Biology subjects

Cohen, S.

Publications and source records attributed to Cohen, S..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

DoggifAI: a transformer based approach for antibodycaninisation

Antibody translation across species offers a compelling strategy to extend the vast and expensive investments in human therapeutic antibodies to veterinary oncology, with applications in both veterinary medicine and comparative oncology. While precise, low-immunogenic treatments are essential for canine cancer care, traditional species conversion methods rely on ad hoc bioinformatics modifications. These methods often implicitly decouple the framework (FR) and complementarity-determining regions (CDRs), ignoring how structural changes in FRs can affect the conformation and function of CDRs. This can compromise binding specificity and require costly high-throughput in vitro screening. To address this, we present DoggifAI, a transformer model that translates non-canine antibody sequences into canine ones by generating species-appropriate framework regions (FRs) based on desired CDRs. This allows the model to better preserve structural compatibility between FRs and CDRs. The model is pretrained in a T5-style text-to-text denoising task on a large multispecies antibody dataset, which allows further finetuning on a much smaller species-specific dataset. DoggifAI generates highly canine-like antibodies and shows promising results in preserving binding specificity. To support further progress in this field, we also release a curated dataset of over 430,000 unique canine antibody chain sequences, significantly expanding the public sequence repertoire. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=127 SRC="FIGDIR/small/656573v1_ufig1.gif" ALT="Figure 1"> View larger version (27K): org.highwire.dtl.DTLVardef@13c10adorg.highwire.dtl.DTLVardef@6afca6org.highwire.dtl.DTLVardef@1f11961org.highwire.dtl.DTLVardef@1b9563a_HPS_FORMAT_FIGEXP M_FIG Graphical Abstract C_FIG HighlightsO_LIWe show that transformer models are capable of generating plausible antibody framework regions based on CDRs C_LIO_LIWe show that resulting framework regions are highly recognisable as coming from the desired species C_LIO_LIWe show promising results for the retention of binding specificity when translating antibody sequences in this way C_LIO_LIWe release a large, high-quality dataset of canine antibody sequences to support future research C_LI

bioinformatics↗

bioRxiv2025

Sexually distinct multi omic responses to progressive endurance exercise training in the rat lung. Findings from MoTrPAC

Despite the lungs being essential for ventilation and aerobic exercise capacity, conventionally the lungs are not thought to adapt to exercise training. Endurance exercise is key to pulmonary rehabilitation programs, which also displays sex-specific differences in therapeutic efficacy. Given the molecular underpinnings of sex-specific lung adaptations to endurance exercise are uncharacterized, we used a multi-omics approach to study sex differences in the lungs of 6-month-old Fischer 344 rats in response to an 8 week progressive endurance treadmill training protocol. This was accomplished by reannotating publicly accessible data from the Molecular Transducers of Physical Activity Consortium (MoTrPAC) and integrating newly-analyzed acetylome data to assess multi-omic sex differences in sedentary and progressively trained states. Female rats displayed enrichment in immune-related features and pathways at the transcriptome and proteome level that were maintained with training. Conversely, in the male rat lung there was an overall decrease in immune pathways following 8 weeks of training. Sexually conserved responses to training included increased enrichment in transcriptomic pathways related to type I alveoli and proteomic pathways related to cilia, and decreased mitochondrial protein acetylation. In both sexes, features known to be enriched in lung diseases were attenuated with training. Together our findings provide novel insight into sex specific responses to endurance exercise training in the rat lung and may offer translational insight into sex-specific differences in lung disease pathogenesis and treatment.

physiology↗

bioRxiv2025

ANALYSIS OF THE IMPACT OF FIVE METHANE MITIGATING FEED ADDITIVES ON MILK PRODUCTION AND ASSOCIATED PARAMETERS ACROSS MULTIPLE COMMERCIAL DAIRY FARMS

