Biology subjects

Li, Y.

Publications and source records attributed to Li, Y..

At least 271 records · Page 15Linked to original sources

bioRxiv2026

Metabolic control of smooth muscle cell phenotype switching in atherosclerosis

The loss of smooth muscle cell (SMC) contractile phenotype contributes to various diseases including atherosclerosis. However, its metabolic basis is not entirely elucidated. Since the transforming growth factor beta (TGF{beta}) signaling is among principal regulators of SMC contractility, we studied metabolic regulation of TGF{beta} signaling in SMCs in vitro and atherosclerotic mouse models and human lesions. We found that TGF{beta} induced Ac-CoA synthetase 2 (ACSS2)-dependent Ac-CoA production, by suppressing pyruvate dehydrogenase kinase 4 (PDK4). This stabilized R-SMADs and TGF{beta} receptor 1, preserving SMC contractile phenotype. SMC-specific PDK4 knockout mimicked the effect of TGF{beta} signaling both metabolically and phenotypically, increasing glucose-derived synthesis of Ac-CoA and SMC contractile phenotype. SMC-specific Pdk4 knockout in ApoE knockout mice reduced atherosclerosis. Furthermore, human specimens demonstrated a strong correlation between PDK4 level and atherosclerosis severity. These findings indicate that continuous TGF{beta} signaling, critical to the maintenance of the normal SMC contractile state and is regulated by PDK4 and carbohydrate metabolism. TeaserReducing PDK4 metabolically restricts aortic plaque growth via TGF{beta}-dependent SMC contractility.

cell biology↗

bioRxiv2026

MTDH-SND1 disruption sensitizes ovarian cancer to ferroptosis and PARP inhibition

BRCA-deficient high-grade serous ovarian cancer is characterized by profound genomic instability and elevated replication-associated DNA damage, rendering these tumors initially sensitive to platinum-based chemotherapy and PARP inhibition. However, despite this vulnerability, most patients ultimately develop resistance, underscoring the need for therapeutic strategies that extend beyond DNA repair-targeted mechanisms. Here, we introduce the MTDH-SND1 complex as a complementary therapeutic target that may expose additional stress vulnerabilities in ovarian cancer cells. We show that pharmacological disruption of the MTDH-SND1 interaction using C26-A6 increases susceptibility to ferroptosis-associated stress, an iron-dependent form of regulated cell death and that BRCA-deficient models are particularly more sensitive to this perturbation. Notably, when combined with PARP inhibition, MTDH-SND1 disruption is associated with increased MHC class I expression in tumor cells, suggesting enhanced tumor visibility to the immune system. Together, these findings support a combination strategy that couples DNA repair disruption with metabolic and immunogenic remodeling in BRCA-deficient ovarian cancer.

cancer biology↗

bioRxiv2026

Gut Microbiota-derived Acetate Safeguards the Colonic Epithelial Acetyl-CoA Reserve to Avert Colonic Senescence

Aging is a major driver of tissue senescence, but little is known about the development of age-unrelated tissue senescence. Here we show that nucleocytosolic acetyl-CoA deficiency in colon epithelial cells, caused by Acly ablation and bacterial depletion, triggers age-independent, p53-dependent colonic senescence leading to severe systemic inflammation. Acetate supplementation, acetate-producing bacterial transplantation, targeted depletion of senescent cells or treatment with lysine deacetylase inhibitors blocks colonic senescence and inflammatory injury. Spontaneous colonic senescence develops following simultaneous deletion of epithelial Acly and Acss2, confirming that microbe-derived acetate maintains the epithelial acetyl-CoA pool via ACSS2 to avert colonic senescence. Mechanistically, acetyl-CoA deficiency deprives a cohort of mitochondrial and nuclear proteins of acetylation, leading to increased oxidative stress and DNA repair stress that trigger cellular senescence. Among these proteins, ATP5F1A-K161 acetylation and H4-K5/8 acetylation are required to protect colonic epithelial cells from developing senescence. These observations unveil a previously unknown mechanism that governs colonic senescence.

physiology↗

bioRxiv2026

A multifunctional polyketide synthase in nematodes produces divergent families of signaling molecules that control different developmental arrests

