Inducible T-Cell Co-Stimulator Emerges During Chronic Liver Disease Progression and Defines a T-Cell–Inflamed Immune State in Hepatocellular Carcinoma

Inducible T-Cell Co-Stimulator Emerges During Chronic Liver Disease Progression and Defines a T-Cell–Inflamed Immune State in Hepatocellular Carcinoma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Inducible T-Cell Co-Stimulator Emerges During Chronic Liver Disease Progression and Defines a T-Cell–Inflamed Immune State in Hepatocellular Carcinoma Grifton Tafadzwa Muchovu, Jean Francois Regis Igiramaboko, Hilarie Uwamahoro, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9060848/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Hepatocellular carcinoma (HCC) arises within chronic liver disease and is molded by complex immune remodeling. Inducible T-cell costimulator (ICOS) is an immune checkpoint associated with T-cell activation, but its timing and role across disease progression remain unclear. This study aimed to clarify the stage-specific regulation of ICOS and determine its role in molding immune architecture during hepatocellular carcinoma development. Methods We conducted staged transcriptomic profiling integrating early metabolic liver dysfunction (GSE89632), advanced inflammatory remodeling and HCC (GSE164760), and external validation in The Cancer Genome Atlas Liver Hepatocellular Carcinoma cohort (TCGA-LIHC). ICOS dynamics were evaluated across disease stages, followed by pathway enrichment, immune signature profiling, multivariable modeling, and unsupervised immune ecosystem stratification. Results ICOS expression was stable in early disease but increased significantly during advanced inflammatory remodeling and malignant transformation. ICOS-high tumors exhibited coordinated enrichment of interferon, costimulatory, inhibitory, and effector T-cell programs, showing a stable T-cell–inflamed immune architecture. In TCGA-LIHC, ICOS expression was independent of viral etiology and tumor stage but remained tightly associated with immune exhaustion signatures after multivariable adjustment. Unsupervised clustering defined an ICOS-enriched immune subtype, distinct from tumor burden and not independently predictive of survival. Conclusions ICOS emerges during advanced liver disease and remains as a structural regulator of immune architecture in HCC. These outcomes position ICOS as a marker of immune ecosystem state rather than tumor progression per se. Looking ahead, ICOS-defined immune states could serve as a practical basis for stratifying patients in future interventional studies, enabling personalized selection for ICOS agonist or antagonist therapies. HCC ICOS immune remodeling TCGA transcriptomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Hepatocellular carcinoma (HCC) is the sixth most diagnosed cancer and the third leading cause of cancer-related death worldwide [ 1 ]. It is driven by both viral hepatitis and the increasing prevalence of metabolic dysfunction–associated steatotic liver disease (MASLD) [ 2 , 3 ]. MASLD can progress to metabolic dysfunction–associated steatohepatitis (MASH), advanced fibrosis, cirrhosis, and ultimately HCC through a multifaceted interaction of metabolic injury, chronic inflammation, and immune dysregulation [ 4 ]. The growing prevalence of MASLD-associated HCC underscores the need to better understand immune regulatory mechanisms that function across disease stages [ 3 , 4 ]. The liver is immunologically unique. It maintains tolerance and mounts adaptive immune responses [ 5 ]. Chronic liver inflammation disturbs this balance, resulting in sustained inflammation and remodeling of the immune microenvironment [ 5 , 6 ]. Previous studies using single-cell transcriptomics have shown substantial heterogeneity among tumor-infiltrating immune cells [ 7 , 8 ]. This includes the coexistence of T-cell activation, dysfunction, and exhaustion states [ 7 , 9 ]. Chronic inflammation is defined by T-cell exhaustion, sustained expression of inhibitory receptors, and transcriptional reprogramming [ 9 ]. These immune adaptations are clinically relevant, as demonstrated by the survival benefit of immune checkpoint blockade in advanced HCC [ 10 ]. Among costimulatory receptors, inducible T-cell co-stimulator (ICOS) has attracted increasing attention [ 11 ]. ICOS belongs to the CD28 family and plays a role in T-cell activation, maintenance, and differentiation [ 11 , 12 ]. Its signalling contributes to follicular helper T-cell development, cytokine production, and immune regulation [ 12 ]. In cancer, ICOS is recognized for its double-edged role [ 13 ]. On one side, ICOS supports classical effector functions, including activation and persistence of T cells within the tumour microenvironment [ 13 ]. On the other hand, ICOS signalling also promotes regulatory T-cell expansion, which can dampen anti-tumour immunity [ 13 , 16 ]. The most decisive findings from recent studies emphasize that ICOS expression marks distinct T-cell states within inflamed tumours, bridging effector activity with regulatory and exhaustion programs [ 15 ]. In HCC, where chronic antigen exposure and inflammatory remodelling are pervasive, ICOS may operate as part of coordinated immune architectures, rather than as an isolated checkpoint molecule [ 16 ]. Despite growing interest in ICOS for use in immunotherapy, its regulation in a spectrum of metabolic liver diseases remains unclear [ 16 , 17 ]. Most investigations evaluate ICOS within established malignancy or experimental models, without resolving whether dysregulation begins during early metabolic dysfunction, arises during fibrotic remodeling, or primarily accompanies malignant transformation [ 16 , 18 ]. A key unanswered question is: at what point during disease progression does ICOS expression diverge from that of its ligand, ICOSLG? Does this dissociation mark a pivotal event in immune microenvironment remodeling, or do ICOS and ICOSLG maintain synchronous transcriptional dynamics across the chronic liver disease spectrum? It is also still unclear whether ICOS reflects tumor burden and anatomical stage, or whether it instead defines a conserved immune microenvironment state independent of tumor progression [ 5 ]. The transcriptional behavior of its ligand, ICOSLG, across progressive liver disease is even less characterized, despite frequent assumptions of coordinated receptor–ligand regulation [ 17 ]. By framing these gaps as concrete, testable questions, this study aims to invite the reader directly into the scientific inquiry underlying our experimental approach. Clarifying the stage-specific emergence of ICOS is necessary for interpreting its biological and translational significance [ 19 ]. Misattributing late-stage immune remodeling features to early pathogenic mechanisms may conceal the true sequence of immune adaptation during hepatocarcinogenesis [ 7 ]. A stage-resolved analysis is therefore necessary to define when ICOS dysregulation arises and whether it reflects structural immune remodeling rather than tumor expansion alone [ 19 , 7 ]. We hypothesized that ICOS dysregulation does not occur during early metabolic dysfunction but instead emerges during advanced inflammatory remodeling and malignant transformation. We further hypothesized that once induced, ICOS becomes embedded within a coordinated T-cell–inflamed immune architecture independent of tumor stage and etiology. To test these hypotheses, we performed, to our knowledge, the first multi-cohort study mapping ICOS emergence across the entire liver disease continuum. This staged multi-cohort transcriptomic investigation integrates early metabolic liver disease datasets, advanced inflammatory and HCC cohorts, and independent validation in the TCGA-LIHC dataset. Through pathway enrichment, immune signature profiling, multivariable modeling, and unsupervised immune ecosystem stratification, we sought to define the temporal emergence and biological positioning of ICOS across chronic liver disease progression and hepatocellular carcinoma. The conceptual framework guiding this study is illustrated in Fig. 1 A. Methodology Study Design and Analytical Scheme This study was conducted as a staged, multi-cohort transcriptomic investigation. The main objective was to define the temporal and stage-dependent regulation of ICOS across metabolic liver disease progression and hepatocellular carcinoma (HCC). It was hypothesized that ICOS dysregulation arises during advanced inflammatory remodeling and malignant transformation, rather than during early metabolic dysfunction. Furthermore, it was posited that, in these later stages, ICOS dysregulation appears associated with coordinated reprogramming of the immune microenvironment. To test these hypotheses, a two-phase analytical scheme was implemented. In the first phase, transcriptomic profiling was performed across early and advanced metabolic liver disease using independent human liver microarray datasets. ICOS and ICOSLG expression dynamics, immune pathway enrichment, and immune gene signature remodeling in hepatocellular carcinoma were evaluated. The second phase validated these outcomes using an independent TCGA-LIHC cohort to assess tumor-stage independence, immune exhaustion coupling, multivariable modeling, interaction testing, survival analysis, and immune ecosystem stratification. Statistical Model All analyses were conducted in Python using pandas, NumPy, SciPy, statsmodels, lifelines, scikit-learn, seaborn, matplotlib, and gseapy. All tests were two-sided, with nominal significance defined as p < 0.05. Where applicable, Benjamini–Hochberg correction was applied. Phase 1: Transcriptomic Analysis of ICOS Regulation in Metabolic Liver Disease Dataset Selection Human liver transcriptomic datasets were retrieved from the Gene Expression Omnibus (GEO). Inclusion criteria were: (i) human liver tissue, (ii) representation of both non-malignant and malignant disease states, (iii) available stage annotation, and (iv) platform-specific probe annotation files. Two datasets met the criteria. GSE89632 (GPL14951; Illumina HumanHT-12) served as the early disease validation cohort. It included Healthy (n = 24), MASLD (n = 20), and MASH (n = 19) samples (n = 63 total). GSE164760 (GPL13667; Affymetrix Human Genome U219) served as the advanced disease discovery cohort and included Healthy (n = 6), NASH (n = 74), Cirrhosis (n = 8), Adjacent tissue (n = 29), and HCC tumors (n = 53) (n = 170 total). Probe Processing and Gene-Level Collapsing Series matrix files were downloaded from GEO, and platform-specific annotation files (GPL14951 and GPL13667) were used for probe-to-gene mapping. Probes without valid gene symbols were excluded. When multiple probes mapped to a single gene, expression values were collapsed using the median probe values. Expression values were retained on the log2-normalized scale. No cross-dataset normalization or batch correction was performed, as each dataset was analyzed independently. Disease Stage Classification Stage groups were derived from curated GEO annotations. For the GSE89632 cohort, samples were grouped as Healthy, MASLD, or MASH. In GSE164760, samples were categorized as Healthy, NASH, Cirrhosis, Adjacent non-tumoral tissue, or HCC tumor. GSM accession identifiers were used as unique keys for merging metadata with gene expression matrices. ICOS and ICOSLG Expression Analysis ICOS and ICOSLG expression values were extracted from gene-level matrices. Distributional assessment confirmed non-normal expression. For this reason, Kruskal–Wallis testing was used for multi-group comparisons. In GSE89632, expressions were compared across Healthy, MASLD, and MASH. As for the GSE164760, cohort comparisons were done across Healthy, NASH, Cirrhosis, Adjacent, and Tumor groups. Pairwise median differences were calculated to quantify directional shifts across stages. ICOS Stratification in HCC