative defined the binding mode and identified additional optimization vectors. Mac1 conservation across the betacoronavirus family gives any inhibitor a built-in pan-family antiviral profile. (3) Xin Wang, Guanying Zhang, and colleagues with Xiangyang Chi introduce AENCS (Antibody Evolution Nexus with Causal-Driven Simulation), a computational antibody engineering framework applied to the problem of SARS-CoV-2 JN.1 immune escape. JN.1 evades the clinical antibody S309 through an N354-linked glycan that blocks the epitope. AENCS integrates molecular dynamics simulation with causal inference to identify mutations that recruit the glycan as a structural contact rather than route around it. Without explicit prior knowledge of N354 as a binding opportunity, the causal analysis converged on mutations that stabilize the glycan in a defined conformation. The engineered antibody ACC01 achieved ~24-fold improved neutralization of JN.1 over the S309 parent. Cryo-EM confirmed ACC01 engages the N354 glycan as a positive binding anchor — a mechanistic inversion of the viral immune-evasion strategy. N354 glycosylation is conserved across recent variants; ACC01 maintained potent activity against the NB.1.8.1 lineage. (4) Veronika Pister, Zuzana Tatarova, and colleagues with Oliver Jonas, Ernest Fraenkel, and Shawn Davidson introduce a spatial pharmacology platform pairing a microdevice for localized multi-drug delivery with simultaneous cyclic immunofluorescence (CyCIF) and MALDI mass-spectrometry imaging in intact tumor tissue. The paired CyCIF-MALDI dataset spans 1.5 million cells across 27 MMTV-PyMT breast cancer tumor sections and nine treatment programs. Metabolic signatures from MALDI reliably predicted immune phenotype in neighboring tissue, establishing lipid and metabolite distributions as a proxy for local immune organization. The dominant polarization axis was between CSF1R+ tumor-associated macrophages and MPO+ infiltrating myeloid cells concentrated near zones of drug-induced tumor-cell death. Drug-resistant regions harbored a lipid-associated macrophage-like population identifiable by metabolic signature before any immune marker appeared. The platform enables paired, pharmacological-concentration-matched assessment of how multiple agents simultaneously reshape the metabolic-immune microenvironment within a single intact tumor. 2026-06-27-elips-mac1-aencs-spatial-pharma Sat, 27 Jun 2026 12:00:00 +0000 453 Four bioRxiv stories, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Oraskovich, Lewis et al. with Schaffer (UC Berkeley) — ELiPS (Expression-Linked Promoter Selection) screens 1M+ synthetic promoter variants via AAV transduction; Golden Gate assembly of random TFBS combinations + minimal promoter + GFP + unique 3' barcode; identifies ~250 bp ubiquitous promoters surpassing CMV and CAG in vitro; one variant drives B-domain-deleted Factor VIII in vivo at liver-specific benchmark levels. Scalable framework for compact promoter discovery addressing the AAV 5 kb packaging limit. (2) Peng, Chakraborty et al. with Lin (Cornell/HHMI) — direct-to-biology campaign (amide-coupling, no-purification mix-and-read FP assay) converts GS-441524 (~10 µM SARS-CoV-2 Mac1 binder) to KP-S54 (44 nM Mac1 IC50, 91 nM MERS-CoV Mac1 IC50); co-crystal structure defines binding mode and optimization vectors. Mac1 ADP-ribosylhydrolase inhibition as an innate-immunity-restoring pan-betacoronavirus antiviral strategy. (3) Wang, Zhang et al. with Chi — AENCS (molecular dynamics + causal inference) identifies mutations that convert the SARS-CoV-2 JN.1 N354 glycan shield into a binding anchor for ACC01; ~24-fold improved JN.1 neutralization over S309 parent; cryo-EM confirms glycan conformation stabilization; activity maintained against NB.1.8.1. Glycan shields as co-optable antibody contact surfaces. (4) Pister, Tatarova et al. with Jonas/Fraenkel/Davidson — spatial pharmacology platform (microdevice multi-drug delivery + CyCIF + MALDI-MSI); 1.5M cells across 27 tumor sections and 9 treatment programs; metabolic signatures predict immune phenotypes; CSF1R+ TAM vs. MPO+ myeloid polarization axis near drug-induced death zones; LAM-like populations in drug-resistant regions detectable by metabolic signature alone. false Two bioRxiv preprints read in full and two from the abstract (bodies not yet rendered). (1) Ke, Zhang, Fang, Zhao, Zhu, Xu, and colleagues with Longxing Cao introduce IMUSE, a strategy for navigating de novo enzyme space by coupling ultra-high-throughput droplet screening with machine learning. De novo enzyme design can now generate libraries whose structural diversity far exceeds what can be tested conventionally, but the absence of evolutionary metadata means physics-based scorers cannot reliably distinguish active from inactive designs. IMUSE resolves the data gap by screening millions of droplet-encapsulated de novo zinc hydrolase designs using single-droplet double-emulsion fluorescence-activated cell sorting, generating a synthetic fitness landscape analogous to what natural evolution provides for native enzymes. X-ray crystal structures of selected active designs confirmed computational models to sub-angstrom precision. Those structures and functional labels train a graph neural network fusing ESM-2 sequence embeddings with AlphaFold-3 residue-level confidence scores and a catalytic-pocket attention module. Applied prospectively to a second-generation library with expanded structural diversity, the model quintupled the active-hit rate to 30% (vs. ~6% in the initial library), and two selected designs displayed catalytic proficiency approaching native enzymes. In sequence-space navigation, the model guided selection of triple mutants with 2.6- to 4.8-fold catalytic efficiency gains. The central contribution: synthetic fitness landscapes built from ultra-high-throughput screening supply the missing evolutionary data for AI-guided de novo enzyme engineering, transforming a stochastic search into deterministic navigation. (2) Zhao, Xu, Huang, Han, Xie, and colleagues report catalytic-center rewiring of Fanzor2 from Acanthamoeba polyphaga mimivirus — a eukaryotic RNA-guided endonuclease in the TnpB/Cas12 evolutionary lineage — to acquire robust collateral trans-cleavage activity. Fanzors naturally lack the bystander cleavage that Cas12a uses for diagnostics because their RuvC-II active site contains a displaced catalytic glutamate (noncanonical D-altE-D arrangement) rather than the canonical D-E-D triad of TnpBs. Restoring the canonical triad by two substitutions (G416E to recreate the RuvC-II glutamate; E467P to relieve steric constraint) activates robust trans-cleavage of ssDNA reporters while attenuating site-specific cis-cleavage. Systematic scanning of position-467 side-chain volume showed that smaller residues create active-site space for the reconstructed glutamate to adopt catalytically competent geometry for nonspecific reporter cleavage. The engineered variant requires minimal target-adjacent sequence context and is activated by as few as seven nucleotides of guide-target complementarity (vs. 14 for Cas12a). Coupled with recombinase-aided amplification, it drives FINDER, a Fanzor-based diagnostic platform: limit of detection below 2 copies per microliter for HPV16/18, complete concordance with qPCR in clinical specimens, single-guide-RNA coverage of 11/14 high-risk HPV subtypes, and — with the mismatch-sensitive variant ApmFz2-EA — single-nucleotide variant genotyping confirmed against Sanger sequencing in 23 sheep blood samples. (3) Alexandria Fiorenza, Mahika Anand, and Karolin Luger (University of Colorado Boulder / HHMI) — abstract only — define an enzymatically inactive PARP-1 chromatin binding mode. The zinc-binding domains Zn1, Zn2, Zn3, and the BRCT domain cooperatively engage nucleosomal linker DNA and compact undamaged chromatin without triggering catalytic activation; this binding mode is insensitive to PARP inhibitors. The model is that PARP-1 associates with the genome in an inactive surveillance state and transitions to catalytic activation only upon encountering a DNA lesion. For drug design: the inactive chromatin-associated PARP-1 pool is structurally characterized here for the first time and is invisible to olaparib, niraparib, rucaparib, and related inhibitors that engage the active, damage-recruited conformation. (4) Min Jae Lee, Jason Hunt, and colleagues with Mark Hochstrasser (Yale University) and Jason Carlyon (Virginia Commonwealth University) — abstract only — identify selective small-molecule inhibitors of OtDUB, the deubiquitylating enzyme that Orientia tsutsugamushi uses to subvert host ubiquitin-dependent pathways during scrub typhus infection. A fluorescence-based DUB activity screen of a chemically diverse library identified amiloride hydrochloride and gentisic acid as OtDUB inhibitors at low dosage with selectivity over the related Wolbachia CidB and yeast Ulp1 DUBs. Computational docking predicts both compounds engage residues near the OtDUB catalytic pocket consistent with competitive inhibition, supported by enzyme kinetic analysis. In O. tsutsugamushi-infected cells, amiloride — an FDA-approved potassium-sparing diuretic — reduced total cellular deubiquitylating activity and intracellular bacterial load. Scrub typhus kills tens of thousands per year across South and Southeast Asia; drug resistance is documented. The OtDUB selectivity for amiloride over host DUBs indicates a pharmacologically distinct active-site pocket and establishes pathogen-encoded DUBs as a druggable anti-infective target class for genetically intractable obligate intracellular bacteria. 2026-06-26-imuse-fanzor2-finder-parp1-inactive-otdub Fri, 26 Jun 2026 12:00:00 +0000 362 Two bioRxiv stories read in full, two from the abstract (bodies not yet rendered): (1) Ke, Zhang, Fang et al. with Cao — IMUSE couples ultra-high-throughput sdDE-FACS screening of millions of droplet-encapsulated de novo zinc hydrolases with a structure-sequence graph neural network (ESM-2 + AlphaFold-3 pLDDT + catalytic-pocket attention) trained on the resulting synthetic fitness landscape. The model quintuples the active-hit rate to 30% on a second-generation library with expanded structural diversity, and selected designs reach native-enzyme catalytic proficiency. Synthetic fitness landscapes supply the evolutionary data gap that blocks AI-guided de novo enzyme engineering. (2) Zhao, Xu, Huang et al. — catalytic-center rewiring of Acanthamoeba polyphaga mimivirus Fanzor2 (G416E + E467P) restores canonical D-E-D RuvC-II triad, activating robust trans-cleavage (attenuating cis-cleavage); position-467 side-chain volume governs active-site geometry for collateral reporter cleavage. FINDER platform (Fanzor + recombinase-aided amplification): sub-2 copies/uL HPV detection, complete qPCR concordance in clinical specimens, single-guide-RNA coverage of 11/14 high-risk HPV subtypes, SNV genotyping by mismatch-sensitive variant confirmed in 23 samples. Compact eukaryotic CRISPR-diagnostic alternative with relaxed TAM requirements and 7-nt activation threshold. (3) Fiorenza, Anand, Luger (CU Boulder / HHMI) — PARP-1 Zn1+Zn2+Zn3+BRCT domains cooperatively bind nucleosomal linker DNA to compact undamaged chromatin in a catalytically inactive, PARP-inhibitor-insensitive mode; structural basis for genome surveillance in an idle pre-lesion state invisible to olaparib/niraparib/rucaparib. (4) Lee, Hunt et al. with Hochstrasser (Yale) and Carlyon (VCU) — fluorescence-based screen identifies amiloride hydrochloride and gentisic acid as selective OtDUB inhibitors; amiloride (FDA-approved diuretic) reduces total cellular DUB activity and intracellular Orientia tsutsugamushi load in infected cells; selectivity over Wolbachia CidB and yeast Ulp1 argues for pharmacologically distinct bacterial pocket — proof of concept for pathogen-encoded DUBs as anti-infective drug target class in genetically intractable obligate intracellular bacteria. false Four bioRxiv preprints, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Ahmed M. Elshazly, Janakiram R. Vangala and colleagues with Senthil K. Radhakrishnan at Virginia Commonwealth University report tumor-selective Mcl1 degradation by an autophagy-targeting chimera (AUTAC). Mcl1 is a major driver of therapeutic resistance in hematologic malignancies, but direct Mcl1 inhibition has failed clinically because Mcl1 is also essential for cardiomyocyte survival. The AUTAC approach routes Mcl1 to lysosomal rather than proteasomal degradation. In multiple myeloma cell lines and primary cardiomyocytes, AUTAC induced robust cytotoxicity and Mcl1 loss in tumor cells while sparing cardiac cells; in vivo, AUTAC reduced tumor Mcl1 without measurably affecting cardiac Mcl1. The mechanistic basis for selectivity is not engineered binding affinity — it is that cardiac cells express substantially less of the p62/SQSTM1, TRAF6, and UBC13 autophagy delivery machinery required for AUTAC activity, and accumulate less compound intracellularly than myeloma cells. AUTAC enhanced the antitumor activity of carfilzomib and venetoclax, including in resistant models, without worsening cardiotoxicity. Lysosomal degradation modalities can exploit a cellular-context therapeutic window that neither inhibitors nor proteasome-targeting degraders can access. (2) Caiqin Wang, Chen-Ting Ma and colleagues with Jennifer N. Rauch at the University of Massachusetts Amherst — together with Eduard Sergienko, Steven H. Olson, and Michael R. Jackson at Scripps Research — report small-molecule inhibitors of the tau–LRP1 interaction identified by high-throughput screening. The cell-to-cell spread of tau aggregates that drives Alzheimer's disease progression is mediated by the endocytic receptor LRP1, specifically its ligand-binding domain 4 (BD4). The team engineered and purified LRP1-BD4 as a scalable biochemical reagent, built three orthogonal binding assays — fluorescence polarization, split luciferase complementation, and time-resolved FRET — each yielding consistent nanomolar binding affinities and demonstrating competitive displacement by tau, RAP, and a known peptide ligand. An HTS campaign identified candidate inhibitors; selected compounds reduced tau uptake in cells, phenocopying competitive inhibition. This establishes LRP1-BD4 as a biochemically tractable, druggable interface for blocking tau spread — acting upstream of the receiving neuron rather than targeting aggregation. (3) Simon C. Pitchford, K. Miraz Rahman and colleagues at King's College London report the rational design of a pathway-selective platelet P2Y1 receptor inverse agonist. Platelets activate through two distinct G-protein branches: the Gq-PLC axis mediates aggregation and hemostasis, while the Gα12/13-Rho-GTPase axis mediates inflammatory platelet chemotaxis. Current antiplatelet drugs suppress both, increasing bleeding risk. Using molecular docking to identify binding-pocket residues distinguishing the two pathways, the team synthesized a series evolving from nucleotide analogue KMR-82-13 to non-nucleotide lead KSN-159-27. KSN-159-27 displayed characteristics of a pathway-selective inverse agonist at P2Y1R toward Gα12/13-mediated pathways while leaving Gq activation intact; in vivo, it suppressed inflammatory cell recruitment while preserving bleeding time and ADP-induced thromboembolic responses — in contrast to the neutral P2Y1R antagonist MRS2500. First demonstration of rationally designed biased P2Y1R pharmacology for separating platelet anti-inflammatory activity from hemostatic risk. (4) Yanan Wang, Jessica Q. Ma and colleagues with Nicholas A. Kotov at the University of Michigan report MICAL-inspired chiral decavanadate nanoclusters that modulate actin dynamics without disrupting filament architecture. MICAL enzymes oxidize specific methionine residues in actin to regulate turnover, but their large size and cofactor requirements limit practical use. Redox-active decavanadate nanoclusters with tartaric acid — atomically precise inorganic clusters — oxidize the same three methionine residues in globular actin (Met-44, Met-47, Met-176) without cofactors. Met-176 oxidation specifically blocks the backdoor segment to prevent depolymerization, while cryo-EM confirms no structural disturbance to actin filaments. The clusters are membrane-permeable, biocompatible, slowed actin dynamics in live NG108-15 cells, and their docking into actin is computationally predictable — opening inorganic nanocluster chemistry as a pharmacologically orthogonal handle on actin dynamics. 2026-06-25-autac-mcl1-tpd-tau-lrp1-hts-p2y1r-biased-mical-vanadate-actin Thu, 25 Jun 2026 12:00:00 +0000 369 Four bioRxiv stories, all v1 posted in the last 24–48 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Elshazly, Vangala and colleagues with Radhakrishnan (Virginia Commonwealth University) — AUTAC-mediated Mcl1 degradation routes the target to lysosomal rather than proteasomal degradation and achieves tumor-selective killing that spares primary cardiomyocytes and murine heart tissue in vivo. Cardiac sparing is not engineered binding selectivity — it arises from lower p62/SQSTM1 + TRAF6 + UBC13 autophagy machinery expression and reduced intracellular drug accumulation in heart versus myeloma cells. AUTAC enhances carfilzomib + venetoclax activity in resistant models without worsening cardiotoxicity. Lysosomal degradation can access a cellular-context therapeutic window unavailable to inhibitors or proteasome-targeting degraders for a target whose direct inhibition has failed clinically due to on-target cardiac toxicity. (2) Wang, Ma and colleagues with Rauch (UMass Amherst) and Sergienko + Olson + Jackson (Scripps Research) — HTS campaign against LRP1 ligand-binding domain 4, the endocytic interface mediating prion-like tau spread in Alzheimer's disease. Three orthogonal assays (FP, split luciferase, TR-FRET) validated nanomolar binding and competitive displacement by tau, RAP, and a peptide ligand. Hit compounds reduce tau cellular uptake, establishing LRP1-BD4 as a druggable interface for blocking tau propagation upstream of the receiving neuron. (3) Pitchford, Rahman and colleagues (King's College London) — rational design of KSN-159-27, a non-nucleotide pathway-selective inverse agonist at platelet P2Y1R. Compound suppresses Gα12/13-Rho-GTPase-mediated inflammatory platelet chemotaxis while leaving Gq-PLC-mediated aggregation and hemostasis intact; in vivo reduces inflammatory cell recruitment without altering bleeding time or thromboembolic responses. First biased P2Y1R pharmacology separating anti-inflammatory from antithrombotic platelet activity. (4) Wang, Ma, Kotov and colleagues (University of Michigan) — chiral decavanadate nanoclusters with tartaric acid as cofactor-free MICAL-mimetic nano-enzymes. Oxidize Met-44, Met-47, Met-176 in globular actin; Met-176 oxidation blocks the backdoor segment to prevent depolymerization without disrupting filament structure by cryo-EM. Membrane-permeable, biocompatible, active in live cells, computationally dockable. Inorganic nanocluster chemistry as a pharmacologically orthogonal actin-modulation handle. Four fresh bioRxiv preprints — two abstract-only as of this morning (bodies not yet rendered) and flagged in-script as such. (1) Kevin M. Wasko, Mara Maker and colleagues with Jennifer A. Doudna at UC Berkeley engineer T cells as a programmable delivery system for genome-editing enzymes. The premise is that cell-type targeting in therapeutic editing is currently solved badly — LNP tropism and AAV serotype give population-level enrichment, not logic — and that T cells, which already do tissue infiltration and programmed antigen recognition, can carry the editing machinery themselves. Engineered T-lymphoid cells produce and transfer enzymes to target cells in response to a programmable ligand using complex spatiotemporal logic, with both cell-contact-dependent and cell-contact-independent delivery modes, and the circuit is demonstrated in primary human T cells. If it generalizes, adoptive cell therapy and in vivo gene editing collapse into one modality where the cell product is both the engineered effector and the delivery surface for editing the patient's own tissue. (2) Pooja Srinivas, Victor Adebomi and colleagues with Gaurav Bhardwaj and Joseph D. Mougous at the University of Washington use de novo protein design to generate orally available miniprotein inhibitors of Bacteroides fragilis toxin (BFT), a metzincin protease that cleaves E-cadherin, triggers proinflammatory signaling, and has been linked to colorectal cancer. The miniproteins bind distal to the BFT active site and physically block toxin-mediated engagement of the claudin-4 receptor. Disulfide-stabilized variants delivered to the cecum, in drinking water, or via in situ secretion from an engineered live biotherapeutic all neutralized the toxin and prevented BFT-associated gut pathology including tumor formation in mice. The headline is not just that designed miniproteins worked in vivo but that they survived gut conditions well enough to be dosed orally and through a live bug — a mechanism-specific, non-antibiotic intervention that disarms a pathogen without depleting the rest of the microbiome. (3) Jingyi Yu and colleagues confine liquid-liquid phase separation to the plasma-membrane interface itself by co-condensing a cell-penetrating peptide (CPP) with silk fibroin (SF). CPP-to-SF stoichiometry tunes the condensate between liquid-like (crosses the membrane) and solid-like (coats it). Cryo-TEM, STED, FRAP, live-cell tracking, and in situ cryo-FIB/cryo-TEM resolve nucleation at the membrane within 30 s and intracellular entry within 5 min — orders of magnitude faster than prior phase-separation delivery — and the same condensates ferry small molecules, nucleic acids, and antibodies at roughly 10× lower CPP dose. By acquiring cytoplasmic access and then re-exiting, the same interfacial condensates support transcellular traversal across the intact cornea in vivo, lowering intraocular pressure with a 4.2× lower betaxolol dose and delivering otherwise excluded siRNA into the anterior chamber. Phase state, not cargo identity, is the design knob, and the membrane interface becomes a programmable determinant of barrier permeability. (4) A Shenzhen Celconta-led team introduces CLEAN-V (CAR-Less ER-Anchor Vector) for in vivo CAR-T. The known non-clinical safety liability of in-vivo-delivered CAR lentiviruses is that the CAR protein expressed in the packaging cell is incorporated into the virion envelope, leading to off-target transduction of the tumor cells the therapy is meant to attack. CLEAN-V adds an ER-retention signal to the CAR during packaging so it never reaches the cell surface and is excluded from budding virions. Particles show near-complete loss of CAR-mediated tumor-cell transduction, integrate seamlessly into existing third-generation LVV four- or five-plasmid workflows, and still produce CAR-T cells with preserved phenotype and function. A clean, manufacturing-compatible engineering fix for the most consequential off-target failure mode of in vivo CAR-T. 2026-06-24-doudna-t-cell-delivery-mougous-bhardwaj-bft-miniprotein-interfacial-condensate-clean-v Wed, 24 Jun 2026 12:00:00 +0000 423 Four bioRxiv stories, two of them abstract-only as of this morning (bodies not yet rendered) and flagged in-script as such: (1) Wasko, Maker et al. with Doudna (UC Berkeley) — T cells engineered as a programmable delivery system for genome-editing enzymes. In response to a programmable ligand, engineered T-lymphoid cells produce and transfer editing machinery to target cells with complex spatiotemporal logic and both cell-contact-dependent and -independent delivery modes; circuit demonstrated in primary human T cells. Collapses adoptive cell therapy and in vivo gene editing into one modality where the cell product is both engineered effector and targeted delivery surface, with cell-type specificity coming from intercellular recognition rather than biophysics. (2) Srinivas, Adebomi et al. with Bhardwaj/Mougous (University of Washington) — de novo designed orally available miniprotein inhibitors of Bacteroides fragilis toxin (a metzincin protease that cleaves E-cadherin and is linked to colorectal cancer) bind distal to the active site and block claudin-4 receptor binding. Disulfide-stabilized variants administered directly to the cecum, in drinking water, or secreted in situ by an engineered live biotherapeutic all neutralized the toxin and prevented BFT-associated gut pathology including tumor formation in mice. Mechanism-specific, non-antibiotic disarmament of a pathogen with miniproteins robust enough for oral and live-bug delivery. (3) Yu et al. — phase-tuned interfacial condensates: co-condensation of a cell-penetrating peptide with silk fibroin at the plasma-membrane interface; CPP-to-SF stoichiometry tunes liquid-like (crosses) versus solid-like (coats); cryo-TEM/STED/in situ cryo-FIB/cryo-TEM resolve nucleation within 30 s and intracellular entry within 5 min; delivers small molecules, nucleic acids, and antibodies at ~10× lower CPP dose; transcellular traversal across intact cornea in vivo lowers intraocular pressure at 4.2× lower betaxolol dose and ferries siRNA into the anterior chamber. Phase state, not cargo identity, is the design knob. (4) Shenzhen Celconta — CLEAN-V CAR-less ER-anchor vector for in vivo CAR-T. ER-retention signal on the CAR keeps it off the packaging-cell surface so it is excluded from budding virions; particles show near-complete loss of off-target CAR-mediated tumor-cell transduction (the headline non-clinical safety liability of in vivo CAR-T), drop into existing third-generation LVV four- or five-plasmid workflows, and still produce CAR-T cells with preserved phenotype and function. Manufacturing-compatible engineering fix for the in vivo CAR-T delivery platform. Four fresh bioRxiv preprints — two abstract-only as of this morning (bodies not yet rendered) and flagged in-script as such. (1) Kevin M. Wasko, Mara Maker and colleagues with Jennifer A. Doudna at UC Berkeley engineer T cells as a programmable delivery system for genome-editing enzymes. The premise is that cell-type targeting in therapeutic editing is currently solved badly — LNP tropism and AAV serotype give population-level enrichment, not logic — and that T cells, which already do tissue infiltration and programmed antigen recognition, can carry the editing machinery themselves. Engineered T-lymphoid cells produce and transfer enzymes to target cells in response to a programmable ligand using complex spatiotemporal logic, with both cell-contact-dependent and cell-contact-independent delivery modes, and the circuit is demonstrated in primary human T cells. If it generalizes, adoptive cell therapy and in vivo gene editing collapse into one modality where the cell product is both the engineered effector and the delivery surface for editing the patient's own tissue. (2) Pooja Srinivas, Victor Adebomi and colleagues with Gaurav Bhardwaj and Joseph D. Mougous at the University of Washington use de novo protein design to generate orally available miniprotein inhibitors of Bacteroides fragilis toxin (BFT), a metzincin protease that cleaves E-cadherin, triggers proinflammatory signaling, and has been linked to colorectal cancer. The miniproteins bind distal to the BFT active site and physically block toxin-mediated engagement of the claudin-4 receptor. Disulfide-stabilized variants delivered to the cecum, in drinking water, or via in situ secretion from an engineered live biotherapeutic all neutralized the toxin and prevented BFT-associated gut pathology including tumor formation in mice. The headline is not just that designed miniproteins worked in vivo but that they survived gut conditions well enough to be dosed orally and through a live bug — a mechanism-specific, non-antibiotic intervention that disarms a pathogen without depleting the rest of the microbiome. (3) Jingyi Yu and colleagues confine liquid-liquid phase separation to the plasma-membrane interface itself by co-condensing a cell-penetrating peptide (CPP) with silk fibroin (SF). CPP-to-SF stoichiometry tunes the condensate between liquid-like (crosses the membrane) and solid-like (coats it). Cryo-TEM, STED, FRAP, live-cell tracking, and in situ cryo-FIB/cryo-TEM resolve nucleation at the membrane within 30 s and intracellular entry within 5 min — orders of magnitude faster than prior phase-separation delivery — and the same condensates ferry small molecules, nucleic acids, and antibodies at roughly 10× lower CPP dose. By acquiring cytoplasmic access and then re-exiting, the same interfacial condensates support transcellular traversal across the intact cornea in vivo, lowering intraocular pressure with a 4.2× lower betaxolol dose and delivering otherwise excluded siRNA into the anterior chamber. Phase state, not cargo identity, is the design knob, and the membrane interface becomes a programmable determinant of barrier permeability. (4) A Shenzhen Celconta-led team introduces CLEAN-V (CAR-Less ER-Anchor Vector) for in vivo CAR-T. The known non-clinical safety liability of in-vivo-delivered CAR lentiviruses is that the CAR protein expressed in the packaging cell is incorporated into the virion envelope, leading to off-target transduction of the tumor cells the therapy is meant to attack. CLEAN-V adds an ER-retention signal to the CAR during packaging so it never reaches the cell surface and is excluded from budding virions. Particles show near-complete loss of CAR-mediated tumor-cell transduction, integrate seamlessly into existing third-generation LVV four- or five-plasmid workflows, and still produce CAR-T cells with preserved phenotype and function. A clean, manufacturing-compatible engineering fix for the most consequential off-target failure mode of in vivo CAR-T. 2026-06-24-doudna-t-cell-delivery-mougous-bhardwaj-bft-miniprotein-interfacial-condensate-clean-v Wed, 24 Jun 2026 12:00:00 +0000 363 Four bioRxiv stories, two of them abstract-only as of this morning (bodies not yet rendered) and flagged in-script as such: (1) Wasko, Maker et al. with Doudna (UC Berkeley) — T cells engineered as a programmable delivery system for genome-editing enzymes. In response to a programmable ligand, engineered T-lymphoid cells produce and transfer editing machinery to target cells with complex spatiotemporal logic and both cell-contact-dependent and -independent delivery modes; circuit demonstrated in primary human T cells. Collapses adoptive cell therapy and in vivo gene editing into one modality where the cell product is both engineered effector and targeted delivery surface, with cell-type specificity coming from intercellular recognition rather than biophysics. (2) Srinivas, Adebomi et al. with Bhardwaj/Mougous (University of Washington) — de novo designed orally available miniprotein inhibitors of Bacteroides fragilis toxin (a metzincin protease that cleaves E-cadherin and is linked to colorectal cancer) bind distal to the active site and block claudin-4 receptor binding. Disulfide-stabilized variants administered directly to the cecum, in drinking water, or secreted in situ by an engineered live biotherapeutic all neutralized the toxin and prevented BFT-associated gut pathology including tumor formation in mice. Mechanism-specific, non-antibiotic disarmament of a pathogen with miniproteins robust enough for oral and live-bug delivery. (3) Yu et al. — phase-tuned interfacial condensates: co-condensation of a cell-penetrating peptide with silk fibroin at the plasma-membrane interface; CPP-to-SF stoichiometry tunes liquid-like (crosses) versus solid-like (coats); cryo-TEM/STED/in situ cryo-FIB/cryo-TEM resolve nucleation within 30 s and intracellular entry within 5 min; delivers small molecules, nucleic acids, and antibodies at ~10× lower CPP dose; transcellular traversal across intact cornea in vivo lowers intraocular pressure at 4.2× lower betaxolol dose and ferries siRNA into the anterior chamber. Phase state, not cargo identity, is the design knob. (4) Shenzhen Celconta — CLEAN-V CAR-less ER-anchor vector for in vivo CAR-T. ER-retention signal on the CAR keeps it off the packaging-cell surface so it is excluded from budding virions; particles show near-complete loss of off-target CAR-mediated tumor-cell transduction (the headline non-clinical safety liability of in vivo CAR-T), drop into existing third-generation LVV four- or five-plasmid workflows, and still produce CAR-T cells with preserved phenotype and function. Manufacturing-compatible engineering fix for the in vivo CAR-T delivery platform. Four fresh bioRxiv preprints — two abstract-only as of this morning (bodies not yet rendered) and flagged in-script as such. (1) Kevin M. Wasko, Mara Maker and colleagues with Jennifer A. Doudna at UC Berkeley engineer T cells as a programmable delivery system for genome-editing enzymes. The premise is that cell-type targeting in therapeutic editing is currently solved badly — LNP tropism and AAV serotype give population-level enrichment, not logic — and that T cells, which already do tissue infiltration and programmed antigen recognition, can carry the editing machinery themselves. Engineered T-lymphoid cells produce and transfer enzymes to target cells in response to a programmable ligand using complex spatiotemporal logic, with both cell-contact-dependent and cell-contact-independent delivery modes, and the circuit is demonstrated in primary human T cells. If it generalizes, adoptive cell therapy and in vivo gene editing collapse into one modality where the cell product is both the engineered effector and the delivery surface for editing the patient's own tissue. (2) Pooja Srinivas, Victor Adebomi and colleagues with Gaurav Bhardwaj and Joseph D. Mougous at the University of Washington use de novo protein design to generate orally available miniprotein inhibitors of Bacteroides fragilis toxin (BFT), a metzincin protease that cleaves E-cadherin, triggers proinflammatory signaling, and has been linked to colorectal cancer. The miniproteins bind distal to the BFT active site and physically block toxin-mediated engagement of the claudin-4 receptor. Disulfide-stabilized variants delivered to the cecum, in drinking water, or via in situ secretion from an engineered live biotherapeutic all neutralized the toxin and prevented BFT-associated gut pathology including