To improve their chances of reproductive success, nematodes not only must arrest their development in response to adverse growth conditions but also must quickly recover if conditions improve. A polyketide synthase (PKS)-nonribosomal peptide synthetase (NRPS) hybrid assembly line that is expressed in the canal-associated neurons (CANs) of Caenorhabditis elegans promotes recovery from starvation-induced larval arrest. Here, we show that in the predatory nematode Pristionchus pacificus this assembly line produces a suite of secondary metabolites, including a family of hybrid polyketide-nonribosomal peptides known as the nemamides, the related nemaketides, and a family of ascarylose-modified polyketides named ascarenes. Depending on the starter unit that is loaded onto the PKS, the assembly line can produce dramatically different downstream products. Whereas the nemamides promote recovery from starvation-induced larval arrest, the ascarenes inhibit development of the dauer larval stage and promote recovery. This dichotomy suggests that the PKS-NRPS megasynthetase serves as a signaling hub in the CANs, controlling multiple developmental events. The PKS-NRPS assembly line is highly conserved across many nematode species, and identification of these chemical signals will help to elucidate the signaling pathways that control development in the worm and lead to novel anthelmintics.

biochemistry↗

bioRxiv2026

GLP-1 analogs restore inflammatory, mitochondrial and intercellular signaling networks in the SncaG51D/G51D knock-in mouse model of Parkinson's disease

Parkinsons disease (PD) is a neurodegenerative disorder characterized by a prolonged prodromal stage that culminates in motor deficits. Current PD therapies primarily alleviate symptoms, underscoring the need for disease-modifying strategies. Glucagon-like peptide-1 (GLP-1) analogs showed early promise as candidate disease modifiers, but recent clinical results have been inconsistent, and their mechanism of action remains poorly defined. Here, we employed our SncaG51D/G51D knock-in mouse model to investigate the effects of subcutaneously administered GLP-1 analogs, semaglutide and lixisenatide. Both analogs reversed motor and non-motor deficits and reduced gliosis and detergent-insoluble -synuclein. Bulk and single-nuclei transcriptomics together with CellChat-based intercellular communication analysis revealed that GLP-1 analogs normalize early striatal mitochondrial and inflammatory dysregulation and restore neuregulin (NRG) and neurexin (NRXN) signaling networks to wild-type levels. Treatment was effective when initiated either before or shortly after symptom onset, defining an early therapeutic window for GLP-1 analog therapy in PD.

neuroscience↗

bioRxiv2026

SUITPy: A Python-based toolbox for the analysis of cerebellar functional and anatomical imaging data across the human lifespan

The human cerebellum plays a central role in motor, emotional, and cognitive functions, and is implicated in many brain disorders. To improve the analysis of functional and anatomical imaging from the cerebellum, we introduce SUITPy, an improved and fully revised Python implementation of the widely used SUIT toolbox. For this new version, we developed a U-Net based model to automatically isolate the cerebellum from adjacent cortical tissue, which achieves higher fidelity than existing algorithms. The isolation works robustly without manual corrections for imaging data across the lifespan. We show that isolation and subsequent normalization to a cerebellum-only template lead to a more precise alignment of cerebellar structures across participants compared to normalization using a whole-brain template. We also show the utility of the cerebellar mask to prevent contamination of cerebellar functional data from surrounding cortical structures. The toolbox also provides functionality for visualizing cerebellar data on a flatmap, along with a range of anatomical and functional cerebellar atlases, thereby offering an essential tool that enables accurate cerebellar analysis across the lifespan.

neuroscience↗

bioRxiv2026

Time Restricted Feeding Mitigates High-Fat-Diet Induced Sleep Disruption and Amplifies NREM Substates