HCC tumors (n = 53) in GSE164760 were dichotomized into ICOS-High and ICOS-Low groups using the median ICOS expression cutoff to preserve balanced group sizes and maximize statistical power for downstream analyses. The decision to use the median rather than a fixed percentile or quartile cutoff was based on the relatively small cohort size, which enabled reliable comparisons between groups of equal size while minimizing potential bias from arbitrary thresholds. Importantly, the median split approach has been commonly employed in transcriptomic analysis to facilitate reproducibility and clear reporting. Group separation was confirmed using the Mann–Whitney U test. Pathway Enrichment Analysis Pre-ranked Gene Set Enrichment Analysis (GSEA) was performed using the MSigDB Hallmark 2020 collection. Genes were ranked by the delta-median expression difference between ICOS-High and ICOS-Low tumors. Enrichment was performed using 1,000 permutations, with gene set size thresholds of 15–500 genes. False discovery rate (FDR) q < 0.25 was considered significant. Immune Signature Profiling Curated immune gene panels representing T-cell activation, exhaustion, costimulatory checkpoints, inhibitory checkpoints, and cytokine–chemokine signaling were assembled from immuno-oncology literature. For each tumor sample, signature scores were calculated as the mean log2 expression of genes within each panel. Group comparisons were performed using Mann–Whitney U testing with Benjamini–Hochberg correction. In addition to group-level comparisons, tumor-level visualization of immune architecture was performed using the full gene expression matrix. Selected immune regulatory genes were extracted, standardized using z-score normalization, and hierarchically clustered to evaluate coordinated immune remodeling across ICOS-defined tumor subsets. Phase 2: Validation, Multivariable Modeling, and Immune Ecosystem Stratification in TCGA-LIHC TCGA Data Acquisition and Harmonization RNA sequencing data (RNASeqV2, log2-normalized) and clinical annotations were obtained from the UCSC Xena portal for the TCGA-LIHC dataset. Clinical variables included overall survival and pathological tumor stage. Molecular and clinical data were merged using the first twelve characters of TCGA barcodes. Samples lacking stage or survival data were excluded from stage-adjusted and survival analyses. All analyses used provided log2-normalized expression values without further transformation. Tumor stage was extracted from and harmonized into an ordinal numeric variable (Stage I–IV). Substages were collapsed into primary-stage categories to preserve ordering and ensure adequate sample sizes per group. Etiology and Stage Association Tumors were stratified into viral and non-viral groups using viral hepatitis serology annotations. Differences in ICOS expression between etiologic groups were assessed using the Mann–Whitney U test. Associations between ICOS expression and tumor stage were evaluated using Spearman's rank correlation and Kruskal–Wallis testing. Differences in correlation strength were evaluated using Fisher’s r-to-z transformation. Immune Functional Score Construction Composite transcriptional scores were constructed to quantify immune architecture. The exhaustion score was defined as the mean log2 expression of PDCD1, CTLA4, LAG3, TIGIT, and HAVCR2. The Treg score was calculated as the mean expression of FOXP3, IL2RA, and CTLA4. The CD8 score was computed using the mean expression of CD8A and CD8B. A Treg/CD8 balance metric was calculated by subtracting the CD8 score from the Treg score. The calculations were computed using vectorized operations. Multivariable Modeling of Exhaustion Architecture Ordinary least squares (OLS) regression was used to evaluate whether ICOS independently predicts exhaustion architecture. The primary model specification was: Exhaustion score ~ ICOS expression + Stage numeric Regression coefficients (β), 95% confidence intervals, R², and F-statistics were calculated using statsmodels. To test stage-dependent modulation of immune coupling, an interaction term was introduced: Exhaustion score ~ ICOS expression + Stage numeric + (ICOS expression × Stage numeric) A non-significant interaction term was interpreted as evidence of stage-independent ICOS–exhaustion coupling. Survival Modeling Cox proportional hazards regression was performed using overall survival as the endpoint. Two prespecified models were fitted: Model 1: OS ~ ICOS expression + Stage numeric Model 2: OS ~ ICOS expression + Stage numeric + Exhaustion score Hazard ratios (HR), 95% confidence intervals, Wald p-values, concordance index (C-index), and partial Akaike Information Criterion (AIC) were calculated using lifelines. Kaplan–Meier survival curves were generated using median ICOS dichotomization and compared using the log-rank test. Immune Ecosystem Stratification To identify immune architectural subtypes, unsupervised k-means clustering was performed using standardized ICOS expression, exhaustion score, Treg score, and CD8 score. Features were z-score normalized using StandardScaler prior to clustering. Cluster number (k = 2–5) was assessed using silhouette score analysis. The optimal solution was selected based on the maximal silhouette coefficient. Cluster-level mean values of immune features were calculated to characterize biological subtypes. Principal component analysis (PCA) was applied to visualize cluster separation in reduced-dimensional space. Kaplan–Meier analysis was used to evaluate survival differences across immune clusters. Results Section 1: Stage-Dependent Transcriptional Remodeling of ICOS Across Metabolic Liver Disease and HCC Results Overview The temporal emergence of ICOS dysregulation during the progression of metabolic liver disease was delineated using a staged analytical approach that integrated early metabolic dysfunction (GSE89632) and advanced inflammatory remodeling in hepatocellular carcinoma (GSE164760). This section is organized into three components: (i) evaluation of ICOS and ICOSLG expression in early metabolic disease, (ii) stage-resolved transcriptional dynamics across advanced liver pathology and HCC, and (iii) immune pathway and gene-level remodeling in ICOS-stratified tumors. Collectively, these analyses establish the timing of ICOS dysregulation and clarify whether it represents isolated gene activation or coordinated immune reprogramming. All figures and Supplementary tables are summarized in Table S0. ICOS and ICOSLG Remain Stable During Early Metabolic Dysfunction ICOS and ICOSLG expression were evaluated across Healthy (n = 24), MASLD (n = 20), and MASH (n = 19) liver samples in GSE89632 to determine whether ICOS dysregulation initiates during early metabolic perturbation. ICOS expression demonstrated no significant difference across early disease states (Kruskal–Wallis H = 0.842, p = 0.656). Median expression values were comparable across groups, with overlapping interquartile ranges and no directional shift (Fig. 1 B). Similarly, ICOSLG expression remained transcriptionally stable (H = 0.969, p = 0.615; Fig. 1 C). Descriptive statistics are in Table S1 . These findings demonstrate that neither ICOS nor its ligand, ICOSLG, undergoes transcriptional modulation during early metabolic liver disease, indicating that dysregulation is not initiated during the initial steatotic or inflammatory transitions. ICOS Becomes Dysregulated During Advanced Inflammatory Remodeling and Malignant Transformation In contrast to early disease stability, ICOS expression varied substantially across advanced pathological stages in GSE164760 (Healthy n = 6; NASH n = 74; Cirrhosis n = 8; Adjacent n = 29; Tumor n = 53). The Kruskal–Wallis test showed significant stage-dependent variation (H = 12.617, p = 0.013; Table S1 ). Median ICOS expression progressively increased from non-malignant liver to cirrhosis and adjacent tissue, with sustained elevation in tumor samples (Fig. 2 A). These results suggest a progressive immunological shift rather than a single event. Directional pairwise median comparisons (Fig. 2 B) revealed upward shifts from NASH to cirrhosis and from cirrhosis to adjacent tissue. Tumor samples displayed heterogeneous but persistently elevated ICOS levels relative to early-stage metabolic disease. Full pairwise statistics are reported in Table S2 . Kernel density analysis also supported a rightward distributional shift in cirrhotic and tumor samples relative to healthy liver (Fig. 2 C). Adjacent tissue exhibited intermediate expression patterns, consistent with progressive microenvironment remodeling during chronic inflammation. Importantly, ICOSLG did not vary substantially across advanced stages (H = 3.443, p = 0.486; Table S1 ) indicating receptor–ligand divergence during disease progression. Collectively, these findings show that ICOS dysregulation emerges during advanced inflammatory remodeling and persists into hepatocellular carcinoma, whereas ICOSLG remains transcriptionally stable. This divergence implies receptor-specific immune regulatory amplification rather than coordinated ligand upregulation. ICOS-High Tumors showed Coordinated Immune Pathway and Gene-Level Remodeling. HCC tumors were stratified into ICOS-High (n = 27) and ICOS-Low (n = 26) groups using a median cutoff to assess whether ICOS dysregulation remains associated with wider immune microenvironment remodeling. Pre-ranked GSEA showed significant enrichment of immune-related Hallmark pathways in ICOS-High tumors (normalized enrichment scores > 1.5; FDR q < 0.25), including interferon signaling, inflammatory response, and TNFα signaling via NF-κB (Fig. 3 A). Complete enrichment statistics are provided in Table S3 and Table S4 . To quantify coordinated immune remodeling at the programmatic level, curated immune gene panels were used for evaluation. ICOS-High tumors demonstrated elevated signature scores across costimulatory checkpoints, inhibitory checkpoints, T-cell activation, T-cell exhaustion, and cytokine–chemokine programs (Fig. 3 B). The most pronounced difference was observed in costimulatory signaling (Δ median = 4.58; FDR = 0.0025), while inhibitory and activation programs showed concordant directional shifts (Table S5 ). To evaluate immune architecture at the tumor level, a focused immune regulatory gene panel was extracted from the full expression matrix. After z-score normalization and hierarchical clustering, ICOS-High tumors exhibited coordinated upregulation of costimulatory receptors (ICOS, TNFRSF9, CD28), inhibitory checkpoints (PDCD1, CTLA4, LAG3, TIGIT, HAVCR2), effector activation genes (IFNG, GZMB, PRF1), and inflammatory chemokines (CXCL9, CXCL10, CCL5) (Fig. 3 C). Importantly, clustering demonstrated that these programs rise in concert rather than independently, indicating that ICOS dysregulation shows a coordinated immune regulatory state instead of isolated gene activation. These data establish that ICOS enrichment in HCC is embedded within a wider T-cell–inflamed transcriptional architecture characterized by simultaneous activation and checkpoint engagement. Section 2: External Validation, Architectural Stability, and Immune Ecosystem Stratification in TCGA-LIHC Results Overview Multi-layer validation in the TCGA-LIHC cohort was performed to determine whether ICOS-associated immune remodeling observed during metabolic liver disease progression persists within established malignancy and reflects conserved immune architecture rather than tumor burden. This validation framework interrogated (i) etiology and stage independence of ICOS expression, (ii) multivariable modeling of the ICOS–exhaustion axis and overall survival, and (iii) ecosystem-level immune stratification using unsupervised clustering. Collectively, these analyses establish whether ICOS functions as a tumor burden marker or as a structural determinant of immune architecture. ICOS Is