tumor formation in mice. The headline is not just that designed miniproteins worked in vivo but that they survived gut conditions well enough to be dosed orally and through a live bug — a mechanism-specific, non-antibiotic intervention that disarms a pathogen without depleting the rest of the microbiome. (3) Jingyi Yu and colleagues confine liquid-liquid phase separation to the plasma-membrane interface itself by co-condensing a cell-penetrating peptide (CPP) with silk fibroin (SF). CPP-to-SF stoichiometry tunes the condensate between liquid-like (crosses the membrane) and solid-like (coats it). Cryo-TEM, STED, FRAP, live-cell tracking, and in situ cryo-FIB/cryo-TEM resolve nucleation at the membrane within 30 s and intracellular entry within 5 min — orders of magnitude faster than prior phase-separation delivery — and the same condensates ferry small molecules, nucleic acids, and antibodies at roughly 10× lower CPP dose. By acquiring cytoplasmic access and then re-exiting, the same interfacial condensates support transcellular traversal across the intact cornea in vivo, lowering intraocular pressure with a 4.2× lower betaxolol dose and delivering otherwise excluded siRNA into the anterior chamber. Phase state, not cargo identity, is the design knob, and the membrane interface becomes a programmable determinant of barrier permeability. (4) A Shenzhen Celconta-led team introduces CLEAN-V (CAR-Less ER-Anchor Vector) for in vivo CAR-T. The known non-clinical safety liability of in-vivo-delivered CAR lentiviruses is that the CAR protein expressed in the packaging cell is incorporated into the virion envelope, leading to off-target transduction of the tumor cells the therapy is meant to attack. CLEAN-V adds an ER-retention signal to the CAR during packaging so it never reaches the cell surface and is excluded from budding virions. Particles show near-complete loss of CAR-mediated tumor-cell transduction, integrate seamlessly into existing third-generation LVV four- or five-plasmid workflows, and still produce CAR-T cells with preserved phenotype and function. A clean, manufacturing-compatible engineering fix for the most consequential off-target failure mode of in vivo CAR-T. 2026-06-24-doudna-t-cell-delivery-mougous-bhardwaj-bft-miniprotein-interfacial-condensate-clean-v Wed, 24 Jun 2026 12:00:00 +0000 363 Four bioRxiv stories, two of them abstract-only as of this morning (bodies not yet rendered) and flagged in-script as such: (1) Wasko, Maker et al. with Doudna (UC Berkeley) — T cells engineered as a programmable delivery system for genome-editing enzymes. In response to a programmable ligand, engineered T-lymphoid cells produce and transfer editing machinery to target cells with complex spatiotemporal logic and both cell-contact-dependent and -independent delivery modes; circuit demonstrated in primary human T cells. Collapses adoptive cell therapy and in vivo gene editing into one modality where the cell product is both engineered effector and targeted delivery surface, with cell-type specificity coming from intercellular recognition rather than biophysics. (2) Srinivas, Adebomi et al. with Bhardwaj/Mougous (University of Washington) — de novo designed orally available miniprotein inhibitors of Bacteroides fragilis toxin (a metzincin protease that cleaves E-cadherin and is linked to colorectal cancer) bind distal to the active site and block claudin-4 receptor binding. Disulfide-stabilized variants administered directly to the cecum, in drinking water, or secreted in situ by an engineered live biotherapeutic all neutralized the toxin and prevented BFT-associated gut pathology including tumor formation in mice. Mechanism-specific, non-antibiotic disarmament of a pathogen with miniproteins robust enough for oral and live-bug delivery. (3) Yu et al. — phase-tuned interfacial condensates: co-condensation of a cell-penetrating peptide with silk fibroin at the plasma-membrane interface; CPP-to-SF stoichiometry tunes liquid-like (crosses) versus solid-like (coats); cryo-TEM/STED/in situ cryo-FIB/cryo-TEM resolve nucleation within 30 s and intracellular entry within 5 min; delivers small molecules, nucleic acids, and antibodies at ~10× lower CPP dose; transcellular traversal across intact cornea in vivo lowers intraocular pressure at 4.2× lower betaxolol dose and ferries siRNA into the anterior chamber. Phase state, not cargo identity, is the design knob. (4) Shenzhen Celconta — CLEAN-V CAR-less ER-anchor vector for in vivo CAR-T. ER-retention signal on the CAR keeps it off the packaging-cell surface so it is excluded from budding virions; particles show near-complete loss of off-target CAR-mediated tumor-cell transduction (the headline non-clinical safety liability of in vivo CAR-T), drop into existing third-generation LVV four- or five-plasmid workflows, and still produce CAR-T cells with preserved phenotype and function. Manufacturing-compatible engineering fix for the in vivo CAR-T delivery platform. Four bioRxiv preprints, all v1 posted in the last 24–36 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Somaya Abdel-Rahman and Moustafa Gabr at Weill Cornell Medicine introduce HTS-Oracle X, a multimodal deep-learning platform that fuses ChemBERTa SMILES embeddings with extended RDKit descriptors through bidirectional cross-attention, trains on continuous biophysical binding signals rather than binary hit/miss labels, and uses Monte Carlo Dropout for uncertainty-adjusted compound selection. Trained on 45,760 Dianthus TRIC-screened compounds per target under scaffold-aware cross-validation, the model was applied prospectively to a 100,160-compound Enamine library against CD28, TIM-3, and VISTA — three immune-checkpoint targets the field has repeatedly failed to crack with small molecules because of their shallow, featureless binding surfaces. Of 150 model-selected compounds, 45 dose-response confirmed (30%% hit rate, 234–408× enrichment over random prospective baselines, 16 sub-micromolar). Top hits HX-CD28-1, HX-TIM3-1, and HX-VISTA-1 sit at binding affinities of 233, 249, and 345 nM respectively, with on-target functional activity in immune-cell and tumor co-culture assays. Training on real biophysical signals + uncertainty-weighted selection + immediate dose-response and functional validation give the platform a believable prospective claim on a class of targets where retrospective-benchmark-trained AI campaigns have repeatedly failed in the wet lab. (2) Huizhe Wang, Shiju Gu and colleagues with Chongzhao Ran at the Athinoula A. Martinos Center (Mass General/Harvard Medical School) introduce ADLumin-5, a self-photosensitizing chemiluminescent small molecule that degrades misfolded proteins in the brain using light it generates itself. The compound binds aggregate and chemiluminescently produces its own visible light, which sensitizes singlet oxygen to photo-oxidize and photodegrade the protein on contact — bypassing the brain-light-penetration barrier that has limited external-excitation photodynamic approaches. In vitro the chemistry oxidizes and degrades β-amyloid, tau, α-synuclein, and TDP-43, confirmed by LC-MS, MALDI-MS, and Western; for β-amyloid, oxidation reduced cellular toxicity. Four-month longitudinal treatment of 5xFAD mice reduced β-amyloid accumulation by in vivo molecular imaging. Because the chemiluminescence is itself the readout, ADLumin-5 is also an imaging agent for the aggregate it is destroying — a photo-theranostic closing the loop between engagement, action, and readout in one molecule. A chemistry-first, aggregate-class-agnostic alternative to enzyme- or antibody-mediated clearance of misfolded proteins. (3) Juan Estevez-Gallego, Pavel Filipcik and colleagues with Michel Steinmetz at the Paul Scherrer Institut resolve microtubule structures at 1.9–2.2 Å by cryo-EM in four different nucleotide states, revealing the roles of amino acids, ions, and hundreds of water molecules in the GTP hydrolysis cycle. Compaction of the GTP-bound lattice into a more stable pre-hydrolysis state activates a specific catalytic water molecule for γ-phosphate cleavage; hydrolysis intermediates are stabilized by strengthened lattice contacts; release of reaction products in the final GDP state disrupts those contacts and destabilizes the lattice. An atomistic framework for microtubule dynamic instability — and a lattice-state-specific water + ion network as a structural starting point for designing nucleotide-state-selective small molecules against a target class anchoring clinical chemistry from vinca alkaloids to taxanes to maytansinoid ADC payloads. (4) Paul Lukacs and colleagues with Kurt Thorn at Arrepath report real-world results of machine-learning-guided phenotypic HTS for antibiotics against E. coli. Large-scale Learning-to-Rank models were trained on public and proprietary datasets with two simultaneous objectives — maximize phenotypic inhibition and minimize human-cell cytotoxicity. The screen targeted a hyperpermeable lptD- E. coli strain and evaluated pre-plated libraries plus structurally novel cherry-picked compounds, with full hit confirmation, profiling, cytotoxicity counter-screening, and MOA determination. Results: doubled hit rate, 3× fewer toxic hits, activity gains holding against both WT and lptD-, and ML predictive power retained on compounds structurally dissimilar from training data. Industrial validation that ML is now reliably useful as a hit-rate and toxicity filter on a Gram-negative antibiotic campaign. Competing interest declared — all authors are or were employed by Arrepath, which has filed patents on the technologies described. 2026-06-23-hts-oracle-x-checkpoint-adlumin5-photo-theranostic-mt-gtp-arrepath-ml-hts Tue, 23 Jun 2026 12:00:00 +0000 442 Four bioRxiv stories, all v1 posted in the last 24–36 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Abdel-Rahman and Gabr (Weill Cornell Medicine) — HTS-Oracle X multimodal deep-learning platform (bidirectional cross-attention fusion of ChemBERTa SMILES embeddings + extended RDKit descriptors, trained on continuous biophysical binding signals with Monte Carlo Dropout uncertainty quantification) on 45,760 Dianthus TRIC compounds per target, prospectively screened 100,160-compound Enamine library against CD28, TIM-3, VISTA — 45/150 dose-response confirmed (30%% hit rate, 234–408× enrichment, 16 sub-µM), with top hits HX-CD28-1 (KD 233 nM), HX-TIM3-1 (KD 249 nM), HX-VISTA-1 (KD 345 nM) showing on-target functional activity in immune-cell and tumor co-culture. Believable prospective AI platform for small-molecule immune-checkpoint modulators against shallow PPIs that have repeatedly resisted both conventional HTS and retrospective-benchmark-trained AI. (2) Wang, Gu and colleagues with Ran (Mass General/HMS Martinos Center) — ADLumin-5 self-photosensitizing chemiluminescent small molecule binds misfolded-protein aggregates and generates its own visible light, sensitizing singlet oxygen for photo-oxidation/degradation of β-amyloid, tau, α-synuclein, TDP-43 (LC-MS/MALDI/Western-confirmed); reduces Aβ toxicity in cells; 4-month longitudinal 5xFAD treatment reduces Aβ accumulation by in vivo imaging. Chemiluminescence serves as aggregate imaging readout — closing the loop in one photo-theranostic, bypassing the brain-light-penetration barrier that limits external-excitation photodynamic approaches. (3) Estevez-Gallego, Filipcik and colleagues with Steinmetz (Paul Scherrer Institut) — cryo-EM microtubule structures at 1.9–2.2 Å in four nucleotide states resolve hundreds of waters, ions, and side chains driving GTP hydrolysis. Lattice compaction into a pre-hydrolysis state activates a catalytic water for γ-phosphate cleavage; intermediates stabilized by strengthened contacts; release of products in the GDP state disrupts contacts and destabilizes the filament. Atomistic framework for dynamic instability and lattice-state-resolved structural starting point for nucleotide-state-selective small molecules against vinca/taxane/maytansine ADC-payload target space. (4) Lukacs et al. with Thorn (Arrepath) — real-world ML-guided phenotypic HTS against E. coli using Learning-to-Rank models trained on public + proprietary data with simultaneous objectives (maximize phenotypic inhibition, minimize human-cell cytotoxicity); hyperpermeable lptD- mutant screen + cherry-picked novel compounds. Doubled hit rate + 3× fewer toxic hits vs. unguided screening, gains holding against WT and lptD-, ML predictive power retained on structurally dissimilar compounds, full pipeline through hit confirmation + profiling + cytotoxicity + MOA. Industrial validation that ML is now reliably useful as a hit-rate + toxicity filter on a Gram-negative antibiotic campaign. Four bioRxiv preprints, all v1 posted in the last 24 hours with bodies not yet rendered — summaries are abstract-grounded. (1) Patrick R. A. Zanon, Bachuki Shashikadze and colleagues with Henrik Daub at NEOsphere Biotechnologies discover NE26394, a first-in-class cereblon (CRBN)-directed molecular glue degrader that selectively eliminates CDK4. The discovery engine is a high-throughput mass-spectrometry-based proteomics screen across thousands of CRBN-directed compounds, reading out neosubstrate degradation directly rather than relying on serendipitous medicinal chemistry. NE26394 is selective for CDK4 over CDK6 — something the clinical CDK4/6 inhibitors palbociclib/ribociclib/abemaciclib have never delivered, and the source of the on-target CDK6 neutropenia that limits their dosing. Mechanistically, CRBN does not recognize CDK4 directly; CDK4 is handed to the ligase by endogenous INK4-family tumor-suppressor proteins (p16/p15/p18/p19), which are co-recruited into the ternary complex. The degrader reads a cellular state (co-substrate availability), not just a sequence. In CDK4-dependent cancer models, NE26394 phenocopies the anti-proliferative RB-E2F pathway perturbation of the clinical inhibitors. CDK4-over-CDK6 selectivity, degradation over substrate-competitive inhibition, and a rational MS-based discovery workflow — well aligned with the targeted-protein-degradation roadmap. (2) Sophia Nguyen, Isaac Pinner and colleagues with Blagojce Jovcevski and John Bruning at the University of Adelaide solve the first crystal structures of Aspergillus fumigatus mannitol-2-dehydrogenase (apo 1.8 Å, NADH-bound 2.1 Å), mapping a large positively-charged central cavity with Phe45 π-stacking the nicotinamide ring, and characterize 1,4-benzoquinone as a cell-active covalent cysteine-modifying inhibitor (IC50 = 1.2 ± 0.2 nM) that modifies up to five cysteine residues on the enzyme and synergizes with the front-line antifungal voriconazole in whole-cell assays. Mannitol biosynthesis has no human counterpart, and the absence of structural information has held back medicinal chemistry on the pathway; the structure plus covalent-warhead-on-cysteine playbook is the foundation a real antifungal campaign needs against a pipeline-barren target space. (3) Hayden L. Struckman, Ian Field and colleagues with Jonathan R. Silva at Washington University in Saint Louis recast cardiac stress-granule coarsening as the actual pathological inflection point for electrophysiological dysfunction. Under acute and chronic oxidative stress, stress granules localize preferentially to z-lines and intercalated discs; early nascent granules are broadly cytoprotective. Progressive merging (coarsening) disrupts α-actinin and L-type calcium channel (Cav1.2) nanodomains and shortens action-potential duration through a microtubule-dependent mechanism. Nocodazole arrest of coarsening preserves nanodomain integrity and restores AP morphology. The therapeutic window is on granule dynamics, not granule existence; microtubule-mediated coarsening, rather than the condensates themselves, is the druggable handle for limiting proarrhythmic remodeling under cardiac oxidative stress. (4) Emma E. Steffke, Lela Latifi and colleagues with Masaki Terabe at the National Cancer Institute use a heterologous ChAdOx1 + modified vaccinia Ankara prime-boost vaccination to confer therapeutic efficacy against orthotopic checkpoint-blockade-refractory SB28 murine glioblastoma. Vaccination is effective against both the established antigen P1A and a newly identified GBM-associated antigen Gpr149, and adding ICB provides no additional benefit. Systemic vaccination drives robust brain infiltration of antigen-specific CD103+CD69+CD8+ tissue-resident memory (TRM)-like T cells that are polyfunctional and durable in tumor-controlled brains; intracranial adoptive transfer of these TRM-like cells alone protects naive recipients from subsequent tumor challenge. Places the therapeutic burden on driving brain-resident memory directly with a clinically usable viral-vector platform — a rationale for clinical evaluation in an indication where systemic ICB has repeatedly failed. 2026-06-22-cdk4-ne26394-mgd-mannitol2dh-antifungal-sg-coarsening-arrhythmia-chadox1-mva-gbm-trm Mon, 22 Jun 2026 12:00:00 +0000 408 Four bioRxiv stories, all v1 posted in the last 24 hours with bodies not yet rendered — summaries are abstract-grounded: (1) Zanon, Shashikadze and colleagues with Daub (NEOsphere Biotechnologies) — NE26394, first-in-class cereblon-directed molecular glue degrader selective for CDK4 over CDK6 (sidesteps the on-target CDK6 neutropenia limiting palbociclib/ribociclib/abemaciclib), identified by a high-throughput MS-based proteomics screen across thousands of CRBN-directed compounds. CRBN recognition of CDK4 is mediated by co-recruitment of endogenous INK4-family tumor-suppressor proteins (p16/p15/p18/p19) into the ternary complex — the degrader reads cellular state, not sequence. Phenocopies the anti-proliferative RB-E2F pathway perturbation of clinical CDK4/6 inhibitors in CDK4-dependent cancer models. CDK4-selective degradation plus rational MS-based discovery workflow aligned with the TPD roadmap. (2) Nguyen, Pinner and colleagues with Jovcevski/Bruning (University of Adelaide) — first crystal structures of Aspergillus fumigatus mannitol-2-dehydrogenase (apo 1.8 Å, NADH-bound 2.1 Å) with Phe45 π-stacking the nicotinamide ring; 1,4-benzoquinone is a cell-active covalent cysteine-modifying inhibitor (IC50 = 1.2 ± 0.2 nM) modifying up to five Cys residues and synergizing with voriconazole. Foundation for medicinal chemistry against a fungal-specific mannitol-biosynthesis pathway with no human counterpart, in a pipeline-barren target space. (3) Struckman, Field and colleagues with Silva (Washington University in Saint Louis) — stress-granule coarsening identified as the pathological inflection point for cardiac electrophysiological dysfunction. SGs localize preferentially to z-lines and intercalated discs under acute and chronic oxidative stress; early nascent granules are cytoprotective; progressive merging disrupts α-actinin and Cav1.2 nanodomains and shortens AP duration via a microtubule-dependent mechanism. Nocodazole arrest of coarsening preserves nanodomain integrity and restores AP morphology — therapeutic window is on dynamics, not existence; microtubule-mediated coarsening is the druggable handle for proarrhythmic remodeling. (4) Steffke, Latifi and colleagues with Terabe (National Cancer Institute) — heterologous ChAdOx1 + modified vaccinia Ankara prime-boost vaccination drives therapeutic efficacy against orthotopic checkpoint-blockade-refractory SB28 murine glioblastoma, effective against P1A and the newly identified antigen Gpr149; ICB adds nothing. Systemic vaccination drives brain CD103+CD69+CD8+ TRM-like infiltration; intracranial adoptive transfer of TRM-like cells alone protects naive recipients from subsequent challenge. Rationale for clinical evaluation of viral-vector vaccination in a GBM indication where systemic ICB has repeatedly failed. Four fresh bioRxiv preprints. (1) Anne Kaditya Snyder, Filomena Tedesco and colleagues with Gary E. Ward (University of Vermont), Nicholas J. Westwood (University of St Andrews), and Julia Schaletzky (UC Berkeley) identify UCB-9721 as a potent inhibitor of MyoA, the unconventional class XIV myosin motor that drives the gliding motility apicomplexan parasites use to invade host cells. The top hit from a >50,000-compound screen against Toxoplasma MyoA actin-activated ATPase is ~40× more potent than the prior best probe KNX-002 in both enzyme and parasite-motility assays, with no detectable mammalian-cell toxicity and cross-genus activity against Plasmodium falciparum, Cryptosporidium parvum, and Babesia duncani. Comparative docking, targeted TgMyoA mutagenesis, and a directed SAR series pinpoint the sulfonamide group and its water-network-mediated hydrogen bonds to R249 and E275 inside the TgMyoA pocket as the source of the potency gain. Establishes a medicinal-chemistry-ready scaffold against a neglected-tropical-disease target with no close mammalian homologue and rare pan-phylum activity. (2) David Z. Bushhouse, Jiawei Fu, and Julius B. Lucks (Northwestern University) describe tug-of-war (TOW) transcriptional riboswitches, a modular architecture that removes the long-standing sequence-overlap constraint between the ligand-binding aptamer and the expression platform. A ligand-stabilized RNA structure placed immediately upstream of an intrinsic terminator suppresses termination by physically competing with the nascent terminator hairpin for occupancy of the RNA polymerase exit channel. Four working switches were built by mixing and matching aptamers with terminators, and bioinformatic + experimental validation shows ~10% of natural ZTP riboswitches — almost all in gram-positive bacteria — already use this antitermination mechanism. A modular platform for engineered ligand-controlled transcription in synthetic biology and cell-and-gene-therapy circuits. (3) Barsha Dash, Xiaocui He and colleagues with Patrick G. Hogan and Anjana Rao (La Jolla Institute for Immunology) characterize IκBδ — an atypical IκB family member encoded by the NFAT-target gene Nfkbid — as a single transgenic lever that simultaneously addresses the two cardinal CD8 T-cell deficits in solid tumors. Nfkbid loss impairs tumor-infiltrating lymphocyte (TIL) accumulation and exacerbates tumor growth, while ectopic IκBδ overexpression expands TILs, lowers exhaustion-associated transcription factors and inhibitory receptors, elevates cytotoxic-molecule production, and improves tumor control. Domain dissection maps the two phenotypes to separable regions: the shared ankyrin-repeat core that contacts NF-κB is sufficient for TIL accumulation, while the ~150-residue N-terminal region is required for the effector-function/anti-exhaustion program. Domain-specific engineering handle for CAR-T programs that have been chasing accumulation and exhaustion as separate bottlenecks. (4) Ben E. Clifton, Bakhytzhan Akdavletov, Prashant Jain, and Paola Laurino (OIST) show that a spontaneous asparagine-to-succinimide cyclization is structurally encoded and functionally essential in the SAR11 marine bacterium Pelagibacter solute-binding protein SAR11_0655. High-resolution X-ray crystallography and LC-MS/MS independently confirm the modification at Asn269; an N269A substitution lowers melting temperature by >20 °C and reduces binding affinity for the substrate 5-oxoproline by 15–50×. Across SAR11 homologues both Asn and Asp are common at position 269, and the N269D variant also partially cyclizes into the same succinimide. Reframes succinimide from a degradation liability into a stable, encodable design handle for engineering thermostability and affinity without engineered disulfides, crosslinks, or non-canonical amino acids. 2026-06-21-myoa-ucb9721-apicomplexan-tow-riboswitches-ikbdelta-til-succinimide Sun, 21 Jun 2026 12:00:00 +0000 353 Four bioRxiv stories: (1) Snyder, Tedesco and colleagues with Ward (Vermont), Westwood (St Andrews), Schaletzky (UC Berkeley) — UCB-9721, identified from a >50,000-compound screen against Toxoplasma MyoA actin-activated ATPase, is ~40× more potent than the prior probe KNX-002 in both enzyme and parasite-motility assays with no detectable mammalian-cell toxicity, and is active against Plasmodium falciparum, Cryptosporidium parvum, and Babesia duncani. Comparative docking, TgMyoA mutagenesis, and directed SAR pinpoint the sulfonamide and its water-network-mediated H-bonds to R249 + E275 as the source of potency. Medicinal-chemistry-ready scaffold against the class XIV myosin motor — a neglected-tropical-disease target with no close mammalian homologue and pan-phylum activity. (2) Bushhouse, Fu, Lucks (Northwestern) — tug-of-war (TOW) transcriptional riboswitches remove the historical aptamer/expression-platform sequence-overlap constraint by using a ligand-stabilized upstream RNA structure to sterically suppress intrinsic termination via the RNA polymerase exit channel. Four modular switches built from diverse aptamer-terminator pairs; bioinformatic validation shows ~10% of natural ZTP riboswitches (mostly gram-positives) already use this mechanism. Modular platform for ligand-controlled transcription in cell-and-gene-therapy circuits. (3) Dash, He and colleagues with Hogan and Rao (La Jolla Institute for Immunology) — IκBδ (NFAT-target gene Nfkbid) is a single transgenic lever that overcomes the two cardinal CD8 T-cell deficits in solid tumors: loss impairs TIL accumulation and worsens growth; ectopic overexpression expands TILs, lowers exhaustion-associated TFs and inhibitory receptors, and elevates cytotoxic-molecule production. Shared ankyrin-repeat core (NF-κB contact) drives accumulation; ~150-residue N-terminal region drives effector function/anti-exhaustion. Domain-specific engineering handle for CAR-T programs. (4) Clifton, Akdavletov, Jain, Laurino (OIST) — a spontaneous Asn-to-succinimide cyclization in Pelagibacter SAR11_0655 is structurally encoded and functionally essential. X-ray + LC-MS/MS confirm modification at Asn269; N269A drops Tm by >20 °C and binding affinity for 5-oxoproline by 15–50×. Asn and Asp both common at the position across homologues; N269D partially cyclizes into the same succinimide. Reframes succinimide from degradation liability to encodable design handle for engineering thermostability + affinity without disulfides, crosslinks, or non-canonical amino acids. Four bioRxiv preprints from the past 48 hours, all v1 with bodies not yet rendered — summaries are abstract-grounded. (1) Ruwan W. Gunasekara, Lin Zhang and colleagues with Jaime Grutzendler at Yale University introduce ExACT, a platform for cell-type-selective intracellular delivery in the brain that exploits endogenous membrane transporters. In vivo screening of combinatorial fluorescent small-molecule libraries in mouse brain identified chemistries whose uptake is dictated by transporter expression, yielding compounds preferentially entering neurons, astrocytes, pericytes, and endothelial cells. One series was strongly selective for brain and retinal endothelium via Slco1a4-mediated uptake; the selectivity principle transferred to the human orthologue SLCO1A2, highly expressed on brain endothelium and oligodendrocytes, and was confirmed in a humanized mouse model and in human iPSC-derived oligodendrocytes. Ectopic SLCO1A2 expression in neurons via gene therapy creates a synthetic entry port, conferring ExACT-conjugate uptake on otherwise inaccessible cells. Bifunctional compounds linking transporter-targeting motifs to antisense oligonucleotides or small-molecule drugs retained pharmacological activity while conferring transporter-dependent cell-type selectivity. A chemistry-first alternative to AAV-tropism and transcytosis-receptor-antibody routes for precision CNS pharmacotherapy. (2) Ryan H. Abdella, Christopher A. Belica and colleagues with Hideki Aihara and Reuben S. Harris at the University of Minnesota report the cryo-EM structure of wildtype full-length APOBEC3B in complex with its natural antagonist BORF2 — the large subunit of the Epstein-Barr virus ribonucleotide reductase. APOBEC3B is an endogenous cytidine deaminase that mutates ssDNA and is a principal driver of mutagenesis, tumor evolution, therapy resistance, and metastasis across many cancers. The structure shows the tandem N-terminal and catalytic C-terminal domains bridging a novel BORF2 dimer interface in a geometry that distinguishes A3B from the related restriction factor APOBEC3G. Mutational analyses implicate the unique NTD-CTD interaction as an intrinsic regulator of deaminase activity, and BORF2 neutralization operates by capturing A3B into sequestered aggregates through a matrix of primary CTD contacts, secondary NTD contacts, and additional A3B CTD-CTD dimerization. Structural platform for the long-sought A3B-selective-over-A3G inhibitor design and for BORF2-mimetic sequestration strategies. (3) Francesca Gasparin, Jiawang Qiu and colleagues with Miguel Angel Pleitez and Vasilis Ntziachristos at TU Munich present MIRACLES (Mid-IR Assessment of Conformation in Lipids by Ensemble Sensing), a label-free imaging modality for cell membrane potential. The plasma membrane's transient electric field shifts the conformations of lipid acyl chains, which changes their mid-infrared vibrational signature; mid-IR optoacoustic detection turns that signature into a single-cell Vm readout without dye, without electrode, and without genetically encoded indicator. Proof-of-concept tracks glucose-stimulated insulin secretion in β-cells in real time at single-cell resolution. Clears in one shot the dye-perturbation, genetic-indicator-perturbation, and patch-clamp-throughput constraints that have defined functional phenotypic screening on electrogenic cells (β-cells, cardiomyocytes, neurons) for drug discovery. (4) Maximilian Seidel, Shengdi Li and colleagues with Lars M. Steinmetz at EMBL, Judith Frydman at Stanford, Martin Beck, and Mikhail Savitski introduce UbSeRP, ubiquitin-linkage-selective ribosome profiling that maps co-translational ubiquitination (CTU) across the yeast translatome. The endogenous substrate set of ribosome-associated quality control (RQC) expands from a small handful of known cases to thousands. Only a subset of identified stalled-ribosome disomes engages RQC, and biophysical features of the nascent chain predict which ones do — a real codebook for which translation problems get resolved by RQC and which slip through. Widespread RQC-independent CTU is implicated in protein complex assembly, reframing nascent-chain ubiquitination as not just error correction but also organization of how chains assemble together. In a chronological aging model, RQC engagement drops and proteasomal handling of nascent proteins shifts to RQC-independent pathways — a translation-step mechanism for age-related proteostatic decline. 