The effects of diet quality and timing on sleep quality remain poorly understood, particularly at the level of sleep microarchitecture. Traditional visual scoring captures only coarse sleep stages, overlooking the marked heterogeneity of electroencephalographic (EEG) patterns in non-rapid eye movement (NREM) sleep of mice. Here, we apply a pipeline that combines EEG feature extraction with unsupervised machine-learning-based clustering to resolve discrete NREM substates and ask how a high-fat diet (HFD) and time- restricted feeding (TRF) affect sleep microarchitectures. HFD increases sleep latency and sleep fragmentation; both abnormalities were ameliorated by active phase TRF. Clustering of 10s epochs identified two high-amplitude NREM substates sensitive to TRF: Cluster 1, enriched in low-delta power and peaking early in the light phase (ZT 0-6), consistent with canonical slow-wave sleep, and Cluster 6, characterized by elevated alpha, sigma, and beta power and peaking in the latter half of the light phase (ZT6-12). TRF increases the frequency of both NREM substates, particularly within longer uninterrupted sleep episodes during the light phases. These findings introduce an objective framework for quantifying murine sleep microarchitecture and show that aligning caloric intake with the circadian active window mitigates HFD-induced macro-level sleep disruption while selectively enhancing two physiologically distinct NREM substates. Significance StatementTime-restricted eating - targeting food intake to a defined window during the circadian active phase - confers well-established metabolic benefits, but its impact on sleep is largely underexplored. Using continuous EEG/EMG recordings, we show that an active- phase eating window mitigates high-fat-diet-induced sleep disruption in mice. We employed a novel machine-learning pipeline, further revealing that timed eating selectively increases distinct NREM substates, demonstrating that "when we eat" fine-tunes the macro- and microarchitecture of sleep. These insights lay the foundation for future translational studies and clinical trials aimed at harnessing timed eating to enhance both metabolic and sleep health.

animal behavior and cognition↗

bioRxiv2026

Type III Druantia two-component antiphage defense depends on the DruH-DruE interaction for halting phage DNA cyclization and replication

Bacteria have evolved multiple immune systems to resist phage invasion, however, only a small part of the defensive mechanisms have been clearly uncovered. In this study, we report a type III Druantia two-component defense system, DruH-DruE, identified from Mycobacterium smegmatis. The DruH-DruE prevents phage DNA cyclization and replication.DruE can be replaced from the defense system by either homolog in M. tuberculosis or M. smegmatis. The physical interaction between this two components is essential for fighting against phage infection. Mutations in the interaction sites led to the loss of phage-defending function of the system. The broad-spectrum antiphage ability of the defense system could be activated by the small tail protein Gp25 of phage A10ZJ24. This study fills a major gap in current knowledge of antiphage mechanism of type III Druantia defense system, expanding our understanding of the immune mechanisms in prokaryotic cells.

microbiology↗

bioRxiv2026

cfMIND: A read-level methylation framework for accurate non- invasive disease detection using cell-free DNA

Plasma cell-free DNA (cfDNA) emerged as a promising non-invasive biomarker for cancers. However, reliable detection remains challenging due to the low abundance of tumor-derived cfDNA fragments and the dilution of informative methylation signals when aggregated into region-level features. Here, we propose a novel approach cfMIND, an efficient and robust machine-learning framework that leverages stratified read-level methylation signals to preserve rare cell-type-specific information and enhance detection sensitivity. By avoiding information loss inherent to conventional aggregation strategies, cfMIND enables more accurate and stable inference across diverse conditions. cfMIND is compatible with various cfDNA methylation sequencing technologies and cancer types. Across multiple cancer datasets (n = 868), cfMIND achieves high performance (AUROC up to 0.966) and maintains strong accuracy even at ultra-low sequencing depth (0.2x) and in early-stage cancers. Notably, cfMIND demonstrates exceptional robustness, generalizing effectively across cohorts and platforms without the need for model retraining. These results highlight its potential utility in heterogeneous experimental and clinically relevant settings. Beyond cancer detection, cfMIND is readily extendable to non-malignant diseases, as demonstrated by its ability to capture disease-associated methylation alterations in amyotrophic lateral sclerosis (ALS). Functional investigations on cfMIND-identified features further reveal enrichment in key regulatory regions implicated in disease pathogenesis and recapitulate tissue- and single-cell-level methylation and transcriptional programs underlying tumor biology. Collectively, cfMIND represents a significant advancement in the field, offering a broadly applicable, functionally interpretable, and high-resolution framework for non-invasive disease detection.

bioinformatics↗

bioRxiv2026

Investigating the Dynamic Relationship Between Anxiety and Spatial Memory Using Autonomous Ecological Momentary Assessment

Anxiety has been extensively studied in relation to memory, yet its dynamic association with spatial episodic memory in naturalistic clinical settings remains largely unexplored. We developed an anxiety-spatial-memory EMA protocol (asm-EMA) and deployed it in 30 epilepsy patients undergoing inpatient EEG monitoring, delivering combined momentary anxiety ratings and a validated spatial memory task pseudo-randomly every 90-150 minutes across multiple days. Subject-level asm-EMA means and session-to-session variability both correlated significantly with standard neuropsychological assessments, supporting the clinical validity of our design. Elevated within-person STAI-6 was selectively associated with faster retrieval responses, yet spatial memory accuracy was independent of all three anxiety measures, suggesting a shift in response strategy rather than memory impairment. Within-day anxiety showed short-term carryover between consecutive sessions, with little persistence beyond the next session. The asm-EMA protocol provides a feasible, autonomous framework for capturing moment-to-moment anxiety-memory dynamics in naturalistic settings.