Independent of Etiology and Tumor Stage in TCGA HCC We first examined whether ICOS expression differed between viral and non-viral hepatocellular carcinoma. ICOS expression distributions were comparable across etiologic groups (Mann–Whitney p = 0.65), with overlapping medians and interquartile ranges (Fig. 4 A). The processed dataset supporting this analysis is provided in Table S6 . Despite the absence of etiologic differences in ICOS abundance, ICOS strongly correlated with immune exhaustion score in both viral (ρ = 0.75, p < 0.001) and non-viral tumors (ρ = 0.72, p < 0.001) (Fig. 4 B–C). A direct statistical comparison using Fisher’s r-to-z transformation showed no significant difference in correlation strength (z = 0.73, p = 0.47), confirming the conservation of the ICOS–exhaustion axis across etiologies. The full comparative statistics are reported in Table S7 . We next evaluated whether ICOS expression escalates with tumor burden. Spearman correlation revealed no association between ICOS and pathologic stage (ρ = 0.03, p = 0.37). Stage-stratified scatter visualization (Fig. 4 D) further demonstrated that ICOS–exhaustion coupling continues across stages without stage-dependent amplification. Stage-specific correlation coefficients are provided in Table S8 . In multivariable regression adjusting for tumor stage (Exhaustion score ~ ICOS expression + Stage numeric), ICOS remained a strong independent predictor of exhaustion architecture (β = 0.555, p < 0.001), whereas stage was not significant (p = 0.89). The model explained 49.8% of the variance in exhaustion (R² = 0.498). Detailed regression coefficients are provided in Table S10 and cross-validated in Table S9 . These data collectively show that ICOS structures immune exhaustion architecture independently of etiology and tumor stage, indicating architectural stability rather than tumor-burden dependence. ICOS Does Not Independently Predict Overall Survival Multivariable Cox proportional hazards models were constructed to determine whether ICOS-mediated immune remodeling is associated with clinical outcomes. In stage-adjusted survival analysis (OS ~ ICOS expression + Stage numeric), ICOS was not associated with overall survival (HR ≈ 1.00, p = 0.87), whereas tumor stage remained a significant predictor (HR ≈ 1.20 per stage increment, p = 0.01; concordance index = 0.55). Kaplan–Meier analysis confirmed no survival difference between ICOS-high and ICOS-low tumors (log-rank p > 0.05; Fig. 5 A). In the fully adjusted model containing exhaustion score (OS ~ ICOS expression + Stage numeric + Exhaustion score), neither ICOS (p = 0.93) nor exhaustion (HR ≈ 0.98, p = 0.77) independently predicted survival, whereas stage remained dominant (p = 0.01; concordance index = 0.52). Coefficients are provided in Table S10 and Table S11 . These relationships are summarized visually in Fig. 5 B, which demonstrates that ICOS and exhaustion hazard ratios cross unity in both models. Together, these data indicate that ICOS-mediated immune polarization reflects tumor immune ecology instead of intrinsic tumor aggressiveness. ICOS Marks T-Cell–Inflamed Tumors Without Preferential Regulatory Skewing Transcriptional scores for regulatory T cells (Treg), CD8 cytotoxic T cells, and immune exhaustion were constructed to dissect immune composition. ICOS expression correlated strongly with both Treg score (ρ = 0.566, p ≈ 1.2 × 10⁻¹¹²) and CD8 score (ρ = 0.503, p ≈ 2.7 × 10⁻¹¹¹). Correlation statistics are provided in Table S12 . Fisher’s r-to-z comparison demonstrated no significant difference between Treg and CD8 correlations (z = 0.105, p = 0.917), and ICOS did not correlate with the Treg/CD8 balance metric (ρ = −0.023, p = 0.34). The immune architecture correlation matrix (Fig. 5 C) illustrates coordinated upregulation of ICOS, exhaustion, Treg, and CD8 programs, with no preferential skewing toward suppressive dominance. These data demonstrate that ICOS marks globally T-cell–inflamed tumors rather than selectively regulatory-biased microenvironments. Immune Ecosystem Stratification Identifies a T-Cell–Inflamed ICOS-High Subtype Unsupervised k-means clustering was performed by standardized ICOS, exhaustion, Treg, and CD8 scores to evaluate whether ICOS defines higher-order immune states beyond linear associations. Silhouette analysis supported a two-cluster solution (silhouette score = 0.36). Cluster classifications are provided in Table S13 . Cluster 1 exhibited coordinated elevation of ICOS (mean 5.15), exhaustion (6.29), Treg (5.11), and CD8 (7.02), defining a T-cell–inflamed immune subtype. Cluster 0 demonstrated uniformly lower immune feature values (Fig. 6 B). Principal component analysis confirmed partition along PC1 (Fig. 6 A). Despite a distinct immune profile, cluster membership did not independently forecast survival (log-rank p = 0.26; Fig. 6 C). This data supports the notion that ICOS defines immunological framework states rather than prognostic tumor biology in untreated TCGA HCC. Discussion ICOS as a Stage-Emergent Architect of Immune Remodeling in HCC This multi-cohort transcriptomic study shows that ICOS functions not solely as a marker of tumour burden or prognosis, but as a stage-emergent regulator of immune architecture in hepatocellular carcinoma (HCC). By combining early metabolic liver disease datasets, advanced inflammatory remodelling cohorts, and external validation in TCGA-LIHC, this study formulates a model in which ICOS dysregulation arises during advanced inflammatory transformation. Subsequently, ICOS stabilizes following malignancy and defines a conserved T-cell–inflamed immune ecosystem within established tumours. The results position ICOS as a structural determinant of the tumour microenvironment rather than a linear immune checkpoint molecule [ 11 , 13 ]. ICOS Emerges Late in Disease Evolution An important finding of this study is the temporal specificity of ICOS dysregulation. In early metabolic liver disease (MASLD and MASH), both ICOS and its ligand ICOSLG remained transcriptionally stable. Despite metabolic stress and inflammation, there was no evidence of progressive ICOS induction. These results suggest that ICOS is not involved in the initial immune perturbation associated with steatosis or early inflammatory activation, which are mainly driven by innate immune signaling and metabolic injury [ 3 , 4 ]. In contrast, ICOS expression increased significantly during advanced pathological states, including cirrhosis, adjacent non-tumoral tissue, and HCC. The distributional shift across these stages was gradual rather than sudden. These observations support a model of progressive immune remodeling rather than oncogenic activation alone [ 19 ]. ICOSLG did not mirror this pattern, indicating receptor-specific amplification rather than coordinated ligand upregulation. These outcomes suggest that ICOS upregulation is linked to chronic inflammatory remodeling and the transition toward a tumor-permissive immune microenvironment [ 5 , 6 ]. ICOS-High Tumors Exhibit Coordinated Immune Activation and Exhaustion When HCC tumors were stratified by ICOS expression, ICOS-high tumors showed coordinated enrichment of interferon signaling, inflammatory response pathways, and TNFα–NFκB signaling. At the gene level, ICOS-high tumors showed simultaneous upregulation of costimulatory receptors, inhibitory checkpoints, cytotoxic effector genes, and inflammatory chemokines, with these programs rising together. This shows a synchronized T-cell–inflamed transcriptional state consistent with prior single-cell observations in HCC [ 7 , 8 ]. Previous studies have characterized ICOS expression within established tumors or individual immune cell subsets, however our analysis demonstrates that ICOS-associated immune architecture emerges progressively during disease evolution and remains stable after malignant transformation. Simultaneous elevations in activation and exhaustion signatures were observed, consistent with chronic antigen exposure and sustained immune engagement [ 9 , 14 ]. The results support the interpretation that ICOS marks an immune microenvironment characterized by sustained T-cell activation alongside compensatory checkpoint regulation. Instead of representing immune suppression alone, ICOS appears to be embedded within a dynamic regulatory balance, a phenomenon well described in chronic inflammatory and cancer settings [ 9 , 14 ]. Immune Architecture Is Independent of Etiology and Tumor Stage Validation in TCGA-LIHC clarified that ICOS expression is independent of both viral etiology and pathological tumor stage. ICOS did not differ between viral and non-viral tumors and did not increase with anatomical stage progression. This dissociation is biologically meaningful. Tumor stage reflects macroscopic growth and invasion, whereas ICOS reflects the immune microenvironment state [ 6 ]. This suggests that once established, ICOS-associated immune architecture remains stable throughout tumor evolution. Despite this stage independence, ICOS consistently correlated with immune exhaustion across etiologies and stage strata. Multivariable modeling confirmed that ICOS independently predicts exhaustion architecture even after adjusting for tumor stage. Together, these data show that the ICOS–exhaustion axis is embedded within the tumor microenvironment and maintained irrespective of tumor burden, consistent with recognized models of exhaustion stability in chronic disease [ 14 ]. ICOS Shows a T-Cell–Inflamed State Without Preferential Regulatory Skewing A key question is whether ICOS enrichment reflects preferential expansion of regulatory T cells. Prior studies have shown that ICOS can be expressed by both effector and regulatory T-cell subsets, depending on context [ 11 , 13 ]. However, our data does not support selective regulatory dominance. ICOS correlated strongly with both Treg and CD8 cytotoxic signatures, and no association was observed with the Treg/CD8 balance metric. Unsupervised clustering identified two immune ecosystems: an immune-low subtype and a T-cell–inflamed subtype characterized by coordinated elevation of ICOS, exhaustion, Treg, and CD8 programs. ICOS was enriched within the inflamed subtype but did not skew immune balance toward suppressive predominance. This is consistent with reports that ICOS functions as a context-dependent modulator of immune activation rather than an inherently suppressive signal [ 13 , 16 ]. Thus, ICOS expression at the bulk tumor level likely reflects global T-cell engagement instead of selective regulatory expansion, consistent with tumor-infiltrating lymphocyte heterogeneity described in HCC [ 7 ]. These findings suggest that ICOS-defined immune states may represent biologically relevant contexts for future ICOS-targeted immunotherapies, where modulation of the ICOS pathway could potentially amplify pre-existing T-cell–inflamed tumor microenvironments. Immune Remodeling Does Not Independently Predict Survival Despite clear immune stratification, ICOS expression and immune cluster membership did not independently predict overall survival in TCGA. Tumor stage remained the major determinant of prognosis, consistent with established epidemiological and clinical observations in HCC [ 1 , 2 ]. This highlights a biological separation between immune architecture and anatomical tumor progression. In largely untreated cohorts such as TCGA, tumor burden is expected to dominate survival outcomes [ 6 ]. Immune polarization may instead influence therapeutic responsiveness rather than baseline prognosis, notably in the context of immune checkpoint blockade [ 10 , 17 ]. Therefore, ICOS-defined immune states should be interpreted as markers of the immune context rather than as indicators of intrinsic tumor aggressiveness. A Hierarchical Model of Immune Remodeling The results support a hierarchical framework: ICOS is not induced during early metabolic