Safety & toxicology
15 articles
Factor XI/XIa inhibition pursued for low-bleeding anticoagulation
- Factor XI and its active form XIa amplify thrombin generation and drive pathological thrombosis, yet contribute little to normal hemostasis, which makes them a candidate target for anticoagulants with lower bleeding risk.
- The review covers FXI/XIa structure, structure-guided design and the main therapeutic modalities, with critical appraisal of clinical evidence across thrombotic indications.
- Open issues include selectivity, disease-dependent efficacy and FXI roles beyond coagulation, all bearing on long-term safety and clinical positioning.
FXI's appeal rests on human genetics (congenital deficiency, little spontaneous bleeding) more than on animal models, and that is the credible part. What I'd probe is the exposure ceiling: if the efficacy window is narrow and trough inhibition drifts, you lose the bleeding advantage. Off-target serine protease selectivity against FXa, thrombin and kallikrein decides whether the margin is real.
PFOS found in human pancreas, disrupts insulin secretion
- In organ donors, PFOS was found in nearly all pancreas samples, far more often than PFHxS or PFOA; mouse data also showed PFOS accumulating in pancreas.
- In cell lines and islets, prolonged high-dose PFOS suppressed insulin secretion, while acute exposure boosted it in mouse and some human islets.
- A GLP-1 receptor antagonist partly blunted the acute surge, and in mice PFOS raised insulin only alongside glucose, so the model chosen shapes the result.
Biphasic data are hard to build on: acute PFOS raised insulin, prolonged high doses suppressed it, and the mouse response needed glucose present. Pancreatic concentrations measured alongside these doses are missing, so the human relevance of the in vitro exposures stays open. The partial GLP-1R antagonist block is a lead worth following, though likely one arm of the mechanism.
BPA, DBP and PFOA mix acts additively at ERα
- A biomonitoring-informed mixture of BPA, DBP and PFOA was tested as a model of environmental estrogen co-exposure.
- Each chemical and the mixture promoted MCF-7 cell proliferation, and the mixture's dose-response stayed within the additivity range of both concentration and response addition models, leaning toward concentration addition.
- ER transcription and cell-free biosensor assays showed ERα agonism and fitted concentration addition better, pointing to receptor activation as central to the additive effect.
Additivity in MCF-7 proliferation is what three ERα agonists should produce, so this reads as confirmation rather than surprise. PFOA and DBP are weak or indirect ER actors, though, and MCF-7 over-reports estrogenic output. With no measured free internal concentrations or in vivo uterotrophic data, whether biomonitoring-level exposure reaches a receptor-relevant range stays unanswered.
GVHD prophylaxis choice reshapes T cell recovery after transplant
- Post-transplant cyclophosphamide and tacrolimus/methotrexate prophylaxis were compared for T cell recovery in phase 3 trial patients using longitudinal T cell receptor sequencing.
- Cyclophosphamide depleted T cells broadly, so recovery restarted from fewer founder cells and left a narrower repertoire for a long period.
- Lower diversity tracked with less chronic GVHD, but early loss of pathogen-specific clones was linked to higher severe-infection risk.
The same depletion that seems to buy less chronic GVHD also strips early pathogen coverage, so a narrower repertoire reads as a trade-off, not a win. This is a human cohort readout: association, not mechanism. For anyone designing a prophylaxis combination, the prize is sparing protective clones while pruning alloreactive ones, and nothing here shows that separation can be achieved.
CYP2C9 drives Δ8-THC conversion to its psychoactive 11-hydroxy metabolite
- In human liver microsomes, Δ8-THC gave several monohydroxylated products, led by the psychoactive 11-hydroxy metabolite; CYP2C9 emerged as main producer.
- Ring hydroxylation came mainly from CYP3A, side-chain oxidation from CYP2D6, CYP1A2 and CYP3A4.