neuroscience↗

bioRxiv2026

CatIF-RL: Activity-Oriented Enzyme Sequence Design by Steered Inverse Protein Folding

Protein inverse folding models are designed to generate amino acid sequences compatible with a given backbone structure, but they are not explicitly optimized for specific biological functions. Here, we present CatIF-RL, a framework that steers a graph-based denoising diffusion inverse folding model toward designing enzyme variants with enhanced catalytic activity. CatIF-RL first adapts the inverse folding model to enzyme structural data, then introduces activity-oriented preference signals using predicted catalytic constant (kcat) as the optimization objective, enabling specialization through generative dataset curation and group-relative policy optimization (GRPO). This process iteratively shifts the sequence distribution toward higher predicted kcat while constraining sequence divergence to sequences that remain compatible with the input structure. On the independent benchmark, CatIF-RL achieves an approximately four-fold increase in predicted kcat relative to native enzymes, substantially outperforming representative inverse folding methods, while maintaining sequence recovery (0.55) and structural fidelity, and supporting motif-preserving partial sequence design. CatIF-RL establishes a practical framework for activity-oriented enzyme design and provides a generalizable strategy for steering structure-conditioned protein generation toward functional optimization.

bioinformatics↗

bioRxiv2026

Intranuclear polyglycine aggregation drives neurodegeneration through epigenetic repression of chromatin accessibility and transcription

Polyglycine (polyG) proteins translated from expanded GGC trinucleotide repeats are implicated in a growing group of neuromuscular degenerative disorders characterized by intranuclear inclusions, yet the pathogenic importance of aggregate localization and the mechanisms underlying polyG-induced neurodegeneration remain unclear. Here we show that intranuclear polyG aggregates are markedly more pathogenic than cytoplasmic aggregates in cellular and mouse models. Intranuclear aggregation causes greater cell death, more severe behavioral deficits and neuropathology, and earlier mortality. Mechanistically, intranuclear polyG aggregates impair nascent RNA synthesis and are associated with a transcriptionally repressive chromatin state marked by globally reduced chromatin accessibility, decreased H3K27 acetylation, and increased HDAC3 expression across cellular, mouse, and human disease tissue. Using a light-inducible system, we further show that this transcriptional impairment depends on insoluble intranuclear aggregate formation rather than diffuse polyG alone. Pharmacological HDAC inhibition partially restores histone acetylation and transcriptional output and ameliorates behavioral and pathological abnormalities. Together, these findings identify intranuclear polyG aggregates as the more pathologically relevant species and uncover epigenetic repression of chromatin accessibility and transcription as a potentially common mechanism underlying polyG diseases.

neuroscience↗

bioRxiv2026

Spatiotemporal analysis of autism gene enrichment implicates cortex, thalamus, and hypothalamus

Autism spectrum disorder (ASD) is a highly heritable neurodevelopmental disorder. Sequencing analyses have identified 185 ASD-associated genes, which implicate neurons, but the specific brain regions through which these neurons influence neurodevelopment remain unclear. Here, we integrate over one million single-cell RNA sequencing profiles from 20 regions of the developing human brain (4-23 post-conceptual weeks) using a new framework, STARMAPS (Sparse Task-specific Analysis for Revealing Molecular Associations in Particular Single-cell datasets). STARMAPS accounts for coordinated regional and developmental perturbations in gene expression, enabling robust cross-region comparison. We replicate prior findings that ASD-associated genes are enriched in excitatory neurons during mid-fetal development, and we extend these results to reveal distinct spatial signatures. Across 26 excitatory neuron subtypes, six clusters showed significant enrichment for ASD-associated genes. These clusters localize to both cortical and subcortical regions, including the motor, temporal, and visual cortex, as well as the thalamus and hypothalamus. Our findings support a major role for excitatory neurons across distributed brain circuits, implicating previously underappreciated subcortical structures in ASD etiology. By providing a statistically rigorous framework for spatiotemporal integration of single-cell data, STARMAPS enables refined mapping of molecular vulnerability across the developing human brain.