The effect of methane-mitigating feed additives on dairy cows has been widely explored; however, confusing conclusions have been reached due to factors such as the inclusion of different doses and experimental conditions in which the additives are tested, or even a small sample size. We present the first extensive study assessing the effects of methane-mitigating feed additives on milk production across several commercial dairy farms. This study used a previously developed predictive AI-driven model based on microbiome samples; the model predicts farms where a significant reduction of methane emissions is expected due to the applied feed additives. Thus, in this study, each feed additive was supplied to a large number of farms, widely distributed across different climatic areas in Israel. The data analysis followed two simulated scenarios: (1) a naive approach, where feed additives are supplied indiscriminately, and (2) an optimized approach, where feed additives are supplied only to farms with a high likelihood of being positively impacted in terms of reduced enteric methane emissions (50% of the farms). The results show that each feed additive significantly increased milk production compared to the control groups. This increase in milk production was significantly higher in the optimized scenario. Other related parameters such as somatic cells were also improved. Our results suggest that the feed additives positively affect milk production, reaching a maximum expression when the AI-driven model is applied.

zoology↗

bioRxiv2025

Integrated single-cell analyses of affinity-tested B-cells enable the identification of a gene signature to predict antibody affinity.

Advancements in single-cell technologies and deep sequencing have revealed the vast B-cell repertoire arising from immunisation, enhancing the number of antibodies available for testing. However, selecting the highest affinity antibodies from many sequences is not a straightforward feat, as mechanisms sustaining affinity maturation and related markers remain under-studied. Here, we generated datasets of antigen-specific B-cells after mouse immunisation as well as re-analysed public data to identify a novel transcriptomic signature, "High Signature" (HS), with predictive power for high-affinity antibodies. HS, derived by integrating antibody sequences, gene expression, and affinity measurements, enabled sub-nanomolar affinity antibody selection without sequence pre-analysis. Notably, HS-expressing B-cells were 2.5 times more likely to yield high-affinity antibodies than randomly picked cells. Mechanistically, we identified RUVBL2, an AAA+ ATPase, as a primary HS modulator, suggesting its involvement in affinity maturation. Furthermore, HS applied to human PBMC data enriched high-affinity antibody expression, underscoring its potential in antibody discovery.

immunology↗

bioRxiv2024

APOE4 alters the lipid droplet proteome and modulates droplet dynamics

Excess lipid droplet (LD) accumulation is associated with several pathological states, including Alzheimers disease (AD). However, the mechanism(s) by which changes in LD composition and dynamics contribute to pathophysiology of these disorders remains unclear. Apolipoprotein E (ApoE) is a droplet associated protein with a common risk variant (E4) that confers the largest increase in genetic risk for late-onset AD. E4 is associated with both increased neuroinflammation and excess LD accumulation. In the current study, we sought to quantitatively profile the lipid and protein composition of LDs between the neutral E3 and risk variant E4, to gain insight into potential LD-driven contributions to AD pathogenesis. Targeted replacement mice expressing human E3 or E4 were injected with saline or lipopolysaccharide (LPS), and after 24 hours, hepatic lipid droplets were isolated for proteomic and lipidomic analyses. Lipidomics revealed a shift in the distribution of glycerophospholipids in E4 LDs with a concomitant increase in phosphatidylcholine species, and overall, the baseline profile of E4 LDs resembled that of the LPS-treated groups. Quantitative proteomics showed that LDs from E4 mice are enriched for proteins involved in protein/vesicle transport but have decreased levels of proteins involved in fatty acid {beta}-oxidation. Interestingly, proteins associated with LDs showed substantial overlap with previously published lists of AD postmortem tissue and microglia omics studies, suggesting a potential role for LDs in modulating AD risk or progression. Given this, we exposed primary microglia from the same E3 or E4 mice to exogenous lipid, inflammatory stimulation, necroptotic N2A cells (nN2A), or a combination of treatments to evaluate LD formation and its impact on the cells immune state. Microglia from E4 mice accumulated more LDs in every condition tested - at baseline and following addition of fatty acids, LPS stimulation, or nN2As. E4 microglia also secreted significantly more cytokines (TNF, IL-1{beta}, IL-10) than E3 microglia in the control, oleic acid, and nN2A treatment conditions, yet showed a blunted response to LPS. In sum, these results suggest that E4 microglia accumulate more LDs compared to E3 microglia and that E4 is associated with a basal LD composition that resembles a pro-inflammatory cell. Together with the high overlap of the LD proteome with established AD-associated datasets, these data further support the idea that alterations in LD dynamics, particularly within microglia, may contribute to the increased risk for AD associated with APOE4.