- CYP2C9 handled Δ8-THC less efficiently than Δ9-THC, though C11 hydroxylation matters more for Δ8; loss-of-function genotypes were predicted to sharply cut 11-hydroxy formation, moving clearance to CYP3A4/CYP2C19.
Microsomes can tell you which enzyme makes the psychoactive metabolite; they can't tell you how a person's exposure will split. They give no in vivo exposure to the 11-hydroxy species, and the genotype effect is a prediction, not an observation. Clinical PK in CYP2C9 poor metabolizers, plus 2C9 inhibitor interaction data, would have to come before anyone calls this a risk factor.
Characterization of Cytochrome P450-Mediated Metabolism of Δ8-tetrahydrocannabinol in vitro ↗
PFOS accumulates in nearly all donor pancreases and alters insulin secretion in a model-dependent manner
- Three legacy perfluoroalkyl acids were quantified in plasma and pancreas from human organ donors; PFOS was found in almost every pancreas, while PFOA appeared only rarely.
- A mouse model likewise showed PFOS reaching pancreas at higher levels than PFOA.
- In rat INS-1 cells, mouse islets and human islets, prolonged high-dose PFOS generally suppressed insulin release, whereas acute exposure raised it, partly blunted by a GLP-1 receptor antagonist.
- In mice, acute PFOS raised plasma insulin only alongside a glucose challenge.
Do the pancreatic concentrations measured in donors resemble the high doses that suppress secretion in INS-1 cells and islets? That mismatch decides how much the in vitro result transfers. Acute stimulation turning into prolonged suppression also suggests a time-dependent, biphasic response, and the mouse effect appearing only with a glucose challenge points to a context-dependent phenotype rather than overt beta-cell toxicity.
Preprint links macrophage oxysterol 25-hydroxycholesterol to heightened statin-related kidney injury in cell, mouse and organoid models
- Examines whether 25-hydroxycholesterol, an oxysterol released by activated macrophages, makes kidney cells more vulnerable to statin toxicity.
- Atorvastatin and fluvastatin killed proximal tubular cells when paired with 25-HC, and statin-treated mice developed kidney injury after a small inflammatory challenge.
- Reduced protein prenylation disturbed autophagosome-lysosome function; mevalonate or macrophage depletion lessened injury, and kidney organoids showed higher KIM-1.
A statin plus an inflammatory oxysterol is a plausible toxic pairing: blocking the mevalonate pathway curbs prenylation, and mevalonate rescue fits an on-target mechanism. Still, in vitro and mouse statin exposures often exceed free plasma levels, so dose relevance is unproven, and this preprint is not peer reviewed. Human association data linking 25-HC, inflammation and kidney outcomes are needed to bridge the translational gap.
Macrophage-Derived 25-Hydroxycholesterol Drives Statin-Associated Acute Kidney Injury ↗
Pyridazino-indoles replace the phenol in pyridazine NLRP3 inhibitors
- Indole-fused pyridazines were developed to replace the phenol group of earlier phenol-pyridazine NLRP3 inhibitors.
- Cryo-EM and modeling suggest the indole keeps the key contact with Ala228 in the binding site.
- Two lead compounds reached low-nanomolar potency with good solubility and permeability, and were profiled in mouse pharmacokinetic studies.
- hERG inhibition ended development, though the compounds remain useful in vivo tools; a reactive metabolite audit is also reported.
Replacing the phenol with an indole-fused pyridazine removes phenol liabilities while keeping the Ala228 contact, and potency and solubility look healthy. The cautionary note is hERG ending a series this strong: did the flat, lipophilic fused ring cause it? I would treat these compounds as in vivo tools, not a template to copy without scrutiny.
Pyridazino-Indoles as Phenol Bioisosteres for Potent Inhibition of NLRP3 ↗
Engineered polyene M16 distinguishes ergosterol from cholesterol, keeping antifungal activity while sparing host cells and kidneys in mice
- Mandimycin M16, built on a 5 + 1 polyene scaffold, was designed to overcome amphotericin B's poor discrimination between fungal and mammalian sterols.