dysfunction. It emerges during advanced inflammatory remodeling and malignant transformation. Within the established HCC, ICOS defines a stable T-cell–inflamed immune ecosystem. This immune architecture is independent of tumor stage and etiology. Immune ecosystem stratification does not independently determine survival in untreated cohorts. This model clarifies the apparent paradox between disease-stage emergence and tumor-stage independence. ICOS is induced during inflammatory transformation instead of anatomical tumor expansion, consistent with progressive immune adaptation described in chronic liver disease and cancer immunology [ 5 , 6 , 14 ]. Limitations This study is based on bulk transcriptomic data and therefore cannot resolve cell-type-specific ICOS expression or spatial immune organization. As a direct next step, single-cell and spatial profiling approaches—such as single-cell RNA sequencing and multiplex imaging—should be employed to dissect the cellular sources of ICOS and map its distribution within the liver tumor microenvironment [ 7 , 8 ]. Additionally, trajectory inference methods could unravel the dynamics of immune cell state transitions during disease progression [ 7 ]. Another limitation is that the TCGA dataset predominantly reflects untreated tumors. Future studies should specifically examine immunotherapy-treated patient cohorts to determine whether ICOS-defined immune ecosystems can predict treatment responsiveness, especially in the context of immune checkpoint blockade, where T-cell inflamed states regularly forecast response [ 10 , 17 ]. By prioritizing these directions, future research can address the current gaps and directly test the translational relevance of ICOS-associated immune states. Conclusion To sum up, ICOS functions as a stage-emergent architect of immune remodeling in hepatocellular carcinoma. It is induced during advanced inflammatory progression, persists following malignant transformation, and defines a conserved T-cell–inflamed tumor ecosystem independent of tumor stage and etiology. By reframing ICOS as a structural regulator of immune ecology rather than a simple checkpoint marker or prognostic factor, this study provides a mechanistic model linking chronic inflammatory liver disease to stable immune architecture in HCC. Declarations Conflict of Interest The authors declare that they have no competing interests. Clinical Trial Number Clinical Trial Number not applicable Funding Declaration No funding was received for this project Availability of Data and Materials All data used in the project is publicly available on GEO repository. All supplementary data and figs that support this project are uploaded on Zenedo and can be accessed via the following link. https://doi.org/10.5281/zenodo.18903688 Author Contributions All authors contributed equally to the conception, design, and execution of the study. The overall manuscript was drafted and critically revised by all authors for important intellectual content. All authors read and approved the final manuscript. Informed Consent Informed consent was not required for this study, as it involved secondary analysis of publicly available, anonymised data. References Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, Jemal A (2024) Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Cancer J Clin 74(3):229–263. https://doi.org/10.3322/caac.21834 McGlynn KA, Petrick JL, El-Serag HB (2021) Epidemiology of hepatocellular carcinoma. Hepatology 73(S1):4–13. https://doi.org/10.1002/hep.31288 Younossi ZM, Kalligeros M, Henry L (2024) Epidemiology of metabolic dysfunction-associated steatotic liver disease. Clin Mol Hepatol 31(Suppl):S32. https://doi.org/10.3350/cmh.2024.0431 Estes C, Razavi H, Loomba R, Younossi Z, Sanyal AJ (2018) Modeling the epidemic of nonalcoholic fatty liver disease demonstrates an exponential increase in burden of disease. Hepatology 67(1):123–133. https://doi.org/10.1002/hep.29466 Pfister D, Núñez NG, Pinyol R, Govaere O, Pinter M, Szydlowska M, Gupta R, Qiu M, Deczkowska A, Weiner A et al (2021) NASH limits anti-tumour surveillance in immunotherapy-treated hepatocellular carcinoma. Nature 592(7854):450–456. https://doi.org/10.1038/s41586-021-03362-0 Llovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, Lencioni R, Koike K, Zucman-Rossi J, Finn RS (2021) Hepatocellular carcinoma. Nat Reviews Disease Primers 7:6. https://doi.org/10.1038/s41572-020-00240-3 Zhang Q, He Y, Luo N, Patel SJ, Gao R, Modak M, Carotta S, Haslinger C, Kind D, Van den Brink MRM et al (2019) Landscape and dynamics of single immune cells in hepatocellular carcinoma. Cell 179(4):829–845. https://doi.org/10.1016/j.cell.2019.10.003 Sun Y, Wu L, Zhong Y, Zhou K, Hou Y, Wang Z, Zhang Y, Wang Y, Zhang Y, Yuan J et al (2021) Single-cell landscape of the ecosystem in early-relapse hepatocellular carcinoma. Cell 184(2):404–421. https://doi.org/10.1016/j.cell.2020.11.041 Beltra JC, Manne S, Abdel-Hakeem MS, Kurachi M, Giles JR, Chen Z, Casella V, Ngiow SF, Khan O, Huang Y et al (2020) Developmental relationships of four exhausted CD8 T cell subsets reveal underlying transcriptional and epigenetic control mechanisms. Immunity 52(5):825–841. https://doi.org/10.1016/j.immuni.2020.04.014 Finn RS, Qin S, Ikeda M, Galle PR, Ducreux M, Kim TY, Kudo M, Breder V, Merle P, Kaseb AO et al (2020) Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma. N Engl J Med 382(20):1894–1905. https://doi.org/10.1056/NEJMoa1915745 Wikenheiser DJ, Stumhofer JS (2016) ICOS co-stimulation: Friend or foe? Front Immunol 7:304. https://doi.org/10.3389/fimmu.2016.00304 Crotty S (2019) T follicular helper cell biology: a decade of discovery and diseases. Immunity 50(5):1132–1148. https://doi.org/10.1016/j.immuni.2019.04.011 Amatore F, Gorvel L, Olive D (2020) Role of inducible co-stimulator (ICOS) in cancer immunotherapy. Expert Opin Biol Ther 20(2):141–150. https://doi.org/10.1080/14712598.2020.1693540 Blank CU, Haining WN, Held W, Hogan PG, Kallies A, Lugli E et al (2019) Defining ‘T cell exhaustion’. Nat Rev Immunol 19(11):665–674. https://doi.org/10.1038/s41577-019-0221-9 Sade-Feldman M, Yizhak K, Bjorgaard SL, Ray JP, de Boer CG, Jenkins RW et al (2019) Defining T cell states associated with response to checkpoint immunotherapy in melanoma. Cell 176(1–2):404–419. https://doi.org/10.1016/j.cell.2018.12.034 Nikanjam M, Kato S, Nishizaki D et al (2026) Inducible T-cell co-stimulator (ICOS) and ICOS ligand: Dealing with a two-faced cancer immunoregulatory system. Cancer Med 15(1):e71467. https://doi.org/10.1002/cam4.71467 Shen KY, Zhu Y, Xie SZ et al (2024) Immunosuppressive tumor microenvironment and immunotherapy of hepatocellular carcinoma: Current status and perspectives. J Hematol Oncol 17:25. https://doi.org/10.1186/s13045-024-01549-2 Khan O, Giles JR, McDonald S et al (2019) TOX transcriptionally and epigenetically programs CD8 T cell exhaustion. Nature 571(7764):211–218. https://doi.org/10.1038/s41586-019-1325-x Ma L, Hernandez MO, Zhao Y, Mehta M, Tran B, Kelly M et al (2019) Tumor cell biodiversity drives microenvironmental reprogramming in liver cancer. Cancer Cell 36(4):418–430. https://doi.org/10.1016/j.ccell.2019.08.007 Additional Declarations No competing interests reported. Supplementary Files TabelS1.xlsx TableS0.xlsx TableS3.xlsx TableS2.xlsx TableS4.xlsx TableS5.xlsx TableS8.xlsx TableS6.xlsx TableS9.xlsx TableS10.xlsx TableS7.xlsx TableS11.xlsx TableS12.xlsx TableS13.xlsx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9060848","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":609836011,"identity":"db921665-9a42-4724-9570-3cfb5ee4b655","order_by":0,"name":"Grifton Tafadzwa Muchovu","email":"data:image/png;base64,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","orcid":"","institution":"Università del Piemonte Orientale","correspondingAuthor":true,"prefix":"","firstName":"Grifton","middleName":"Tafadzwa","lastName":"Muchovu","suffix":""},{"id":609836012,"identity":"ff1f7de5-df82-4750-9cec-5670fd6e6993","order_by":1,"name":"Jean Francois Regis Igiramaboko","email":"","orcid":"","institution":"Trenholm State Community College","correspondingAuthor":false,"prefix":"","firstName":"Jean","middleName":"Francois Regis","lastName":"Igiramaboko","suffix":""},{"id":609836013,"identity":"2d76d3b9-9130-4723-8523-12463f437d8a","order_by":2,"name":"Hilarie Uwamahoro","email":"","orcid":"","institution":"Tuskegee University","correspondingAuthor":false,"prefix":"","firstName":"Hilarie","middleName":"","lastName":"Uwamahoro","suffix":""},{"id":609836014,"identity":"f7f89e66-b0cd-4353-a554-6ce85ee59223","order_by":3,"name":"Concorde Niyigaba Isange","email":"","orcid":"","institution":"Università del Piemonte Orientale","correspondingAuthor":false,"prefix":"","firstName":"Concorde","middleName":"Niyigaba","lastName":"Isange","suffix":""},{"id":609836015,"identity":"2be93f3d-b99c-4d4f-855b-e15fbea3a670","order_by":4,"name":"Pouya Bozorgpouryazdi","email":"","orcid":"","institution":"Università del Piemonte Orientale","correspondingAuthor":false,"prefix":"","firstName":"Pouya","middleName":"","lastName":"Bozorgpouryazdi","suffix":""}],"badges":[],"createdAt":"2026-03-07 21:23:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9060848/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9060848/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105308932,"identity":"d51e8451-ecc1-41f0-a3e0-0ce6d2a2aa54","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2551057,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual framework and early-stage transcriptional stability of ICOS in metabolic liver disease (GSE89632).\u003cstrong\u003e (A)\u003c/strong\u003e Conceptual schematic summarizing the study rationale and central hypothesis that ICOS emerges during advanced inflammatory remodeling and stabilizes a T-cell–inflamed immune ecosystem independent of tumor stage and etiology. \u003cstrong\u003e(B)\u003c/strong\u003e ICOS expression across Healthy (n=24), MASLD (n=20), and MASH (n=19) liver samples. No significant differences were observed (Kruskal–Wallis H = 0.842, p = 0.656).\u003cstrong\u003e(C)\u003c/strong\u003e ICOSLG expression across the same early disease states. No significant differences were detected (Kruskal–Wallis H = 0.969, p = 0.615). Expression values represent log2-normalized microarray intensities. Boxes denote interquartile range (IQR), center lines indicate medians, and whiskers represent 1.5× IQR. Statistical testing was performed using Kruskal–Wallis analysis.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/e28dbda4256337d1d6ac843e.png"},{"id":105308945,"identity":"5e5913ab-0b69-4927-bbfc-f11524cc184a","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":585816,"visible":true,"origin":"","legend":"\u003cp\u003eStage-dependent emergence of ICOS dysregulation during advanced inflammatory remodeling and hepatocellular carcinoma (GSE164760). (A) Boxplot of ICOS expression across Healthy (n=6), NASH (n=74), Cirrhosis (n=8), Adjacent non-tumoral tissue (n=29), and HCC tumor samples (n=53). ICOS expression varied substantially across stages (Kruskal–Wallis H = 12.617, p = 0.013). (B) Pairwise median directional shifts illustrating progressive elevation of ICOS from NASH to cirrhosis and adjacent tissue. Full pairwise statistics are provided in Supplementary Table S3. (C) Density distribution plot demonstrating rightward shift of ICOS expression in cirrhosis and tumor samples relative to healthy liver. Expression values are log2-normalized. These findings show that ICOS dysregulation emerges during advanced inflammatory remodeling and persists to malignancy.