physiology↗

bioRxiv2024

Potent antitumor activity of a designed interleukin-21 mimic

Long-standing goals of cancer immunotherapy are to activate cytotoxic antitumor T cells across a broad range of affinities while dampening suppressive regulatory T (Treg) cell responses, but current approaches achieve these goals with limited success. Here, we report a de novo IL-21 mimic, 21h10, designed to have augmented stability and high signaling potency in both humans and mice. In multiple animal models and in ex vivo human melanoma patient derived organotypic tumor spheroids (PDOTS), 21h10 showed robust antitumor activity. 21h10 generates significantly prolonged STAT signaling in vivo compared with native IL-21, and has considerably stronger anti-tumor activity. Toxicities associated with systemic administration of 21h10 could be mitigated by TNF blockade without compromising antitumor efficacy. In the tumor microenvironment, 21h10 induced highly cytotoxic antitumor T cells from clonotypes with a range of affinities for endogenous tumor antigens, robustly expanding low-affinity cytotoxic T cells and driving high expression of interferon- (IFN-) and granzyme B compared to native IL-21, while increasing the frequency of IFN-+ Th1 cells and reducing that of Foxp3+ Tregs. As 21h10 has full human/mouse cross-reactivity, high stability and potency, and potentiates low-affinity antitumor responses, it has considerable translational potential.

biochemistry↗

bioRxiv2024

Neurons and astrocytes have distinct organelle signatures and responses to stress

Neurons and astrocytes play critical yet divergent roles in brain physiology and neurological conditions. Intracellular organelles are integral to cellular function. However, an in-depth characterization of organelles in live brain cells has not been performed. Here, we used multispectral imaging to simultaneously visualize six organelles - endoplasmic reticulum (ER), lysosomes, mitochondria, peroxisomes, Golgi, and lipid droplets - in live primary rodent neurons and astrocytes. We generated a dataset of 173 Z-stack and 99 time-lapse images accompanied by quantitative analysis of 1418 metrics (the "organelle signature"). Comparative analysis revealed clear cell-type specificity in organelle morphology and interactions. Neurons were characterized by prominent mitochondrial composition and interactions, while astrocytes contained more lysosomes and lipid droplet interactions. Additionally, neurons displayed a more robust organelle response than astrocytes to acute oxidative or ER stress. Our data provide a systems-level characterization of neuron and astrocyte organelles that can be a reference for understanding cell- type-specific physiology and disease.

cell biology↗

bioRxiv2024

Endogenous neuronal DNA double-strand breaks are not sufficient to drive brain aging and neurodegeneration

Loss of genomic information due to the accumulation of somatic DNA damage has been implicated in aging and neurodegeneration1-3. Somatic mutations in human neurons increase with age4, but it is unclear whether this is a cause or a consequence of brain aging. Here, we clarify the role of endogenous, neuronal DNA double-strand breaks (DSBs) in brain aging and neurodegeneration by generating mice with post-developmental inactivation of the classical non-homologous end-joining (C-NHEJ) core factor Xrcc4 in forebrain neurons. Xrcc4 is critical for the ligation step of C-NHEJ and has no known function outside of DSB repair5,6. We find that, unlike their wild-type counterparts, C-NHEJ-deficient neurons accumulate high levels of DSB foci with age, indicating that neurons undergo frequent DSBs that are typically efficiently repaired by C-NHEJ across their lifespan. Genome-wide mapping reveals that endogenous neuronal DSBs preferentially occur in promoter regions and other genic features. Analysis of 3-D genome organization shows intra-chromosomal clustering and loop extrusion of neuronal DSB regions. Strikingly, however, DSB accumulation caused by loss of C-NHEJ induces only minor epigenetic alterations and does not significantly affect gene expression, 3-D genome organization, or mutational outcomes at neuronal DSBs. Despite extensive aging-associated accumulation of neuronal DSBs, mice with neuronal Xrcc4 inactivation do not show neurodegeneration, neuroinflammation, reduced lifespan, or impaired memory and learning behavior. We conclude that the formation of spontaneous neuronal DSBs caused by normal cellular processes is insufficient to cause brain aging and neurodegeneration, even in the absence of C-NHEJ, the principal neuronal DSB repair pathway.