2026-06-20-exact-transporter-delivery-apobec3b-cryoem-miracles-vm-imaging-ubserp-rqc Sat, 20 Jun 2026 12:00:00 +0000 335 Four bioRxiv stories, all v1 with bodies not yet rendered — summaries are abstract-grounded: (1) Gunasekara, Zhang and colleagues with Grutzendler (Yale University) — ExACT platform exploits endogenous membrane transporters for cell-type-selective intracellular brain delivery. In vivo combinatorial fluorescent small-molecule screening identifies chemistries with transporter-controlled uptake into neurons, astrocytes, pericytes, endothelial cells; one series selectively enters brain and retinal endothelium via Slco1a4 and its human orthologue SLCO1A2 (also in oligodendrocytes) in a humanized mouse + iPSC-derived oligodendrocytes. Ectopic SLCO1A2 in neurons via gene therapy creates a synthetic entry port. Bifunctional compounds linking transporter motifs to ASOs or small-molecule drugs retain pharmacology and gain transporter-dependent cell selectivity. Chemistry-first alternative to AAV tropism and transcytosis-receptor-antibody routes. (2) Abdella, Belica and colleagues with Aihara and Harris (University of Minnesota) — cryo-EM structure of full-length wildtype APOBEC3B in complex with EBV antagonist BORF2 (ribonucleotide reductase large subunit). The tandem NTD + catalytic CTD bridge a novel BORF2 dimer interface in a geometry distinct from APOBEC3G; mutational analyses implicate the unique NTD-CTD interaction as intrinsic regulator of deaminase activity; BORF2 neutralization operates by capturing A3B into sequestered aggregates through primary CTD + secondary NTD + CTD-CTD contacts. Structural platform for the long-sought A3B-over-A3G-selective inhibitor design and BORF2-mimetic sequestration strategies in oncology. (3) Gasparin, Qiu and colleagues with Pleitez and Ntziachristos (Technical University of Munich) — MIRACLES mid-IR optoacoustic membrane-potential imaging detects lipid acyl-chain conformational changes under the plasma-membrane electric field dynamics as a label-free, dye-free, electrode-free single-cell Vm readout. Proof-of-concept tracks glucose-stimulated insulin secretion in β-cells in real time at single-cell resolution. Clears in one shot the dye-perturbation, genetic-indicator-perturbation, and patch-clamp-throughput constraints of functional phenotypic screening on electrogenic cells. (4) Seidel, Li and colleagues with Steinmetz (EMBL), Frydman (Stanford), Beck, Savitski — UbSeRP ubiquitin-linkage-selective ribosome profiling maps co-translational ubiquitination across the yeast translatome. Endogenous RQC substrate set expands from few to thousands; only a subset of stalled-ribosome disomes engages RQC and nascent-chain biophysical features predict which ones; widespread RQC-independent CTU is implicated in protein complex assembly; in chronological aging, RQC engagement drops and nascent-protein handling shifts to RQC-independent proteasomal pathways — translation-step mechanism for age-related proteostatic decline. Four bioRxiv preprints from the past 48 hours touching early-amyloid detection chemistry, structure-guided antibody alternatives, proteasomal control of transcription, and a piece of drug repurposing on a glycosylation target. (1) Karolina M. Kleczko, Daniel Gestaut and colleagues with Judith Frydman at Stanford University introduce Q-DOAS, a proximity-quenching plate-reader assay that meaningfully extends what can be measured in protein aggregation. Thioflavin T, the workhorse for amyloid kinetics for fifty years, reports only on mature cross-beta fibrils and is essentially blind to the early oligomers that most current disease models implicate. Q-DOAS site-specifically conjugates a single BODIPY-TMR dye to a target protein and reads self-assembly as real-time quenching — the dye signal drops as monomers come together, well before any cross-beta structure forms. Validated on mutant huntingtin exon 1 and the A53T alpha-synuclein variant, used to dissect mutational effects on early oligomerization, run as a plate-reader primary screen for protein and small-molecule inhibitors, and used to quantify seeding activity in cellular and mouse Huntington's-disease models. Critically, Q-DOAS detects amyloid seeds directly from Parkinson's-disease patient cerebrospinal fluid without any amplification step — a real difference from seed-amplification workflows like RT-QuIC. For Huntington's, the synucleinopathies, and aggregation-focused medicinal-chemistry campaigns, the cleanest available primary assay. (2) Xinya Wang, Mukhtara Perera and colleagues with Shabih Shakeel at the Walter and Eliza Hall Institute of Medical Research present a 2.6-angstrom cryo-EM structure of LRRC15 bound to the antibody-drug-conjugate samrotamab vedotin, currently in phase 1 trials. LRRC15 is a leucine-rich-repeat membrane protein massively upregulated on cancer-associated fibroblasts in desmoplastic tumors (pancreatic, breast, head and neck) where it is associated with therapy resistance. The complex shows samrotamab engaging a membrane-proximal lateral epitope, leaving the canonical concave LRR surface — the face where any LRR receptor normally docks its signaling partners — entirely exposed. The authors then computationally designed de novo mini-protein binders against that concave surface and pulled out multiple nanomolar binders. The next-generation playbook for cancer-associated fibroblasts is to engage the signaling-competent face directly with small designed proteins, either blocking stromal crosstalk or carrying orthogonal payloads — a clean illustration of how de novo binder design now slots directly into an experimental structure pipeline within a single paper. (3) Petar Jevtic, Samuel R. Witus and colleagues with Michael Rape at UC Berkeley describe an unusual mode of proteasomal regulation of transcription. The Groucho-TLE co-repressors sit at transcription start sites holding genes off; Rape's group shows that the cullin-RING ligase SCF-FBXL14 ubiquitinates them not by recognizing their sequence, but by recognizing the act of TLE-mediated chromatin presentation itself. The ligase ubiquitinates whatever the TLEs are presenting, including the TLEs themselves, and the proteasome strips the complex off DNA — continuous cycling that keeps transcription start sites accessible to activators. Stem cells lose lineage specification when the circuit is broken, and cancer-associated TLE1 mutations that block FBXL14 recruitment become interpretable as loss-of-cycling rather than loss-of-repression. FBXL14 looks like a node where one could rebalance transcription in cancers carrying these TLE alterations, and the broader principle — an E3 ligase that reads presentation rather than substrate sequence — is worth tracking because it implies many more presentation-dependent degrons than are currently annotated. (4) Sajini K. Khal, Nathan A. Linhart and colleagues with Michael Boyce at Duke University identify the classical alcohol-deterrent drug disulfiram as a cell-active covalent inhibitor of UDP-galactose 4'-epimerase (GALE), the central enzyme balancing the UDP-galactose/UDP-glucose and UDP-GalNAc/UDP-GlcNAc pools. A coupled luminescence-based high-throughput screen pulled out disulfiram, biochemistry localized inhibition to covalent modification of Cys153 by the reactive diethyldithiocarbamate metabolite, and cellular treatment phenocopied genetic GALE deletion — knocking down terminally sialylated glycans, mucin-type O-glycans, and properly glycosylated mucin-domain glycoproteins — with galactose supplementation rescuing. The therapeutic angle is mucin hypersecretion in muco-obstructive lung disease and mucinous tumors, where nothing currently exists on the nucleotide-sugar side. Beyond disulfiram itself, an HTS-compatible assay is now in place for a proper GALE medicinal-chemistry campaign. 2026-06-19-qdoas-amyloid-lrrc15-minibinder-tle-fbxl14-gale-disulfiram Fri, 19 Jun 2026 12:00:00 +0000 293 Four bioRxiv stories: (1) Kleczko, Gestaut and colleagues with Frydman (Stanford University) — Q-DOAS, a proximity-quenching plate-reader assay that site-specifically conjugates a single BODIPY-TMR dye to a target protein and reads self-assembly as real-time quenching, detecting pre-amyloid oligomers of mutant huntingtin exon 1 and A53T alpha-synuclein well before cross-beta structure forms. Compatible with mechanistic kinetic modeling, plate-reader inhibitor screens (protein + small molecule), and quantitative seeding-activity measurement in HD cell and mouse models. Detects amyloid seeds directly from Parkinson's-disease CSF without amplification — a real difference from RT-QuIC. Fills a gap that ThT (blind to early oligomers) and seed-amplification workflows have left open in neurodegenerative-disease drug discovery. (2) Wang, Perera and colleagues with Shakeel (Walter and Eliza Hall Institute of Medical Research) — 2.6-angstrom cryo-EM structure of the LRRC15-samrotamab complex defining the first high-resolution view of the cancer-associated-fibroblast marker LRRC15 bound by the phase-1 antibody-drug conjugate samrotamab vedotin. Samrotamab engages a lateral membrane-proximal epitope, leaving the canonical concave LRR surface fully exposed. The authors then computationally designed de novo mini-protein binders against that concave surface with nanomolar affinity. Next-generation playbook for cancer-associated fibroblasts in desmoplastic tumors (pancreatic, breast, head and neck) that goes beyond ADC payload delivery to engage the signaling-competent face directly with designed small proteins. (3) Jevtic, Witus and colleagues with Rape (UC Berkeley) — SCF-FBXL14 ubiquitinates TLE/Groucho co-repressors by recognizing the act of TLE-mediated chromatin presentation rather than target sequence, so proteasomal cycling continuously strips co-repressors off transcription start sites and frees them for activator binding. Disruption blocks stem-cell lineage specification, and cancer-associated TLE1 mutations that impair FBXL14 recruitment are reframed as loss-of-cycling rather than loss-of-repression. Implies a broader class of presentation-dependent degrons and a therapeutic node for cancers carrying TLE alterations. (4) Khal, Linhart and colleagues with Boyce (Duke University) — HTS identifies disulfiram (FDA-approved alcohol-deterrent) as a cell-active covalent inhibitor of UDP-galactose 4'-epimerase (GALE) acting via diethyldithiocarbamate-mediated modification of Cys153. Phenocopies genetic GALE deletion (reduced terminally sialylated glycans, mucin-type O-glycans, mucin-domain glycoprotein glycosylation) with galactose supplementation rescuing. Suggests GALE modulation as a strategy for mucin hypersecretion in muco-obstructive lung disease and mucinous cancers, and puts an HTS-compatible assay in place for a proper GALE medicinal-chemistry campaign. Four bioRxiv stories spanning a chemistry-first platform for mapping where small molecules bind on RNA, a diffusion model that finally bakes linker-length and physicochemical constraints into PROTAC generation, a closed-loop active-learning pipeline for lipid nanoparticles with 84-fold cell-type selectivity improvement, and the first standalone activity-based probe for E2 ubiquitin-conjugating enzymes. (1) Lieuwe L. Rietveld, Weiyao Wu and colleagues with Zeshi Li at Utrecht University introduce SOLiD-MaP (Singlet Oxygen footprinting on RNA in a Ligand-Directed manner for Mutational Profiling). RNA-targeted small molecules are emerging as a therapeutic modality, but defining where on the RNA a ligand binds remains hard because existing crosslinking and covalent-capture chemistries depend on ligand-specific probe design, warhead compatibility, and local reaction geometry. SOLiD-MaP decouples the warhead from the binding event: the ligand carries a photosensitizer that generates singlet oxygen locally, and aniline acts as an efficient nucleophile that captures it on nearby ribonucleotides. A soluble quencher constrains the diffusion radius of singlet oxygen, giving a dial for spatial resolution. A pairwise reverse-transcription-stop assay and mutational profiling with NGS read out ligand-proximal positions. Validated on the Mango-II aptamer with thiazole-orange-derived photosensitizers. Orthogonal, ligand-class-agnostic chemistry for RNA mechanism-of-action work that lets a chemical biology program iterate on RNA-targeting chemotypes without rebuilding the assay each time. (2) Boyu Shi, Jiawei Liu and colleagues with Fuyi Li at the University of Adelaide propose DesignMaster, a diffusion-based generative framework for PROTAC design that explicitly conditions on linker length and physicochemical properties. Structure-based ternary modeling and fragment-based linker generators both tend to ignore the length control and drug-like-property envelope that determines whether a ternary complex will actually form in cells. DesignMaster uses an E(3)-equivariant graph transformer with a gated multi-condition fusion module that injects linker length and physicochemical constraints throughout the diffusion trajectory. On PROTAC-DB v2 and v3 it beats SOTA baselines by 3.2%% in validity and 34.4%% in recovery, and in a case study targeting 6W7O it cut linker RMSD against the reference PROTAC BCPyr by 51.78%%. The generator now respects the medicinal-chemistry constraints PROTAC chemists already enforce by hand — a step toward AI linker design that is actually drop-in for degrader programs. (3) Wei Hong Toh, Lucia Cheng and colleagues with Hai-Quan Mao at Johns Hopkins introduce FALCON (Framework for Active Learning-driven Compositional Optimization of Nanoparticles), a closed-loop pipeline combining iterative in-vivo screening, surrogate modeling, and multi-objective optimization for LNP composition. In B-cell-targeted validation, FALCON-optimized particles achieved a 1.8-fold increase in splenic B-cell transfection in vivo over reference compositions. When the objective was switched to selectivity, FALCON delivered an 84-fold improvement in selective splenic-B-cell-over-liver transfection, with spleen-tropic behavior holding across factorial panels of ionizable and helper lipid chemistries. In vaccine studies these particles drove higher IgG2c titers and a Th1-biased profile. Redeployed for myeloid-cell-selective delivery, FALCON recovered enhanced selectivity both across and within spleen and liver compartments. Credible replacement for the brute-force screening loop that has defined LNP cell-selective optimization. (4) Wenshe Liu and Sanjay Chanda at Texas A&M report Ubiquitin Chloromethylketone (UbCMK), the first standalone activity-based probe for E2 ubiquitin-conjugating enzymes. E2s sit at the center of ubiquitin signaling but their catalytic cysteines are weakly nucleophilic with elevated pKa values; existing ubiquitin probes mostly target deubiquitinases and the one reported E2-targeting probe requires E1-dependent activation and captures only a limited panel. DFT identified chloromethylketone as an electrophilic warhead with a low activation barrier specifically for weakly nucleophilic thiolates. UbCMK, synthesized by activated-cysteine-based protein ligation, irreversibly labeled multiple E2s and cysteine DUBs; activity-based protein profiling and quantitative proteomics in HEK293T lysates enriched a broad E2 set including many enzymes inaccessible to prior probes, and it also engages E1s. The probe was used to quantify endogenous E2 mobilization across oxidative, proteotoxic, inflammatory, metabolic, lipid-oxidative, and genotoxic stress. Single chemical handle for direct E2 selectivity interrogation that TPD programs have been missing. 2026-06-18-solid-map-rna-designmaster-protac-falcon-lnp-ubcmk-e2 Thu, 18 Jun 2026 12:00:00 +0000 377 Four bioRxiv stories: (1) Rietveld, Wu and colleagues with Zeshi Li (Utrecht University) — SOLiD-MaP (Singlet Oxygen footprinting on RNA in a Ligand-Directed manner for Mutational Profiling) is a photoproximity labeling platform for mapping small-molecule binding sites on RNA. The ligand carries a photosensitizer that generates singlet oxygen locally; aniline acts as an efficient nucleophile that captures it on nearby ribonucleotides; a soluble quencher constrains the diffusion radius. A pairwise RT-stop assay and mutational profiling with NGS read out ligand-proximal positions. Validated on the Mango-II aptamer with thiazole-orange-derived photosensitizers. Orthogonal, ligand-class-agnostic chemistry for RNA-targeted small-molecule mechanism-of-action work — avoids the per-compound warhead engineering of crosslinking/covalent-capture methods. (2) Shi, Liu and colleagues with Fuyi Li (University of Adelaide) — DesignMaster, a diffusion-based generative framework for PROTAC design built on an E(3)-equivariant graph transformer with a gated multi-condition fusion module that injects linker length and physicochemical constraints throughout the diffusion trajectory. On PROTAC-DB v2 and v3 it beats SOTA by 3.2%% in validity and 34.4%% in recovery; targeting 6W7O it cut linker RMSD against reference BCPyr by 51.78%%. Bakes the medicinal-chemistry constraints PROTAC chemists already enforce by hand into the generator — drop-in AI linker design for degrader programs. (3) Toh, Cheng and colleagues with Hai-Quan Mao (Johns Hopkins University) — FALCON (Framework for Active Learning-driven Compositional Optimization of Nanoparticles), a closed-loop pipeline combining iterative in-vivo screening, surrogate modeling, and multi-objective optimization for LNP composition. 1.8-fold increase in splenic B-cell transfection vs reference; 84-fold improvement in selective splenic-B-cell-over-liver transfection when optimizing for selectivity, with spleen-tropic behavior across factorial panels of ionizable + helper lipid chemistries. Vaccine studies show higher IgG2c titers and Th1-biased profile. Redeployed for myeloid-cell-selective delivery with enhanced selectivity across and within spleen/liver. Active-learning replacement for brute-force LNP screening. (4) Liu, Chanda (Texas A&M University) — Ubiquitin Chloromethylketone (UbCMK), first standalone activity-based probe for E2 ubiquitin-conjugating enzymes. DFT identified CMK as electrophilic warhead with low activation barrier against weakly nucleophilic E2 catalytic cysteines (high pKa). Synthesized by activated-cysteine protein ligation; irreversibly labels multiple E2s and cysteine DUBs and also engages E1s. Activity-based proteomics in HEK293T lysates enriches broad E2 set including previously inaccessible enzymes. Used to quantify endogenous E2 mobilization across oxidative, proteotoxic, inflammatory, metabolic, lipid-oxidative, and genotoxic stress. Single chemical handle for direct E2 selectivity interrogation that TPD programs have been missing. Four bioRxiv stories spanning a drug-repurposing-derived antitubercular inducing alanine auxotrophy through irreversible inhibition of a host-inaccessible Mtb metabolic enzyme, an all-atom generative binder-design model from MIT that produces proteins and peptides across modalities and target classes in a single unified system, the cGMP bioprocess paper for the N332-GT5 germline-targeting HIV vaccine now in first-in-human dosing in HVTN144 out of Bill Schief's Scripps program, and a de novo protein design pipeline that produces polymorph-specific mini-protein binders for α-synuclein amyloid fibrils. (1) Mohamed-Ali Shalaby, Naveen Beeralingappa and colleagues with Daniel Flaherty at Purdue University and Matthew Chojnacki identify a new metabolic vulnerability in Mycobacterium tuberculosis. A drug-repurposing screen returned TI-374 with sub-micromolar potency against Mtb; resistance mutation mapping, supplementation rescue, and direct biochemistry converged on alanine aminotransferase (AlaA, encoded by Rv0337c) as the target, with TI-374 acting as an irreversible inhibitor. Inhibition forces alanine auxotrophy. Critically, TI-374 remained active against intracellular Mtb in an ex vivo macrophage model despite excess endogenous alanine, suggesting Mtb cannot scavenge alanine from the host, and genetic deletion of alaA attenuates Mtb survival in a murine infection model. Medicinal chemistry then took TI-374 to TI-801 with low-nanomolar potency. Host-inaccessible metabolic vulnerability combined with knockout phenotype and a low-nanomolar covalent chemical probe — the validation arc the antitubercular pipeline has needed. (2) Hannes Stark, Frederic Faltings and colleagues with Tommi Jaakkola, Regina Barzilay, Gabriele Corso, Bill DeGrado, David Sabatini and others at MIT introduce BoltzGen (version 2 posted 2026-06-16), an all-atom generative model unifying binder design with structure prediction in a single network. Inference is controlled by a flexible specification language over covalent bonds, structure constraints, and binding sites. Eight diverse design campaigns across 26 targets validate functional and affinity outcomes; modalities range from nanobodies to disulfide-bonded peptides, targets from disordered proteins to small molecules. The headline result is nanobody binders for targets with low similarity to anything in the training set with a known bound complex — a real test of generalization rather than recall. Weights, data, and code are released. Architectural argument that folding-grade structural reasoning inside the generative process produces mode-defensible binders across the modalities drug discovery cares about, from a single open model. (3) Sam Pallerla, Sandeep Uplekar and colleagues with the Bill Schief group at The Scripps Research Institute and collaborators at IAVI report end-to-end bioprocess development and cGMP manufacturing of N332-GT5 gp140 — the germline-targeting HIV-1 envelope trimer designed to initiate broadly neutralizing antibody lineages and currently in first-in-human dosing in HVTN144. A stable CHO clone built on ATUM's Leap-In transposon platform reached titers above 200 mg/L and held genetic stability through 60 population doublings. The process scaled from Ambr250 mini-bioreactors to 200-L single-use systems with consistent quality, and a three-step purification (affinity capture, multimodal polishing, viral clearance) yielded better than 99%% trimer with native-like antigenicity, site-specific glycan occupancy matching the design, and viral clearance exceeding 18-log and 11-log reductions. Scalable, reproducible manufacturing paradigm for structurally complex HIV-1 envelope immunogens at clinical scale — the part of the sequential germline-targeting program that has to land structurally faithful or the downstream B-cell lineage maturation thesis cannot be tested. (4) Ahmed Sadek, Nolwen Rey and colleagues with Bruno Correia at EPFL present AmyBind, a computational pipeline for de novo design of mini-protein binders targeting polymorph-specific structural features of amyloid fibrils. Two α-synuclein fibril-specific binders were developed; one exhibited polymorph specificity. Neither inhibited fibril elongation in vitro, consistent with their designed lateral-surface binding mode. In cell-based models both binders showed polymorph-specific colocalization without altering fibril uptake or seeding, supporting their utility as fibril-selective labeling agents. Polymorph specificity becomes a designable property of de novo mini-proteins — a near-term route to fibril-selective imaging probes for differential diagnosis of synucleinopathies (where distinct polymorphs may correspond to distinct clinical syndromes) and a longer-term path to polymorph-selective amyloid therapeutics that small-molecule PET tracers and pan-amyloid antibodies have struggled to deliver. 2026-06-17-alaa-mtb-auxotrophy-boltzgen-binder-n332-gt5-hvtn144-amybind-asyn Wed, 17 Jun 2026 12:00:00 +0000 372 Four bioRxiv stories: (1) Shalaby, Beeralingappa and colleagues with Flaherty (Purdue University) and Chojnacki — drug-repurposing screen identifies TI-374 with sub-micromolar anti-Mtb activity acting as an irreversible inhibitor of alanine aminotransferase AlaA (encoded by Rv0337c), inducing alanine auxotrophy. Medicinal chemistry yields TI-801 with low-nanomolar potency; activity is retained against intracellular Mtb in ex vivo macrophages despite excess endogenous alanine (Mtb cannot scavenge host alanine), and genetic deletion of alaA attenuates Mtb survival in a murine infection model. Host-inaccessible metabolic vulnerability plus knockout phenotype plus low-nanomolar covalent chemical probe — a validation arc the antitubercular pipeline has needed. (2) Stark, Faltings and colleagues with Jaakkola, Barzilay, Corso (Massachusetts Institute of Technology, v2 2026-06-16) — BoltzGen, an all-atom generative model unifying binder design with structure prediction in a single network. Inference is controlled by a flexible specification language over covalent bonds, structure constraints, and binding sites. Validated functionally and by affinity across 26 targets in 8 design campaigns; modalities span nanobodies to disulfide-bonded peptides; targets span disordered proteins to small molecules. Nanobody binders identified for targets with low similarity to anything in the training set with a known bound complex — a real test of generalization rather than recall. Open weights, data, and code. (3) Pallerla, Uplekar and colleagues with Schief (Scripps Research / IAVI) — end-to-end bioprocess and cGMP manufacturing of N332-GT5 gp140, the germline-targeting HIV-1 envelope trimer now in HVTN144 first-in-human dosing. Stable CHO clone via ATUM Leap-In transposon reaches >200 mg/L titers with genetic stability through 60 population doublings; scale-up from Ambr250 to 200-L single-use systems with three-step purification (affinity capture, multimodal polishing, viral clearance) yields >99%% trimeric purity with preserved quaternary structure, native-like antigenicity, site-specific glycan occupancy matching design, and viral clearance exceeding 18-log/11-log reductions. Scalable, reproducible manufacturing paradigm for structurally complex HIV-1 envelope immunogens at clinical scale. (4) Sadek, Rey and colleagues with Correia (EPFL) — AmyBind computational pipeline for de novo design of mini-protein binders targeting polymorph-specific structural features of amyloid fibrils. Two α-synuclein fibril-specific binders developed, one polymorph-discriminating; neither inhibits elongation in vitro (designed lateral-surface binding), and both show polymorph-specific colocalization in cell models without altering uptake or seeding. Polymorph specificity becomes a designable property of de novo mini-proteins — near-term application as fibril-selective imaging probes for differential diagnosis of synucleinopathies and a longer-term path to polymorph-selective amyloid therapeutics that small-molecule PET tracers and pan-amyloid antibodies have struggled to deliver. Four bioRxiv stories spanning a fundamentally new mode of targeted protein degradation that bypasses the E3 ligase landscape entirely, an mRNA-delivered intracellular nanobody chaperone that rescues F508del-CFTR through a previously unseen open-channel conformation, a billion-scale phenotypic virtual screen at Genentech with prospective wet-lab confirmation of novel kinase-pathway scaffolds, and a chemical-biology assay showing that lipid-driven membrane permeation follows opposite rules in mammalian and Gram-negative cells. (1) Soh-Hyun Park, Yejin Jang and colleagues with Min Jae Lee at Seoul National University College of Medicine develop Protea-Tac (version 3 posted 2026-06-13), a heterobifunctional protein degrader composed of a ubiquitin receptor and an intracellular antibody. The Ub-receptor half integrates the degrader directly into 26S proteasomes without altering their functional integrity, and the intrabody half recruits a chosen substrate into proximity with the protease machinery — sufficient to drive degradation without any ubiquitin chemistry on the target. The platform is demonstrated across c-Fos, BRD4, Flag-TDP43, HA-tau, and GFP-ODC, and shows in vivo anti-tumor efficacy from c-Fos degradation in a mouse model. The platform is modular (intrabody is the swappable element), ubiquitin-independent, and target-specific. The substrate space accessible to Protea-Tac is bounded by intrabody specificity rather than by the tissue distribution + intrinsic activity + chemical tractability of the E3 ligase landscape that constrains conventional PROTACs and molecular glues — redrawing the map for TPD substrates that the E3-based platform has visibly failed to reach, at the cost of a protein modality with attendant delivery challenges. (2) Maud Overtus, Tea Rubil and colleagues with Cédric Govaerts at the Université Libre de Bruxelles, John Hunt at Columbia, Marcus Mall at the Charité, and David Sheppard at Bristol (version 3 posted 2026-06-14, revised title) deliver as lipid-nanoparticle mRNA a nanobody — T2a — that thermally stabilizes CFTR via high-affinity binding to nucleotide-binding domain 1. T2a synergizes with clinically-approved correctors to substantially improve F508del-CFTR maturation, plasma-membrane expression, and channel activity. Single-channel recordings show T2a promotes both full-open and sub-conductance gating and protects against thermal deactivation. Because T2a binding to NBD1 prevents the canonical ATP-driven NBD1-NBD2 dimerization that drives normal channel opening, observing channel activity in T2a's presence implied a non-canonical active conformation — which cryo-EM confirms by revealing NBD1 binding an alternative site on CFTR incompatible with NBD dimerization. New paradigm for intracellular-nanobody-as-drug delivered by mRNA-LNP for protein-misfolding disease, with an unexpected active-channel conformation directly relevant to next-generation corrector chemistry. (3) Allen Wu, Hugo Yao and colleagues with David Richmond at Genentech build PhenoCompass, a multimodal co-embedding model aligning chemical structures with Cell Painting high-content phenotypic imaging, trained on the JUMP morphology consortium dataset across more than 100,000 compounds. Retrospective validation on historical biochemical screens confirms PhenoCompass ranks compounds by biochemical target engagement, not merely visual similarity. The prospective phenotypic virtual screen ran across 3.8 billion Enamine REAL compounds for PI3K/mTOR pathway inhibitors, nominating 11 novel structurally diverse compounds with 54-fold enrichment over training-set hits, and orthogonal FOXO3A reporter + direct kinase-inhibition assays confirmed 7 structurally novel inhibitors with distinct mechanisms of action. Establishes high-content morphology-grounded virtual screening at billion-compound library scale with real prospective wet-lab confirmation of scaffold-novel hits — and shows that the cellular morphological signature aggregates over diverse molecular targets converging onto the same downstream pathway state. A credible response to the longstanding throughput ceiling on phenotypic drug discovery. (4) David Coffin, Sushmita Bhandari and colleagues with Marcos Pires at the University of Virginia apply the CHAMP (Chloroalkane Azide-based Membrane Penetration) assay to a systematically designed library of lipid conjugates in both HeLa and E. coli cells. CHAMP pairs a minimally disruptive azide tag with a cytosolically anchored HaloTag to quantitatively measure cytosolic accumulation. In mammalian cells, larger and more hydrophobic lipid conjugates — medium-chain, cyclized, heteroatom-containing — are preferentially internalized, consistent with the classical lipidation playbook. In E. coli, the same modifications routinely reduce accumulation. Charge, scaffold composition, and the specific envelope barriers of the diderm system are resolved. The work provides a quantitative explanation for the long-standing scarcity of lipidated Gram-negative antimicrobials, exposes the limits of lipophilicity-driven optimization in antibacterial medicinal chemistry, and gives empirical design rules that translate directly into go-no-go criteria for hydrophobic modifications. 2026-06-15-protea-tac-ub-independent-cftr-nanobody-mrna-phenocompass-champ-lipidation Mon, 15 Jun 2026 12:00:00 +0000 401 Four bioRxiv stories: (1) Park, Jang and colleagues with Lee (Seoul National University College of Medicine, v3 2026-06-13) — Protea-Tac heterobifunctional protein degrader composed of a ubiquitin receptor fused to an intracellular antibody recruits substrates directly into the 26S proteasome without ubiquitination of the target. Demonstrated against c-Fos, BRD4, Flag-TDP43, HA-tau, GFP-ODC with in vivo anti-tumor efficacy from c-Fos degradation in a mouse model. Modular ubiquitin-independent TPD platform whose substrate space is bounded by intrabody specificity rather than the E3 ligase landscape that constrains conventional PROTACs and molecular glues. (2) Overtus, Rubil and colleagues with Govaerts (Université Libre de Bruxelles), Hunt (Columbia), Mall (Charité), Sheppard (Bristol) (v3 2026-06-14, revised title) — lipid-nanoparticle-delivered mRNA encoding nanobody T2a that thermally stabilizes CFTR via high-affinity binding to NBD1 synergizes with clinically-approved correctors to substantially improve F508del-CFTR maturation + plasma-membrane expression + channel activity. Single-channel recordings show T2a promotes both full-open and sub-conductance gating + protects against thermal deactivation. Cryo-EM reveals a novel open-channel conformation in which NBD1 binds an alternative site on CFTR incompatible with canonical ATP-driven NBD1-NBD2 dimerization. New paradigm for intracellular-nanobody-as-drug delivered by mRNA-LNP for protein-misfolding disease plus an unexpected active CFTR conformation relevant to next-generation corrector design. (3) Wu, Yao and colleagues with Richmond (Genentech) — PhenoCompass multimodal co-embedding model aligning chemical structures + Cell Painting high-content phenotypic imaging trained on the JUMP morphology consortium dataset (>100,000 compounds) performs a prospective phenotypic virtual screen across 3.8 billion Enamine REAL compounds for PI3K/mTOR pathway inhibitors. 11 novel compounds nominated with 54-fold enrichment over training-set hits; orthogonal FOXO3A reporter + direct kinase inhibition confirm 7 structurally novel inhibitors with distinct mechanisms of action. Morphology-grounded virtual screening at billion-compound library scale with real prospective wet-lab confirmation of scaffold-novel hits converging onto a shared pathway signature. (4) Coffin, Bhandari and colleagues with Pires (University of Virginia) — CHAMP chloroalkane azide membrane penetration assay pairing a minimally disruptive azide tag with a cytosolically anchored HaloTag quantitatively measures cytosolic accumulation of a systematically designed library of lipid conjugates in both HeLa and E. coli. Larger and more hydrophobic conjugates (medium-chain, cyclized, heteroatom-containing) are preferentially internalized by mammalian cells but the same modifications routinely reduce E. coli accumulation. Quantitative explanation for the long-standing scarcity of lipidated Gram-negative antimicrobials and empirical design rules that translate directly into go-no-go criteria for hydrophobic modifications in antibacterial medicinal chemistry. Four bioRxiv stories at the interface of chemistry, computation, and cell biology — an honest prospective stress test of AlphaFold 3 as a docking engine, a phenotype-first covalent fragment screen turning up a synthetic-lethal handle on a classic target, de novo proteins that template specific carbonate polymorphs, and a probe-free proximity-labeling chemistry. (1) Kaushal Menon, Avinash Davasam and colleagues with Jiankun Lyu at The Rockefeller University present what is probably the most rigorous prospective test of AlphaFold 3 as a small-molecule discovery engine to date (version 2 posted 2026-06-12). AlphaFold 3 can co-fold proteins with ligands, but whether that translates into finding new chemical matter or merely into re-posing what is already known had remained an open and contested question. On retrospective enrichment against decoys, AlphaFold 3 looks dominant — but the authors show this is largely an artifact of hidden ligand-only biases in computational decoy sets. In three large experimental datasets without such biases (the sigma-2 receptor, the D4 dopamine receptor, and AmpC beta-lactamase, >2,500 tested molecules), the physics-based docking program DOCK3 achieved stronger overall enrichment, while AlphaFold 3 contributed mainly to early enrichment. Out-of-sample pose reproduction on >8,000 post-training-cutoff complexes correlated strongly with training-set similarity — indicating partial memorization rather than learning of general molecular-recognition principles. The headline is a prospective head-to-head against the sigma-2 receptor (novel to AlphaFold 3's training set): AlphaFold 3 achieved a 13%% hit rate and pulled out a 13 nM binder, with a crystal structure confirming a near-native predicted pose, but DOCK3 in parallel delivered roughly twice the hit rate from the same library, with a similar affinity distribution among top hits. Establishes AlphaFold 3 as a complementary engine — particularly as a post-docking filter — rather than a replacement for conventional docking, and provides a realistic envelope for structure-based campaigns. (2) Federica Raguseo, Eloise Fellows and colleagues with Simon Boulton at the Francis Crick Institute report a phenotype-first covalent fragment screen that turns up an unexpected synthetic-lethal handle on a classic drug target. They screened ~500 cysteine-reactive fragments against an isogenic pair of eHAP iCAS9 cells — ATRX wild-type and ATRX-knockout — looking for differential cell killing rather than binding. A single chloroacetamide they call PP12 selectively impaired ATRX-deficient cells. Rather than chasing the binder through chemoproteomics alone, the team cross-referenced competitive click-chemoproteomic hits against genome-wide CRISPR synthetic-lethal datasets and converged on thymidylate synthase (TYMS). Validation came from a co-crystal structure, competition with 5-fluorouracil at the active site, and a measurable drop in cellular dTMP. Mechanistically, TYMS inhibition induces replication stress that is selectively cytotoxic to ATRX-deficient cells in a manner dependent on FAM111A and SLFN11. A generalizable phenotype-first workflow that combines covalent chemoproteomics with CRISPR genetic-interaction data to find context-specific covalent ligands without committing upfront to a target hypothesis. (3) Paul Kwon, Xinting Li and colleagues with David Baker at the University of Washington extend de novo protein design into biomineralization. By fine-tuning RFdiffusion-2 on repeat-protein scaffolds, they designed reconfigurable two-dimensional protein arrays that template specific calcium carbonate polymorphs — calcite from one set of designs and metastable aragonite under nucleation conditions that otherwise produce phase mixtures. They also designed lattice-matched proteins that template cobalt carbonate, and showed that a D3 protein cage confines homogeneous cobalt carbonate nanocrystals inside its interior; the resulting cage-mineral hybrid functioned as an electrocatalyst for alkaline water splitting. Designed protein interfaces are now controlling phase, polymorph, and morphology at the inorganic-organic boundary, opening a credible route to functional protein-mineral composites for sensing, delivery, and catalysis. (4) Changho Lee, Jong-Kwan Lee and colleagues with Tae-Hyuk Kwon and Hyun-Woo Rhee at Ulsan National Institute of Science and Technology develop IDM, an organic photocatalyst that generates hydroxyl and superoxide radicals via water oxidation rather than singlet oxygen, dioxidizing proximal histidines into a persistent lactam-tautomer dioxidized state that remains electrophilic and chemically addressable after cell lysis. This decouples photocatalytic proximity labeling from live-cell probe access, eliminating a major source of spatial bias in dynamic compartments like endosomes and exosomes where probe accessibility is intrinsically nonuniform. Applied to intracellular vesicle trafficking, the workflow (PF-Map) recovers a vesicle-trafficking-related subproteome that probe-dependent workflows had missed — a chemistry-first design choice expanding what proximity proteomics can resolve. 