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/652202643eac8686e8b67b4c.png"},{"id":105564600,"identity":"b0f9b183-2e36-4c5a-b2a4-1d7d3ff475ea","added_by":"auto","created_at":"2026-03-27 12:50:10","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":499839,"visible":true,"origin":"","legend":"\u003cp\u003eICOS-high tumors exhibit coordinated immune pathways and gene-level remodeling. (A) Hallmark Gene Set Enrichment Analysis (GSEA) of ICOS-High versus ICOS-Low tumors. Immune-related pathways, including interferon signaling, inflammatory response, and TNFα signaling via NF-κB, were enriched in ICOS-High tumors (FDR q \u0026lt; 0.25). (B) Immune signature heatmap showing median program-level enrichment across costimulatory checkpoints, inhibitory checkpoints, T-cell activation, T-cell exhaustion, and cytokine–chemokine signaling. (C) Tumor-level immune gene heatmap (z-score normalized) demonstrating coordinated upregulation of costimulatory receptors (ICOS, TNFRSF9, CD28), inhibitory checkpoints (PDCD1, CTLA4, LAG3, TIGIT, HAVCR2), cytotoxic effectors (IFNG, GZMB, PRF1), and inflammatory chemokines (CXCL9, CXCL10, CCL5) in ICOS-High tumors. Together, these panels demonstrate that ICOS enrichment reflects coordinated immune remodeling instead of isolated gene activation.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/cb0dffb279c1d2ffc768b104.png"},{"id":105564529,"identity":"16a8126e-f031-4cf6-8b7e-335771dc9afc","added_by":"auto","created_at":"2026-03-27 12:49:54","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1380429,"visible":true,"origin":"","legend":"\u003cp\u003eICOS-associated immune architecture is independent of viral etiology and tumor stage in TCGA-LIHC. (A) ICOS expression by etiology (viral vs non-viral). No significant difference was observed (Mann–Whitney p = 0.65). (B) Correlation of ICOS with exhaustion score in viral tumors (ρ = 0.75, p \u0026lt; 0.001). (C) Correlation of ICOS with exhaustion score in non-viral tumors (ρ = 0.72, p \u0026lt; 0.001). No significant difference in correlation strength (Fisher’s z = 0.73, p = 0.47). These findings show that the ICOS–exhaustion axis is conserved across etiologies.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/c9d74d2774886e07e491d35d.png"},{"id":105564583,"identity":"6009eb8c-6e12-4cc6-8d87-65abe89e58f8","added_by":"auto","created_at":"2026-03-27 12:50:07","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":629217,"visible":true,"origin":"","legend":"\u003cp\u003eICOS structures immune architecture but does not independently predict survival. (A) Kaplan–Meier survival curves comparing ICOS-High and ICOS-Low tumors (log-rank p \u0026gt; 0.05). (B) Multivariable Cox regression forest plot showing hazard ratios (HR) for ICOS expression and tumor stage. ICOS was not independently associated with survival (HR ≈ 1.00, p \u0026gt; 0.05), whereas stage remained significant. (C) Immune architecture correlation matrix illustrating coordinated upregulation of ICOS, exhaustion, Treg, and CD8 programs. No preferential skewing toward Treg dominance was observed. These data indicate that ICOS reflects immune ecology instead of intrinsic tumor aggressiveness.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/6dad9c6b4458dc988cae1d57.png"},{"id":105564987,"identity":"0e985ef9-0c6d-42c3-b76d-df7162765e64","added_by":"auto","created_at":"2026-03-27 12:51:31","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":865788,"visible":true,"origin":"","legend":"\u003cp\u003eUnsupervised immune ecosystem stratification identifies an ICOS-enriched T-cell–inflamed subtype. (A) Principal component analysis (PCA) of standardized ICOS, exhaustion, Treg, and CD8 scores reveal two immune clusters. (B) Cluster-level immune feature heatmap demonstrating coordinated elevation of ICOS, exhaustion, Treg, and CD8 programs in Cluster 1 relative to Cluster 0. (C) Kaplan–Meier survival analysis by immune cluster showing no independent survival difference (log-rank p = 0.26). Silhouette analysis supported a two-cluster solution (silhouette score = 0.36). These data indicate that ICOS defines immune ecosystem states rather than prognostic tumor biology.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/307fd0d4ab5df86b063a8479.png"},{"id":105569977,"identity":"6b80cce9-e5d0-4de8-a8d8-ff795bcc4e67","added_by":"auto","created_at":"2026-03-27 13:14:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7495647,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/5bf744ff-bcee-4480-9ce6-1467b6de3f85.pdf"},{"id":105564993,"identity":"e88ac557-c781-45d4-88a6-9b637d000ab8","added_by":"auto","created_at":"2026-03-27 12:51:31","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":10495,"visible":true,"origin":"","legend":"","description":"","filename":"TabelS1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/89842a50df1cca40ffef5833.xlsx"},{"id":105564533,"identity":"58b984e9-be19-4987-bd0a-cb310fb61d08","added_by":"auto","created_at":"2026-03-27 12:49:55","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":31388,"visible":true,"origin":"","legend":"","description":"","filename":"TableS0.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/99e2e3fa3ea5f58ad63ccb99.xlsx"},{"id":105308931,"identity":"5337429e-5d0d-4f1c-b7da-0afc8d2ad1a6","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":12714,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/7407013700d064ffeaa6d2e5.xlsx"},{"id":105564809,"identity":"432179a0-2005-431b-8ea6-b8492ab03857","added_by":"auto","created_at":"2026-03-27 12:50:54","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":10542,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/41ab240b7819ebccff2d11ee.xlsx"},{"id":105308940,"identity":"08027912-f0b0-4586-ab2a-25b0c910ab2e","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":20362,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/44763bb1b097ce46b0c6a011.xlsx"},{"id":105564954,"identity":"645fdcf9-3f40-48a9-9961-abdf916d19aa","added_by":"auto","created_at":"2026-03-27 12:51:25","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":10649,"visible":true,"origin":"","legend":"","description":"","filename":"TableS5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/9015de7d218e5a8ed7476377.xlsx"},{"id":105308937,"identity":"bc140d88-991a-4ccd-8030-4b0ab2b84da5","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":10398,"visible":true,"origin":"","legend":"","description":"","filename":"TableS8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/7cb0faebb4faf2d8e9f9ff13.xlsx"},{"id":105564969,"identity":"19fcd90d-6419-49a7-849a-0798f7a306e1","added_by":"auto","created_at":"2026-03-27 12:51:29","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":39188,"visible":true,"origin":"","legend":"","description":"","filename":"TableS6.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/9efb70c78c0f69efe0b28bb5.xlsx"},{"id":105308946,"identity":"15df7e0d-4e98-44bb-b478-1eebb3fe654e","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":10520,"visible":true,"origin":"","legend":"","description":"","filename":"TableS9.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/7f6aaa3e7384c497732011bc.xlsx"},{"id":105308944,"identity":"219c35c4-b424-4eaf-af6d-c34eb54e22b5","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":10604,"visible":true,"origin":"","legend":"","description":"","filename":"TableS10.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/13e542bd97727d474557524a.xlsx"},{"id":105308934,"identity":"1d3c728f-e1a3-4eac-9bd8-82a7c580426e","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":10358,"visible":true,"origin":"","legend":"","description":"","filename":"TableS7.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/0d7259ce91b62147ec871f27.xlsx"},{"id":105565177,"identity":"ecfa88cb-0adc-429a-abd5-6581d2a128aa","added_by":"auto","created_at":"2026-03-27 12:52:15","extension":"xlsx","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":10787,"visible":true,"origin":"","legend":"","description":"","filename":"TableS11.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/a1042af9258581d1844181e9.xlsx"},{"id":105308936,"identity":"65d51325-b229-4efc-8d38-2403c09e57b7","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":10457,"visible":true,"origin":"","legend":"","description":"","filename":"TableS12.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/29e610d0f8495d435c3edd1d.xlsx"},{"id":105308938,"identity":"d7f352ee-3ac7-419a-8a94-268a01727896","added_by":"auto","created_at":"2026-03-24 15:01:07","extension":"xlsx","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":72236,"visible":true,"origin":"","legend":"","description":"","filename":"TableS13.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9060848/v1/7eba1cd8d4ba273307aa0ca5.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eInducible T-Cell Co-Stimulator Emerges During Chronic Liver Disease Progression and Defines a T-Cell–Inflamed Immune State in Hepatocellular Carcinoma\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHepatocellular carcinoma (HCC) is the sixth most diagnosed cancer and the third leading cause of cancer-related death worldwide [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. It is driven by both viral hepatitis and the increasing prevalence of metabolic dysfunction\u0026ndash;associated steatotic liver disease (MASLD) [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. MASLD can progress to metabolic dysfunction\u0026ndash;associated steatohepatitis (MASH), advanced fibrosis, cirrhosis, and ultimately HCC through a multifaceted interaction of metabolic injury, chronic inflammation, and immune dysregulation [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. The growing prevalence of MASLD-associated HCC underscores the need to better understand immune regulatory mechanisms that function across disease stages [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe liver is immunologically unique. It maintains tolerance and mounts adaptive immune responses [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Chronic liver inflammation disturbs this balance, resulting in sustained inflammation and remodeling of the immune microenvironment [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Previous studies using single-cell transcriptomics have shown substantial heterogeneity among tumor-infiltrating immune cells [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. This includes the coexistence of T-cell activation, dysfunction, and exhaustion states [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Chronic inflammation is defined by T-cell exhaustion, sustained expression of inhibitory receptors, and transcriptional reprogramming [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. These immune adaptations are clinically relevant, as demonstrated by the survival benefit of immune checkpoint blockade in advanced HCC [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAmong costimulatory receptors, inducible T-cell co-stimulator (ICOS) has attracted increasing attention [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. ICOS belongs to the CD28 family and plays a role in T-cell activation, maintenance, and differentiation [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Its signalling contributes to follicular helper T-cell development, cytokine production, and immune regulation [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. In cancer, ICOS is recognized for its double-edged role [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. On one side, ICOS supports classical effector functions, including activation and persistence of T cells within the tumour microenvironment [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. On the other hand, ICOS signalling also promotes regulatory T-cell expansion, which can dampen anti-tumour immunity [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. The most decisive findings from recent studies emphasize that ICOS expression marks distinct T-cell states within inflamed tumours, bridging effector activity with regulatory and exhaustion programs [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. In HCC, where chronic antigen exposure and inflammatory remodelling are pervasive, ICOS may operate as part of coordinated immune architectures, rather than as an isolated checkpoint molecule [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite growing interest in ICOS for use in immunotherapy, its regulation in a spectrum of metabolic liver diseases remains unclear [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Most investigations evaluate ICOS within established malignancy or experimental models, without resolving whether dysregulation begins during early metabolic dysfunction, arises during fibrotic remodeling, or primarily accompanies malignant transformation [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. A key unanswered question is: at what point during disease progression does ICOS expression diverge from that of its ligand, ICOSLG? Does this dissociation mark a pivotal event in immune microenvironment remodeling, or do ICOS and ICOSLG maintain synchronous transcriptional dynamics across the chronic liver disease spectrum? It is also still unclear whether ICOS reflects tumor burden and anatomical stage, or whether it instead defines a conserved immune microenvironment state independent of tumor progression [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. The transcriptional behavior of its ligand, ICOSLG, across progressive liver disease is even less characterized, despite frequent assumptions of coordinated receptor\u0026ndash;ligand regulation [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. By framing these gaps as concrete, testable questions, this study aims to invite the reader directly into the scientific inquiry underlying our experimental approach.