neuroscience↗

bioRxiv2024

Dynamic convergence of autism disorder risk genes across neurodevelopment

Over three hundred and seventy-three risk genes, broadly enriched for roles in neuronal communication and gene expression regulation, underlie risk for autism spectrum disorder (ASD) and developmental delay (DD). Functional genomic studies of subsets of these genes consistently indicate a convergent role in neurogenesis, but how these diverse risk genes converge on a smaller number of biological pathways in mature neurons is unclear. To uncover shared downstream impacts between neurodevelopmental disorder (NDD) risk genes, here we apply a pooled CRISPR approach to contrast the transcriptomic impacts of targeting 29 NDD loss-of-function genes across human induced pluripotent stem cell (hiPSC)-derived neural progenitor cells, glutamatergic neurons, and GABAergic neurons. Points of convergence vary between the cell types of the brain and are greatest in mature glutamatergic neurons, where they broadly target not just synaptic and epigenetic, but unexpectedly, mitochondrial biology. The strongest convergent networks occur between NDD genes with common co-expression patterns in the post-mortem brain, biological annotations, and clinical associations, suggesting that convergence may one-day inform patient stratification and treatment. Towards this, ten out of eleven drugs tested that were predicted to reverse convergent signatures in human cells and/or arousal and sensory processing behaviors in zebrafish ameliorated at least one behavioral phenotype in vivo. Altogether, robust convergence in post-mitotic neurons represents a clinically actionable therapeutic window.

genetics↗

bioRxiv2024

Dynamic stress- and inflammatory-based regulation of psychiatric risk loci in human neurons

The immune environment influences neurodevelopment and subsequent clinical trajectories for psychiatric outcomes in childhood and adolescence. Yet it remains unclear if the impact of maternal and fetal immune activation varies with distinct polygenic risk profiles. Therefore, here we catalogue genotype and environment (GxE) interactions, contrasting allele-specific regulatory activity between inflammatory cues. We report a cue-specific neuronal massively parallel reporter assay (MPRA) of 152 loci from genome-wide association study (GWAS) linked to ten brain traits/disorders, empirically dissecting the impact of interleukin-6 (IL-6) and interferon-alpha (IFN) on transcriptional activity. In human induced pluripotent stem cell (hiPSC)-derived glutamatergic neurons, 1,156 active candidate regulatory risk sequences (MPRA-active CRSs) are resolved, including 267 with variant-specific effects (MPRA-emVars) and 61 with variant-by-cytokine interaction effects (interaction MPRA-emVars). Broadly, neuronal immune-mediated regulatory activity is associated with differences in transcription factor binding and chromatin accessibility, the gene targets of which show pleiotropic enrichments for brain, metabolic, and immune disorders. Dynamic genetic regulation mediates neuroimmune effects, informing our understanding of the genomics of psychiatric and neurological traits, mechanisms governing pleiotropy across disorders, and how immune mechanisms mediate genetic risk.

genetics↗

bioRxiv2024

α-Synuclein acts as a cholesteryl-ester sensor on lipid droplets regulating organelle size and abundance.

While aggregated alpha-Synuclein (Syn) is commonly associated with Parkinsons disease, its physiological function as a membrane-binding protein is poorly understood. Here, we show that endogenous Syn binds lipid droplets (LDs) in multiple human cell lines and in stem cell-derived dopaminergic neurons. LD-binding encompasses Syn residues 1-100, which masks their detection by immunofluorescence microscopy, probably explaining the scarcity of similar observations in earlier studies. Syn-LD interactions are highly temperature-sensitive and selective for cholesteryl-ester-rich LDs. They promote the formation of Syn multimers that dissociate from LDs at non-permissive temperatures. Syn remains LD-bound throughout starvation-induced lipolysis, whereas siRNA-knockdown diminishes LD abundance and compromises cell viability upon nutrient depletion, without affecting LD biosynthesis. Reciprocally, excess Syn stimulates LD accumulation in dependence of lipid availability, restricts organelle size and ensures intracellular LD organization, which strictly depends on functional membrane-binding. Supporting a general role of Syn in cellular lipid and cholesterol metabolism, our results point to additional loss-of-function similarities between Parkinsons, Alzheimers and Gauchers disease.

cell biology↗

bioRxiv2024

A tool for CRISPR-Cas9 gRNA evaluation based on computational models of gene expression