- It recognizes the conjugated diene unique to ergosterol, shows no measurable cholesterol binding, and kills fungi rapidly across a panel of WHO priority pathogens.
- In vitro and in mice, hemolysis, cytotoxicity and kidney toxicity were undetectable at the doses tested.
- In disseminated infection models, it beat liposomal amphotericin B at matched doses, with large fungal burden reductions and full survival.
Undetectable nephrotoxicity sounds too clean for a polyene, and I'd want to know how long they dosed. Short mouse regimens rarely reproduce the cumulative tubular damage that limits amphotericin B over weeks in patients. Before taking M16 forward, I'd want repeat-dose tox with exposure data and a workable formulation; the cholesterol-sparing mechanism is convincing, but the safety margin is unproven.
Using induced-fit docking plus FEP to design away from CYP, hERG and PXR
- Anti-targets such as CYP3A4, CYP2D6, hERG and PXR have large, flexible sites that make structure-based design difficult.
- The workflow pairs induced-fit docking with free energy perturbation, enforcing a consensus binding mode across a congeneric series to sample receptor conformations efficiently.
- It was tested on public retrospective datasets and applied prospectively in active drug discovery programs.
Most teams treat hERG and CYP as a post-hoc assay problem; getting usable FEP on these promiscuous pockets would move liability into the design cycle. The consensus-binding-mode restriction is the clever bit, but it also limits you to congeneric series. Prospective hit rates on novel scaffolds will show whether this holds beyond the series it was tuned on.
Dialing hERG out of a DCN1–UBE2M inhibitor without losing potency
- The earlier tool compound NAcM-OPT blocked the DCN1–UBE2M interaction, was orally available and slowed xenograft growth, but carried significant hERG affinity that stalled development.
- The authors systematically modified the scaffold to cut hERG binding while keeping DCN1 engagement.
- Lead compound 110 matched the parent in biochemical and cellular potency, with better physicochemical properties and improved selectivity against hERG.
This is the unglamorous but necessary second act: a validated PPI chemotype that was dead on cardiac safety, rescued by SAR alone. The takeaway for anyone with a basic, lipophilic PPI binder is that hERG can often be engineered out without giving up the pharmacophore, so check your own series before abandoning it.
KPB82, a ceramide-mimetic sulfonamide, is a selective ACER2 inhibitor in a bioRxiv preprint
- ACER2 hydrolyses ceramides to sphingosine and fatty acids, and no pharmacological tool for it existed.
- KPB82, a dodecane sulfonamide carrying a ceramide-like hydroxyphenylpropyl group, inhibited ACER2 potently, with minimal activity on acid and neutral ceramidases and none on ACER1 or ACER3.
- Kinetics were competitive, and docking into an AlphaFold3 model placed it in the substrate pocket near the Zn2+-binding histidines.
- The title reports protection of tumor and primary cells from doxorubicin cytotoxicity.
No ACER2 tool compound until now, so a competitive, ACER1/ACER3-sparing hit matters. A dodecane sulfonamide with a hydroxyphenylpropyl headgroup reads like a ceramide mimic, and the docked pose near the Zn2+ histidines is only an AlphaFold3 hypothesis. I'd want IC50s, an SAR series, and cellular target engagement before trusting the doxorubicin rescue.
Entrance-channel design tames cytotoxicity of the JK-4b NNRTI lead
- JK-4b, a potent NNRTI active against wild-type and clinical HIV-1 variants, was shelved because of high cytotoxicity; the authors used it as a lead for new analogues.
- Structure-guided designs addressing the NNRTI entrance channel gave a series with lower cytotoxicity and improved selectivity.
- Lead compound 9n retained strong potency against common resistance mutants, including double mutants, at levels comparable to etravirine.