\u003c/p\u003e \u003cp\u003eClarifying the stage-specific emergence of ICOS is necessary for interpreting its biological and translational significance [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Misattributing late-stage immune remodeling features to early pathogenic mechanisms may conceal the true sequence of immune adaptation during hepatocarcinogenesis [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. A stage-resolved analysis is therefore necessary to define when ICOS dysregulation arises and whether it reflects structural immune remodeling rather than tumor expansion alone [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe hypothesized that ICOS dysregulation does not occur during early metabolic dysfunction but instead emerges during advanced inflammatory remodeling and malignant transformation. We further hypothesized that once induced, ICOS becomes embedded within a coordinated T-cell\u0026ndash;inflamed immune architecture independent of tumor stage and etiology.\u003c/p\u003e \u003cp\u003eTo test these hypotheses, we performed, to our knowledge, the first multi-cohort study mapping ICOS emergence across the entire liver disease continuum. This staged multi-cohort transcriptomic investigation integrates early metabolic liver disease datasets, advanced inflammatory and HCC cohorts, and independent validation in the TCGA-LIHC dataset. Through pathway enrichment, immune signature profiling, multivariable modeling, and unsupervised immune ecosystem stratification, we sought to define the temporal emergence and biological positioning of ICOS across chronic liver disease progression and hepatocellular carcinoma. The conceptual framework guiding this study is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Methodology","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy Design and Analytical Scheme\u003c/h2\u003e \u003cp\u003eThis study was conducted as a staged, multi-cohort transcriptomic investigation. The main objective was to define the temporal and stage-dependent regulation of ICOS across metabolic liver disease progression and hepatocellular carcinoma (HCC). It was hypothesized that ICOS dysregulation arises during advanced inflammatory remodeling and malignant transformation, rather than during early metabolic dysfunction. Furthermore, it was posited that, in these later stages, ICOS dysregulation appears associated with coordinated reprogramming of the immune microenvironment.\u003c/p\u003e \u003cp\u003eTo test these hypotheses, a two-phase analytical scheme was implemented. In the first phase, transcriptomic profiling was performed across early and advanced metabolic liver disease using independent human liver microarray datasets. ICOS and ICOSLG expression dynamics, immune pathway enrichment, and immune gene signature remodeling in hepatocellular carcinoma were evaluated. The second phase validated these outcomes using an independent TCGA-LIHC cohort to assess tumor-stage independence, immune exhaustion coupling, multivariable modeling, interaction testing, survival analysis, and immune ecosystem stratification.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStatistical Model\u003c/h3\u003e\n\u003cp\u003eAll analyses were conducted in Python using pandas, NumPy, SciPy, statsmodels, lifelines, scikit-learn, seaborn, matplotlib, and gseapy. All tests were two-sided, with nominal significance defined as p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Where applicable, Benjamini\u0026ndash;Hochberg correction was applied.\u003c/p\u003e\n\u003ch3\u003ePhase 1: Transcriptomic Analysis of ICOS Regulation in Metabolic Liver Disease\u003c/h3\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eDataset Selection\u003c/h2\u003e \u003cp\u003eHuman liver transcriptomic datasets were retrieved from the Gene Expression Omnibus (GEO). Inclusion criteria were: (i) human liver tissue, (ii) representation of both non-malignant and malignant disease states, (iii) available stage annotation, and (iv) platform-specific probe annotation files.\u003c/p\u003e \u003cp\u003eTwo datasets met the criteria. GSE89632 (GPL14951; Illumina HumanHT-12) served as the early disease validation cohort. It included Healthy (n\u0026thinsp;=\u0026thinsp;24), MASLD (n\u0026thinsp;=\u0026thinsp;20), and MASH (n\u0026thinsp;=\u0026thinsp;19) samples (n\u0026thinsp;=\u0026thinsp;63 total). GSE164760 (GPL13667; Affymetrix Human Genome U219) served as the advanced disease discovery cohort and included Healthy (n\u0026thinsp;=\u0026thinsp;6), NASH (n\u0026thinsp;=\u0026thinsp;74), Cirrhosis (n\u0026thinsp;=\u0026thinsp;8), Adjacent tissue (n\u0026thinsp;=\u0026thinsp;29), and HCC tumors (n\u0026thinsp;=\u0026thinsp;53) (n\u0026thinsp;=\u0026thinsp;170 total).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eProbe Processing and Gene-Level Collapsing\u003c/h3\u003e\n\u003cp\u003eSeries matrix files were downloaded from GEO, and platform-specific annotation files (GPL14951 and GPL13667) were used for probe-to-gene mapping. Probes without valid gene symbols were excluded. When multiple probes mapped to a single gene, expression values were collapsed using the median probe values. Expression values were retained on the log2-normalized scale. No cross-dataset normalization or batch correction was performed, as each dataset was analyzed independently.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eDisease Stage Classification\u003c/h2\u003e \u003cp\u003eStage groups were derived from curated GEO annotations. For the GSE89632 cohort, samples were grouped as Healthy, MASLD, or MASH. In GSE164760, samples were categorized as Healthy, NASH, Cirrhosis, Adjacent non-tumoral tissue, or HCC tumor. GSM accession identifiers were used as unique keys for merging metadata with gene expression matrices.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eICOS and ICOSLG Expression Analysis\u003c/h3\u003e\n\u003cp\u003eICOS and ICOSLG expression values were extracted from gene-level matrices. Distributional assessment confirmed non-normal expression. For this reason, Kruskal\u0026ndash;Wallis testing was used for multi-group comparisons. In GSE89632, expressions were compared across Healthy, MASLD, and MASH. As for the GSE164760, cohort comparisons were done across Healthy, NASH, Cirrhosis, Adjacent, and Tumor groups. Pairwise median differences were calculated to quantify directional shifts across stages.\u003c/p\u003e\n\u003ch3\u003eICOS Stratification in HCC\u003c/h3\u003e\n\u003cp\u003eHCC tumors (n\u0026thinsp;=\u0026thinsp;53) in GSE164760 were dichotomized into ICOS-High and ICOS-Low groups using the median ICOS expression cutoff to preserve balanced group sizes and maximize statistical power for downstream analyses. The decision to use the median rather than a fixed percentile or quartile cutoff was based on the relatively small cohort size, which enabled reliable comparisons between groups of equal size while minimizing potential bias from arbitrary thresholds. Importantly, the median split approach has been commonly employed in transcriptomic analysis to facilitate reproducibility and clear reporting. Group separation was confirmed using the Mann\u0026ndash;Whitney U test.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003ePathway Enrichment Analysis\u003c/h2\u003e \u003cp\u003ePre-ranked Gene Set Enrichment Analysis (GSEA) was performed using the MSigDB Hallmark 2020 collection. Genes were ranked by the delta-median expression difference between ICOS-High and ICOS-Low tumors. Enrichment was performed using 1,000 permutations, with gene set size thresholds of 15\u0026ndash;500 genes. False discovery rate (FDR) q\u0026thinsp;\u0026lt;\u0026thinsp;0.25 was considered significant.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eImmune Signature Profiling\u003c/h2\u003e \u003cp\u003eCurated immune gene panels representing T-cell activation, exhaustion, costimulatory checkpoints, inhibitory checkpoints, and cytokine\u0026ndash;chemokine signaling were assembled from immuno-oncology literature. For each tumor sample, signature scores were calculated as the mean log2 expression of genes within each panel. Group comparisons were performed using Mann\u0026ndash;Whitney U testing with Benjamini\u0026ndash;Hochberg correction.\u003c/p\u003e \u003cp\u003eIn addition to group-level comparisons, tumor-level visualization of immune architecture was performed using the full gene expression matrix. Selected immune regulatory genes were extracted, standardized using z-score normalization, and hierarchically clustered to evaluate coordinated immune remodeling across ICOS-defined tumor subsets.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003ePhase 2: Validation, Multivariable Modeling, and Immune Ecosystem Stratification in TCGA-LIHC\u003c/h2\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003eTCGA Data Acquisition and Harmonization\u003c/h2\u003e \u003cp\u003eRNA sequencing data (RNASeqV2, log2-normalized) and clinical annotations were obtained from the UCSC Xena portal for the TCGA-LIHC dataset. Clinical variables included overall survival and pathological tumor stage. Molecular and clinical data were merged using the first twelve characters of TCGA barcodes. Samples lacking stage or survival data were excluded from stage-adjusted and survival analyses. All analyses used provided log2-normalized expression values without further transformation.\u003c/p\u003e \u003cp\u003eTumor stage was extracted from and harmonized into an ordinal numeric variable (Stage I\u0026ndash;IV). Substages were collapsed into primary-stage categories to preserve ordering and ensure adequate sample sizes per group.