2026-06-13-af3-docking-covalent-tyms-rfdiff2-biomineralization-his2o-pf-map Sat, 13 Jun 2026 12:00:00 +0000 349 Four bioRxiv stories: (1) Menon, Davasam and colleagues with Lyu (The Rockefeller University, v2 2026-06-12) — prospective head-to-head test of AlphaFold 3 as a small-molecule discovery engine on the sigma-2 receptor (novel to AlphaFold 3's training set) achieves 13%% hit rate and a 13 nM binder with crystal-structure-confirmed near-native predicted pose, but ~half the hit rate of DOCK3 in parallel from the same library. Three large bias-free experimental datasets (sigma-2, D4, AmpC, >2,500 tested molecules) show DOCK3 outperforms AlphaFold 3 in overall enrichment with AlphaFold 3 contributing mainly to early enrichment; >8,000 post-training-cutoff pose-reproduction tests show AlphaFold 3 accuracy depends strongly on training-set similarity. Establishes AlphaFold 3 as a complementary engine and post-docking filter rather than a replacement for physics-based docking, and reframes retrospective decoy benchmarks as largely artifactual. (2) Raguseo, Fellows and colleagues with Boulton (Francis Crick Institute) — phenotype-first covalent fragment screen across ~500 cysteine-reactive fragments against isogenic ATRX-wild-type/ATRX-knockout eHAP iCAS9 cells identifies chloroacetamide PP12 that selectively kills ATRX-deficient cells. Cross-referencing competitive click-chemoproteomic hits against genome-wide CRISPR synthetic-lethal datasets converges on thymidylate synthase (TYMS), validated by co-crystal structure, 5-fluorouracil active-site competition, and a drop in cellular dTMP, with replication-stress cytotoxicity dependent on FAM111A and SLFN11. Generalizable phenotype-then-deconvolute workflow linking covalent chemoproteomics with CRISPR genetic-interaction data for context-specific covalent ligand discovery without upfront target commitment. (3) Kwon, Li and colleagues with Baker (University of Washington) — fine-tunes RFdiffusion-2 on repeat-protein scaffolds to design reconfigurable 2D protein arrays that template specific calcium carbonate polymorphs (calcite vs metastable aragonite under nucleation conditions that otherwise yield phase mixtures) and lattice-matched proteins that template cobalt carbonate, with a D3 protein cage confining homogeneous cobalt carbonate nanocrystals inside its interior to form a cage-mineral hybrid functioning as an alkaline water-splitting electrocatalyst. De novo protein design now controls phase, polymorph, and morphology at the inorganic-organic boundary as a route to functional protein-mineral composites. (4) Lee, Lee and colleagues with Kwon and Rhee (Ulsan National Institute of Science and Technology) — IDM organic photocatalyst generates hydroxyl and superoxide radicals via water oxidation (instead of singlet oxygen + exogenous probe trapping) to dioxidize proximal histidines into a persistent lactam-tautomer dioxidized state that remains electrophilic after cell lysis. Decouples photocatalytic proximity labeling from live-cell probe access and recovers a vesicle-trafficking subproteome that conventional probe-dependent workflows underestimated. Chemistry-first design choice expanding what proximity proteomics can resolve in dynamic compartments like endosomes and exosomes. Three bioRxiv stories spanning a cryo-EM-grounded structural framework for biased small-molecule agonism at a GPCR target for fibrosis and heart failure, a covalent active-site cysteine handle on metallo-beta-lactamases as a route around the zinc-chelation strategy that has dominated and underperformed in NDM inhibition, and a candid diagnosis of why machine-learning-driven antibiotic discovery keeps producing membrane disruptors rather than mechanistically novel compounds. (1) Joshua Osei-Owusu, Tristan Girbau and colleagues with Andrew Kruse at Harvard Medical School solve three cryo-electron microscopy structures of the relaxin receptor RXFP1 — ligand-free, bound to the native peptide hormone relaxin-2, and bound to the small-molecule drug candidate AZD5462 — and show that the two ligand classes activate the receptor by completely different mechanisms. RXFP1 is an unusual GPCR with a long ectodomain that mediates the physiological adaptations of pregnancy and that is being clinically pursued in fibrosis and heart failure, with both protein and small-molecule agonists in mid-stage trials. Relaxin-2 engages the ectodomain and induces a conformational reorganization of the linker domain into a helical secondary structure that propagates activation into the transmembrane bundle. AZD5462 binds inside the transmembrane domain itself and stabilizes a unique active conformation that selectively recruits beta-arrestin. Hydrogen-deuterium-exchange mass spectrometry backs up the structural model. Neither mechanism matches the push-pull activation characterized for the glycoprotein hormone receptors that RXFP1 is structurally related to. Provides an explicit structural framework for designing biased agonists at a target that was historically peptide-only. (2) Jacob Bradley, Karen Calvopina Tapia and colleagues with Adam Roberts at the Liverpool School of Tropical Medicine and Christopher Schofield at the University of Oxford show that the small-molecule 3-bromopyruvate (3-BP) selectively restores the antimicrobial activity of meropenem against carbapenem-resistant clinical and environmental isolates of Escherichia coli, Klebsiella pneumoniae, and Acinetobacter baumannii from Tanzania and Malawi that carry NDM-1 or NDM-5 genes — but not against strains in which resistance is driven by serine beta-lactamases. Mass-spectrometry data on the purified enzymes support a mechanism in which 3-BP forms a covalent bond with an active-site cysteine residue of NDM-1 and NDM-5. No metallo-beta-lactamase inhibitors are approved for clinical use; most reported inhibitors are zinc-ion chelators acting at the active site or in solution, an approach that has historically struggled with selectivity and toxicity. 3-BP itself is not a clinical candidate, but the result legitimizes covalent reactivity as a discovery strategy for the metallo-beta-lactamase family — the same playbook that produced the successful serine-beta-lactamase inhibitors. (3) Christopher Chain, Sara Ghaffari and colleagues with Zemer Gitai at Princeton University build a machine-learning antibiotic discovery pipeline that, instead of post-hoc Tanimoto similarity filtering, employs an XGBoost-based mechanism-of-action classifier up front to prioritize compounds predicted to have mechanisms distinct from known antibiotic classes, combined with graph neural networks for antibacterial activity and toxicity prediction. Applied to the Zinc20 database the approach identified non-toxic antibacterial compounds structurally distinct from known antibiotics — but the majority exhibited membrane-targeting activity with selectivity for bacterial over mammalian cells, and none hit novel protein targets. Systematic analysis traced this limitation to mechanistic bias in the training data rather than model architecture: the activity model preferentially scored compounds resembling specific MoA classes already overrepresented in the training data. Substantial model architecture and training-data enhancements did not overcome this constraint. Reframes the gating problem in ML-driven antibiotic discovery as curation of diverse, mechanistically-annotated chemical-biology training data — not the choice of model — and is a useful, sobering counterweight to generic AI-in-drug-discovery hype. 2026-06-12-rxfp1-relaxin-ndm-3bp-covalent-ml-antibiotic-moa-bias Fri, 12 Jun 2026 12:00:00 +0000 347 Three bioRxiv stories: (1) Osei-Owusu, Girbau and colleagues with Kruse (Harvard Medical School) — three cryo-EM structures of the relaxin receptor RXFP1 (ligand-free, bound to native protein hormone relaxin-2, bound to small-molecule drug candidate AZD5462) reveal divergent activation mechanisms. Relaxin-2 engages the ectodomain and induces helical reorganization of the linker domain backed by HDX-MS; AZD5462 binds inside the transmembrane bundle and stabilizes a unique active conformation that selectively recruits beta-arrestin. Both are distinct from the push-pull mechanism of structurally related glycoprotein hormone receptors. Provides an explicit structural framework for next-generation biased small-molecule agonists at a GPCR target in mid-stage trials for fibrosis and heart failure. (2) Bradley, Calvopina Tapia and colleagues with Roberts (Liverpool School of Tropical Medicine) and Schofield (University of Oxford) — 3-bromopyruvate selectively restores meropenem activity against carbapenem-resistant clinical/environmental E. coli, K. pneumoniae and A. baumannii isolates from Tanzania and Malawi carrying NDM-1 or NDM-5 genes but not strains carrying serine beta-lactamases, with mass-spectrometry data supporting covalent reaction at an active-site cysteine of NDM-1 and NDM-5. Legitimizes covalent cysteine targeting as a discovery strategy for the metallo-beta-lactamase family distinct from the zinc-chelation strategy that has dominated and underperformed in this space. (3) Chain, Ghaffari and colleagues with Gitai (Princeton University) — XGBoost-based mechanism-of-action classifier prioritizes compounds predicted to have mechanisms distinct from known antibiotic classes up front (combined with graph neural networks for activity + toxicity prediction). Applied to the Zinc20 database the approach recovered non-toxic antibacterial compounds structurally distinct from approved antibiotics but with the majority showing membrane-targeting activity and none hitting novel protein targets; systematic analysis traces this to mechanistic bias in the training data rather than model architecture, with substantial model and training-data enhancements failing to overcome the constraint. Reframes the bottleneck in ML-driven antibiotic discovery as curation of diverse, mechanistically-annotated chemical biology rather than modeling. Four bioRxiv stories spanning a previously uncharacterized molecular interface coupling a lupus genetic risk gene to mTORC1-driven cytokine output, a modular biologic-plus-small-molecule platform that brings targeted post-translational modification editing to disordered undruggable proteins, a high-throughput peptide screen yielding an arginine-depleted muscle-tropic delivery vehicle for antisense exon-skippers in Duchenne muscular dystrophy, and a continuous-glucose-monitor-informed redesign of glucagon therapy via thermostable solid microneedle patches. (1) Tin-Yun Chiu, Patrick Hall and colleagues with Christopher Parker at The Scripps Research Institute identify a direct interaction between SLC15A4 (the lupus-associated endolysosomal solute carrier essential for TLR7-9-mediated cytokine production) and LAMTOR1 (core scaffold subunit of the lysosomal Ragulator complex), which couples endolysosomal TLR activation to mTORC1 and downstream inflammatory responses. Structural modeling and site-directed mutagenesis show that an alpha-helix of LAMTOR1 engages the substrate-binding pocket of SLC15A4, and disruption of this interface broadly suppresses TLR7-9-mediated cytokine production, mTOR activation, and IRF5/IRF7 signaling. Most concrete picture yet of a tractable small-molecule binding pocket at the heart of the endolysosomal-TLR-to-mTOR axis in systemic lupus erythematosus and a clean rationale for chemical disruption of the SLC15A4-LAMTOR1 interface as a path to disease-modifying lupus therapy. (2) Feng Shen, Octavia Merino-Chavez and colleagues with Laura Dassama at Stanford University introduce the NanoBridge, a modular platform that uses biologic binders — single-domain antibodies and peptide ligands — to transiently tag a protein of interest with FKBP12(F36V), then deploys heterobifunctional small molecules to recruit endogenous PTM enzymes onto that handle. Across three structurally diverse and largely unstructured targets — KRAS, p53, and BCL11A — the platform achieves rapid, reversible, temporally controlled targeted degradation, phosphorylation, and acetylation of the endogenous protein with no exogenous PTM writer. Generalizable strategy for multiplex PTM editing on proteins that conventional chemically-induced proximity has been unable to reach. (3) Charlotte Farquhar, Nina Dow and colleagues with Bradley Pentelute at the Massachusetts Institute of Technology built a high-throughput charge-based chromatographic enrichment platform capable of screening over 15,000 synthetic peptides — including sequences with noncanonical (abiotic) amino acids — to de novo discover arginine-depleted cell-penetrating peptides for phosphorodiamidate morpholino oligomer (PMO) delivery in Duchenne muscular dystrophy. Four lead candidates gave efficient nuclear PMO delivery with ~10-fold lower in-vitro toxicity than penetratin; the top peptide CXP1 conjugated to a PMO produced greater exon skipping in dystrophic mdx mice than the standard R6G-PMO at equivalent doses, with a clean tissue-level pharmacokinetic-pharmacodynamic relationship. New mechanistically distinct chemical class of muscle-tropic delivery vehicles for the antisense modality. (4) Junwei Li, Caleb Byrne and colleagues with Giovanni Traverso and Robert Langer at the Massachusetts Institute of Technology mined continuous glucose monitor data from 1,135 type-1-diabetes patients across 246.18 patient-years and showed that 20-30%% of hypoglycemic episodes are follow-on events after under-treatment of a prior event, and that independent hypoglycemic episodes last 79-108 minutes. Guided by these kinetics they engineered two thermostable high-density solid glucagon formulations into microneedle patches: an on-demand patch for rapid prevention or treatment of mild daytime hypoglycemia, and an enzyme-driven hypoglycemia-responsive patch for autonomous nighttime glucagon release. Both formats showed good in-vitro stability/loading/release kinetics and reversed hypoglycemia in diabetic animals — a credible path to making glucagon a routine on-demand modality beyond emergency-only injection. 2026-06-11-slc15a4-lamtor-nanobridge-ptm-cxp1-pmo-glucagon-microneedle Thu, 11 Jun 2026 12:00:00 +0000 423 Four bioRxiv stories: (1) Chiu, Hall and colleagues with Parker (The Scripps Research Institute) — identify a direct SLC15A4-LAMTOR1 interaction where the alpha-helix of LAMTOR1 engages the substrate-binding pocket of SLC15A4 (lupus-associated endolysosomal solute carrier essential for TLR7-9 cytokine production), coupling endolysosomal TLR activation to mTORC1 + IRF5/7 signaling + inflammatory cytokine output. Site-directed mutagenesis disrupting this interface broadly suppresses TLR7-9-mediated cytokine production, defining a concrete druggable pocket at the heart of the endolysosomal-TLR-to-mTOR axis in systemic lupus erythematosus. (2) Shen, Merino-Chavez and colleagues with Dassama (Stanford University) — NanoBridge modular chemical-biology platform using single-domain antibody + peptide biologic binders to transiently direct FKBP12(F36V) to unmodified endogenous proteins, enabling multiplex heterobifunctional-small-molecule-driven recruitment of endogenous PTM enzymes (degradation, phosphorylation, acetylation) with rapid, reversible, temporal control and no exogenous PTM writers across three disordered pocket-less targets — KRAS, p53, and BCL11A. Opens targeted PTM editing on proteins that conventional chemically-induced proximity cannot reach. (3) Farquhar, Dow and colleagues with Pentelute (Massachusetts Institute of Technology) — high-throughput charge-based chromatographic enrichment platform screening >15,000 synthetic peptides including abiotic noncanonical amino acids to de-novo discover arginine-depleted cell-penetrating peptides for PMO delivery in Duchenne muscular dystrophy. Four leads with ~10-fold lower in-vitro toxicity than penetratin; top peptide CXP1-PMO conjugate produced greater exon skipping than R6G-PMO at equivalent doses in mdx mice with clean tissue-level PK/PD relationship. New mechanistically distinct chemical class of muscle-tropic antisense delivery vehicles. (4) Li, Byrne and colleagues with Traverso + Langer (Massachusetts Institute of Technology) — mined CGM data from 1,135 T1D patients across 246.18 patient-years showing 20-30%% of hypoglycemic episodes are follow-on from under-treatment + independent hypoglycemic episodes last 79-108 minutes. Engineered two thermostable high-density solid glucagon microneedle patches: on-demand daytime patch for rapid prevention/treatment of mild hypoglycemia, and enzyme-driven hypoglycemia-responsive patch for autonomous nighttime glucagon release. Both showed good in-vitro stability/loading/release and reversed hypoglycemia in diabetic animals — a credible path to make glucagon a routine on-demand modality beyond emergency-only injection. Four bioRxiv stories spanning the first small-molecule chemical-biology validation of a long-pursued myeloid and neuroimmune checkpoint as druggable across both Alzheimer's disease and tumor immunity, a previously hidden ubiquitin-independent proteasome activation feedback loop linking targeted protein degradation to innate inflammatory signaling, a chirality-controlled engineering handle on lipid nanoparticle internal architecture and mRNA translation output, and a single-molecule blood-based liquid biopsy that detects pancreatic ductal adenocarcinoma at stage one. (1) Sherif Abdel-Rahman, Nirmal Murugan and Moustafa Gabr at Weill Cornell Medicine establish ILT3 (LILRB4) as a tractable small-molecule target across two disease contexts in back-to-back preprints. ILT3 is an inhibitory myeloid receptor with confirmed roles both as a tumor immune checkpoint and as a neuroimmune checkpoint in Alzheimer's disease (where it restricts microglial activation and amyloid clearance through ApoE-dependent signaling), and to date has been pursued exclusively as an antibody target. The first paper uses affinity-selection mass spectrometry against recombinant ILT3 to identify LT12, with nanomolar binding confirmed by MST and SPR, a discrete binding pocket mapped by computational modeling and mutagenesis, and disruption of the ILT3-ApoE interaction. In human iPSC-derived microglia, LT12 attenuates SHP1/SHP2 signaling, suppresses NF-kB activation, lowers IL-1-beta secretion, and restores amyloid-beta uptake; in 5xFAD mice it improves cognition and reduces amyloid burden. The second paper uses a Dianthus temperature-related intensity change (TRIC) screen of an 8,961-member Enamine library to identify ICB-7, binding in a hydrophobic D2-domain pocket, disrupting the ILT3-SCG2 interaction, and inhibiting SHP1/SHP2/STAT3 signaling. ICB-7 enhances cytotoxic T-cell killing in patient-derived colorectal carcinoma and AML co-cultures and shows favorable pharmacokinetics and anti-tumor efficacy in the CT26 syngeneic colorectal model. ILT3 is now a small-molecule target with chemical matter from two independent screening platforms, and a single myeloid checkpoint has rigorous preclinical proof-of-concept in both inflammation-driven CNS degeneration and solid tumor immunotherapy. (2) Sebastian Watson, James Williamson and colleagues with Paul Lehner at the University of Cambridge use pulsed-SILAC proteomics to systematically map proteins degraded by the proteasome without prior ubiquitination — ubiquitin-independent proteasomal degradation. They identify the paralogous proteasome activation factors ZFAND5 and ZFAND6: ZFAND5 contains a ubiquitin-independent degron that drives its own rapid turnover by the very proteasome it allosterically activates, coupling activation and activator destruction in a single feedback circuit. ZFAND5/6 plus the autophagy scaffold p62 restrain NF-kB activation downstream of TLR4, traced to degradation of UBCH5c, an abundant E2 ubiquitin-conjugating enzyme that primes ubiquitin chain synthesis. The landscape of proteins susceptible to ubiquitin-independent proteasomal degradation is larger than appreciated and intersects directly with the substrate space for targeted protein degradation, and innate inflammatory signaling sits on top of an E2 enzyme whose abundance is tunable through proteasome activator dynamics. (3) Daniel Aschmann, Renske Knol and colleagues with Alexander Kros at Leiden University show that the stereochemistry of the ionizable lipid in a lipid nanoparticle reshapes its supramolecular architecture and mRNA payload translation output. cKK-E12 — normally used as a stereoisomer mixture — was resolved into trans (R,S/S,R) and cis (R,R/S,S); cis-isomer LNPs have superior physicochemical properties, stability, and protein expression. Systematically varying stereochemistry of ionizable lipid, phospholipid, and cholesterol revealed stereochemistry-dependent differences in uptake and protein expression across six cell lines, zebrafish embryos, and mice. AI-assisted cryo-EM and SAXS trace the functional differences to a stereochemistry-controlled selection between lamellar and inverse hexagonal internal lipid phases. Whole-particle chiral identity — not just the ionizable headgroup — is a previously underexploited engineering handle on internal architecture and translation output. (4) Shuhei Sakamoto, Hiroshi Hiraide and colleagues with Yu Kagami at Cosomil and collaborators across the University of Tokyo and multiple Japanese hospitals report a blood-based liquid biopsy — single-molecule enzyme activity profiling (SEAP) — detecting proteoform-level alterations in circulating pancreas-specific digestive enzymes at single-molecule resolution. A combinatorial classifier across 690 blood samples from multiple hospitals and biobanks detected PDAC stage I-IV at 95.5%% specificity and 75.0%% sensitivity, with 74.2%% sensitivity for stage I-II. In an independent early-stage-enriched cohort it called positive on 54.2%% of stage IA, 64.3%% of stage IB, and 21.4%% of stage 0 pancreatic intraepithelial lesions. PDAC survival is overwhelmingly determined by stage at detection; a single-molecule activity readout that engages with the earliest stages of disease from a blood draw is the kind of platform that, if it holds in prospective trials, would change how this cancer is found. 2026-06-10-ilt3-asms-tric-checkpoint-zfand5-proteasome-feedback-ckk-stereo-lnp-seap-pdac Wed, 10 Jun 2026 12:00:00 +0000 545 Four bioRxiv stories: (1) Abdel-Rahman, Murugan, Gabr (Weill Cornell Medicine) — paired-paper validation that ILT3 (LILRB4), long pursued as an antibody target across both tumor myeloid immune checkpoint and Alzheimer's disease neuroimmune checkpoint biology, is small-molecule druggable across both indications with two orthogonal chemotypes from two independent screening platforms. Affinity-selection mass spectrometry identifies LT12 with nanomolar binding (MST + SPR), discrete binding pocket mapped by computational modeling and mutagenesis, disrupts ILT3-ApoE interaction, attenuates SHP1/SHP2 + suppresses NF-kB + reduces IL-1-beta + restores amyloid-beta uptake in human iPSC-derived microglia, improves cognition + reduces amyloid burden in 5xFAD mice. Dianthus TRIC screen of an 8,961-member Enamine library identifies ICB-7 with D2-domain hydrophobic binding pocket, disrupts ILT3-SCG2 interaction, inhibits SHP1/SHP2/STAT3 signaling, enhances cytotoxic T-cell killing in patient-derived colorectal carcinoma + AML co-cultures, favorable PK + anti-tumor efficacy in CT26 syngeneic colorectal model. (2) Watson, Williamson and colleagues with Lehner (University of Cambridge) — pulsed-SILAC proteomics identifies ZFAND5 + ZFAND6 as ubiquitin-independent proteasome activators containing a ubiquitin-independent degron that drives their own rapid turnover by the very proteasome they allosterically activate, coupling activation + activator destruction in a feedback circuit. ZFAND5/6 + p62 restrain NF-kB activation downstream of TLR4 through degradation of UBCH5c E2 ubiquitin-conjugating enzyme. Expands the catalog of ubiquitin-independent proteasomal degradation intersecting targeted protein degradation substrate space and reveals tunable innate-immune signaling through proteasome activator dynamics. (3) Aschmann, Knol and colleagues with Kros (Leiden University) — cKK-E12 ionizable lipid separated into trans (R,S/S,R) and cis (R,R/S,S) stereoisomers, with cis-isomer LNPs showing superior physicochemical properties + stability + protein expression. Systematically varying stereochemistry of ionizable lipid + phospholipid + cholesterol reveals stereochemistry-dependent differences in uptake + protein expression across six cell lines + zebrafish embryos + mice; AI-assisted cryo-EM + SAXS links functional differences to stereochemistry-controlled selection between lamellar + inverse hexagonal internal lipid phases. Whole-particle chiral identity as a previously underexploited LNP engineering handle. (4) Sakamoto, Hiraide and colleagues with Komatsu/Kagami (Cosomil/University of Tokyo) — single-molecule enzyme activity profiling (SEAP) blood-based liquid biopsy detecting proteoform-level alterations in circulating pancreas-specific digestive enzymes. Combinatorial classifier achieves 95.5%% specificity + 75.0%% sensitivity (74.2%% for stage I-II) across 690 blood samples from multiple hospitals and biobanks; 54.2%% stage IA + 64.3%% stage IB + 21.4%% stage 0 detection in independent early-stage-enriched validation cohort. Single-molecule activity readout engaging the earliest stages of PDAC from a blood draw with potential to change how this cancer is found. Four bioRxiv stories spanning a synthetic protease-activated GPCR platform that rewires pain-inducing protease activity into an analgesic signal in vivo, a thermodynamically grounded drug-target specificity foundation model unifying off-target prediction/repurposing/generative design from sequence, a miniaturized zwitterionic subcutaneous cell-encapsulation platform sustaining a broadly neutralizing anti-HIV antibody for 36 weeks, and a deep-mutational-scanning map of the CFTR variant selectivity and off-target landscape of VX-121 (the new corrector inside Alyftrek). (1) Matthew Ravalin, Nicholas Kalogriopoulos and colleagues with Alice Ting at Stanford build synthetic protease-activated receptors — SynPARs — from engineered, auto-inhibited GPCRs in which proteolytic relief of autoinhibition allows activation by either an exogenous or a tethered agonist. The same scaffold drives transgene expression, real-time fluorescence, or endogenous G-protein signaling, and is modular across diverse secreted proteases. A cell-based library selection optimizes the protease recognition sequence for any target enzyme. In the mouse dorsal root ganglion, where extracellular trypsin produces hyperalgesia, SynPAR-equipped neurons translate the trypsin signal into an analgesic response — a clean demonstration that a pain-inducing protease can be rewired into a pain-suppressing input at the level of individual sensory neurons. Lays the groundwork for a new class of programmable cell-based and gene-based medicines built on a well-characterized GPCR chassis. (2) Sai Reddy at ETH Zurich reports a drug-target Specificity Foundation Model (dtSFM) built on the mathematical identity that transformer softmax attention is isomorphic to the Boltzmann distribution governing binding at thermal equilibrium — prescribing a single sequence-native architecture that computes molecular binding compatibility as a thermodynamic quantity directly from sequence. Trained on 714,747 measured drug-protein interactions across 522,776 compounds and 22,964 proteins, with AlphaFold 3 used as an independent structural verifier sharing no architecture, training data, or representational basis. The encoder retrieves a drug's target and a target's drug at 95% and 89% recall-at-10, and ranks the known off-targets of clinical kinase inhibitors at a median of 30th out of 4,910 genes — the top 0.6% of a proteome-scale screen, validated against a chemoproteomic panel. In the target-to-drug direction it screens the 522,776-compound library against three immunology targets and identifies 46 novel candidates passing AlphaFold 3 structural gating. The cross-attentive generative decoder produces novel molecules against 16 targets, with 850 of 1,200 (71%) designed candidates matching the AlphaFold 3 structural confidence of the approved drug. Unifies off-target prediction, repurposing, and de novo generative design from one sequence-native architecture; wet-lab validation is the immediate next step. (3) Minseok Lee, Bo Wang and colleagues with Minglin Ma at Cornell build a miniaturized subcutaneous cellular-implant platform combining clonal mesenchymal stromal cells engineered to produce PGT121 — a broadly neutralizing anti-HIV-1 antibody — with a zwitterionic polyurethane nanofibrous foreign-body-response-mitigating membrane in a thin cylindrical form factor compatible with applicator-based minimally invasive implantation and retrieval. Subcutaneously implanted in multiple murine models, the device sustained serum PGT121 concentrations for up to 36 weeks, and devices remained functional after cryopreservation — opening an off-the-shelf manufacturing path. The applicator-based implantation and retrieval procedures were demonstrated in a clinically relevant minipig model. The same chassis is target-agnostic: swap the antibody, swap the protein. For passive immunization against HIV and chronic biologics that today require repeat injection in settings where cold chain and clinical infrastructure are limiting, a meaningful advance in the cell-encapsulation modality. (4) Akshat Jhangiani, Joel Olson and colleagues with Jonathan Schlebach at Purdue University use deep mutational scanning across 232 rare CFTR variants to profile the variant selectivity and off-target landscape of VX-121 — the new CFTR corrector inside Alyftrek. VX-121 generally enhances variant expression more strongly than its precursor VX-445 and is most potent toward variants with mutations in the first membrane-spanning domain. Y1032C is identified as a variant with diminished proteostatic and functional selectivity for VX-121 relative to VX-445, traced by computational docking to disruption of favorable interactions between the native tyrosine side chain and VX-121. Photo-crosslinking experiments show that VX-121 avoids a key off-target interaction that burdens VX-445. Provides a rational variant-resolved framework for matching specific rare CFTR alleles to the right corrector chemotype, and demonstrates that VX-121's improved profile reflects a cleaner on-target binding pose rather than just bulk potency. 2026-06-09-synpar-protease-gpcr-dtsfm-zwitterionic-implant-vx121-cftr-dms Tue, 09 Jun 2026 12:00:00 +0000 455 Four bioRxiv stories: (1) Ravalin, Kalogriopoulos and colleagues with Ting (Stanford University) — synthetic protease-activated receptors (SynPARs) from engineered auto-inhibited GPCRs where proteolytic relief of autoinhibition enables activation by exogenous or tethered agonist, modular across diverse secreted proteases with cell-based library selection optimizing protease recognition sequence; in mouse dorsal root ganglion SynPAR-equipped neurons rewire trypsin (a pain-inducing protease) into an analgesic signal, establishing a programmable cell-based/gene-based modality for protease-sensing medicines on a well-characterized GPCR chassis. (2) Reddy (ETH Zurich) — drug-target Specificity Foundation Model (dtSFM) built on the transformer-softmax-attention/Boltzmann-distribution isomorphism computing molecular binding compatibility as a thermodynamic quantity directly from sequence, trained on 714,747 measured drug-protein interactions across 522,776 compounds and 22,964 proteins with AlphaFold 3 as orthogonal structural verifier. 