genomics↗

bioRxiv2026

Hypertension is related to a slower radiotracer removal from lateral ventricles

BackgroundImpairment of brain waste removal contributes to Alzheimers disease etiology and progression. Although hypertension is a risk factor for dementia, little is known about how it affects measures of clearance in human brain. MethodsCross-sectional (n=159) and longitudinal (n=94) analysis of the relationship between blood pressure (BP) and brain clearance. The estimate of brain clearance was measured using positron emission tomography (PET) as the rate of radiotracer (MK-6240) efflux from the lateral ventricles in the 10-30-minute window after tracer injection. We also examined cerebral blood flow, PET-derived tau deposition in the medial temporal lobe, cognition and plasma biomarkers of neurodegeneration. At baseline we compared participants with (n=88) and without (n=71) hypertension. For longitudinal analyses we defined two groups based on systolic BP trajectories from baseline to follow-up: as long-term controlled (n=76) or uncontrolled BP (n=18). ResultsAt baseline, subjects with hypertension had lower ventricular clearance than normotensive controls (Cohens d=0.53, p=0.001). Over the course of the observation period (median 1.85 years) subjects in the uncontrolled BP group experienced a steeper reduction in clearance rates ({beta}=-5.88) than subjects in the controlled BP group ({beta}=-0.81, interaction p=0.039). ConclusionsOur study suggests that hypertension impairs brain clearance of fluids.

neuroscience↗

bioRxiv2026

A CAMKK2-UBR4-19S Proteasome Axis Regulates Chondrocyte Proteostasis and SOX9 Stability

ObjectiveInvestigate how Ca2+/calmodulin dependent protein kinase kinase 2 (CaMKK2) orchestrates a catabolic shift in chondrocytes during early osteoarthritis (OA). MethodsCartilage, osteochondral plugs and chondrocytes were collected from patients undergoing total hip arthroplasty or non-OA donors. SOX9 levels were assessed via immunoblotting or immunohistochemistry (IHC). Sox9 levels were also assessed by IHC in knee joints from wild-type (WT) and Camkk2-/- mice that underwent sham or destabilization of medial meniscus (DMM), with or without STO-609 (0.033 mg/kg) treatment. Co-immunoprecipitation followed by mass spectrometry was performed to identify CaMKK2 interacting proteins in chondrocytes. Kinase assays were analyzed by immunoblotting and phosphosites identified by mass spectrometry. Proteasome function was assessed in murine and human chondrocytes lacking or expressing kinase-active or kinase-inactive CaMKK2. ResultsInhibition or loss of CaMKK2 increased SOX9, whereas the expression of kinase-active, not inactive, CaMKK2 reduced Sox9 in human and mouse OA cartilage. Proteomic analysis of CaMKK2 immunoprecipitates revealed the presence of ubiquitin E3 ligase Ubr4 and the 19S proteasome regulatory particle (RP). CaMKK2 kinase activity was dispensable for its interactions with Ubr4, 19S RP, and Sox9-ubiquitin conjugates, and kinase-inactive CaMKK2 attenuated Sox9 degradation in chondrocytes. Further, CaMKK2 phosphorylated the 19S RP ATPase Psmc5 on Ser136, and an intact kinase increased proteasome activity in chondrocytes. ConclusionsOur findings identify CaMKK2 as a dual-function regulator of chondrocyte UPS with a scaffolding role to assemble UPSUbr4-19S RP around polyubiquitinated proteins such as Sox9, and a catalytic role to enhance proteasome function, potentially through Psmc5 phosphorylation, thereby linking chondrocyte inflammatory signaling to Sox9 degradation and cartilage degeneration.

cell biology↗

bioRxiv2026

Aging impairs control of center of mass during repeated visual perturbations in walking