CRISPR based technologies have revolutionized all biomedical fields as it enables efficient genomic editing. These technologies are often used to silence genes by inducing mutations that are expected to nullify their expression. To this end, dozens of computational tools have been developed to design gRNAs, CRISPRs gene-targeting molecular guide, with high cutting efficiency and no off-target effect. However, these tools do not consider the induced mutations effect on the genes expression, which is the actual objective that should be optimized. This fact can often lead to failures in the design, as an efficient cutting of the DNA does not ensure the desired effect in protein production. Therefore, we developed EXPosition, a computational tool for gRNA design. It is the first tool designed to improve the true objective of using CRISPR: the effect it has on gene expression. To this end, we used predictive deep-learning models for the relevant gene expression steps: transcription, splicing, and translation initiation. We validated our tool by demonstrating that it can classify sites as "silencing" or "non-silencing" better than models that consider only the cutting efficiency. We believe that this tool will significantly improve both the efficiency and accuracy of genome editing endeavors. EXPosition is available at http://www.cs.tau.ac.il/~tamirtul/EXPosition.

synthetic biology↗

bioRxiv2024

Oxytocin signaling regulates maternally-directed behavior during early life

Oxytocin is essential in shaping social behavior across the lifespan. While the role of oxytocin signaling in parental care has been widely investigated, little is known about its function in social behavior during early life. This is partly due to the lack of precise technologies for studying the developing brain. Here, we studied the role of oxytocin in pup social behavior under acute separation from and reunion with the mother. We show that the activity of oxytocin neurons was increased by acute maternal separation and returned to baseline after reunion. Behaviorally, maternally-separated pups emitted more ultrasonic vocalizations upon reunion, which were further modulated by nipple attachment behavior. These effects were attenuated by blocking the oxytocin receptor during maternal separation. To investigate the role of oxytocin neurons with higher precision, we established a method for transcranial optogenetic silencing of neuronal activity in untethered and freely behaving pups. Using this approach, we found that silencing of oxytocin neurons during maternal separation disrupted vocal behavior during separation and reunion in a sex-specific manner. Our findings reveal an important role of oxytocin in context-dependent vocal communication in pups, offering new insights into the mechanisms of social behavior during early life.

neuroscience↗

bioRxiv2024

TDP-43 pathology links innate and adaptive immunity in amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis is the most common fatal motor neuron disease. Approximately 90% of ALS patients exhibit pathology of the master RNA regulator, Transactive Response DNA Binding protein (TDP-43). Despite the prevalence TDP-43 pathology in ALS motor neurons, recent findings suggest immune dysfunction is a determinant of disease progression in patients. Whether TDP-43 pathology elicits disease-modifying immune responses in ALS remains underexplored. In this study, we demonstrate that TDP-43 pathology is internalized by antigen presenting cells, causes vesicle rupture, and leads to innate and adaptive immune cell activation. Using a multiplex imaging platform, we observed interactions between innate and adaptive immune cells near TDP-43 pathological lesions in ALS brain. We used a mass cytometry-based whole-blood stimulation assay to provide evidence that ALS patient peripheral immune cells exhibit responses to TDP-43 aggregates. Taken together, this study provides a novel link between TDP-43 pathology and ALS immune dysfunction, and further highlights the translational and diagnostic implications of monitoring and manipulating the ALS immune response.

cell biology↗

bioRxiv2023

Secondary metabolite profiling of Pseudomonas aeruginosa isolates reveals rare genomic traits

Pseudomonas aeruginosa is a ubiquitous gram-negative opportunistic pathogen with remarkable phylogenetic and phenotypic variability. In this work, we applied classical molecular networking analysis to secondary metabolite profiling data from seven Pseudomonas aeruginosa strains, including five clinical isolates from the lung secretions of people with cystic fibrosis. Combined with whole-genome sequencing, we show that some P. aeruginosa isolates, including nmFLRO1, produce a previously unreported class of acyl putrescines, isolate SH3A does not produce di-rhamnolipids because its genome belongs to phylogenetic clade 5, and the secondary metabolite profile of isolate SH1B reflects a frame-shift mutation in the quorum sensing regulator rhlR. This study highlights for the first time that secondary metabolite profiling provides unique insight into genetic variation of P. aeruginosa. ImportanceSecondary metabolite profiling of Pseudomonas aeruginosa isolates can be used to identify rare genomic variants that impact quorum sensing and metabolite biosynthesis that underlie virulence.