95% drug-to-target and 89% target-to-drug recall-at-10; ranks documented off-targets of clinical kinase inhibitors at median 30th of 4,910 genes (top 0.6%) validated against a chemoproteomic panel; identifies 46 novel candidates against three immunology targets passing AlphaFold 3 gating in target-to-drug screen of a 522,776-compound library; cross-attentive generative decoder designs molecules for 16 targets with 850 of 1,200 (71%) matching the AlphaFold 3 structural confidence of the approved drug. Unifies off-target prediction, repurposing, and generative design from one sequence-native architecture. (3) Lee, Wang and colleagues with Ma (Cornell University) — miniaturized subcutaneous cellular implant combining clonal mesenchymal stromal cells engineered to produce the broadly neutralizing anti-HIV-1 antibody PGT121 with a zwitterionic polyurethane nanofibrous foreign-body-response-mitigating membrane in a thin cylindrical form factor compatible with applicator-based minimally invasive implantation/retrieval; sustained serum PGT121 concentrations for up to 36 weeks across multiple murine models with cryopreservation-compatible off-the-shelf manufacturing and minipig-demonstrated implantation/retrieval procedure. Target-agnostic chassis for chronic biologic delivery in resource-limited settings. (4) Jhangiani, Olson and colleagues with Schlebach (Purdue University) — deep mutational scanning across 232 rare CFTR variants profiling the variant selectivity and off-target landscape of VX-121, the new CFTR corrector inside Alyftrek. VX-121 generally enhances variant expression beyond VX-445 and is most potent toward MSD1 mutations; Y1032C shows diminished selectivity for VX-121 vs VX-445 traced by computational docking to disrupted native-tyrosine side-chain coordination; photo-crosslinking shows VX-121 avoids a key off-target interaction of VX-445. Rational variant-resolved framework for matching rare CFTR alleles to the right corrector chemotype. Four bioRxiv stories spanning a tractable G-protein-coupled receptor sitting upstream of therapy-resistant melanoma, a defined pool of endogenous double-stranded RNAs that arm MDA5 plus a direct DDX3X-mutant cancer connection, a heme-driven tissue-level stress framework for excess extracellular protease activity, and a humanized-mouse demonstration that immune-checkpoint-inhibitor colitis is driven by autoantibodies acting through Fc gamma receptors. (1) Naina Gour, Aishwarya Atakkatan and colleagues with Xinzhong Dong at the Johns Hopkins University School of Medicine identify Mas-Related G-protein-coupled receptor X4 (MRGPRX4) — previously characterized as a sensory-neuron-restricted receptor for cholestatic itch — as an unexpected melanoma driver. MRGPRX4 is upregulated in melanoma and confined to invasive, neural-crest-like and pre-EMT states associated with dedifferentiation and therapy resistance. Ectopic MRGPRX4 expression in mouse melanocytes drove 100%% penetrant, highly metastatic melanoma — demonstration of oncogenic behavior, not just association. The receptor activates PI3K-AKT-MAPK signaling in a basal, ligand-independent manner. Multi-omic profiling links MRGPRX4 expression to a mesenchymal and neural-crest-like program defining a distinct invasive niche, with broad transcriptional overlap with BRAF- and NRAS-driven tumors plus an additional ECM-rich invasive program. The receptor also remodels the tumor microenvironment toward an immunosuppressive, checkpoint-high state enriched in suppressive myeloid cells. Pharmacologic MRGPRX4 inhibition suppressed basal signaling and limited melanoma growth and invasion — nominating a tractable, ligand-independent somatosensory GPCR sitting upstream of the very phenotypic state that drives therapy-resistant melanoma. (2) Jiazhang Chen, Zhe Mo and colleagues with Sun Hur at Boston Children's Hospital and Roger Greenberg at the University of Pennsylvania answer the long-standing question of which endogenous dsRNAs actually arm MDA5 by direct affinity purification of cellular MDA5 filaments followed by RNA sequencing. ADAR1 converts adenosine to inosine in endogenous dsRNAs to prevent excessive MDA5-driven interferon-stimulated gene expression, but only a small fraction of ADAR1 substrates activate MDA5, and cellular MDA5 filaments themselves had never been isolated. Greater than 84%% of dsRNAs suppressed by combined DDX3X helicase and ADAR1 base-editing activity were present in MDA5 filaments — versus less than 1%% of substrates acted on by ADAR1 alone. The immunogenic species are inverted-repeat structures embedded in 3'-UTRs with high base-pair complementarity and long intervening sequences between repeats, with a minor contribution from sense/antisense intermolecular duplexes. Critically, the majority of these dual-substrate immunogenic dsRNAs were hyperedited in DDX3X-mutant cancers. ADAR1 inhibition is one of the most active small-molecule modalities in oncology right now; defining the specific MDA5-loading substrate pool — and a DDX3X-mutant cancer subset where editing is dialed up — gives the field a much sharper handle on patient stratification and on the failure modes of ADAR1 inhibition. (3) Karen Agaronyan, Allison Greaney and colleagues with Ruslan Medzhitov at Yale University frame tissue-level stress as a discrete and conserved injury signature. Using the lung and a panel of diverse proteases, the team identifies a unifying signature of proteolytic stress marked by vascular disruption, red-blood-cell extravasation, and heme release that triggers oxidative stress. Alveolar macrophages act as the primary sensors of this response, activating an NRF2-dependent heme detoxification program, while fibroblasts produce protease inhibitors that limit the damage. Repeated exposure induces tissue adaptation that protects against subsequent injury and infection. For diseases driven by excess extracellular protease activity — chronic obstructive pulmonary disease, rheumatoid arthritis, and a long list of others — this defines a tractable, pharmacologically addressable axis at the level of heme detoxification, NRF2 induction, and protease-inhibitor biology, rather than chasing individual proteases. (4) Irina Voloshyna, Yuri Patskovsky and colleagues with Michelle Krogsgaard at the NYU Langone Medical Center recast immune-checkpoint-inhibitor colitis as an autoantibody-driven, FcgR-dependent process. The group transferred IgG from melanoma patients treated with pembrolizumab, nivolumab, or ipilimumab — with or without severe colitis — into wild-type or humanized-FcgR mice receiving comparable checkpoint blockade. Wild-type recipients showed no colonic changes. Humanized-FcgR mice receiving IgG from colitic patients developed colon inflammation marked by submucosal lymphocyte infiltration, goblet-cell loss, and elevated circulating IL-6, IL-17, and IL-22. scRNA-seq identified an IgG-regulated inflammatory network spanning IFN-gamma-producing ILC1, Th1 and cytotoxic T cells, IL-1-beta M1 macrophages, plasmablasts, and IL-22-producing ILC3 LTi cells. Patient serum autoantibody profiling further identified the chemokine receptors CCR5 and CXCR4 as candidate immune-related autoantigens associated with colitis susceptibility. This reframes immune-related adverse events from generic T-cell collateral damage into a defined autoantibody-FcgR axis — and CCR5 and CXCR4 are both already drugged, opening immediate translational angles for prophylaxis or treatment. 2026-06-08-mrgprx4-melanoma-mda5-dsrna-heme-proteolytic-stress-icb-autoantibody-colitis Mon, 08 Jun 2026 12:00:00 +0000 469 Four bioRxiv stories: (1) Gour, Atakkatan and colleagues with Dong (Johns Hopkins University School of Medicine) — Mas-Related GPCR X4 (MRGPRX4), previously characterized as a sensory-neuron-restricted receptor for cholestatic itch, identified as an unexpected melanoma driver upregulated in invasive, neural-crest-like and pre-EMT states associated with dedifferentiation and therapy resistance. Ectopic MRGPRX4 expression in mouse melanocytes drives 100%% penetrant, highly metastatic melanoma through basal, ligand-independent PI3K-AKT-MAPK activation. Multi-omics links MRGPRX4 to a mesenchymal/neural-crest-like program with broad transcriptional overlap with BRAF/NRAS-driven tumors plus an additional ECM-rich invasive program, plus microenvironmental remodeling toward an immunosuppressive checkpoint-high state enriched in suppressive myeloid cells. Pharmacologic MRGPRX4 inhibition limits melanoma growth and invasion — tractable somatosensory GPCR sitting upstream of the therapy-resistant melanoma phenotype. (2) Chen, Mo and colleagues with Hur (Boston Children's Hospital) and Greenberg (University of Pennsylvania) — affinity-purify cellular MDA5 filaments and sequence their bound RNA to identify the endogenous dsRNAs that actually drive MDA5-mediated interferon signaling. Greater than 84%% of dsRNAs suppressed by combined DDX3X helicase and ADAR1 base-editing activity are present in MDA5 filaments versus less than 1%% of ADAR1-only substrates; dual-substrate immunogenic dsRNAs are inverted repeats embedded in 3'-UTRs with high base-pair complementarity and long intervening sequences plus minor contribution from sense/antisense intermolecular duplexes; majority hyperedited in DDX3X-mutant cancers. Defined MDA5-loading substrate pool and a DDX3X-mutant cancer subset for ADAR1-inhibitor patient stratification and failure-mode analysis. (3) Agaronyan, Greaney and colleagues with Medzhitov (Yale University) — define tissue-level proteolytic stress as a conserved heme-NRF2-protease-inhibitor axis across diverse proteases applied to the lung. Vascular disruption, red-blood-cell extravasation, and heme release trigger oxidative stress sensed by alveolar macrophages via NRF2-dependent heme detoxification, with fibroblasts producing protease inhibitors to limit damage and repeated exposure inducing tissue adaptation protective against subsequent injury and infection. Pharmacologically addressable framework for COPD, rheumatoid arthritis, and other diseases driven by excess extracellular protease activity. (4) Voloshyna, Patskovsky and colleagues with Krogsgaard (NYU Langone Medical Center) — IgG from pembrolizumab/nivolumab/ipilimumab-treated melanoma patients with severe colitis transferred into humanized-FcgR mice receiving comparable ICI therapy drives submucosal lymphocyte infiltration, goblet-cell loss, and elevated IL-6/IL-17/IL-22 while wild-type recipients show no colonic changes. scRNA-seq identifies an IgG-regulated inflammatory network spanning IFN-gamma-producing ILC1, Th1/cytotoxic T cells, IL-1-beta M1 macrophages, plasmablasts, and IL-22-producing ILC3 LTi cells. Patient serum autoantibody profiling nominates CCR5 and CXCR4 chemokine receptors — both already drugged — as candidate autoantigens, recasting irAEs as a defined autoantibody-FcgR axis with immediate translational angles. Four bioRxiv stories spanning a first-in-class small-molecule modality against a conserved herpesvirus replication enzyme, an organ-targeted mRNA gene replacement that reverses polycystic kidney disease, an autonomous agentic-AI-driven enzyme-cascade optimization platform, and a new endothelial resilience pathway for atherosclerosis. (1) Neha Sharma, Xiaoyu Xie and colleagues with Sandra Weller at the University of Connecticut develop the first viral nuclease inhibitors (VNIs) targeting a conserved class of essential herpesvirus replication enzymes. The HSV-1 UL12 gene encodes a 5'-to-3' alkaline exonuclease that collaborates with the single-stranded DNA binding protein ICP8 to drive recombination-dependent viral replication, and orthologs are essential in every other human herpesvirus — UL98 in HCMV and SOX in KSHV are conserved across alpha, beta, and gamma subfamilies. The team solved the first crystal structure of an alphaherpesvirus alkaline nuclease (UL12.5) and used it to drive structure-activity relationships, producing inhibitors with nanomolar to low-micromolar EC50/IC50 values and antiviral activity against HSV-1 and HCMV in cell culture. The same chemistry inhibits the HCMV UL98 and KSHV SOX orthologs, supporting a pan-herpesvirus small-molecule strategy as standalone therapy or in combination with nucleoside analogs against viruses where resistance to existing agents is a growing clinical problem. (2) Joshua Giblin, Karen Lambaren and colleagues with Eun Ji Chung at the University of Southern California develop a collecting-duct-targeted lipid nanoparticle that delivers Pkd2 mRNA directly to the renal collecting-duct epithelia where cysts originate in ADPKD. ADPKD is caused by mutations in PKD1 or PKD2 that leave cells deficient in polycystin-1 or polycystin-2 protein; no approved therapy restores polycystin expression, and tolvaptan — the only approved intervention — targets downstream signaling rather than the genetic defect. The team incorporated a collecting-duct-targeting peptide into the LNP, drove renal accumulation over non-targeted formulations, and in a Pkd2-deficient mouse model repeated administration reversed established cystic disease, restored renal architecture, and reduced fibrotic and inflammatory markers without detectable off-target toxicity. A single dose also reduced cyst burden and improved renal function in the Pkd1-deficient background. First organ-targeted mRNA gene replacement to reverse rather than slow ADPKD in vivo across both genetic backgrounds — credible alternative to AAV-based gene therapy for a disease where dosing-to-organ has been the rate-limiter. (3) Michal Vasina, David Kovar and colleagues with Zbynek Prokop at Masaryk University, working with Stavros Stavrakis and Andrew deMello at ETH Zurich and Stanislav Mazurenko, introduce CascadeMAP, an autonomous closed-loop microfluidic platform that combines high-throughput droplet microfluidics with Bayesian optimization and a multi-agent AI system to optimize enzyme cascades end-to-end. The team validated it on two cascades — a fluorescence-monitored glycerol detection pathway and a label-free Raman-spectroscopy-monitored 1,2,3-trichloropropane degradation pathway — two orthogonal detection modalities. Bayesian optimization identified optimal conditions 3x faster than classical Design of Experiments. The multi-agent AI system automated hypothesis generation across 11 GB of experimental data, pattern recognition, and insight synthesis. Running for 7 days without human intervention, CascadeMAP processed ~220,000 reactions across ~7,400 different conditions — a generalizable framework for autonomous optimization of biocatalytic cascades and metabolic pathways with direct relevance to industrial biocatalysis and synthetic biology pipelines. (4) Iqra Ilyas and colleagues with Suowen Xu at the University of Science and Technology of China — in a multi-institution collaboration spanning Bordeaux, Munich, Utrecht, Sheffield, Bad Nauheim, Charlottesville and Shanghai — identify desert hedgehog (DHH), a canonical ligand of the hedgehog signaling pathway, as a novel intrinsic endothelial resilience factor that protects against atherosclerosis. Most current cardiovascular pharmacotherapies target risk factors that drive disease rather than factors that protect endothelium in the face of those risk factors. scRNA-seq of atheroprone and atheroprotective regions of mouse aorta identified DHH as a mechanoresponsive gene enriched in regions exposed to unidirectional laminar flow, with expression depleted in atherosclerotic arteries and patient serum. Endothelial-specific knockout aggravated atherosclerosis through enhanced endothelial-mesenchymal transition (EndoMT). Mechanistically, DHH directly binds the receptor LRP1 and competitively blocks PAI-1 binding, suppressing downstream canonical TGF-beta and noncanonical AKT/ERK1/2 signaling that drives EndoMT. Recombinant DHH protein and the PAI-1 inhibitor TM5275 each reduced atherosclerosis in ApoE knockout mice — translatable resilience-based pharmacology distinct from lipid-lowering and antihypertensive standard of care. 2026-06-07-hsv-ul12-vni-cdlnp-pkd2-cascademap-dhh-pai1-atherosclerosis Sun, 07 Jun 2026 12:00:00 +0000 459 Four bioRxiv stories: (1) Sharma, Xie and colleagues with Weller (University of Connecticut) — first viral nuclease inhibitors (VNIs) targeting the HSV-1 UL12 alkaline exonuclease essential for recombination-dependent viral replication. First crystal structure of an alphaherpesvirus alkaline nuclease (UL12.5) drives structure-activity relationships; inhibitors active at nanomolar to low-micromolar EC50/IC50 with antiviral activity in cell culture against HSV-1 and HCMV. Same chemistry inhibits the HCMV UL98 and KSHV SOX orthologs across alpha, beta, and gamma subfamilies — pan-herpesvirus small-molecule strategy as standalone or nucleoside-analog combination therapy against resistance-prone viruses. (2) Giblin, Lambaren and colleagues with Chung (USC) — collecting-duct-targeting-peptide-functionalized lipid nanoparticles deliver Pkd2 mRNA to renal collecting-duct epithelia, the predominant site of cyst origin in ADPKD. Repeated administration reverses established cystic disease, restores renal architecture, and reduces fibrotic and inflammatory markers in Pkd2-deficient mice without detectable off-target toxicity; single dose reduces cyst burden and improves renal function in the Pkd1-deficient background. First organ-targeted mRNA gene replacement to reverse rather than slow ADPKD in vivo across both genetic backgrounds. (3) Vasina, Kovar and colleagues with Prokop (Masaryk University), Stavrakis and deMello (ETH Zurich), and Mazurenko — CascadeMAP, an autonomous closed-loop microfluidic platform combining high-throughput droplet microfluidics with Bayesian optimization and a multi-agent AI system to optimize enzyme cascades end-to-end. Validated on fluorescence-monitored glycerol detection and Raman-monitored 1,2,3-trichloropropane degradation pathways. Bayesian optimization 3x faster than Design of Experiments; multi-agent AI automates hypothesis generation across 11 GB of experimental data, pattern recognition, and insight synthesis; 7-day unattended operation processes ~220,000 reactions across ~7,400 conditions. Generalizable framework for autonomous optimization of biocatalytic cascades and metabolic pathways. (4) Ilyas, Wang and colleagues with Xu (University of Science and Technology of China, multi-institution) — desert hedgehog (DHH) identified by scRNA-seq as a mechanoresponsive endothelial resilience factor enriched in atheroprotective aortic regions and depleted in atherosclerotic arteries / patient serum. Endothelial-specific DHH knockout aggravates atherosclerosis through enhanced endothelial-mesenchymal transition; DHH directly binds LRP1 and competitively blocks PAI-1 to suppress canonical TGF-beta and noncanonical AKT/ERK1/2 signaling. Recombinant DHH protein and PAI-1 inhibitor TM5275 each reduce atherosclerosis in ApoE knockout mice — translatable resilience-based pharmacology orthogonal to lipid-lowering and antihypertensive standard of care. Four bioRxiv stories spanning a sequencer-only substrate for dense 3D molecular phenotyping in intact tissue, an expression-increasing ASO modality against a regulatory RNA at the LDL-receptor locus with in-vivo validation, a clinically deliverable CNS-penetrant CYP51-inhibitor repurposing path for TDP-43 proteinopathies, and a systems-scale germline architecture of antibody immunodominance. (1) Nianchao Qian, Rabia Yasser and colleagues with Joshua Weinstein at the University of Chicago introduce cross-linked volumetric DNA microscopy (xVDM), a sequencing-only method that builds dense three-dimensional molecular networks in intact tissue without a microscope. Resolving cell phenotypes in full tissue context has required imaging methods that retain spatial neighborhoods but trade off depth, throughput, and the number of biomolecules captured. xVDM seeds unique molecular identifiers directly into the tissue protein matrix and links them with uniquely labeled DNA bridges, generating a DNA-encoded proximity network from which cell-scale molecular communities are reconstructed computationally. Nucleating networks from the protein matrix produces denser graphs and broader transcriptome recovery than transcript-nucleated approaches. The team mapped genetically annotated 3D networks in intact zebrafish embryos at 12, 18, and 24 hours post-fertilization and extended the framework to human tonsil with antibody-oligonucleotide conjugates. The whole pipeline runs on standard bench reagents and a DNA sequencer — a low-equipment-barrier substrate for 3D molecular phenotyping. (2) Keerthana Gnanapradeepan, Sylvia Manska and colleagues with Daniel Tardiff at CAMP4 Therapeutics provide a working example of a novel ASO modality that increases endogenous gene expression by targeting non-coding regulatory RNAs at promoters and enhancers. The team mapped regulatory RNAs at the LDLR locus in HepG2 cells and human liver using epigenomic profiling and bespoke regRNA capture sequencing, designed ASOs against actionable elements, and showed that several increased LDLR mRNA and protein in cells. A lead ASO administered to a humanized-liver mouse model raised human LDLR mRNA and protein and lowered plasma LDL-cholesterol — the in-vivo translation step that has been the rate-limiter for this whole class of expression-increasing oligonucleotides. The framework — regulatory-element mapping, ASO screening against actionable sequences, humanized in-vivo validation — generalizes directly to haploinsufficient disease loci where restoring half-dose expression is the therapeutic goal. (3) Nathan Kochen, Sahar Zafari and colleagues with Jonathan Sachs at the University of Minnesota nominate posaconazole — an FDA-approved, CNS-penetrant, orally delivered antifungal — as a leading candidate to clear cellular TDP-43 pathology relevant to ALS and frontotemporal dementia. The group had previously shown that ketoconazole, an antifungal CYP51 inhibitor, stabilized TDP-43 native self-interactions and reduced pathology, but ketoconazole's liver toxicity blocks ALS repurposing. Screening seven additional CYP51 inhibitors in their TDP-43 mislocalization-and-aggregation assay identified posaconazole as the strongest hit. Mechanism work shows posaconazole outperforms ketoconazole by inducing stronger autophagy activation and heat-shock-protein upregulation that clear TDP-43, and its effect depends on its known ability to lower cellular cholesterol; predicted binding affinity to human CYP51 across the panel correlates with potency against TDP-43. In a low-dose sodium-arsenite stress model in iPSC-derived motor neurons, posaconazole was active in the nanomolar range — a clinically deliverable, CNS-penetrant repurposing path for sporadic and familial TDP-43 proteinopathies. (4) Ellen Shrock, Eric Sun and colleagues with Stephen Elledge at Harvard and David Baker at the Institute for Protein Design show that germline-encoded antibody features, not somatic mutations, primarily determine which antigenic sites become immunodominant. Using the SARS-CoV-2 receptor-binding domain as the model antigen, the team combined large-scale antibody sequencing with deep mutational scanning and AlphaFold-3 structural modeling. 76%% of RBD-targeting antibodies share conserved gene-segment usage or germline-encoded HCDR3 motifs, and structural analysis pinpoints recurrent germline residues that contact the immunodominant epitopes — mutating these residues abolishes binding. SARS-CoV-2 variant substitutions frequently disrupt these germline contacts and explain immune escape; other variant substitutions open new germline-mediated epitopes, generating new immunodominant clusters. Innate genetic features dominate binding specificity for most of the antibody response — a constraint that rewrites how vaccine designers should think about epitope steering and germline targeting. 2026-06-06-xvdm-volumetric-dna-microscopy-ldlr-regrna-aso-posaconazole-tdp43-immunodominance-germline Sat, 06 Jun 2026 12:00:00 +0000 408 Four bioRxiv stories: (1) Qian, Yasser and colleagues with Weinstein (University of Chicago) — cross-linked volumetric DNA microscopy (xVDM), a sequencing-only method that builds dense 3D molecular networks in intact tissue without a microscope. Seeds unique molecular identifiers into the tissue protein matrix and links them via uniquely labeled DNA bridges to create a dense DNA-encoded proximity network from which cell-scale molecular communities are reconstructed computationally. Nucleating networks from the protein matrix produces denser graphs and broader transcriptome recovery than transcript-nucleated approaches. Genetically annotated 3D networks in intact zebrafish embryos at 12, 18, and 24 hpf; extension to human tonsil via antibody-oligonucleotide conjugates; standard bench reagents and a DNA sequencer — a low-equipment-barrier substrate for 3D molecular phenotyping. (2) Gnanapradeepan, Manska and colleagues with Tardiff (CAMP4 Therapeutics) — ASOs targeting non-coding regulatory RNAs at the LDLR locus increase endogenous LDLR expression in cells and in vivo. Epigenomic mapping and regRNA capture-seq in HepG2 and human liver, ASO screening against actionable elements, lead ASO administered to a humanized-liver mouse model raises human LDLR mRNA and protein and lowers plasma LDL-cholesterol — the in-vivo translation step that has been the rate-limiter for expression-increasing oligonucleotide modalities. Generalizes to haploinsufficient disease loci where restoring half-dose expression is the therapeutic goal. (3) Nathan Kochen, Zafari and colleagues with Sachs (University of Minnesota) — FDA-approved, CNS-penetrant, orally delivered antifungal posaconazole identified as the strongest CYP51-inhibitor hit in a TDP-43 mislocalization-and-aggregation assay across an 8-azole panel. Outperforms hepatotoxic ketoconazole via stronger autophagy activation and HSP upregulation that clear TDP-43; effect is cholesterol-lowering-dependent; predicted CYP51 affinity across the panel correlates with TDP-43 potency. Nanomolar activity in sodium-arsenite-stressed iPSC-derived motor neurons positions posaconazole as a clinically deliverable CNS-penetrant ALS/FTD repurposing candidate for TDP-43 proteinopathies. (4) Shrock, Sun and colleagues with Elledge (Harvard) and Baker (Institute for Protein Design) — systems-scale combination of large-scale antibody sequencing, deep mutational scanning, and AlphaFold-3 structural modeling using SARS-CoV-2 RBD as model antigen shows germline-encoded antibody features rather than somatic mutations drive immunodominance. 76%% of RBD-targeting antibodies display conserved gene-segment usage and/or germline-encoded HCDR3 motifs; structural analysis pinpoints recurrent germline-encoded residues that contact immunodominant epitopes (mutating these residues eliminates binding); SARS-CoV-2 variant substitutions either disrupt germline contacts to drive immune escape or open new germline-mediated epitopes generating new immunodominant clusters. Constrains vaccine-design logic for epitope steering and germline targeting. Four bioRxiv stories spanning a generative substrate for designer transmembrane nanopores, generalized prediction of T-cell-receptor specificity, structural mechanism of proteasome regulation by Parkinsonism-associated proteins, and a patient-specific vascularized tumor-organoid platform for combination immunotherapy. (1) Anvay Philomin, Rashmi Sonigra and colleagues with David Baker at the University of Washington present a rapid generative pipeline for de novo monomeric transmembrane beta-barrel design that combines diffusion-based backbone generation conditioned on beta-barrel structural features with transmembrane-barrel-optimized sequence design. Native homo-oligomeric beta-barrels have been the workhorse of nanopore sensing, sequencing, and separation, but the uniform symmetric lumen limits spatial resolution and the localized interactions required for analyte discrimination and selective filtration; energy-based monomeric designs have been built one at a time but each requires extensive expert input and does not scale. The team characterized 48 designs spanning 10-16 strands with measurable conductances corresponding to pore diameters of 0.7-1.5 nm, and solved crystal structures of two designs that match the designed models at atomic-level agreement. They demonstrated versatility by designing nanopores with copper-binding sites for selective ion sensing, larger pores that support DNA translocation, and longer pores with increased hydrophobic thickness that mediate ion transport across three-dimensional networks of monoglyceride bilayers and synthetic block-copolymer-lipid hybrid membranes. A programmable design substrate for monomeric transmembrane beta-barrels with custom inner-surface chemistry, geometry, and host-membrane compatibility — generative platform for biosensors, single-molecule chemistry, sequencing, separations, and synthetic transmembrane circuits. (2) Lucas Woods, Bradley Neff and colleagues with John Altin at the Translational Genomics Research Institute and Cristian Tomasetti show that AlphaFold 3 predicts MHC/peptide/T-cell-receptor triad structures with sufficient accuracy that generalized prediction of T-cell-receptor specificity from sequence becomes possible against epitopes the model has never seen. The MHC/peptide/T-cell-receptor interface is the most diverse trimolecular interaction known in biology and the central trigger for adaptive immunity; T-cell receptors are now routinely sequenced at scale, but decoding their cognate antigens has remained the bottleneck, and prior in-silico methods could not predict against unseen epitopes — limiting their applicability to a tiny fraction of cases. AlphaFold 3 applied to >9000 T-cell receptors mapped to >1000 distinct epitopes restricted by >70 HLA class I and class II alleles identifies structural features that distinguish cognate triads from controls, and a downstream model achieves median AUCs of 0.81-0.92 on validation triads unseen by AlphaFold 3 or by the feature-selection step. Generalized sequence-to-specificity prediction unlocks T-cell-receptor repertoire decoding, neoantigen-directed cell-therapy design, autoimmune-target identification, and infection-and-vaccine-response analysis at a scale that experimental sorting and tetramer screening cannot reach. (3) Frank Adolf, Ellen Goodall and colleagues with Brenda Schulman at the Max Planck Institute of Biochemistry, John Hanna at Brigham and Women's Hospital, and Wade Harper at Harvard Medical School resolve by cryo-EM the structural mechanism by which the Parkinsonism-associated proteins FBXO7 and PI31 regulate the 20S proteasome core particle. FBXO7 and PI31 variants are linked to rare Parkinsonian syndromes, implicating their dysfunction in neurodegeneration, but how they engage the proteasome at the structural level was not understood. Each protein can associate with the proteasome independently, with multiple domains of FBXO7 contributing; the FBXO7 C-terminal domain engages multiple subunits inside the core particle and blocks the beta-5 peptidase activity, while PI31 engages all three catalytic sites within the 20S core, providing the previously unknown structural basis for beta-1 inhibition. Disease-associated variants in both proteins disrupt proteasome regulation and assembly of the SKP1/FBXO7/PI31 complex. Rewrites the inventory of endogenous proteasome modulators in neurons and opens a structurally tractable axis for small-molecule or molecular-glue interventions to phenocopy or correct disease-associated dysregulation in Parkinson's disease. (4) Naveen Natesh, Soumen Maity and colleagues in Shyni Varghese's group at Duke University present a vascularized patient-specific tumor-organoid platform that combines self-organizing microvascular networks built from beta-2-microglobulin-knockout endothelial cells with patient-derived tumor organoids and autologous tumor-infiltrating lymphocytes — a universal endothelium engineered to minimize allo-recognition and make the platform broadly compatible across patient samples. The team quantitatively measured T-cell infiltration under immune-checkpoint blockade and stratified responders from non-responders consistent with patients' clinical outcomes. Single-cell RNA sequencing identified tumor-intrinsic and immune-associated programs underlying responder versus non-responder status, including tumor-driven hyperangiogenic signaling as a barrier to T-cell extravasation. Pharmacological co-targeting of PD1 and VEGF restored T-cell infiltration in non-responder organoids and shifted them from an immune-excluded to an immune-inflamed state — mechanistic rationale plus patient-level preclinical platform for the design of combination immunotherapy. 2026-06-05-baker-beta-barrel-nanopore-af3-tcr-fbxo7-pi31-proteasome-vascularized-tumor-organoid Fri, 05 Jun 2026 12:00:00 +0000 418 Four bioRxiv stories: (1) Philomin, Sonigra and colleagues with Baker (University of Washington) — rapid generative pipeline for de novo monomeric transmembrane beta-barrel nanopore design combining diffusion-based backbone generation with transmembrane-barrel-optimized sequence design. 