- 9n also showed minimal cytochrome P450 inhibition and good human liver microsome stability, supporting further development.
Rescuing a shelved lead is always more interesting than starting cold: JK-4b already had the potency and resistance coverage, so the real question was whether entrance-channel contacts could decouple that from its cytotoxicity. If 9n's profile holds up, that suggests the toxicity was scaffold-tunable, not baked into the pharmacophore. Matching etravirine on double mutants while staying clean on P450 and microsome stability is what I'd want to see before making more.
Iron(II)-Activatable Conjugates Enable Tumor-Conditional Release of the HDAC Inhibitor Panobinostat
- Hydroxamic acid-based HDAC inhibitors are limited by poor selectivity among metalloenzymes and by genotoxicity risk linked to isocyanate metabolites.
- The study applies a previously reported iron(II)-activatable conjugate platform to cage the pan-HDAC inhibitor panobinostat, aiming for tumor-conditional payload release.
- In multiple myeloma cell line models, the conjugate successfully caged HDAC inhibitory activity, and a bioisosteric control conjugate showed reduced genotoxicity relative to the free hydroxamate.
- Follow-up investigation identified the carboxylesterase CES1 as responsible for unexpected proteolytic instability of the panobinostat conjugate observed both in vitro and in vivo.
- The finding is noted as a focus for future work toward mitigating hydroxamate-associated toxicity and mutagenicity through more robust tumor-conditional activation.
Caging panobinostat's hydroxamate behind an Fe(II) trigger is a sensible fix for the isocyanate-genotoxicity liability that's dogged this chemotype, and the bioisosteric control actually showing lower genotoxicity than free drug is the kind of data I'd want before calling any HDACi "safer." But what grabs me is the CES1 finding: an esterase chewing through the conjugate in vitro and in vivo is the failure mode a clean MM cell assay never warns you about. That's not a footnote — it's the design problem for the next iteration, whether you engineer around the CES1 recognition motif or add orthogonal masking. I'd rather see this reported now than discovered in an IND package; it's a useful cautionary data point for anyone building Fe(II)-activatable payloads beyond this one hydroxamate.
Multimodal AI Framework Targets Drug-Resistant Bacteria While Screening for Toxicity
- Antimicrobial resistance is narrowing treatment options for pathogens like Staphylococcus aureus and Neisseria gonorrhoeae, motivating new computational discovery tools.
- MAPViS, a multimodal deep-learning framework, fuses molecular graph, fingerprint, and descriptor/3D geometric features via gated fusion, pretrained on ~10 million molecules using hierarchical pseudolabel classification and masked graph contrastive learning.
- After fine-tuning, it jointly predicts antibacterial activity against S. aureus and N. gonorrhoeae plus cytotoxicity risk in three human cell lines (HepG2, HSkMC, IMR-90).
- On benchmark and external validation data, MAPViS showed strong early-enrichment performance for S. aureus, with all ten top-ranked candidates confirmed experimentally active, while generalization to N. gonorrhoeae was more modest.
- The authors describe MAPViS as a scalable tool for joint activity-toxicity modeling to support antibiotic virtual screening.
Ligand-based, not structure-based — no target structure in the loop, so I'd worry about applicability domain once you push past its ~10M-molecule pretraining chemotypes into a new target class. The joint activity/cytotox head is the real differentiator: folding HepG2/HSkMC/IMR-90 flags into the same ranked list saves the counterscreen step I'd otherwise run manually — genuine time saved, not just convenience. But the S. aureus (10/10 confirmed) vs. N. gonorrhoeae ("modest") split is the tell: performance tracks training-data density per pathogen, not real antibacterial chemotype intuition. Before trusting a ranked hit list on an unfamiliar target, I'd want per-pathogen calibration and confirmed true negatives, not just top-10 hit rates. Good as a cheap pre-HTS triage filter; not a substitute for structure-based design or real counterscreening.