Sensory degradation with aging can impair balance control, partly by disrupting visual contributions to self-motion estimation. We investigated age-related differences in center of mass (CoM) trajectories controlling in the frontal plane during exposure to repeated visual perturbations in walking. We hypothesized that the older adults would show larger responses to visual perturbations and less adaptation to repeated visual perturbation exposure. We applied three visual perturbations to 14 healthy older (age: 75.0{+/-}2.4) and 16 younger adults (age: 23.4{+/-}3.9) walking on a treadmill: fixating a stationary target with the background moving to the right (MB), tracking a target moving rightward over a stationary background with head rotation (MT-HR), and tracking a moving target with eye movement only (MT-EM). CoM position deviations due to the visual perturbations were assessed. Over the whole trial, the older adults did not show larger CoM position variability than the younger during visual perturbation walking, which was contrary to our hypothesis. During visual perturbation epochs, both age groups deviated in the same direction except MB. In MB, the older adults deviated to an opposite direction after a few perturbation repetitions. Moreover, in MT-HR and MT-EM, the older adults deviated earlier than the younger adults and they deviated more in the MT-HR condition. This indicates that older adults exhibit reduced ability to accurately estimate self-motion through correction by other sensory modalities when exposed to visual perturbations. Over repeated perturbations, the older adults showed decreased CoM deviations in MT-EM, which suggests that they still maintain the capacity to downweight visual information after repeated exposure.

neuroscience↗

bioRxiv2026

An generative-AI framework for target-Specific MicroRNAs towards RNAi-based drug design

MicroRNA (miRNAs) are small non-coding RNAs that regulate gene expression by binding to the target messenger RNA (mRNA), whose versatility has inspired RNA-interference (RNAi)-based drug designs. However, off-target effects lead to unintended gene silencing and toxicity. Existing methods suffer from experimental data scarcity and fail to effectively integrate target specificity into designing de novo small interference RNAs (siRNA). To overcome the above challenges, we present SO_SCPLOWPECIC_SCPLOWMO_SCPLOWIC_SCPLOWR, a specificity-guided generative framework that synthesizes target-conditioned miRNAs. By training on a large experimental data containing 2.2M miRNA-mRNA pairs, SO_SCPLOWPECIC_SCPLOWMO_SCPLOWIC_SCPLOWR minimizes off-target effects with enhanced on-target potency. As a result, SO_SCPLOWPECIC_SCPLOWMO_SCPLOWIC_SCPLOWR-generated miRNAs bind more strongly to the target mRNAs than the observed miRNAs and much less so to off-target mRNAs. We tested SO_SCPLOWPECIC_SCPLOWMO_SCPLOWIC_SCPLOWR on mRNA targets for liver disease, for which 6 FDA-approved siRNA-based drugs were available. SO_SCPLOWPECIC_SCPLOWMO_SCPLOWIC_SCPLOWR recovers binding regions that correspond to FDA-approved siRNA drugs across 3 targets, and demonstrates greater structural specificity for on-target mRNAs than for off-target mRNAs. Together, SO_SCPLOWPECIC_SCPLOWMO_SCPLOWIC_SCPLOWR offers an AI solution to synthesize miRNA-inspired and target-specific siRNA sequences towards RNAi-based drug design. Code/data is available at https://anonymous.4open.science/r/SpeciMiR-278F

genomics↗

bioRxiv2026

ME1 Programs Latent Effector Capacity and Grounds a Mathematical Model of Reversible T Cell Exhaustion

A central paradox in cancer immunotherapy is that patients harboring apparently exhausted or dysfunctional CD8+ T cells can exhibit rapid and durable responses to immune checkpoint inhibition (ICI). Although such rebound responses do not occur in all ICI-treated patients, resolving this paradox is essential for identifying and therapeutically maximizing reversible T cell exhaustion before clinical benefit is achieved. Here, we identify malic enzyme 1 (ME1) as a key molecular determinant of a latent, epigenetically poised effector CD8+ T cell state that is both necessary and sufficient for responsiveness to ICI. Genetic loss of ME1 abolishes therapeutic efficacy despite intact checkpoint blockade, whereas enforced ME1 expression enables robust antitumor responses even in otherwise ICI-resistant tumor models. Chromatin accessibility profiling, together with bulk and single-cell transcriptomic analyses, demonstrates that ME1 preserves effector readiness by maintaining latent effector capacity as a measurable biological state. These findings also experimentally ground a mechanistically interpretable dynamical model of T cell exhaustion, in which latent effector capacity, E(t), evolves according to a first-order differential equation shaped by activation history and decay. By defining latent effector capacity as a quantifiable, history-dependent state variable, our findings further enable the rational design of AI-driven models to predict patient-specific responsiveness and optimize therapeutic strategies in T cell-based immunotherapy. Together, our results redefine exhaustion as a state of masked potential governed by history-dependent latent effector capacity and provide a unified framework explaining rapid functional rebound, therapeutic heterogeneity, and the boundary between reversible and irreversible T cell exhaustion.

immunology↗