48 designs spanning 10-16 strands with measurable conductances at pore diameters of 0.7-1.5 nm; crystal structures of two designs match designed models at atomic-level agreement. Programmable substrate for sensors/sequencing/separations: copper-binding variants for selective ion sensing, larger pores supporting DNA translocation, longer hydrophobic pores mediating transport across block-copolymer-lipid hybrid membranes. (2) Woods, Neff and colleagues with Altin (Translational Genomics Research Institute) and Tomasetti — AlphaFold 3 predicts MHC/peptide/T-cell-receptor triad structures with sufficient accuracy that generalized sequence-to-specificity prediction against unseen epitopes becomes possible. AF3 applied to >9000 TCRs / >1000 epitopes / >70 HLA class I/II alleles; downstream model achieves median AUC 0.81-0.92 on triads unseen by AF3 or feature selection. Unblocks T-cell-receptor repertoire decoding, neoantigen-directed cell therapy, autoimmune target identification, and infection/vaccine response analysis at scales tetramer sorting cannot reach. (3) Adolf, Goodall and colleagues with Schulman (Max Planck Institute of Biochemistry), Hanna (Brigham and Women's), and Harper (Harvard Medical School) — cryo-EM resolves the structural mechanism by which the Parkinsonism-associated proteins FBXO7 and PI31 regulate the 20S proteasome. The FBXO7 C-terminal domain engages multiple core-particle subunits and blocks beta-5 peptidase activity; PI31 engages all three catalytic sites providing the previously unknown structural basis for beta-1 inhibition. Disease-associated variants disrupt proteasome regulation and SKP1/FBXO7/PI31 complex assembly. Rewrites the inventory of endogenous proteasome modulators in neurons and opens a structurally tractable axis for Parkinson's disease intervention. (4) Natesh, Maity and colleagues in Varghese's group (Duke University) — vascularized patient-specific tumor-organoid platform combines self-organizing microvascular networks built from beta-2-microglobulin-knockout endothelial cells with patient-derived tumor organoids and TILs as a universal endothelium minimizing allo-recognition. Quantitative T-cell-infiltration measurement under checkpoint blockade stratifies responders/non-responders consistent with clinical outcomes; scRNA-seq identifies tumor-driven hyperangiogenic signaling as barrier to T-cell extravasation; PD1+VEGF co-targeting restores infiltration in non-responder organoids, shifting them from immune-excluded to immune-inflamed. Patient-level preclinical platform for combination immunotherapy design. Four bioRxiv stories spanning a repurposed antibiotic as the first mutant-selective small-molecule disease-modifying agent for polycythemia vera, virus-free genome-integrating CAR-T manufacturing, a new synthetic-receptor class for extracellular protease activity, and short single-stranded DNA as a programmable intracellular RNA generator. (1) Anastasia Gumerova, Christoph Schaniel, Zhuoying Huang and colleagues with Huihui Li and Ronald Hoffman at the Icahn School of Medicine at Mount Sinai discovered that the FDA-approved oxazolidinone antibiotic linezolid acts as a selective inhibitor of the JAK2 V617F mutation that drives roughly 95%% of polycythemia vera. Clinically available JAK1/2 inhibitors are mutant-allele agnostic — they suppress excess hematopoiesis but spare the mutant stem cell pool and lose efficacy as resistance emerges. In JAK2-mutant erythroleukemia cells, linezolid suppressed proliferation, reduced STAT5 signaling, perturbed the cell cycle, and increased apoptosis without affecting wild-type acute leukemia cells. Computational modeling localized the interaction specifically to V617F mutant kinase but not wild-type JAK2. In a JAK2VF mouse model that faithfully recapitulates human PV, linezolid normalized spleen size and blood counts and selectively depleted the mutant hematopoietic stem and progenitor compartment. In ex vivo colony assays on PV patient PBMCs, mutant colonies were preferentially eliminated while wild-type colonies were preserved, with the most primitive progenitors as preferred targets. Closest stem-cell-depleting mutant-selective small molecule the myeloproliferative neoplasm field has had on a known safety profile — de facto lead structure for next-generation selective JAK2 V617F inhibitors. (2) Kazuya Inoue, Mariko Masui and colleagues with Nao Suzuki-Yamazaki at FUJIFILM Corporation attack the dominant bottleneck in virus-free CAR-T engineering — getting donor DNA into the nucleus. Viral vectors deliver but at cost/safety/scalability constraints; LNPs deliver mRNA well but stall at the cytoplasm for the DNA payload required for stable transgene integration. The team fused a nuclear localization signal directly to either a transposase or a genome-editing nuclease and packaged the system in proprietary on-demand-encapsulation ready-to-use LNPs. The NLS-fused transposase route delivered markedly higher CAR transgene integration in primary human T cells than electroporation, with higher CAR expression and better viability; for TRAC-locus site-specific integration, end-joining-based repair was substantially more efficient than HDR after LNP delivery. Resulting CAR-T cells exhibited potent antigen-specific cytotoxicity. Ready-to-use LNPs plus active nuclear shuttling as a viable virus-free CAR-T manufacturing platform — clean architectural fix for the nuclear-import rate-limiting step. (3) James Tran, Christopher Kuffner and colleagues in John Ngo's group at Boston University introduce ProNotch, a synthetic receptor architecture that converts extracellular protease activity into programmable transcriptional outputs. ProNotch tethers a destabilized mutant of the Notch1 negative regulatory region (NRR) — the autoinhibitory module that normally gates Notch activation — to inhibitory anti-NRR single-chain antibody fragments via protease-cleavable linkers. scFv engagement simultaneously rescues surface trafficking of the destabilized receptor and suppresses basal signaling, until linker cleavage releases the inhibitory scFv and unmasks ligand-independent receptor activation. Linker substitution reprograms protease specificity across diverse enzymes; tandem substrate repeats raise sensitivity without raising background; single-chain designs implement AND and OR logic gates that integrate multiple protease inputs into a single transcriptional output. ProNotch detected endogenous matrix-metalloproteinase-14 activity from cancer cell lines in cis and in trans and drove protease-dependent cell-state transitions in C3H/10T1/2 fibroblasts. The scFv-NRR module also worked as a soluble protease-activated pro-antibody that conditionally inhibited DLL4-driven Notch signaling and ligand-independent activation of mutant Notch1 in a T-cell acute lymphoblastic leukemia line. New input modality for engineered cell therapies and conditional biologics in solid tumors, fibrosis, and inflammation — disease microenvironments where extracellular protease activity is the most reliable spatial marker. (4) Ahmed Mahas, Roberta Perrotta and colleagues in George Church's group at Harvard Medical School establish that short single-stranded DNA oligonucleotides can drive intracellular transcription of functional RNAs in mammalian cells when paired with orthogonal bacteriophage RNA polymerases. The trick is geometric: embed the phage promoter inside a short hairpin so the recognition site is locally double-stranded on an otherwise single-stranded template, recovering robust polymerase engagement. The platform produces functional CRISPR guides that drive adenine base editing at reporter and endogenous loci and CRISPR-based transcriptional activation across multiple human cell lines, replacing cloning workflows, complex RNA synthesis, and viral-vector delivery with directly synthesizable ssDNA inputs — compact, synthetically accessible inputs for programmable intracellular RNA production with obvious utility across synthetic biology and genome engineering. 2026-06-04-linezolid-jak2v617f-nls-lnp-cart-pronotch-protease-church-ssdna-rna Thu, 04 Jun 2026 12:00:00 +0000 445 Four bioRxiv stories: (1) Gumerova, Schaniel, Huang and colleagues with Li and Hoffman (Icahn School of Medicine at Mount Sinai) — FDA-approved oxazolidinone antibiotic linezolid acts as a selective inhibitor of the JAK2 V617F mutation driving ~95%% of polycythemia vera. Suppresses proliferation/STAT5 signaling and induces apoptosis in JAK2VF erythroleukemia cells without affecting wild-type AML lines; computational modeling localizes binding to mutant V617F kinase but not wild-type JAK2. In a JAK2VF PV mouse model normalizes spleen size + blood counts and selectively depletes mutant HSPCs; in ex vivo PBMC colony assays from PV patients, mutant colonies are preferentially eliminated while wild-type are preserved, with primitive progenitors as the preferred targets. Closest stem-cell-depleting mutant-selective small molecule the myeloproliferative neoplasm field has had on a known safety profile. (2) Inoue, Masui and colleagues with Suzuki-Yamazaki (FUJIFILM Corporation) — virus-free CAR-T platform combining NLS-fused transposase or genome-editing nuclease active cytoplasm-to-nucleus shuttling of donor DNA with proprietary on-demand-encapsulation ready-to-use LNPs. Higher CAR transgene integration than electroporation in primary human T cells with higher CAR expression and better viability; TRAC-locus site-specific integration is substantially more efficient through end-joining than HDR after LNP delivery; resulting CAR-T cells exhibit potent antigen-specific cytotoxicity. Architectural fix for the nuclear-import rate-limiting step of non-viral CAR-T. (3) Tran, Kuffner and colleagues in Ngo's group (Boston University) — ProNotch synthetic receptor tethers a destabilized Notch1 NRR mutant to inhibitory anti-NRR scFvs via protease-cleavable linkers. scFv engagement simultaneously rescues surface trafficking and suppresses basal signaling until linker cleavage releases the scFv and initiates ligand-independent signaling. Linker substitution reprograms protease specificity; tandem substrate repeats raise sensitivity without background; single-chain designs implement AND/OR logic gates integrating multi-protease inputs. Detects endogenous MMP-14 activity in cis and in trans; drives protease-dependent cell-state transitions in C3H/10T1/2 fibroblasts; soluble scFv-NRR pro-antibody conditionally inhibits DLL4-driven signaling and ligand-independent mutant Notch1 activation in T-ALL. New input modality for engineered cell therapies and conditional biologics in solid tumors, fibrosis, and inflammation. (4) Mahas, Perrotta and colleagues in Church's group (Harvard Medical School) — short ssDNA oligonucleotides drive intracellular transcription of functional RNAs in mammalian cells when paired with orthogonal bacteriophage RNAPs. Embedding phage promoter in a short hairpin creates a locally double-stranded recognition site on otherwise single-stranded template, recovering robust polymerase engagement. Functional CRISPR guides drive adenine base editing at reporter and endogenous loci, plus CRISPR-based transcriptional activation across human cell lines, replacing cloning, complex RNA synthesis, and viral vectors with directly synthesizable ssDNA inputs. Four bioRxiv stories spanning extrahepatic LNP retargeting, metabolite-to-chromatin biochemistry, ribosome engineering for non-natural polymers, and pathology-targeted reversal of muscle fibrosis. (1) Joseph Bufton, Thomas Green, and colleagues with Benjamin Carter at Hone Bio Limited (Bristol) introduce NanoPilot, a modular antibody-anchor fusion-protein platform that competitively blocks the apolipoprotein E (ApoE) binding site of low-density-lipoprotein receptor (LDLR) to redirect preformulated lipid nanoparticles away from hepatocytes. The construct is added to existing LNP formulations in 10 minutes with two pipetting steps. An anti-CD3ε NanoPilot increased T-cell transfection 40-fold and reduced monocyte transfection 10-fold in human peripheral blood mononuclear cells in vitro; in immunocompetent mouse models, the same construct achieved 30–40%% splenic and hepatic T-cell transfection with bulk liver accumulation reduced 3-fold. An anti-c-Kit version enhanced delivery to a hematopoietic-stem-cell-like cell line in co-culture. A generalizable extrahepatic retargeting layer compatible with in-vivo CAR-T and hematopoietic-stem-cell editing programs without reformulating the LNP. (2) Sarah Graff and colleagues with Simone Sidoli and Julio Maianti at Albert Einstein College of Medicine establish the first direct mechanistic link between a TCA-cycle metabolite and JmjC histone-demethylase active sites. Quantitative proteomics and peptide pull-down assays identify multiple Jumonji-domain-containing histone demethylases — including KDM4D and KDM6B, which erase the repressive H3K9me2/3 and H3K27me2/3 marks — as direct interactors of succinyl-lysine histone peptides, with dose-dependent inhibition of demethylase catalysis. Supraphysiological sodium succinate treatment of HepG2/C3A hepatoma cells raises histone succinylation alongside H3K9me2/3 and H3K27me2/3, with reduced transcriptional output, and CUT&Tag and ChIP-MS demonstrate co-occupancy of succinylation marks with the repressive methylation at common loci. Defines a biochemical handle on how succinate accumulation (e.g., from SDH loss in paraganglioma/pheochromocytoma and metabolic stress states) alters the epigenome, and surfaces a new endogenous ligand class for JmjC chemical-biology programs. (3) Avery Kent, Chandrima Majumdar and colleagues with Alanna Schepartz and Jamie Cate at UC Berkeley engineer the ribosome exit tunnel to substantially improve consecutive incorporation of structurally challenging non-natural monomers — the kind of constrained or non-canonical backbones required to build orally bioavailable peptide therapeutics and non-canonical biopolymers but on which the wild-type ribosome stalls. Identifies the exit tunnel (not the peptidyl-transferase center) as the tunable bottleneck for genetic-code-reprogramming and ribosomal display, with direct relevance to any group working on reprogrammed translation. (4) Nasim Kajabadi and colleagues with Fabio Rossi at the University of British Columbia, together with Urs Häfeli, Paul Kostenuik, and Robert Young, develop a pathology-targeted prostaglandin-E-receptor-4 (EP4) agonist conjugate that homes to damaged muscle and reverses established fibrosis in a rat model of Duchenne muscular dystrophy. Addresses the adolescent-onset therapeutic gap where existing interventions fail to reverse fibro-fatty replacement and depletion of the regenerative niche, and where systemic EP4 agonism alone could not penetrate or persist at diseased sites. Nominates tissue-homing EP4 agonism as a translational target axis for the broader fibrotic-disease space where structural damage has outpaced regenerative capacity. 2026-06-03-nanopilot-apoe-lnp-succinyl-jmjc-ribosome-tunnel-ep4-dmd Wed, 03 Jun 2026 12:00:00 +0000 386 Four bioRxiv stories: (1) Bufton, Carter and colleagues (Hone Bio / Bristol) — NanoPilot modular antibody-anchor fusion-protein platform competitively blocks the ApoE binding site of LDLR to redirect preformulated lipid nanoparticles away from hepatocytes; added in 10 min with two pipetting steps. Anti-CD3ε NanoPilot: 40-fold T-cell transfection boost + 10-fold monocyte reduction in human PBMCs in vitro and 30–40%% splenic/hepatic T-cell transfection with 3-fold reduced bulk liver accumulation in immunocompetent mice; anti-c-Kit variant enhances HSC-like cell line delivery in co-culture. Generalizable extrahepatic retargeting layer compatible with in-vivo CAR-T and HSC editing programs without reformulating the LNP. (2) Graff, Sidoli, Maianti and colleagues (Albert Einstein) — quantitative proteomics and peptide pull-down identify multiple JmjC histone demethylases (including KDM4D and KDM6B, erasers of H3K9me2/3 and H3K27me2/3) as direct interactors of succinyl-lysine histone peptides with dose-dependent catalytic inhibition; supraphysiological sodium succinate in HepG2/C3A raises histone succinylation, H3K9me2/3, and H3K27me2/3 with reduced transcriptional output, and CUT&Tag/ChIP-MS shows co-occupancy of succinylation with the repressive methylation. First direct mechanistic link between a TCA-cycle metabolite and JmjC active sites with implications for SDH-loss cancer epigenetics and a new endogenous ligand class for JmjC chemical biology. (3) Kent, Majumdar, Schepartz, Cate and colleagues (UC Berkeley) — engineered ribosome exit-tunnel variants substantially improve consecutive incorporation of structurally challenging non-natural monomers; identifies the exit tunnel (not the peptidyl-transferase center) as the tunable bottleneck for genetic-code-reprogramming and ribosomal display of oral-bioavailable peptide therapeutics and non-canonical biopolymers. (4) Kajabadi, Rossi and colleagues with Häfeli, Kostenuik, and Young (UBC) — pathology-targeted prostaglandin-E-receptor-4 agonist conjugate homes to damaged muscle in a rat DMD model and reverses established fibrosis, addressing the adolescent-onset therapeutic gap and nominating tissue-homing EP4 agonism as a translational axis for broader fibrotic-disease settings. Four bioRxiv stories spanning systemic CNS oligonucleotide delivery, HLA-supertype-spanning peptide immunotherapy, synthetic morphogenesis design principles, and a heterologous-expression quality-control finding with algorithmic fix. (1) Guillaume Jacquot, Marion David, and colleagues with Michel Khrestchatisky at Vect-Horus and Aix-Marseille University demonstrate that camelid-derived single-domain antibodies (VHHs) against transferrin receptor 1 conjugated 1:1 to tool siRNAs enable cross-species systemic delivery of nucleic-acid therapeutics into deep brain regions. The conjugates produce rapid intracellular uptake and potent RNAi at low nanomolar concentrations in neural cells, and systemic dosing in wild-type mice, hTfR1 transgenic mice, and non-human primates yields rapid TfR-dependent distributional clearance with efficient functional uptake in deep brain structures, durable 50-80% target knockdown at both mRNA and protein levels, and ED50 below 1 mg/kg siRNA. A fit-for-purpose platform for systemic CNS oligonucleotide therapeutics at clinically relevant doses. (2) Wyatt Blackson and colleagues with Nikolaos Sgourakis at Children's Hospital of Philadelphia, Carl June at Penn, and Possu Huang at Stanford develop xTRACeRs, a generalized framework for antigenic-peptide targeting across HLA-I polymorphism that combines a peptide conformational prediction tool (PepPred) with the TRACeR-I cross-HLA binding-protein engineering system. xTRACeRs validated across clinically relevant peptide antigens presented in five common HLA-A/B/C supertypes; cryo-EM structures of xTRACeR-pHLA complexes for an oncofetal PRAME antigen and a neuroblastoma-specific PHOX2B antigen reveal mechanisms for navigating polymorphic HLA surface residues while making extensive peptide-specific contacts. Two xTRACeRs implemented as CAR-T cells demonstrate potent, specific killing. Overcoming allotype restriction lifts a key barrier in HLA-targeted immunotherapy by expanding patient coverage. (3) Satoshi Toda, Guoye Guan, and colleagues with Wendell Lim at UCSF's Cell Design Institute construct synthetic pattern-forming circuits with juxtacrine synNotch receptor interactions providing local self-activation and diffusible competitor molecules providing long-range inhibition. Reaction-diffusion alone yields heterogeneous cell states but no robust patterns; coupling the circuit to induction of cadherin-driven local adhesion and sorting produces well-defined patterns. Theoretical analysis shows mechanical coupling dramatically expands the parameter window over which patterns form. Implication: evolved morphogenetic systems likely integrate signaling with mechanics, and synthetic morphogenesis programs for tissue engineering and organoid design should plan to do the same. (4) Mohd Ahmad and colleagues with Grzegorz Kudla at MRC Human Genetics Unit, University of Edinburgh, uncover that conventional codon optimization causes pervasive missplicing as a common failure mode of heterologous expression. Across systematically designed libraries of >5000 synthetic reporter genes and natural human cDNAs in human cell lines, most variants are at least partially spliced, with spliced isoforms often dominating to reduce protein output or ablate expression. Sequence-determinant analysis reveals a hierarchical architecture: GC content sets baseline mRNA, local sequence features control splicing, and tissue-specific codon adaptation to tRNA pools fine-tunes translation. Predictive models, benchmarked against existing optimization tools, are paired with a splice-aware optimization algorithm that substantially improves transgene performance — directly relevant to mRNA therapeutics, viral-vector gene therapy, cell engineering, and recombinant protein production. 2026-06-02-tfr1-sdab-sirna-brain-htracer-hla-peptide-lim-morphogenesis-splice-aware Tue, 02 Jun 2026 12:00:00 +0000 420 Four bioRxiv stories: (1) Jacquot, David, and colleagues with Khrestchatisky (Vect-Horus / Aix-Marseille University) — camelid VHHs against transferrin receptor 1 conjugated 1:1 to tool siRNAs enable cross-species systemic delivery of nucleic-acid therapeutics into deep brain regions. Wild-type mice, hTfR1 transgenic mice, and non-human primates show rapid TfR-dependent distributional clearance, efficient functional uptake in deep brain structures, and durable 50-80% mRNA and protein knockdown with ED50 below 1 mg/kg siRNA. Translational platform for systemic CNS oligonucleotide therapeutics at clinically relevant doses. (2) Blackson and colleagues with Sgourakis (CHOP), June (Penn), and Huang (Stanford) — xTRACeRs framework combines PepPred peptide-conformational prediction with the TRACeR-I cross-HLA binding-protein engineering system to develop binders compatible across five common HLA-A/B/C supertypes. Cryo-EM structures of xTRACeR-pHLA complexes for PRAME and PHOX2B antigens reveal how binders navigate polymorphic HLA surfaces while making extensive peptide contacts; two xTRACeRs implemented as CAR-T cells show potent specific killing. Expands patient coverage of HLA-targeted immunotherapy across supertypes. (3) Toda, Guan, and colleagues with Lim (UCSF Cell Design Institute) — synthetic pattern-forming circuit using juxtacrine synNotch local self-activation plus diffusible competitor long-range inhibition produces heterogeneous cell states but no robust patterns; coupling to cadherin-driven cell adhesion and sorting yields well-defined patterns. Theoretical analysis shows mechanical coupling substantially expands the parameter window for pattern formation. Integration of signaling with mechanics as a design principle for synthetic morphogenesis in tissue engineering and organoid design. (4) Ahmad and colleagues with Kudla (MRC Human Genetics Unit, Edinburgh) — systematic libraries of >5000 synthetic reporter genes and natural human cDNAs show conventional codon optimization causes pervasive missplicing as a common failure mode of heterologous expression, with spliced isoforms often dominating to reduce protein output or ablate expression. Hierarchical regulation: GC content sets baseline mRNA, local sequence controls splicing, codon-tRNA adaptation tunes translation. Predictive models plus a splice-aware optimization algorithm substantially improve transgene performance across mRNA therapeutics, viral gene therapy, cell engineering, and recombinant protein production. Four bioRxiv stories spanning a new pharmacological-chaperone modality for tumor suppressors, non-catalytic HDAC6 chemical biology, chemogenetic ion-channel modulation, and machine-learning-guided drug discovery inside biomolecular condensates. (1) Mohamed Fouad, Erica Golemis and colleagues in John Karanicolas's group at Fox Chase Cancer Center hypothesized that a substantial subset of VHL missense mutations driving sporadic clear cell renal cell carcinoma operate by thermodynamically destabilizing the folded protein product rather than disrupting catalytic function, shifting the unfolded-to-folded pVHL ratio and abolishing activity. Computational structure-based virtual screening against the folded native conformation of pVHL plus biophysical binding assays and cellular function readouts converged on CP4.29, which selectively stabilizes the native fold of mutant pVHL and restores wild-type pVHL activity inside cells carrying the mutation. The explicit pharmacological-chaperone framework for tumor suppressors — most prior chaperone programs targeted lysosomal storage enzymes, almost none have rescued a folded-state tumor suppressor in cancer — opens the long tail of destabilizing missense mutations across other tumor suppressors currently considered undruggable. (2) Sasheen Gordon and colleagues in George Burslem's group at the University of Pennsylvania develop small-molecule inhibitors of the HDAC6 zinc-finger ubiquitin-binding domain (ZnF-UBD), targeting the second active surface of HDAC6 that mediates the protein-protein interactions coupling deacetylation to ubiquitin-dependent degradation — functionally separable from deacetylase catalysis but unaddressed by every clinical-stage HDAC6 inhibitor. Starting from a known quinazolinone scaffold, a modeling-guided scaffold hop produces a novel phthalazinone series with improved physicochemical properties, and machine-learning co-folding affinity predictions reproduce the experimental IC50 rank order. Logic: dissect ubiquitin-binding-domain pharmacology from deacetylase pharmacology so the two halves of HDAC6 biology can be drugged independently — relevant across cell-fate and proteostasis programs where blunt catalytic inhibition has been disappointing. (3) Leila Khajoeinejad and colleagues in Paul Riegelhaupt's group at Weill Cornell Medicine engineer chemogenetically activatable two-pore-domain potassium (K2P) channels for direct manipulation of membrane potential, where the existing pharmacological toolkit lacks high-affinity, subtype-selective modulators. Inserting the FRB domain of mTOR into a short cytoplasmic loop of the TREK1 K2P channel produces fusion channels that, upon rapamycin or non-immunomodulatory rapalog binding, recruit endogenous FKBP to form an FRB/rapamycin/FKBP ternary complex within the channel core. Cryo-EM structures show the ternary complex rigidifies the loop between transmembrane segments 2 and 3 into an active channel conformation, and electrophysiology shows nanomolar rapamycin produces clean activation across multiple K2P isoforms. A chemogenetic platform for direct potassium-channel-based control of membrane potential built on an orally bioavailable approved-drug scaffold — research tool plus starting point for cell-type-restricted electrical-state engineering in therapeutics. (4) Aman Khambhawala and colleagues in Jeetain Mittal's group at Texas A&M University deliver the first interpretable machine-learning framework for predicting selective small-molecule partitioning across distinct biomolecular condensates. Using experimental partition coefficients for FDA-approved drugs and metabolites across four condensate systems, they train a graph-based model with multi-task pretraining and per-condensate fine-tuning, plus evidential uncertainty quantification and atom-level attribution. The model outperforms descriptor-based baselines and reveals that partitioning is not governed by a universal chemical rule — the same scaffold is read differently by different condensate microenvironments. Scoring ~1.7M ChEMBL drug-like molecules identifies a chemically diverse space of predicted condensate-selective partitioners with confidence mapping of where new measurements would be most informative. Condensate partitioning becomes a chemically learnable property with a useful prioritization layer over commercial screening libraries. 2026-05-31-vhl-stabilizer-hdac6-znf-rapamycin-k2p-condensate-partitioning Sun, 31 May 2026 12:00:00 +0000 362 Four bioRxiv stories: (1) Fouad/Golemis and colleagues in John Karanicolas's group (Fox Chase Cancer Center) — computational structure-based virtual screening against the folded native conformation of pVHL plus biophysical and cellular validation delivers CP4.29, a small-molecule pharmacological chaperone that selectively stabilizes mutant pVHL and restores wild-type pVHL activity in cells carrying VHL missense mutations driving clear cell renal cell carcinoma. The first explicit pharmacological-chaperone framework for a folded-state tumor suppressor in a cancer context — most prior chaperone programs targeted lysosomal storage enzymes — with implications for the long tail of destabilizing missense mutations across other tumor suppressors currently considered undruggable. (2) Gordon and colleagues in Burslem's group (Penn) — quinazolinone-originated phthalazinone series targeting the HDAC6 zinc-finger ubiquitin-binding domain PPI with ubiquitin, decoupling the second active surface of HDAC6 from the catalytic deacetylase domains every clinical-stage inhibitor addresses; modeling-guided scaffold hop yields improved physicochemical properties, and ML co-folding affinity predictions reproduce experimental IC50 rank order. Dissects ubiquitin-binding-domain pharmacology from deacetylase pharmacology so the two halves of HDAC6 biology can be drugged independently. (3) Khajoeinejad and colleagues in Riegelhaupt's group (Weill Cornell Medicine) — inserting the FRB domain of mTOR into a short cytoplasmic loop of the TREK1 K2P channel produces fusion channels that recruit endogenous FKBP via rapamycin or non-immunomodulatory rapalogs to form an FRB/rapamycin/FKBP ternary complex inside the channel core; cryo-EM shows the complex rigidifies the TM2/TM3 loop into an active conformation, and electrophysiology shows nanomolar activation across multiple K2P isoforms. Chemogenetic platform for direct potassium-channel-based control of membrane potential on an approved-drug scaffold. (4) Khambhawala and colleagues in Mittal's group (Texas A&M) — interpretable graph-based ML framework with multi-task pretraining, per-condensate fine-tuning, evidential uncertainty quantification, and atom-level attribution predicts FDA-approved-drug partition coefficients across four biomolecular condensates, outperforming descriptor baselines and revealing that identical scaffolds are read differently by different condensate microenvironments; scores 1.7M ChEMBL molecules for condensate-selective partitioners with confidence mapping for wet-lab prioritization. Condensate partitioning becomes a chemically learnable property as a prioritization layer over commercial screening libraries. Four bioRxiv stories spanning generative protein design as a new intracellular modality, alternative-splicing-driven target biology, GPCR structural pharmacology, and AI-driven multi-property protein design. (1) Federica Vacca, Daniel Marston, and colleagues in Mihai Azoitei's group at Duke developed a generative protein design platform for engineering selective intracellular inhibitors of Dbl-family guanine nucleotide exchange factors — the main activators of RhoA-family GTPases whose subcellular kinetics have been essentially impossible to dissect with small molecules. Engineered miniproteins showed high affinity and remarkable specificity across the family with clean discrimination between closely related members, modulated GEF activity in vitro and in living cells, and were coupled to a light-activated module for optogenetic control. Generative protein design as a target-class-spanning intracellular modulator platform for signaling targets small molecules have not addressed. (2) Ng Bin Yi Lewis with Polly Chen and Leilei Chen at the Cancer Science Institute of Singapore reveal an unexpected ADAR1 isoform repertoire driven by intron-13 retention via an evolutionarily conserved weaker 5' splice site, generating previously uncharacterized truncated p90 and p130 isoforms lacking the C-terminal deaminase-domain portion. Retention is autoregulated by ADAR1 antagonizing hnRNPA1 binding in an editing-independent manner. p90 is predominantly cytoplasmic and represses PKR/eIF2alpha activation by sequestering dsRNA — canonical immune-suppressive ADAR1 function without any deaminase activity. p90 is elevated in most colorectal tumors versus matched normal, correlates with hnRNPA1, and drives accelerated xenograft growth versus p110. Implies ADAR1 deaminase-targeting and dsRNA-binding inhibitors will have non-equivalent therapeutic profiles depending on isoform dominance. (3) Daniel Isom and colleagues at the University of Miami Miller School use the DCyFIR humanized yeast platform — which isolates extracellular pH from cytoplasmic confounders — to show mu, delta, and kappa opioid receptors are proton-gated coincidence detectors with agonist efficacy falling steeply as pH drops into ranges seen in inflamed and ischemic tissue. pHinder calculations and variant profiling pinpoint a single conserved aspartate, D149 (Ballesteros-Weinstein 3.32), whose protonation breaks the salt bridge anchoring orthosteric opioid ligands. The pH sensor is preserved across every mu-OR alternative splice variant retaining a continuous seven-transmembrane fold (more than twenty variants profiled). Structural anchor for inflammation-associated opioid analgesic failure and a defined target for pH-biased agonist medicinal chemistry. (4) Lijun Liu and colleagues with Jianquan Zhao at the Institute of Computing Technology, Chinese Academy of Sciences introduce OCDesign, an objective-curriculum-guided framework for multi-property protein design that introduces objectives sequentially in a defined curriculum rather than jointly optimizing all properties one-shot. Using antibody-binding protein A as a test system, one-shot Pareto-style optimization yielded zero functional designs while staged optimization — solubility and structural consistency first, then binding affinity, then alkaline resistance — produced multi-property functional proteins with substantially fewer wet-lab experiments. Objective ordering is a first-class hyperparameter in high-dimensional protein design and a determinant of whether the design problem has feasible solutions at all. 2026-05-30-de-novo-gef-inhibitors-adar1-p90-isoform-proton-opioid-receptor-ocdesign-curriculum Sat, 30 May 2026 12:00:00 +0000 359 Four bioRxiv stories: (1) Vacca/Marston and colleagues in Mihai Azoitei's group (Duke) — generative protein design platform engineers high-affinity, family-selective miniprotein inhibitors of Dbl-family RhoA GEFs that block catalytic activity in vitro and in cells; one inhibitor coupled to a light-activated module enables optogenetic on-demand control of intracellular signaling. Target-class-spanning intracellular modulator platform for signaling targets small molecules have not addressed. (2) Ng with Polly Chen and Leilei Chen (Cancer Science Institute of Singapore) — ADAR1 intron-13 retention via an evolutionarily conserved weaker 5' splice site generates previously uncharacterized truncated p90 and p130 isoforms lacking the C-terminal deaminase domain; retention is autoregulated through ADAR1-antagonized hnRNPA1 binding (editing-independent). p90 is predominantly cytoplasmic, represses PKR/eIF2alpha by dsRNA sequestration (canonical immune-suppressive ADAR1 function without any deaminase activity), is elevated in most colorectal tumors versus matched normal, and drives accelerated xenograft growth versus p110. ADAR1 deaminase-targeting and dsRNA-binding inhibitor programs will have non-equivalent therapeutic profiles depending on isoform dominance. (3) Isom lab (Miami Miller) — DCyFIR humanized yeast platform shows mu/delta/kappa opioid receptors are proton-gated coincidence detectors with agonist efficacy falling steeply as pH drops into inflamed/ischemic ranges; pHinder calculations and variant profiling identify a single conserved aspartate D149 (Ballesteros-Weinstein 3.32) whose protonation breaks the orthosteric salt bridge. pH sensor preserved across more than twenty mu-OR splice variants retaining a continuous 7-TM fold. Structural anchor for inflammation-associated opioid analgesic failure and a defined target for pH-biased agonist medicinal chemistry. (4) Liu/Zhao (Institute of Computing Technology, CAS) — OCDesign curriculum-guided multi-property protein design framework: introducing objectives sequentially (solubility/structure → binding affinity → alkaline resistance) on protein A yields multi-property functional designs with substantially fewer wet-lab experiments where one-shot Pareto-style optimization yields zero. Objective ordering is a first-class hyperparameter in high-dimensional protein design. Four bioRxiv stories spanning a landmark cell-type-resolved proteomic resource, two pieces of biocatalytic chemistry directly applicable to medicinal chemistry, and a delivery innovation for an entire approved drug class. (1) Christoph Weiss, Erika Sjoestedt and colleagues in Matthias Mann's group at the Max Planck Institute of Biochemistry, with the Human Protein Atlas team, used Deep Visual Proteomics on tissue from a healthy female donor to build a cell-type-resolved proteomic atlas covering 27 cell populations across 14 tissues, quantifying roughly two-thirds of all human protein-coding genes with up to 8,500 proteins per population. The proteome partitions bimodally into a universal housekeeping core and highly specialized cell-type programs. Integration with the Human Protein Atlas single-cell RNA resource shows messenger-RNA-to-protein concordance is governed by pathway rather than cellular identity. As a working example, cancer-testis antigens are surfaced in oocytes that are invisible to bulk profiling — the kind of finding that immediately seeds immuno-oncology target nomination. The atlas is openly accessible and is going to become a default reference for cell-type-resolved target discovery. (2) Mingfeng Liu, Masao Ohashi, Wenchao Han and Qixin Zhou with Ken Houk and Yi Tang at UCLA characterized two fungal cytochrome P450 monooxygenases — PpnB and NseB from the penipratynolene and sinuxylamide biosynthetic pathways — that catalyze oxidative C(sp2)-demethylation of O-prenyl-L-tyrosine to yield O-homoallenyl-L-tyrosine and O-but-2-ynyl-L-tyrosine respectively, installing allene and internal alkyne functional groups from a common five-carbon prenyl precursor. DFT calculations, isotopic labeling, heterologous expression, and biotransformation establish a mechanism involving selective C-C bond cleavage followed by a product-determining hydrogen atom abstraction step. A third pathway P450, PpnD, acts as an oxidative isomerase converting the terminal allene into a terminal alkyne. New catalytic repertoire for cytochrome P450 enzymes with direct relevance to medicinal chemistry (alkyne pharmacophores) and bioorthogonal conjugation (click chemistry). (3) Kim Anh Vu, Hanlin Zhang and colleagues with Jonathan Watts, Erik Sontheimer, Scot Wolfe, and the Khvorova lab at UMass Chan used iterative structure-guided position-wise optimization to systematically introduce chemical modifications at each position of SpyCas9 gRNAs — going beyond conventional 2'-fluoro/2'-O-methyl repertoire to include 2'-amino-RNA, 4'-thio-RNA, and extended nucleic acid (exNA) — yielding guides where 90-100%% of nucleotides are sugar- or backbone-modified that consistently maintain or enhance editing efficacy in cells and in vivo. Validated across nuclease editing, base editing, and other Cas9-based tools. Resolves the long-standing trade-off between stability/delivery flexibility and editing activity for therapeutic CRISPR. (4) Gilles Jacquot, Camille Hervy and colleagues in Alexandre Detappe's group at Institut Strauss / Gustave Roussy / Strasbourg (with Mark Tibbitt at ETH and Sarah Cianférani) developed TACT, an injectable diacetyl-L-tartaric-anhydride-functionalized chitosan hydrogel that confines ADCs within a protective subcutaneous depot. Compatible with clinically approved ADC formulations without modification; DAR-dependent release kinetics. In direct head-to-head with recombinant hyaluronidase, TACT prevents the vesicant-driven tissue necrosis hyaluronidase causes while preserving IV-equivalent antitumor efficacy. In non-human primates, SC TACT achieves 78%% relative bioavailability for total trastuzumab, cuts peak free-DM1 catabolite exposure 7.6-fold versus IV, and produces only transient cutaneous reactions. Validated across T-DM1, T-DXd, a custom cleavable MMAE series, and enfortumab vedotin (Nectin-4 MMAE). Depot-mediated confinement as a route-of-administration unlock for an entire drug class. 