microbiology↗

bioRxiv2023

OOCYTE AND EMBRYO CULTURE UNDER OIL PROFOUNDLY ALTERS EFFECTIVE CONCENTRATIONS OF SMALL MOLECULE INHIBITORS

Culture of oocytes and embryos in media under oil is a cornerstone of fertility treatment, and extensively employed in experimental investigation of early mammalian development. It has been noted anecdotally by some that certain small molecule inhibitors might lose activity in oil-covered culture systems, presumably by drug partitioning into the oil. Here we took a pseudo-pharmacological approach to appraise this formally. Using different culture dish designs with defined media:oil volume ratios, we show that the EC50 of the widely employed microtubule poison nocodazole shifts as a function of the media:oil ratio, such that nocodazole concentrations that prevent cell division in oil-free culture fail to in oil-covered media drops. Relatively subtle changes in culture dish design lead to measurable changes in EC50. This effect is not specific to one type of culture oil, and can be readily observed both in oocyte and embryo culture experiments. We subsequently applied a similar approach to a small panel of widely employed cell cycle-related inhibitors, finding that most lose activity in standard oil-covered oocyte/embryo culture systems. Our data suggest that loss of small molecule activity in oil-covered oocyte and embryo culture is a widespread phenomenon with potentially far-reaching implications for data reproducibility, and we recommend avoiding oil-covered culture for experiments employing inhibitors/drugs wherever possible.

developmental biology↗

bioRxiv2023

Contact-FP: a dimerization-dependent fluorescent protein toolkit for visualizing membrane contact site dynamics

Membrane contact sites (MCSs) are sites of close apposition between two organelles used to exchange ions, lipids, and information. Cells respond to changing environmental or developmental conditions by modulating the number, extent, or duration of MCSs. Because of their small size and dynamic nature, tools to study the dynamics of MCSs in live cells have been limited. Dimerization-dependent fluorescent proteins (ddFPs) targeted to organelle membranes are an ideal tool for studying MCS dynamics because they reversibly interact to fluoresce specifically at the interface between two organelles. Here, we build on previous work using ddFPs as sensors to visualize the morphology and dynamics of MCSs. We engineered a suite of ddFPs called Contact-FP that targets ddFP monomers to lipid droplets (LDs), the endoplasmic reticulum (ER), mitochondria, peroxisomes, lysosomes, plasma membrane, caveolae, and the cytoplasm. We show that these probes correctly localize to their target organelles. Using LDs as a test case, we demonstrate that Contact-FP pairs specifically localize to the interface between two target organelles. Titration of LD-mitochondria ddFPs revealed that these sensors can be used at high concentrations to drive MCSs or can be titrated down to minimally perturb and visualize endogenous MCSs. We show that Contact-FP probes can be used to: (1) visualize LD-mitochondria MCS dynamics, (2) observe changes in LD-mitochondria MCS dynamics upon overexpression of PLIN5, a known LD-mitochondrial tether, and (3) visualize two MCSs that share one organelle simultaneously (e.g., LD-mitochondria and LD-ER MCSs). Contact-FP probes can be optimized to visualize MCSs between any pair of organelles represented in the toolkit.

cell biology↗

bioRxiv2023

Cancer immunotherapy responses persist after lymph node resection

Lymphatic transport facilitates the presentation of cancer antigens in tumor-draining lymph nodes (tdLNs), leading to T cell activation and the generation of systemic anti-cancer immune surveillance. Surgical removal of tdLNs to control cancer progression is routine in clinical practice. However, whether removing tdLNs impairs immune checkpoint blockade (ICB) is still controversial. Our analysis demonstrates that melanoma patients remain responsive to PD-1 checkpoint blockade after regional LN dissection. We were able to recapitulate the persistent response to ICB after regional LN resection in murine melanoma and mammary carcinoma models. Mechanistically, soluble antigen is diverted to distant LNs after tdLN dissection. Consistently, robust ICB responses in patients with head and neck cancer after primary tumor and tdLN resection correlated with the presence of reactive LNs in distant sites. These findings indicate that distant LNs sufficiently compensate for the removal of direct tdLNs and sustain the response to ICB.

cancer biology↗