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eEtiology and Stage Association\u003c/h2\u003e \u003cp\u003eTumors were stratified into viral and non-viral groups using viral hepatitis serology annotations. Differences in ICOS expression between etiologic groups were assessed using the Mann\u0026ndash;Whitney U test. Associations between ICOS expression and tumor stage were evaluated using Spearman's rank correlation and Kruskal\u0026ndash;Wallis testing. Differences in correlation strength were evaluated using Fisher\u0026rsquo;s r-to-z transformation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eImmune Functional Score Construction\u003c/h2\u003e \u003cp\u003eComposite transcriptional scores were constructed to quantify immune architecture. The exhaustion score was defined as the mean log2 expression of PDCD1, CTLA4, LAG3, TIGIT, and HAVCR2. The Treg score was calculated as the mean expression of FOXP3, IL2RA, and CTLA4. The CD8 score was computed using the mean expression of CD8A and CD8B. A Treg/CD8 balance metric was calculated by subtracting the CD8 score from the Treg score. The calculations were computed using vectorized operations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eMultivariable Modeling of Exhaustion Architecture\u003c/h2\u003e \u003cp\u003eOrdinary least squares (OLS) regression was used to evaluate whether ICOS independently predicts exhaustion architecture. The primary model specification was:\u003c/p\u003e \u003cp\u003eExhaustion score\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric\u003c/p\u003e \u003cp\u003eRegression coefficients (β), 95% confidence intervals, R\u0026sup2;, and F-statistics were calculated using statsmodels. To test stage-dependent modulation of immune coupling, an interaction term was introduced:\u003c/p\u003e \u003cp\u003eExhaustion score\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric + (ICOS expression \u0026times; Stage numeric)\u003c/p\u003e \u003cp\u003eA non-significant interaction term was interpreted as evidence of stage-independent ICOS\u0026ndash;exhaustion coupling.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eSurvival Modeling\u003c/h2\u003e \u003cp\u003eCox proportional hazards regression was performed using overall survival as the endpoint. Two prespecified models were fitted:\u003c/p\u003e \u003cp\u003eModel 1: OS\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric\u003c/p\u003e \u003cp\u003eModel 2: OS\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric\u0026thinsp;+\u0026thinsp;Exhaustion score\u003c/p\u003e \u003cp\u003eHazard ratios (HR), 95% confidence intervals, Wald p-values, concordance index (C-index), and partial Akaike Information Criterion (AIC) were calculated using lifelines. Kaplan\u0026ndash;Meier survival curves were generated using median ICOS dichotomization and compared using the log-rank test.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eImmune Ecosystem Stratification\u003c/h2\u003e \u003cp\u003eTo identify immune architectural subtypes, unsupervised k-means clustering was performed using standardized ICOS expression, exhaustion score, Treg score, and CD8 score. Features were z-score normalized using StandardScaler prior to clustering. Cluster number (k\u0026thinsp;=\u0026thinsp;2\u0026ndash;5) was assessed using silhouette score analysis. The optimal solution was selected based on the maximal silhouette coefficient.\u003c/p\u003e \u003cp\u003eCluster-level mean values of immune features were calculated to characterize biological subtypes. Principal component analysis (PCA) was applied to visualize cluster separation in reduced-dimensional space. Kaplan\u0026ndash;Meier analysis was used to evaluate survival differences across immune clusters.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eSection 1: Stage-Dependent Transcriptional Remodeling of ICOS Across Metabolic Liver Disease and HCC\u003c/h2\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003eResults Overview\u003c/h2\u003e \u003cp\u003eThe temporal emergence of ICOS dysregulation during the progression of metabolic liver disease was delineated using a staged analytical approach that integrated early metabolic dysfunction (GSE89632) and advanced inflammatory remodeling in hepatocellular carcinoma (GSE164760). This section is organized into three components: (i) evaluation of ICOS and ICOSLG expression in early metabolic disease, (ii) stage-resolved transcriptional dynamics across advanced liver pathology and HCC, and (iii) immune pathway and gene-level remodeling in ICOS-stratified tumors. Collectively, these analyses establish the timing of ICOS dysregulation and clarify whether it represents isolated gene activation or coordinated immune reprogramming. All figures and Supplementary tables are summarized in Table S0.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eICOS and ICOSLG Remain Stable During Early Metabolic Dysfunction\u003c/h2\u003e \u003cp\u003eICOS and ICOSLG expression were evaluated across Healthy (n\u0026thinsp;=\u0026thinsp;24), MASLD (n\u0026thinsp;=\u0026thinsp;20), and MASH (n\u0026thinsp;=\u0026thinsp;19) liver samples in GSE89632 to determine whether ICOS dysregulation initiates during early metabolic perturbation.\u003c/p\u003e \u003cp\u003eICOS expression demonstrated no significant difference across early disease states (Kruskal\u0026ndash;Wallis H\u0026thinsp;=\u0026thinsp;0.842, p\u0026thinsp;=\u0026thinsp;0.656). Median expression values were comparable across groups, with overlapping interquartile ranges and no directional shift (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Similarly, ICOSLG expression remained transcriptionally stable (H\u0026thinsp;=\u0026thinsp;0.969, p\u0026thinsp;=\u0026thinsp;0.615; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC). Descriptive statistics are in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. These findings demonstrate that neither ICOS nor its ligand, ICOSLG, undergoes transcriptional modulation during early metabolic liver disease, indicating that dysregulation is not initiated during the initial steatotic or inflammatory transitions.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eICOS Becomes Dysregulated During Advanced Inflammatory Remodeling and Malignant Transformation\u003c/h2\u003e \u003cp\u003eIn contrast to early disease stability, ICOS expression varied substantially across advanced pathological stages in GSE164760 (Healthy n\u0026thinsp;=\u0026thinsp;6; NASH n\u0026thinsp;=\u0026thinsp;74; Cirrhosis n\u0026thinsp;=\u0026thinsp;8; Adjacent n\u0026thinsp;=\u0026thinsp;29; Tumor n\u0026thinsp;=\u0026thinsp;53).\u003c/p\u003e \u003cp\u003eThe Kruskal\u0026ndash;Wallis test showed significant stage-dependent variation (H\u0026thinsp;=\u0026thinsp;12.617, p\u0026thinsp;=\u0026thinsp;0.013; Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Median ICOS expression progressively increased from non-malignant liver to cirrhosis and adjacent tissue, with sustained elevation in tumor samples (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). These results suggest a progressive immunological shift rather than a single event. Directional pairwise median comparisons (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB) revealed upward shifts from NASH to cirrhosis and from cirrhosis to adjacent tissue. Tumor samples displayed heterogeneous but persistently elevated ICOS levels relative to early-stage metabolic disease. Full pairwise statistics are reported in Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eKernel density analysis also supported a rightward distributional shift in cirrhotic and tumor samples relative to healthy liver (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). Adjacent tissue exhibited intermediate expression patterns, consistent with progressive microenvironment remodeling during chronic inflammation.\u003c/p\u003e \u003cp\u003eImportantly, ICOSLG did not vary substantially across advanced stages (H\u0026thinsp;=\u0026thinsp;3.443, p\u0026thinsp;=\u0026thinsp;0.486; Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e) indicating receptor\u0026ndash;ligand divergence during disease progression.\u003c/p\u003e \u003cp\u003eCollectively, these findings show that ICOS dysregulation emerges during advanced inflammatory remodeling and persists into hepatocellular carcinoma, whereas ICOSLG remains transcriptionally stable. This divergence implies receptor-specific immune regulatory amplification rather than coordinated ligand upregulation.\u003c/p\u003e \u003cp\u003e \u003cem\u003eICOS-High Tumors showed Coordinated Immune Pathway and Gene-Level Remodeling.\u003c/em\u003e \u003c/p\u003e \u003cp\u003eHCC tumors were stratified into ICOS-High (n\u0026thinsp;=\u0026thinsp;27) and ICOS-Low (n\u0026thinsp;=\u0026thinsp;26) groups using a median cutoff to assess whether ICOS dysregulation remains associated with wider immune microenvironment remodeling. Pre-ranked GSEA showed significant enrichment of immune-related Hallmark pathways in ICOS-High tumors (normalized enrichment scores\u0026thinsp;\u0026gt;\u0026thinsp;1.5; FDR q\u0026thinsp;\u0026lt;\u0026thinsp;0.25), including interferon signaling, inflammatory response, and TNFα signaling via NF-κB (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Complete enrichment statistics are provided in Table \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e and Table \u003cspan refid=\"MOESM4\" class=\"InternalRef\"\u003eS4\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo quantify coordinated immune remodeling at the programmatic level, curated immune gene panels were used for evaluation. ICOS-High tumors demonstrated elevated signature scores across costimulatory checkpoints, inhibitory checkpoints, T-cell activation, T-cell exhaustion, and cytokine\u0026ndash;chemokine programs (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). The most pronounced difference was observed in costimulatory signaling (Δ median\u0026thinsp;=\u0026thinsp;4.58; FDR\u0026thinsp;=\u0026thinsp;0.0025), while inhibitory and activation programs showed concordant directional shifts (Table \u003cspan refid=\"MOESM5\" class=\"InternalRef\"\u003eS5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo evaluate immune architecture at the tumor level, a focused immune regulatory gene panel was extracted from the full expression matrix. After z-score normalization and hierarchical clustering, ICOS-High tumors exhibited coordinated upregulation of costimulatory receptors (ICOS, TNFRSF9, CD28), inhibitory checkpoints (PDCD1, CTLA4, LAG3, TIGIT, HAVCR2), effector activation genes (IFNG, GZMB, PRF1), and inflammatory chemokines (CXCL9, CXCL10, CCL5) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC).\u003c/p\u003e \u003cp\u003eImportantly, clustering demonstrated that these programs rise in concert rather than independently, indicating that ICOS dysregulation shows a coordinated immune regulatory state instead of isolated gene activation.\u003c/p\u003e \u003cp\u003eThese data establish that ICOS enrichment in HCC is embedded within a wider T-cell\u0026ndash;inflamed transcriptional architecture characterized by simultaneous activation and checkpoint engagement.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eSection 2: External Validation, Architectural Stability, and Immune Ecosystem Stratification in TCGA-LIHC\u003c/h2\u003e \u003cdiv id=\"Sec26\" class=\"Section4\"\u003e \u003ch2\u003eResults Overview\u003c/h2\u003e \u003cp\u003eMulti-layer validation in the TCGA-LIHC cohort was performed to determine whether ICOS-associated immune remodeling observed during metabolic liver disease progression persists within established malignancy and reflects conserved immune architecture rather than tumor burden. This validation framework interrogated (i) etiology and stage independence of ICOS expression, (ii) multivariable modeling of the ICOS\u0026ndash;exhaustion axis and overall survival, and (iii) ecosystem-level immune stratification using unsupervised clustering. Collectively, these analyses establish whether ICOS functions as a tumor burden marker or as a structural determinant of immune architecture.