2026-05-29-cell-proteomic-atlas-allene-alkyne-p450-modified-grna-tact-adc-hydrogel Fri, 29 May 2026 12:00:00 +0000 366 Four bioRxiv stories: (1) Weiss/Sjoestedt and colleagues in Matthias Mann's group (Max Planck Biochemistry) with the Human Protein Atlas — Deep Visual Proteomics atlas across 27 cell populations and 14 tissues quantifies ~2/3 of all human protein-coding genes (up to 8,500 per population); proteome partitions bimodally into universal core and cell-type-specialized programs; RNA-protein concordance is pathway-driven not identity-driven; cancer-testis antigens surfaced in oocytes invisible to bulk profiling. Open-access foundational resource for cell-type-resolved target discovery. (2) Liu/Ohashi/Han/Zhou with Houk and Tang (UCLA) — two fungal cytochrome P450s PpnB and NseB catalyze oxidative C(sp2)-demethylation of O-prenyl-L-tyrosine to install allenes and internal alkynes via selective C-C cleavage + product-determining H atom abstraction; DFT + isotope labeling + heterologous expression nail the mechanism; a third pathway P450 PpnD isomerizes terminal allene to terminal alkyne. New enzymatic editing strategy with direct relevance to alkyne medicinal-chemistry pharmacophores and bioorthogonal click-chemistry conjugation. (3) Vu/Zhang and colleagues with Watts/Sontheimer/Wolfe/Khvorova (UMass Chan) — iterative structure-guided position-wise SpyCas9 gRNA modification beyond conventional 2'-fluoro/2'-O-methyl (adding 2'-amino-RNA, 4'-thio-RNA, exNA) yields 90-100%% sugar/backbone-modified guides that consistently maintain or enhance editing in cells and in vivo across nuclease, base-editing, and other Cas9 tools. Resolves the stability/delivery vs activity trade-off for therapeutic CRISPR. (4) Jacquot/Hervy and colleagues in Detappe's group (Strasbourg/Gustave Roussy) with Tibbitt (ETH) and Cianférani — TACT, a diacetyl-tartaric-anhydride-functionalized chitosan hydrogel, confines ADCs in a subcutaneous depot, compatible with approved formulations; vs hyaluronidase prevents vesicant-driven tissue necrosis while preserving IV-equivalent antitumor efficacy; in NHP delivers 78%% relative bioavailability and 7.6-fold lower peak free-DM1 exposure. Validated across T-DM1, T-DXd, custom MMAE series, and enfortumab vedotin — route-of-administration unlock for an entire drug class. Four bioRxiv stories spanning a long-undruggable small-GTPase chemical biology target, a primate-validated next-generation gene therapy vector, a tactical degrader-design discovery, and a macrocyclization-driven dark-kinome probe. (1) Elena De Vita and colleagues in Ed Tate's group at Queen Mary University of London report the first cell-active, rationally designed covalent inhibitors of the RAB27A/B small-GTPase-effector protein-protein interaction — historically intractable due to the lack of well-defined pockets outside the conserved GTP binding site and the large RAB27-effector PPI surface. An electrophile-first biochemical screen identified a non-conserved cysteine (Cys123) flanking the effector interface, and X-ray crystallography-guided design produced a novel class of acrylamide covalent inhibitors that enantioselectively target RAB27A/B-Cys123 in cells. Potency and selectivity were confirmed by biochemical, cellular, chemical proteomics, and melanocyte phenotype recapitulation assays alongside a matched inactive enantioprobe control. Critically, the team shows that the legacy literature compound Nexinhib-20 is non-selectively reactive and toxic, with implications for re-reading prior RAB27 biology. The paper delivers the first toolbox of cell-active chemical probes for a long-studied vesicle-trafficking, metastasis, and inflammation target. (2) Youguang Luo and colleagues at Qilu Pharmaceutical engineered muscle-tropic AAV capsids using a rational-design-plus-directed-evolution platform they call REACH, then benchmarked the lead candidate M1 head-to-head against AAV9 and MyoAAV in non-human primates. Systemic dosing showed M1 delivers >10-fold higher skeletal muscle transduction than AAV9 and 2-3-fold higher than MyoAAV, with a 183-fold reduction in liver distribution and significant de-targeting from dorsal root ganglia (11-fold), lung (27-fold), spleen (2-fold), and kidney (2-fold) — addressing two of the field's hard ceilings (liver and DRG toxicity) for muscular dystrophy gene therapy at once. (3) Niclas Olsson and colleagues in Kirill Bersuker's group at Calico Life Sciences used global ubiquitination profiling to map the lysines on EGFR and HER2 kinase domains that get ubiquitinated in cells treated with heterobifunctional degraders. VHL- and Cereblon-recruiting degraders converge on a common subset of accessible lysines; mutagenesis shows the number of available ubiquitination sites quantitatively specifies the extent of RTK degradation, and introduction of non-native ubiquitination sites mostly fails to rescue clearance of lysine-deficient RTK mutants — productive ubiquitination is geometrically constrained, not just a function of ternary complex formation. A concrete design rule for target prioritization in heterobifunctional degrader programs. (4) Antonia Zerva, Thomas Hanke, and Stefan Knapp at Goethe University Frankfurt addressed the long-standing lack of high-quality chemical tools for HIPK4, an understudied dark kinome member with genetic links to spermiogenesis and cutaneous squamous cell carcinoma. Starting from the broadly promiscuous kinase inhibitor bosutinib, systematic macrocyclization of the quinoline scaffold yielded AZ137 — an 11 nM biochemical IC50, 76 nM cellular EC50 HIPK4 inhibitor with exceptional selectivity across three orthogonal kinase panels, high solubility, and no detectable cytotoxicity. Cellular activity was confirmed in HIPK4-dependent F-actin remodeling assays. Paired with a matched negative control, the probe set provides the foundational framework for pharmacological validation of HIPK4 as a therapeutic target. 2026-05-28-rab27-covalent-muscle-aav-protac-ubiquitin-sites-hipk4-macrocycle Thu, 28 May 2026 12:00:00 +0000 310 Four bioRxiv stories: (1) De Vita/Tate lab (Queen Mary) — first cell-active, rationally designed covalent inhibitors of the RAB27A/B-effector protein-protein interaction. Electrophile-first screen plus crystallography-guided optimization yields a stereoselective acrylamide series targeting a non-conserved Cys123 flanking the PPI interface; potency and selectivity confirmed by chemical proteomics and melanocyte phenotype recapitulation against a matched inactive enantioprobe control. Legacy compound Nexinhib-20 reclassified as non-selectively reactive and toxic. First high-quality chemical toolbox for a long-undruggable vesicle-trafficking, metastasis, and inflammation target. (2) Luo et al. (Qilu Pharmaceutical) — REACH-engineered muscle-tropic AAV capsid M1 delivers >10-fold higher skeletal muscle transduction than AAV9, 2-3-fold over MyoAAV, with 183-fold lower liver distribution and 11/27/2/2-fold de-targeting from DRG/lung/spleen/kidney in non-human primates. Primate-validated capsid addressing liver and DRG toxicity ceilings simultaneously for muscular dystrophy gene therapy. (3) Olsson/Bersuker lab (Calico Life Sciences) — global ubiquitination profiling shows VHL- and Cereblon-recruiting degraders converge on a common subset of accessible lysines on EGFR/HER2 kinase domains; the number of available sites quantitatively specifies degradation extent, and introduction of non-native lysines mostly fails to rescue clearance of lysine-deficient mutants. Productive ubiquitination is geometrically constrained — a concrete design rule for heterobifunctional degrader target prioritization. (4) Zerva/Hanke/Knapp (Goethe University Frankfurt) — macrocyclization of bosutinib's quinoline scaffold yields AZ137, an 11 nM biochemical IC50, 76 nM cellular EC50 HIPK4 inhibitor with exceptional selectivity across three orthogonal kinase panels, high solubility, and no cytotoxicity; cellular activity confirmed in HIPK4-dependent F-actin remodeling. First high-quality chemical probe set for the understudied HIPK4 dark kinome member, paired with a matched negative control for target validation. Four bioRxiv stories spanning a new non-viral delivery modality, a metabolic synthetic-lethal pairing for leukemia, a new mode of cGAS activation, and a route-of-administration unlock for ADCs. (1) Brian Lash, Daniel Strebinger, Michael Segel and colleagues in Feng Zhang's group at the Broad Institute reconstitute the Drosophila melanogaster Arc1 (dArc1) retroelement-derived capsid entirely from purified recombinant protein components and in vitro transcribed RNA, creating a fully defined, cell-free assembled protein nanoparticle system as a third lane for in vivo macromolecular delivery alongside lipid nanoparticles and AAV. Engineering the dArc1 RNA-binding domain enables efficient encapsulation of mRNA payloads and Cas9 ribonucleoproteins. The capsids enter mammalian cells through a direct interaction with the surface receptor SORCS2, which is upregulated in regenerating and dystrophic tissue — providing biologically built-in tropism for muscle disease. In the mdx mouse model of Duchenne muscular dystrophy, a single intramuscular injection of dArc1 capsids carrying Cas9 gene editors produced up to 18%% exon skipping and restored dystrophin expression. Implications: avoids pre-existing anti-AAV immunity, is in principle redoseable, and reframes manufacturing as a recombinant-protein-purification + RNA-assembly problem rather than viral-titer biology. (2) Mathew Austin, Shahidul Patel, and colleagues in Bela Patel's group at Barts Cancer Institute identify the Philadelphia chromosome-positive subgroup of adult B-ALL as a stereotypical arginine auxotroph driven by low expression of argininosuccinate synthase (ASS1-low) — analyzing >550 adult B-ALL cases by transcriptomics. Pegargiminase, a pegylated arginine deiminase already in clinical development for other tumors, induced robust apoptosis in Ph+ ALL cell lines and primary samples in an ASS1-dependent manner, was effective as monotherapy in independent Ph+ ALL PDX, and mechanistically acts through endoplasmic reticulum stress orthogonal to tyrosine kinase inhibition. In non-genetic TKI-resistance models — the leading clinical problem in this subtype — pegargiminase + TKI combinations robustly eradicated TKI-resistant leukemia where TKI alone failed, establishing a metabolic synthetic-lethal pairing for a chemotherapy-free regimen in a subset that desperately needs one. (3) Shen Yang, Shaobin Wu, and colleagues in Jenny Ting's group at UNC Chapel Hill with Jeff Sohn (Hopkins) and Jack Griffith (UNC) show that DNA with unpaired bubble regions (Bu-DNA) — generated during transcription, recombination, replication, and stressed mitochondrial DNA — causes cGAS hyperactivation. Cryo-EM and single-molecule FRET reveal that two cGAS molecules clamp onto a single Bu-DNA by bending the duplex into a V-shape with the unpaired region as a hinge, forming a 2-cGAS-to-1-DNA complex that binds the catalytic domain much more tightly than fully paired DNA but suppresses condensation — limiting oligomeric state while hyperactivating catalysis. Hyperactivation is observed when Bu-DNA is embedded in linear, circular, plasmid, and mitochondrial DNA substrates. Pattern diversity within the ligand reshapes the receptor's oligomeric state and tunes output, opening a structural mode of selectivity for cGAS-STING modulators. (4) Gilles Jacquot, Camille Hervy, and colleagues in Alexandre Detappe's group at Institut Strauss / Gustave Roussy / Strasbourg (with Mark Tibbitt at ETH and Sarah Cianférani) developed TACT, an injectable diacetyl-L-tartaric-anhydride-functionalized chitosan hydrogel that confines ADCs within a protective subcutaneous depot, addressing the vesicant-tissue-necrosis problem that has kept every approved ADC IV-only. TACT is compatible with clinically approved ADC formulations without drug-product modification and provides drug-to-antibody-ratio-dependent release kinetics with a quantitative relationship to in vivo absorption timing. In direct comparison, recombinant human hyaluronidase (rHuPH20) co-formulated with vesicant ADCs caused severe tissue necrosis whereas TACT prevented macroscopic injury while preserving antitumor efficacy comparable to IV; TUNEL staining showed TACT attenuated peri-depot apoptotic injury 3-fold relative to T-DM1 alone and 2-fold relative to rHuPH20. In non-human primates, SC TACT achieved 78%% relative bioavailability for total trastuzumab, reduced peak circulating free DM1 catabolite exposure 7.6-fold versus IV, and produced only transient, self-resolving cutaneous reactions. Validated across T-DM1 and T-DXd (non-cleavable MCC + cleavable peptide linkers), a custom cleavable MMAE series, and enfortumab vedotin (Nectin-4 MMAE) — a route-of-administration unlock for an entire drug class. 2026-05-26-darc1-capsid-ass1-arginine-ph-all-cgas-bubble-dna-tact-hydrogel-adc Tue, 26 May 2026 12:00:00 +0000 273 Four bioRxiv stories: (1) Lash/Strebinger/Segel and colleagues in Feng Zhang's group (Broad) — in vitro reconstitution of the Drosophila Arc1 retroelement capsid from purified recombinant protein and IVT RNA gives a fully defined, cell-free assembled protein nanoparticle for mRNA and Cas9 RNP delivery; RNA-binding-domain engineering enables efficient cargo loading; mammalian cell entry via the SORCS2 surface receptor (upregulated in regenerating and dystrophic tissue) gives built-in tropism; intramuscular injection in mdx mice delivers Cas9 to muscle, producing up to 18%% exon skipping and dystrophin restoration. A genuinely new third lane for in vivo macromolecular delivery alongside LNP and AAV — pre-existing-immunity-free, redoseable, and manufacturable as recombinant protein + RNA assembly. (2) Austin/Patel and colleagues in Bela Patel's group (Barts Cancer Institute) — Philadelphia chromosome-positive adult B-ALL is a stereotypical ASS1-low arginine auxotroph across >550 cases; pegargiminase (clinical-stage pegylated arginine deiminase) induces ER-stress-mediated apoptosis orthogonal to TKI in PDX, and pegargiminase + TKI combinations robustly eradicate non-genetic-TKI-resistant leukemia where TKI alone fails — a metabolic synthetic-lethal pairing for a chemotherapy-free regimen in adult Ph+ ALL. (3) Yang/Wu and colleagues in Jenny Ting's group (UNC Chapel Hill) with Sohn (Hopkins) and Griffith (UNC) — DNA with unpaired bubble regions (Bu-DNA, generated during transcription/recombination/replication and in stressed mitochondrial DNA) hyperactivates cGAS via a previously unrecognized 2-cGAS-to-1-DNA complex that bends Bu-DNA into a V-shape using the unpaired region as a hinge; binds the catalytic domain much more tightly than paired DNA but suppresses condensation. Pattern diversity within the ligand reshapes oligomeric state and output — a structural mode of selectivity for cGAS-STING modulators. (4) Jacquot/Hervy and colleagues in Alexandre Detappe's group (Strasbourg / Gustave Roussy / Institut Strauss) with Tibbitt (ETH) and Cianférani — TACT, a diacetyl-tartaric-anhydride-functionalized chitosan hydrogel, confines ADCs within a subcutaneous depot and replaces excipient-mediated dispersion. Compatible with approved ADC formulations without modification; vs hyaluronidase co-formulation it prevents the vesicant-driven tissue necrosis hyaluronidase causes while preserving IV-equivalent antitumor efficacy; in NHP, 78%% relative bioavailability for total trastuzumab, 7.6-fold lower peak free-payload exposure; validated across T-DM1, T-DXd, and enfortumab vedotin. Route-of-administration unlock for an entire drug class. Four bioRxiv stories spanning a new cell-therapy architecture, a senolytic strategy for residual disease, a chemical-biology rewriting of nucleoside antimetabolite biosynthesis, and a non-toxic alternative to calcineurin catalytic inhibition. (1) Tushar Nichakawade with the Kinzler, Papadopoulos and Pardoll groups at Johns Hopkins introduce MANTIS — Multi-ANtigen Triggered Immune Synapse — a fundamentally new AND-gated CAR architecture that uses the physical dimensions of an extracellular blocking domain, rather than chimeric intracellular signaling logic, to spatially couple two antigen binding events into a single T cell activation. The CAR carries a bulky extracellular domain that prevents engagement of the second antigen at the immune synapse; first-antigen engagement sheds this domain, unveiling a free CAR of the right size to engage a second antigen via kinetic segregation and trigger activation only when both antigens co-occur. The mechanism turns the well-known spatial constraints of kinetic-segregation T cell triggering into a programmable input gate, expanding the set of safely targetable antigen pairs and giving cell engineers a tunable physical knob — receptor dimensions — alongside the chemical and signaling knobs already in use. (2) Zhe Wang and Yulin Li at Houston Methodist with Stanford and Baylor collaborators identify lysosomal ferrous iron accumulation as a conserved hallmark and an actionable vulnerability of therapy-induced senescence — the residual-disease state that fuels recurrence after chemotherapy or radiation. Across diverse tumor models the senescent tumor cell population is markedly hypersensitive to ferroptosis induction; mechanistically, TFEB activates heme oxygenase one, ferrous iron accumulates in the lysosome, and the cell becomes primed for iron-dependent lipid peroxidation death. In breast cancer PDX models, sequential ferroptosis induction after chemotherapy delays recurrence, and dual inhibition of GPX4 and FSP1 — the parallel ferroptosis defense arms — produces often-complete eradication of residual tumors with no overt toxicity, giving a structurally rational senolytic dual-target drug-development thesis. (3) Chen-Fan Lee, Tianyu Zhou and Michael Freeman at the University of Minnesota with Wilfred van der Donk solve the long-mysterious biosynthesis of sinefungin, a SAM-mimetic nucleoside antimetabolite. The defining C-C bond is forged not by a PLP enzyme (the standard hypothesis for thirty years) but by a vitamin B12-dependent radical SAM enzyme that homolytically substitutes adenosylcobalamin to transfer the adenosyl group whole, with cobalamin catalytically consumed — a rare instance. A cryptic phosphorylation/dephosphorylation gates substrate flow, and two peptide aminoacyl-tRNA ligases (PEARLs) — previously known as peptide-modifying enzymes — install alanines onto the nucleoside scaffold, with one of them iteratively elongating; pathway reconstituted as a defined enzymatic cascade, expanding the biocatalytic toolkit for amino-acid-nucleoside conjugates. (4) Adi Cohen, Maya Gabay and Maayan Gal at Tel Aviv University report cell-penetrating peptide conjugates of a C16orf74-derived peptide that selectively block the calcineurin-NFAT protein-protein interaction in human T cells without inhibiting calcineurin's broader catalytic activity. Cyclosporine A and tacrolimus block calcineurin's catalytic site directly, accounting for the bulk of their toxicity. TAT- and polyarginine-conjugated C16orf74 enters cells, inhibits NFAT nuclear translocation, suppresses IL-2 transcription at low-micromolar concentrations, and shows no cytotoxicity up to 50 µM — proof-of-concept that PPI-selective interface inhibition can substitute for catalytic-site inhibition in clinical immunosuppression. 2026-05-25-mantis-logic-car-tis-ferroptosis-senolysis-sinefungin-radical-sam-cn-nfat-peptide Mon, 25 May 2026 12:00:00 +0000 446 Four bioRxiv stories: (1) Nichakawade/Kinzler/Papadopoulos/Pardoll groups (Johns Hopkins) — MANTIS, an AND-gated CAR architecture using the physical dimensions of an extracellular blocking domain to spatially regulate CAR activation: first-antigen engagement sheds the bulky blocker, licensing a free CAR to engage a second antigen via kinetic-segregation triggering and activate the T cell only when both antigens co-occur. Programmable physical input gate that expands the safely targetable antigen-pair space in CAR T cell therapy. (2) Wang/Li (Houston Methodist with Stanford, Baylor) — therapy-induced senescent tumor cells accumulate lysosomal ferrous iron via a TFEB-HO1 axis, become hypersensitive to ferroptosis, and are durably eradicated by sequential dual inhibition of GPX4 and FSP1 after chemotherapy in breast cancer PDX models with no overt toxicity — mechanism-specific senolytic strategy for residual disease. (3) Lee/Zhou/Freeman with van der Donk (Minnesota) — sinefungin biosynthesis solved: a vitamin B12-dependent radical SAM enzyme installs the defining C-C bond via homolytic substitution that catalytically consumes adenosylcobalamin, a cryptic phosphorylation/dephosphorylation gates substrate flow, and two PEARL peptide aminoacyl-tRNA ligases extend the nucleoside scaffold with alanines (one iteratively). Pathway reconstituted as a defined enzymatic cascade — biocatalytic toolkit for amino-acid-nucleoside conjugates expanded. (4) Cohen/Gabay/Gal (Tel Aviv) — TAT- and polyarginine-conjugated C16orf74 peptides cross the cell membrane, block the calcineurin-NFAT PPI without inhibiting calcineurin catalysis, suppress NFAT nuclear translocation and IL-2 transcription in human T cells at low-µM doses, and show no cytotoxicity to 50 µM — proof-of-concept for PPI-selective alternatives to cyclosporine and tacrolimus in clinical immunosuppression. Four bioRxiv stories spanning a new RAS modality, a new protease chemical biology platform, a mechanically programmed cell therapy, and a structural redefinition of mitochondrial copper biology. (1) Matthew Mitcheltree and colleagues at Merck Sharp and Dohme report that mirror-image mRNA display — screening L-peptide libraries against the chemically synthesized D-protein target then resynthesizing hits as their mirror image to give proteolytically stable all-D ligands of the native target — delivered nanomolar all-D macrocyclic peptide binders that engage a previously unexploited cryptic back pocket on KRAS distinct from the switch-two pocket addressed by every approved and clinical RAS therapy. The ligands bind KRAS in both GDP- and GTP-bound states with equal affinity, exploit a single-residue difference among RAS isoforms to bind KRAS selectively over HRAS and NRAS, and inhibit oncogenic signaling in KRAS-mutant cells through a mechanism not dependent on switch-two engagement — a structurally orthogonal partner for combination with sotorasib/adagrasib/pan-RAS macrocycles and a fallback after switch-two resistance, with the broader implication that mirror-image display extends proteolytically stable peptide pharmacology to cytosolic targets currently inaccessible. (2) Maria Barrueco and Maximilian Lakemeyer at the University of Bath introduce covalent macrocyclic activity-based probes (cmABPs) that selectively label non-catalytic residues in serine protease active sites, generalizing protease chemoproteomics beyond the conserved catalytic serine that dominates current ABPs. Phage display of macrocyclic peptides combined with systematic SuFEx (sulfur-six-fluoride) electrophile scanning positions the warhead to engage alternative nucleophiles — active-site lysines and tyrosines — and applied to plasma kallikrein and urokinase plasminogen activator yielded probes with rapid and complete covalent labeling, with small changes in electrophile chemistry producing tuneable kinetics; selectivity across closely related serine proteases — historically the major obstacle for catalytic-serine ABPs because that residue is identical across family members — becomes accessible when the warhead engages a residue that varies. (3) Yu-Chia Chen and Kyle Vining at Penn show that the stress-relaxation rate of the matrix in which dendritic cells are cultured durably programs their functional state even after the mechanical cue is removed. Fast-relaxing matrices license cells with enhanced antigen presentation, faster migration, and higher T cell-recruiting chemokine expression; slow-relaxing matrices bias toward pro-inflammatory cytokine secretion. The programming is conserved across human and murine cells, transduced through PI3K signaling and branched-actin polymerization downstream of integrated IL-4 and GM-CSF signaling, and adoptive transfer of fast-relaxing-matrix-licensed DCs generated higher draining-lymph-node CD8 T cell responses than DCs cultured on conventional plastic — a manufacturing-stage lever for DC therapy separate from antigen-loading and maturation-cocktail levers. (4) Sourik Ghosh and Will Wei at Penn show that mitochondrial copper, delivered by the inner-membrane transporter SLC25A3, is required to maintain the stability of the lipoylated E2 subunits of mitochondrial dehydrogenases and the lipoylation enzymes LIPT1 and LIPT2. Copper directly engages the reduced lipoyl moiety as shown by chemical probes and synthetic peptides — acting as a structural cofactor of the lipoylation machinery, not through electron transport chain function — and copper depletion destabilizes the machinery, drops PDH and OGDH activity, rewires the TCA cycle, depletes aspartate, suppresses mTORC1, and limits proliferation; selective copper redelivery to mitochondria restores everything. Provides the mechanistic backbone the cuproptosis literature has been missing and a more refined target hypothesis for copper-ionophore and copper-chelator programs in oncology. 2026-05-24-kras-d-peptide-back-pocket-sufex-protease-dc-mechanics-mito-copper Sun, 24 May 2026 12:00:00 +0000 497 Four bioRxiv stories: (1) Mitcheltree/Boo/Boyer and colleagues (Merck Sharp and Dohme) — mirror-image mRNA display delivers all-D macrocyclic peptide ligands binding a cryptic back pocket on KRAS distinct from switch-two, with GDP/GTP state-agnostic nanomolar affinity, single-residue isoform selectivity over HRAS/NRAS, and inhibition of oncogenic signaling in KRAS-mutant cells via a mechanism not requiring switch-two engagement — structurally orthogonal partner for sotorasib/adagrasib/pan-RAS macrocycles and a fallback after switch-two resistance; mirror-image display extends proteolytically stable peptide pharmacology to inaccessible cytosolic targets. (2) Barrueco/Lakemeyer (Bath) covalent macrocyclic activity-based probes combine phage-display macrocycles with SuFEx (sulfur-six-fluoride) electrophile scanning to selectively label non-catalytic active-site lysines and tyrosines on plasma kallikrein and urokinase plasminogen activator — small electrophile changes produce tuneable covalent kinetics, and selectivity across closely related serine proteases becomes accessible because the warhead engages a residue that varies among family members. (3) Chen/Vining (Penn) matrix stress-relaxation rate durably programs DC functional state — fast-relaxing matrices license migratory antigen-presenting cells with elevated T cell-recruiting chemokines, slow-relaxing matrices bias toward pro-inflammatory cytokine secretion; mechanism through PI3K signaling and branched-actin polymerization downstream of IL-4 + GM-CSF, conserved across human and murine DCs, and adoptive transfer of fast-relaxing-matrix-licensed DCs gives higher antigen-specific CD8 T cell responses in vivo — manufacturing-stage lever for DC therapy. (4) Ghosh/Wei (Penn) mitochondrial copper delivered by SLC25A3 is a direct structural cofactor of lipoylated TCA cycle dehydrogenases and LIPT1/LIPT2 lipoylation enzymes via direct engagement of the reduced lipoyl moiety; copper depletion destabilizes the machinery, drops PDH/OGDH activity, rewires the TCA cycle, depletes aspartate, suppresses mTORC1, and limits proliferation independent of ETC inhibition — mechanistic backbone the cuproptosis literature was missing and a refined target hypothesis for copper ionophore/chelator oncology programs. Four bioRxiv stories spanning new oncology modality, checkpoint target biology, mRNA delivery, and antigen-specific tolerance. (1) Nancy Ratner's group at Cincinnati Children's Hospital Medical Center with Yi Zheng reports that PKC-delta agonism selectively kills NF1-deficient tumor cells through pyroptosis — PKC-delta phosphorylates KRAS at a site that traps it in an inactive state, the resulting collapse of RAS signaling in the NF1-null context triggers caspase activation and gasdermin-E cleavage, and NF1-intact cells with NF1's brake on the RAS pathway are spared, giving a structural selectivity window. PKC-delta agonism suppresses growth of NF1-deficient plexiform neurofibroma and malignant peripheral nerve sheath tumor models in vivo — indications where surgery is the only meaningful intervention and MEK inhibitors do not durably control disease — and the use of immunogenic pyroptosis rather than apoptosis as the executor opens an adaptive-immunity engagement axis against residual tumor. (2) Panyu Fei and Wei Chen at Zhejiang University with Michael Dustin at Oxford used crystallography plus single-molecule imaging to show that PD-1's inhibitory function on T cells is mechanistically driven by SHP2's structural scaffolding, not its phosphatase activity — pPD-1 accelerates SHP2's transition into the activated conformation by >94-fold, and activated SHP2 directly dismantles phase-separated pTCR and pCD28 signaling condensates independently of catalysis. This reframes next-generation PD-1 modulators around non-catalytic SHP2 conformational locking and condensate-protective protein-protein-interaction inhibitors, and recontextualizes existing allosteric SHP2 inhibitors developed for tumor-cell-intrinsic RAS in the checkpoint context. (3) Gregg Duncan's group at the University of Maryland reports that adding the clinical lung surfactant Poractant alfa to ionizable LNPs yields an order-of-magnitude higher particle concentration at assembly, sharper pH sensitivity in the endosomal range, faster cytosolic mRNA release, higher reporter expression with no toxicity in vitro, and enhanced pulmonary mCherry expression in mice. Using a clinical-grade surfactant already approved for neonatal RDS as an LNP excipient is a low-friction path into IND-enabling work for inhaled mRNA programs in CF, AAT deficiency, respiratory infection, and pulmonary oncology. (4) Tomoyoshi Yamano's group at the WPI Nano Life Science Institute at Kanazawa University reports Tol-mRNA, a modular mRNA construct co-encoding a target antigen with PD-L1 and TGF-beta as regulatory cues that drives antigen-specific Treg induction while suppressing effector responses to the same antigen; reduced inflammation and clinical scores in autoimmune disease models, prevented allergic symptoms and dampened type-2 inflammation in food allergy models, and generated antigen-specific Tregs in human PBMC. Modular antigen-agnostic tolerance modality with autoimmune, allergy, and transplant applications, structurally analogous to how LNP-mRNA vaccines became an antigen-agnostic immunization modality. 2026-05-23-pkcdelta-nf1-pyroptosis-pd1-shp2-scaffold-surfactant-lnp-tol-mrna Sat, 23 May 2026 12:00:00 +0000 386 Four bioRxiv stories: (1) Ratner lab (Cincinnati Children's) with Zheng — PKC-delta agonism selectively induces gasdermin-E pyroptosis in NF1-deficient tumors through KRAS phosphorylation that traps the GTPase inactive; NF1-intact cells are spared, providing a synthetic-lethal-style selectivity window. In vivo suppression of plexiform neurofibroma and MPNST models, two indications without durable medical therapy; immunogenic pyroptosis as the executor opens an adaptive-immunity engagement axis. (2) Fei/Chen (Zhejiang) with Dustin (Oxford) — crystallography + single-molecule imaging show pPD-1 accelerates SHP2 conformational activation >94-fold and that activated SHP2's structural domains dismantle phase-separated pTCR and pCD28 signaling condensates independent of phosphatase activity; reframes next-generation PD-1 modulators around scaffolding/conformational pharmacology and recontextualizes existing allosteric SHP2 inhibitors in the checkpoint context. (3) Duncan lab (Maryland) — clinical lung surfactant Poractant alfa as an LNP excipient gives order-of-magnitude higher particle yield, sharper endosomal pH sensitivity, faster cytosolic mRNA release, higher in vitro expression with no toxicity, and enhanced pulmonary mCherry expression in mice; low-friction IND-enabling path for inhaled mRNA in CF, AAT deficiency, respiratory infection, and pulmonary oncology. (4) Yamano lab (Kanazawa NanoLSI) — Tol-mRNA modular construct co-encoding antigen with PD-L1 and TGF-beta drives antigen-specific Treg induction while suppressing effector responses to the same antigen; reduces disease in autoimmune and food allergy models, generates antigen-specific Tregs in human PBMC. Antigen-agnostic tolerance modality analogous to how LNP-mRNA vaccines became an antigen-agnostic immunization modality. Four bioRxiv stories spanning target biology, neuroimmune AD modality, AI-driven de novo peptide design, and biologics delivery. (1) Neil McDonald's group at the Francis Crick Institute with Yves-Alain Barde and the ZEB85 team solved the structural basis for human TrkB activation by both BDNF and the selective small-protein agonist ZEB85, identifying a sulfated tyrosine at residue 400 in the TrkB extracellular domain that is indispensable for ZEB85 activation but is not required for BDNF binding (BDNF binding extends into the juxtamembrane segment); AlphaFold modeling and biophysical analysis show BDNF displaces ZEB85 through an overlapping but distinct epitope, and the sulfotyrosine is absent in TrkA and TrkC — the molecular handle behind ZEB85 selectivity and a defined target site for new TrkB-specific protein ligands. (2) Moustafa Gabr's group at Weill Cornell ran a high-throughput CETSA screen of ~40,000 compounds against the inhibitory immune receptor ILT3 (LILRB4) on Alzheimer's microglia and pulled out IB15C, a sub-micromolar binder validated by MST, SPR, docking, and site-directed mutagenesis; in human iPSC-derived microglia IB15C disrupts the ILT3-ApoE interaction, reduces SHP1/2 phosphatase signaling, suppresses cytokine secretion, and enhances amyloid uptake, with favorable PK and safety — small-molecule disinhibition lever for microglia complementing TREM2 agonist antibodies. (3) Mads Nygaard's group at Gubra reports an end-to-end de novo peptide discovery pipeline taken from AlphaFold2 hallucination (ColabDesign) through in vivo proof of concept against the obesity target MC4R; >5,000 linear and head-to-tail cyclic candidates generated, 74% of linear and 23% of cyclic actives, lead cyclic at 340 nM despite lacking the canonical melanocortin activation motif; deep mutational scanning + half-life extender scanning + combinatorial library identified an alternative APWR activation motif, and a P5 substitution gave E5P with 6.7 nM potency on hMC4R; 10 nmol icv E5P reduced acute food intake in mice — de novo design reaching receptor-pocket peptide leads with non-natural activation pharmacophores end-to-end. (4) Omid Veiseh's group at Rice University with Mick Diehl report a retrievable, immunomodulatory cell-encapsulation device that acts as a sustained mAb factory from a single subcutaneous implant — an immunocompetent-screened chemically modified alginate formulation sustains stable serum 3BNC117 (HIV neutralizing) titers for 1 year in mice, scRNA-seq shows a local anti-inflammatory pro-resolving niche that attenuates fibrosis, the same allogeneic chassis produces 13 different mAbs including ipilimumab, pembrolizumab, adalimumab, and PGT121, and a retrievable macrodevice enables on-demand termination and re-implantation for dose tuning; subcutaneous NHP implants gave stable ipilimumab for >6 months with no toxicity, no anti-drug antibodies, and dose-proportional exposure across a three-dose escalation — single-administration mAb delivery platform. 2026-05-22-trkb-sulfotyrosine-ilt3-microglia-mc4r-de-novo-peptide-cell-mab-factory Fri, 22 May 2026 12:00:00 +0000 378 Four bioRxiv stories: (1) McDonald lab (Francis Crick) with Barde and the ZEB85 team — structural basis for TrkB activation by BDNF and the selective protein agonist ZEB85 reveals a sulfated tyrosine at TrkB Y400 indispensable for ZEB85 but not BDNF activation; BDNF binding extends into the juxtamembrane region and displaces ZEB85 via an overlapping epitope; sulfotyrosine is absent in TrkA/TrkC — the molecular handle behind TrkB selectivity and a defined site for new TrkB-specific ligands relevant to neuroprotection programs. (2) Gabr lab (Weill Cornell) HT-CETSA screen of ~40k compounds against ILT3/LILRB4 yielded IB15C (sub-µM binder, MST/SPR/docking/mutagenesis-validated) that disrupts ILT3-ApoE, reduces SHP1/2, suppresses cytokine secretion, and enhances amyloid uptake in human iPSC-derived microglia with favorable PK/safety — small-molecule disinhibition complementing TREM2 agonist antibodies in AD. (3) Gubra (Nygaard lab) de novo cyclic MC4R peptide via AlphaFold2 hallucination (ColabDesign) — >5,000 candidates, lead cyclic 340 nM with no canonical melanocortin motif, deep mutational scanning + combinatorial maturation identified APWR alternative activation motif, P5 substitution gave E5P at 6.7 nM on hMC4R, 10 nmol icv reduced acute food intake in mice — end-to-end de novo design through in vivo POC with non-natural activation pharmacophore. (4) Veiseh lab (Rice) with Diehl — retrievable immunomodulatory alginate cell encapsulation device sustains 3BNC117 titers for 1 year in mice (scRNA-seq shows pro-resolving niche attenuating fibrosis), produces 13 mAbs from an allogeneic chassis including ipilimumab, pembrolizumab, adalimumab, PGT121, retrievable for on-demand termination and dose tuning; subcutaneous NHP implants gave stable ipilimumab for >6 months with no toxicity, no ADAs, and dose-proportional exposure across a three-dose escalation. Four bioRxiv stories spanning targeted molecular imaging, glycobiology-driven IBD targets, GPCR conformational pharmacology, and biosensing. (1) Larry Marnett's group at Vanderbilt with Robert Coffey and Craig Duvall raised an alpaca-derived library of 73 COX-2-specific nanobody clones and engineered the lead candidate F9 with two surface lysine-to-glutamine mutations (K45Q, K77Q) so a fluorophore (ROX) conjugates cleanly at the amino terminus without disrupting COX-2 binding; the conjugate has a 17.9 min plasma half-life, no acute toxicity through 40 mg/kg, and lights up COX-2-positive adenomas selectively on noninvasive fluorescence endoscopy in the AOM/DSS mouse colorectal model — chemistry-controlled nanobody conjugation that converts decades of COX-2 mechanistic work into a screening probe for the early lesions (aberrant crypt foci, hyperplastic polyps, microadenomas) that conventional colonoscopy misses. (2) Jennifer Kohler at UT Southwestern with Lance Hooper's group identify B3GNT7 as the critical glycosyltransferase that builds sulfated polyLacNAc (keratan sulfate) extensions on colonic mucin O-glycans; uniquely among B3GNT family members it prefers a sulfated acceptor substrate, expression is dramatically reduced in ulcerative colitis tissue, the enzyme is required for polyLacNAc-modified mucus in a human goblet cell model, and intestinal B3gnt7 deficiency makes mice markedly more susceptible to chemically induced colitis and enteric infection — nominates a specific druggable glycan layer upstream of barrier failure in IBD, with B3GNT7 stabilization or downstream sulfated polyLacNAc supplementation as new therapeutic axes. (3) Irina Kufareva's group at the Skaggs School at UC San Diego ran 10-microsecond all-atom MD simulations on ACKR3 and CXCR4 with CXCL12 and the antagonist variant [P2G]CXCL12: CXCR4 transitions cleanly between active and inactive states with ligand identity, while ACKR3 holds a persistently active TM7 and a variable, loosely constrained TM6 and lacks the residue interaction networks in the TM core that gate full activation in CXCR4 — the receptor is conformationally pre-disposed to a beta-arrestin-competent geometry but cannot lock in a G-protein-competent one; predictions validated by BRET on ACKR3 mutants. Concrete structural-dynamics handle on what ACKR3 bias means for HIV, cancer metastasis, and regenerative-medicine programs. (4) Tom Soh's lab at Stanford reports the CAFET (charge-amplified FET) aptamer biosensor: a strand-displacement aptamer switch releases a hybridized displacement strand from within the Debye length of the transistor surface on target binding, with an optional charge label for further amplification — picomolar detection of 3-hydroxykynurenine and progesterone in undiluted human plasma, four orders of magnitude below the aptamer KD. Modular, aptamer-agnostic architecture moves the bottleneck for new analytes to aptamer selection. 2026-05-21-cox2-nanobody-endoscopy-b3gnt7-glycans-ackr3-dynamics-cafet-aptamer Thu, 21 May 2026 12:00:00 +0000 308 Four bioRxiv stories: (1) Marnett, Coffey, and Duvall labs (Vanderbilt) build the first COX-2-targeted nanobody for in-vivo molecular endoscopy of colorectal adenomas — alpaca-derived 73-clone library, lead F9 with two surface K-to-Q mutations enables clean N-terminal fluorophore conjugation, 17.9 min plasma half-life, no toxicity through 40 mg/kg, tumor-selective imaging in the AOM/DSS mouse model. Chemistry-controlled nanobody conjugation turns decades of COX-2 work into a screening probe for the early lesions that colonoscopy misses. (2) Kohler and Hooper labs (UT Southwestern) identify B3GNT7 as the critical glycosyltransferase that builds sulfated polyLacNAc (keratan sulfate) on colonic mucin O-glycans — uniquely prefers a sulfated acceptor, expression dramatically reduced in ulcerative colitis, required for polyLacNAc-modified mucus in human goblet cells, intestinal deficiency drives colitis and enteric infection susceptibility in mice. Druggable glycan layer upstream of barrier failure in IBD. (3) Kufareva lab (UCSD Skaggs) 10-microsecond MD simulations show ACKR3 holds persistently active TM7 and loosely constrained TM6 and lacks the TM-core residue interaction networks that gate CXCR4 activation — receptor is conformationally pre-disposed to beta-arrestin geometry but cannot lock G-protein geometry; BRET-validated on ACKR3 mutants. Concrete structural handle on ACKR3 bias for HIV, cancer metastasis, regenerative medicine. (4) Soh lab (Stanford) CAFET charge-amplified FET aptamer biosensor with strand-displacement switch achieves picomolar detection of 3-hydroxykynurenine and progesterone in undiluted human plasma — four orders below aptamer KD; modular architecture moves bottleneck to aptamer selection. Four bioRxiv stories pushing modality, delivery, screening, and supply. (1) Jim Wells' group at UCSF extends targeted protein degradation outside the cell with KineTACs — bispecific degraders that link an anti-LRP8 arm to a cytokine receptor that traffics to the lysosome — to deplete the selenium uptake receptor LRP8 from cancer cell surfaces; loss of LRP8 collapses selenium import, drops GPX4 and other selenoproteins, and sensitizes tumors to ferroptosis. The conceptual move beyond ferroptosis is that an extracellular degrader can reprogram an intracellular translational dependency, opening nutrient-acquisition receptors as a general resistance node. (2) Joost Sluijter's group at UMC Utrecht built NEO-TOP-EVs, an extracellular vesicle platform reverse-engineered from native EV biogenesis — PI(4,5)P2 plasma-membrane targeting, ESCRT-mediated scission, and self-assembly-driven cargo clustering — to deliver Cas9 and adenine base editor ribonucleoproteins with no nucleic acid template; PCSK9 splice-site disruption in vitro lowered PCSK9 and rebounded LDL receptor activity, and proof-of-concept in vivo Cre delivery to mouse liver demonstrates protein-based editing without viral or lipid-nanoparticle re-dosing. (3) Steven Banik's group at Stanford defines ratiometric transcriptional activation: fusing a protein of interest to a transcriptional inhibitor lets the cell sense the inhibitor-to-transcription-factor ratio, turning any degradation event — ligase-, molecular-glue-, or autophagy-driven — into a scalable transcriptional output (reporter, survival, barcode); applied to oncogenic targets the platform yielded new molecular glue degraders and offers a generic discovery engine for induced proximity. (4) Keisuke Kaji's group at Edinburgh report pre-cHep, a culture condition that converts primary human hepatocytes into a proliferating precursor expanding over one-million-fold in thirty days while retaining mouse liver repopulation parity with the starting cells — three-dimensional spheroid differentiation in seven days restores albumin and CYP activity, and CRISPR-tractability enables isogenic disease modelling; addresses the chronic supply, variability, and cost problems in PHH-based ADME and hepatotoxicity workflows. 2026-05-20-kinetac-lrp8-ferroptosis-ev-base-editing-ratiometric-degraders-million-fold-hepatocytes Wed, 20 May 2026 12:00:00 +0000 253 Four bioRxiv stories: (1) Wells lab (UCSF) KineTAC bispecific degraders against LRP8 — extracellular degradation of the selenium uptake receptor LRP8 collapses selenium import, drops GPX4 and other selenoproteins, and sensitizes cancer cells to ferroptosis; opens nutrient-acquisition receptors as a general resistance node accessible from outside the cell. (2) Sluijter lab (UMC Utrecht) NEO-TOP-EVs engineered extracellular vesicle platform combining PI(4,5)P2 membrane targeting, ESCRT-mediated scission, and self-assembly-driven cargo clustering — efficient delivery of Cas9 and adenine base editor RNPs with no nucleic acid template; in vitro PCSK9 splice-site disruption drops PCSK9 and rebounds LDLR activity, and proof-of-concept in vivo Cre delivery to mouse liver demonstrates protein-only editing. (3) Banik lab (Stanford) ratiometric transcriptional activation — fusing POI to a transcriptional inhibitor lets a cell sense degradation as a transcriptional output (reporter/survival/barcode); mechanism-agnostic across ligase, molecular glue, and autophagy degraders, multiplexable across pools, and yields new molecular glue degraders against oncogenic targets — generic discovery engine for induced proximity. (4) Kaji lab (Edinburgh) pre-cHep precursor of chemically expanded hepatocytes — >10^6-fold expansion in 30 days, mouse liver repopulation parity with primary human hepatocytes, 3D spheroid differentiation in 7 days restores albumin and CYP activity, CRISPR-tractable; scalable, reproducible, editable PHH alternative for ADME and hepatotoxicity workflows. Four bioRxiv stories on the target and modality side of drug discovery. (1) Adam Antebi's group at the Max Planck Institute for Biology of Ageing identifies the short-chain dehydrogenase DHS-26/DHRS1 as a previously uncharacterized effector of an mTOR-steroid longevity axis: deletion of the TORC1 regulator raga-1/RRAGA in C. elegans enhances production of the bile-acid-like hormone dafachronic acid and extends life span through the DAF-12 nuclear receptor (FXR homolog), with mTOR- and FXR-regulation of DHRS1 conserved in mouse — opens a pharmacological route to mTOR-like life-span effects without direct mTOR inhibition, with FXR as an already-clinical adjacent target. (2) Anna Qiao's group at Tianjin University reports three cryo-EM structures of the broadly promiscuous bitter taste receptor TAS2R14 bound to 3,5-diiodosalicylic acid, flufenamic acid, and aristolochic acid at 2.39–2.69 Å — transmembrane helix 6 emerges as the regulatory hub with a 12 Å lever-like agonist-triggered rearrangement, and cholesterol hemisuccinate is identified as a new orthosteric agonist that selectively potentiates DA-mediated signaling ~5-fold via orthosteric-allosteric coupling. Critical for the airway smooth-muscle relaxation (asthma) and adipocyte browning programs increasingly targeting TAS2Rs beyond the tongue. (3) Christopher Ahern's group at the University of Iowa optimizes suppressor tRNA payloads for UGA premature termination codons — the single most common rare-variant pathogenic class in neurodevelopmental disease — using a luciferase-UGA reporter mouse with non-invasive transcranial bioluminescence readout: self-complementary AAV2/9 with the tRNA in a minimal 100 bp context produces broad efficacious rescue, multiplexing and native introns enable dose-sparing at low titers, and tRNA-seq confirms no perturbation of endogenous tRNA pools or charging; a delivery framework ready for UGA-driven neurodevelopmental disorders. (4) Mariane Le Fur's group at the Martinos Center at Mass General presents a bioorthogonal substrate plus turn-on fluorophore that becomes fluorescent only after click reaction with lysyl oxidase's allysine product — two commercial reagents, under two hours, compatible with second-harmonic generation and immunofluorescence on the same section, validated across healthy and fibrotic mouse and human tissue. Closes a routine assay gap for anti-fibrotic and LOX-inhibitor target-engagement readouts. 2026-05-19-longevity-axis-bitter-receptor-stop-codon-repair-fibrosis-probe Tue, 19 May 2026 12:00:00 +0000 311 Four bioRxiv stories on targets and modalities: (1) Antebi lab (Max Planck Institute for Biology of Ageing) identifies DHS-26/DHRS1 as a downstream effector of an mTOR-FXR longevity axis — raga-1/RRAGA deletion in C. elegans drives production of dafachronic acid (bile-acid-like hormone) and life-span extension through DAF-12 (FXR homolog); mouse DHRS1 also mTOR- and FXR-regulated, suggesting a pharmacological route to mTOR-like longevity without direct mTOR inhibition. (2) Qiao lab (Tianjin) cryo-EM of TAS2R14-Gi with three agonists (DA/FA/AA) at 2.4–2.7 Å — TM6 is the regulatory hub (12 Å lever-like rearrangement), and cholesterol hemisuccinate identified as a new orthosteric agonist that selectively potentiates DA signaling ~5-fold via orthosteric-allosteric coupling; relevant for TAS2R programs in asthma airway smooth muscle and adipocyte browning. (3) Ahern lab (Iowa) optimizes suppressor tRNA delivery for UGA premature stop codons (most common rare-variant pathogenic class in neurodevelopmental disease) in mammalian brain — luciferase-UGA reporter mouse with non-invasive bioluminescence, scAAV2/9 with tRNA in 100 bp minimal context, multiplexing and native introns enable dose-sparing; no perturbation of endogenous tRNA pools or charging. (4) Le Fur lab (MGH Martinos) bioorthogonal substrate plus turn-on fluorophore for lysyl oxidase activity — two reagents, under two hours, compatible with SHG and immunofluorescence, validated on healthy and fibrotic mouse/human tissue; closes a target-engagement assay gap for anti-fibrotic and LOX-inhibitor programs. Four bioRxiv stories on the manufacturing and modality side of drug discovery. (1) Bernd Schnabl's group at UC San Diego reports GR-7, a non-absorbable microbial bile salt hydrolase inhibitor that reprograms cecal bile acid pools without systemic host exposure; in mouse MASH models, early intervention suppresses liver fibrosis and late-stage dosing reduces hepatic steatosis and inflammation through restored gut barrier function and rewired hepatic bile acid metabolism — a non-systemic small molecule drugging a bacterial enzyme to indirectly modulate FXR/TGR5 signaling, structurally distinct from FGF21, THR-beta, and GLP-1/glucagon programs crowding the field. (2) Pingjing Zhang's group at Shanghai Cell Therapy Group engineered chimeric T7 RNA polymerases by fusing mutant catalytic domains to thermostable DNA-binding domains (Sso7d, MC1), achieving salt tolerance to 270 mM NaCl, dsRNA contamination below 0.001%, and a 50% circular RNA yield improvement to 15 mg/mL in simple batch transcription — a manufacturing-side fix for the reactogenicity bottleneck particularly relevant to circRNA programs. (3) Kate Galloway's group at MIT systematically dissected how genetic organization and component choice in two-gene lentiviral vectors carrying RNA-based regulatory devices (ribozymes, splicing switches) affect both production titer and downstream expression, arriving at design rules that lift titer over 30-fold versus naive configurations — a usable lentiviral platform with embedded RNA-level control for cell therapy programs wanting ligand-responsive activation or feedback regulation. (4) Jay Vornhagen's group at Indiana University School of Medicine screened an antiviral compound library and identified six FDA-approved antivirals with previously uncharacterized antibacterial activity against Klebsiella pneumoniae with strain-specific potency and context-dependent activity under simulated gut conditions — a phenotype-first repurposing path toward gut decolonization, currently an unmet need despite gut colonization preceding most invasive Klebsiella disease. 2026-05-18-microbiome-mash-rna-manufacturing-lentiviral-circuits-antiviral-repurposing Mon, 18 May 2026 12:00:00 +0000 238 Four bioRxiv stories on manufacturing and modality: (1) Schnabl lab (UCSD) GR-7, a non-absorbable microbial bile salt hydrolase inhibitor that reprograms cecal bile acids without systemic host exposure — early dosing suppresses liver fibrosis, late dosing reduces steatosis and inflammation in mouse MASH; non-systemic small molecule drugging a bacterial enzyme to indirectly modulate FXR/TGR5, distinct from FGF21/THR-beta/GLP1-glucagon programs. (2) Zhang lab (Shanghai Cell Therapy Group) chimeric T7 RNAP fusing mutant catalytic domains to thermostable DNA-binding domains (Sso7d, MC1) — salt tolerance to 270 mM NaCl, dsRNA <0.001%, 50% circRNA yield improvement to 15 mg/mL in simple batch transcription. (3) Galloway lab (MIT) design rules for two-gene lentiviral vectors with embedded RNA-based regulatory devices (ribozymes, splicing switches) lift titer >30-fold versus naive configurations — usable lentiviral platform with RNA-level control for cell therapy. (4) Vornhagen lab (Indiana) screened antiviral library, identified six FDA-approved antivirals with antibacterial activity against Klebsiella pneumoniae, strain-specific and context-dependent under simulated gut conditions — phenotype-first repurposing path toward Klebsiella gut decolonization, currently unmet. Four bioRxiv stories from a quiet Sunday cycle, all posted May 13-14. (1) Kyle Daniels' group at Stanford built a library of synthetic cytokine receptors with programmable extracellular ligand-binding domains paired with custom intracellular signaling modules, then used them to polarize primary human macrophages into user-defined functional states; validated a two-state mathematical model predicting polarization outcomes from receptor composition, and demonstrated a 30-fold reduction in tumor burden in vivo — programmable cell therapy at the macrophage level that sidesteps the antigen-escape and persistence problems plaguing T-cell programs in solid tumors. (2) Ruiz, Marambaud, Escande et al. (Institut Pasteur de Montevideo) identify nitazoxanide — a long-approved oral antiparasitic with decades of safety data — as a BMP9-ALK1-SMAD pathway activator that restores SMAD1/5/8 signaling and ID1 expression in patient endothelial cells carrying HHT mutations, reduces vascular malformations in mouse models, and concurrently suppresses mTOR to correct a counterproductive metabolic cascade; generic-approved oral drug for hereditary hemorrhagic telangiectasia compresses translational timelines from years to months. (3) Lorena Saelices (UT Southwestern) used cryo-EM to solve brain-derived ATTR fibril structures from V30M and V30G patients and finds CNS folds are distinct from the peripheral fibrils characterized for the last decade — as tafamidis, acoramidis, and TTR silencers prolong peripheral survival, ATTR deposits emerge in long-surviving patients' brains, raising the mechanistic question of whether stabilizers translate to CNS and the design question of whether brain-fold-specific small molecules or antibodies are needed for clearance. (4) Tomas Helikar's group (Nebraska) introduces MechAInistic, an LLM-guided multi-agent layer over genome-scale constraint-based metabolic models — natural-language questions convert into executable workflows for perturbation analysis, pathway comparison, and target nomination; case studies on immune cell models identify mitochondrial rewiring in rheumatoid arthritis and nominate drug targets for multiple sclerosis — a template for pairing agentic AI with mechanistic systems biology rather than purely black-box statistical models. 2026-05-17-programmable-macrophages-repurposing-brain-amyloid-agentic-ai Sun, 17 May 2026 12:00:00 +0000 236 Four bioRxiv stories from Sunday May 17: (1) Daniels lab (Stanford) synthetic cytokine receptors with programmable extracellular ligand-binding and intracellular signaling modules polarize primary human macrophages into user-defined states, validated by a two-state model and 30-fold tumor burden reduction in vivo — programmable cell therapy at the macrophage level. (2) Ruiz/Marambaud/Escande (Institut Pasteur de Montevideo) identify nitazoxanide, an FDA-approved antiparasitic, as a BMP9-ALK1-SMAD activator that restores SMAD1/5/8 signaling and ID1 expression in HHT-mutant endothelium, reduces vascular malformations in mouse models, and suppresses mTOR — generic oral drug compressing HHT translational timelines from years to months. (3) Saelices (UT Southwestern) cryo-EM shows brain-derived ATTR fibrils from V30M and V30G patients adopt CNS-specific folds distinct from peripheral fibrils — as TTR stabilizers and silencers prolong peripheral survival, CNS ATTR demands brain-fold-specific therapeutic design. (4) Helikar lab (Nebraska) MechAInistic — LLM-guided multi-agent layer over genome-scale constraint-based metabolic models converts natural-language queries into executable workflows for perturbation/pathway analysis and target nomination, case studies identify mitochondrial rewiring in RA and drug targets in MS. Four stories on selectivity, drug-discovery baselines, and what cell engineering looks like delivered in vivo. (1) Neel Shah's group at Columbia (bioRxiv, May 14) presents substrate-derived covalent peptide inhibitors of tyrosine kinases that abandon the conserved ATP pocket entirely — an optimized substrate peptide carries an electrophile to a non-conserved cysteine on the kinase surface, demonstrated for Src and FGFR1, with covalent peptides that discriminate an oncogenic FGFR1 mutant from wild-type and retain activity against a clinically relevant resistance mutation; a fundamentally new binding mode for kinase drugs. (2) The Shoichet lab at UCSF (bioRxiv, May 13) systematically introduced single-atom changes to 18 lead molecules across 6 targets (257 random analogs) and found ~11% improved potency >=10-fold — a sobering baseline for medicinal-chemistry campaigns; the more potent analogs also had systematically worse PK (metabolic stability, plasma free fraction), so potency gains often don't survive to in vivo. (3) BioAge Labs walked Fierce Biotech through its multi-disease strategy for BGE-102, an oral, brain-penetrant NLRP3 inhibitor with a novel chemotype and non-canonical binding site — Phase 1 showed >=85% median hsCRP reductions at 60/120 mg with brain exposure above target IC90; Phase 2 CV proof-of-concept this half, DME Phase 1b/2a mid-year, CNS to follow; one molecule chasing inflammation across three organ systems. (4) Azalea Therapeutics' ASGCT 2026 late-breaker (May 15) reported the first in-primate demonstration of in vivo TRAC-CAR T cell engineering with site-specific integration at the TRAC locus — a dual-vector platform (cell-selective enveloped delivery vector + AAV homology template) generated CAR T cells in vivo in 6 rhesus macaques from a single IV dose without lymphodepletion, with complete B-cell depletion in blood, lymph nodes, and bone marrow; complements yesterday's Tessera random-integration story with a site-specific architecture. 2026-05-16-selectivity-baselines-and-in-vivo-cart Sat, 16 May 2026 12:00:00 +0000 255 Four stories: (1) Shah lab (Columbia) bioRxiv May 14 — substrate-derived covalent peptide inhibitors of tyrosine kinases that bypass the conserved ATP pocket by placing an electrophile on a substrate-mimetic peptide targeted to a non-conserved cysteine; demonstrated for Src and FGFR1 including discrimination of an oncogenic FGFR1 mutant from wild-type and activity against a clinically relevant resistance mutation. (2) Shoichet lab (UCSF) bioRxiv May 13 — random single-atom perturbations of 18 lead molecules across 6 targets yielded ~11% rate of >=10-fold potency improvement, a sobering medicinal-chemistry baseline; the more potent analogs systematically had worse PK. (3) BioAge BGE-102, oral brain-penetrant NLRP3 inhibitor with novel chemotype and non-canonical binding site — Phase 1 >=85% median hsCRP reduction at 60/120 mg, brain exposure above target IC90; CV Phase 2 this half, DME Phase 1b/2a mid-year, CNS planned. (4) Azalea Therapeutics ASGCT 2026 late-breaker May 15 — first-in-primate in vivo TRAC-CAR T cell engineering via dual vector (cell-selective enveloped delivery + AAV homology template) with site-specific integration at the TRAC locus; 6 rhesus macaques, single IV dose, no lymphodepletion, complete B-cell depletion in blood, lymph nodes, and bone marrow; complements yesterday's Tessera random-integration story with a site-specific architecture. Four stories pointing in one direction: the modalities and tools we use to make medicines are getting more programmable. Tessera Therapeutics at ASGCT 2026 shows a single IV infusion of RNA Gene Writer in a targeted LNP edits ~85% of circulating blood cells in their sickle cell model with stable 40-60% long-term HSC editing (peak ~76% in NHPs to 84 days), plus first permanent CAR transgene integration in NHPs from one dose of T-cell-targeted LNP (~60% T cells edited, B cell clearance in blood and lymph nodes) — no apheresis, no busulfan, no ex vivo manufacturing, enabled by ~52-fold bone-marrow-over-liver LNP biodistribution. FDA approves Arvinas/Pfizer's vepdegestrant (Veppanu) for ESR1-mutated ER+/HER2- advanced breast cancer — first heterobifunctional protein degrader ever approved; VERITAC-2 showed median PFS 5 mo vs 2.1 mo on fulvestrant (HR 0.57) with ORR 19% vs 4%, oral once-daily vs fulvestrant intramuscular partial degradation. Andreotti group bioRxiv preprint argues covalent warheads are pharmacodynamic, not just chemistry: acrylamide vs 2-butynamide drive distinct long-range BTK conformational dynamics with measurably different downstream signaling, so warhead choice should be a matrix decision not a default. OpenAI introduces GPT-Rosalind, a domain-specific life sciences reasoning model — leading BixBench scores, beats GPT-5.4 on 6/11 LABBench-2 tasks, trusted-access deployment with Amgen, Moderna, Allen Institute, Thermo Fisher; frontier vendors are now optimizing for scientific workflows, shifting build-vs-buy on internal tooling. 2026-05-15-inaugural-briefing Fri, 15 May 2026 12:00:00 +0000 226 Four stories on programmable modalities: (1) Tessera Therapeutics at ASGCT 2026 shows a single IV infusion of RNA Gene Writer in targeted LNP edits ~85% of circulating blood cells in sickle cell model (40-60% durable HSC editing, peak ~76% NHP to 84 days) plus first permanent CAR transgene integration in NHPs from one dose of T-cell-targeted LNP (~60% T cells, B cell clearance), enabled by ~52-fold bone-marrow-over-liver LNP biodistribution; (2) FDA approves Arvinas/Pfizer vepdegestrant (Veppanu) May 1 for ESR1-mutated ER+/HER2- advanced breast cancer — first PROTAC ever approved, VERITAC-2 median PFS 5 vs 2.1 mo (HR 0.57), ORR 19% vs 4%, oral vs intramuscular fulvestrant; (3) Andreotti group bioRxiv preprint shows covalent warheads (acrylamide vs 2-butynamide) drive distinct BTK conformational dynamics and downstream signaling — warhead choice is pharmacodynamic, not just chemistry; (4) OpenAI introduces GPT-Rosalind domain-specific life sciences reasoning model with leading BixBench scores, beats GPT-5.4 on 6/11 LABBench-2 tasks, trusted-access with Amgen, Moderna, Allen Institute, Thermo Fisher.