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eICOS Is Independent of Etiology and Tumor Stage in TCGA HCC\u003c/h2\u003e \u003cp\u003eWe first examined whether ICOS expression differed between viral and non-viral hepatocellular carcinoma. ICOS expression distributions were comparable across etiologic groups (Mann\u0026ndash;Whitney p\u0026thinsp;=\u0026thinsp;0.65), with overlapping medians and interquartile ranges (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). The processed dataset supporting this analysis is provided in Table \u003cspan refid=\"MOESM6\" class=\"InternalRef\"\u003eS6\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDespite the absence of etiologic differences in ICOS abundance, ICOS strongly correlated with immune exhaustion score in both viral (ρ\u0026thinsp;=\u0026thinsp;0.75, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and non-viral tumors (ρ\u0026thinsp;=\u0026thinsp;0.72, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB\u0026ndash;C). A direct statistical comparison using Fisher\u0026rsquo;s r-to-z transformation showed no significant difference in correlation strength (z\u0026thinsp;=\u0026thinsp;0.73, p\u0026thinsp;=\u0026thinsp;0.47), confirming the conservation of the ICOS\u0026ndash;exhaustion axis across etiologies. The full comparative statistics are reported in Table \u003cspan refid=\"MOESM7\" class=\"InternalRef\"\u003eS7\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eWe next evaluated whether ICOS expression escalates with tumor burden. Spearman correlation revealed no association between ICOS and pathologic stage (ρ\u0026thinsp;=\u0026thinsp;0.03, p\u0026thinsp;=\u0026thinsp;0.37). Stage-stratified scatter visualization (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD) further demonstrated that ICOS\u0026ndash;exhaustion coupling continues across stages without stage-dependent amplification. Stage-specific correlation coefficients are provided in Table \u003cspan refid=\"MOESM8\" class=\"InternalRef\"\u003eS8\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eIn multivariable regression adjusting for tumor stage (Exhaustion score\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric), ICOS remained a strong independent predictor of exhaustion architecture (β\u0026thinsp;=\u0026thinsp;0.555, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), whereas stage was not significant (p\u0026thinsp;=\u0026thinsp;0.89). The model explained 49.8% of the variance in exhaustion (R\u0026sup2; = 0.498). Detailed regression coefficients are provided in Table \u003cspan refid=\"MOESM10\" class=\"InternalRef\"\u003eS10\u003c/span\u003e and cross-validated in Table \u003cspan refid=\"MOESM9\" class=\"InternalRef\"\u003eS9\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThese data collectively show that ICOS structures immune exhaustion architecture independently of etiology and tumor stage, indicating architectural stability rather than tumor-burden dependence.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eICOS Does Not Independently Predict Overall Survival\u003c/h2\u003e \u003cp\u003eMultivariable Cox proportional hazards models were constructed to determine whether ICOS-mediated immune remodeling is associated with clinical outcomes. In stage-adjusted survival analysis (OS\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric), ICOS was not associated with overall survival (HR\u0026thinsp;\u0026asymp;\u0026thinsp;1.00, p\u0026thinsp;=\u0026thinsp;0.87), whereas tumor stage remained a significant predictor (HR\u0026thinsp;\u0026asymp;\u0026thinsp;1.20 per stage increment, p\u0026thinsp;=\u0026thinsp;0.01; concordance index\u0026thinsp;=\u0026thinsp;0.55). Kaplan\u0026ndash;Meier analysis confirmed no survival difference between ICOS-high and ICOS-low tumors (log-rank p\u0026thinsp;\u0026gt;\u0026thinsp;0.05; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn the fully adjusted model containing exhaustion score (OS\u0026thinsp;~\u0026thinsp;ICOS expression\u0026thinsp;+\u0026thinsp;Stage numeric\u0026thinsp;+\u0026thinsp;Exhaustion score), neither ICOS (p\u0026thinsp;=\u0026thinsp;0.93) nor exhaustion (HR\u0026thinsp;\u0026asymp;\u0026thinsp;0.98, p\u0026thinsp;=\u0026thinsp;0.77) independently predicted survival, whereas stage remained dominant (p\u0026thinsp;=\u0026thinsp;0.01; concordance index\u0026thinsp;=\u0026thinsp;0.52). Coefficients are provided in Table \u003cspan refid=\"MOESM10\" class=\"InternalRef\"\u003eS10\u003c/span\u003e and Table \u003cspan refid=\"MOESM11\" class=\"InternalRef\"\u003eS11\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThese relationships are summarized visually in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB, which demonstrates that ICOS and exhaustion hazard ratios cross unity in both models. Together, these data indicate that ICOS-mediated immune polarization reflects tumor immune ecology instead of intrinsic tumor aggressiveness.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003eICOS Marks T-Cell\u0026ndash;Inflamed Tumors Without Preferential Regulatory Skewing\u003c/h2\u003e \u003cp\u003eTranscriptional scores for regulatory T cells (Treg), CD8 cytotoxic T cells, and immune exhaustion were constructed to dissect immune composition.\u003c/p\u003e \u003cp\u003eICOS expression correlated strongly with both Treg score (ρ\u0026thinsp;=\u0026thinsp;0.566, p\u0026thinsp;\u0026asymp;\u0026thinsp;1.2 \u0026times; 10⁻\u0026sup1;\u0026sup1;\u0026sup2;) and CD8 score (ρ\u0026thinsp;=\u0026thinsp;0.503, p\u0026thinsp;\u0026asymp;\u0026thinsp;2.7 \u0026times; 10⁻\u0026sup1;\u0026sup1;\u0026sup1;). Correlation statistics are provided in Table \u003cspan refid=\"MOESM12\" class=\"InternalRef\"\u003eS12\u003c/span\u003e. Fisher\u0026rsquo;s r-to-z comparison demonstrated no significant difference between Treg and CD8 correlations (z\u0026thinsp;=\u0026thinsp;0.105, p\u0026thinsp;=\u0026thinsp;0.917), and ICOS did not correlate with the Treg/CD8 balance metric (ρ = \u0026minus;0.023, p\u0026thinsp;=\u0026thinsp;0.34).\u003c/p\u003e \u003cp\u003eThe immune architecture correlation matrix (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC) illustrates coordinated upregulation of ICOS, exhaustion, Treg, and CD8 programs, with no preferential skewing toward suppressive dominance. These data demonstrate that ICOS marks globally T-cell\u0026ndash;inflamed tumors rather than selectively regulatory-biased microenvironments.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eImmune Ecosystem Stratification Identifies a T-Cell–Inflamed ICOS-High Subtype\u003c/h3\u003e\n\u003cp\u003eUnsupervised k-means clustering was performed by standardized ICOS, exhaustion, Treg, and CD8 scores to evaluate whether ICOS defines higher-order immune states beyond linear associations. Silhouette analysis supported a two-cluster solution (silhouette score\u0026thinsp;=\u0026thinsp;0.36). Cluster classifications are provided in Table \u003cspan refid=\"MOESM13\" class=\"InternalRef\"\u003eS13\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eCluster 1 exhibited coordinated elevation of ICOS (mean 5.15), exhaustion (6.29), Treg (5.11), and CD8 (7.02), defining a T-cell\u0026ndash;inflamed immune subtype. Cluster 0 demonstrated uniformly lower immune feature values (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB). Principal component analysis confirmed partition along PC1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDespite a distinct immune profile, cluster membership did not independently forecast survival (log-rank p\u0026thinsp;=\u0026thinsp;0.26; Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC). This data supports the notion that ICOS defines immunological framework states rather than prognostic tumor biology in untreated TCGA HCC.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec32\" class=\"Section2\"\u003e \u003ch2\u003eICOS as a Stage-Emergent Architect of Immune Remodeling in HCC\u003c/h2\u003e \u003cp\u003eThis multi-cohort transcriptomic study shows that ICOS functions not solely as a marker of tumour burden or prognosis, but as a stage-emergent regulator of immune architecture in hepatocellular carcinoma (HCC). By combining early metabolic liver disease datasets, advanced inflammatory remodelling cohorts, and external validation in TCGA-LIHC, this study formulates a model in which ICOS dysregulation arises during advanced inflammatory transformation. Subsequently, ICOS stabilizes following malignancy and defines a conserved T-cell\u0026ndash;inflamed immune ecosystem within established tumours. The results position ICOS as a structural determinant of the tumour microenvironment rather than a linear immune checkpoint molecule [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cdiv id=\"Sec33\" class=\"Section3\"\u003e \u003ch2\u003eICOS Emerges Late in Disease Evolution\u003c/h2\u003e \u003cp\u003eAn important finding of this study is the temporal specificity of ICOS dysregulation. In early metabolic liver disease (MASLD and MASH), both ICOS and its ligand ICOSLG remained transcriptionally stable. Despite metabolic stress and inflammation, there was no evidence of progressive ICOS induction. These results suggest that ICOS is not involved in the initial immune perturbation associated with steatosis or early inflammatory activation, which are mainly driven by innate immune signaling and metabolic injury [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn contrast, ICOS expression increased significantly during advanced pathological states, including cirrhosis, adjacent non-tumoral tissue, and HCC. The distributional shift across these stages was gradual rather than sudden. These observations support a model of progressive immune remodeling rather than oncogenic activation alone [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. ICOSLG did not mirror this pattern, indicating receptor-specific amplification rather than coordinated ligand upregulation. These outcomes suggest that ICOS upregulation is linked to chronic inflammatory remodeling and the transition toward a tumor-permissive immune microenvironment [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec34\" class=\"Section3\"\u003e \u003ch2\u003eICOS-High Tumors Exhibit Coordinated Immune Activation and Exhaustion\u003c/h2\u003e \u003cp\u003eWhen HCC tumors were stratified by ICOS expression, ICOS-high tumors showed coordinated enrichment of interferon signaling, inflammatory response pathways, and TNFα\u0026ndash;NFκB signaling. At the gene level, ICOS-high tumors showed simultaneous upregulation of costimulatory receptors, inhibitory checkpoints, cytotoxic effector genes, and inflammatory chemokines, with these programs rising together. This shows a synchronized T-cell\u0026ndash;inflamed transcriptional state consistent with prior single-cell observations in HCC [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Previous studies have characterized ICOS expression within established tumors or individual immune cell subsets, however our analysis demonstrates that ICOS-associated immune architecture emerges progressively during disease evolution and remains stable after malignant transformation.\u003c/p\u003e \u003cp\u003eSimultaneous elevations in activation and exhaustion signatures were observed, consistent with chronic antigen exposure and sustained immune engagement [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. The results support the interpretation that ICOS marks an immune microenvironment characterized by sustained T-cell activation alongside compensatory checkpoint regulation. Instead of representing immune suppression alone, ICOS appears to be embedded within a dynamic regulatory balance, a phenomenon well described in chronic inflammatory and cancer settings [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e\n\u003ch3\u003eImmune Architecture Is Independent of Etiology and Tumor Stage\u003c/h3\u003e\n\u003cp\u003eValidation in TCGA-LIHC clarified that ICOS expression is independent of both viral etiology and pathological tumor stage. ICOS did not differ between viral and non-viral tumors and did not increase with anatomical stage progression. This dissociation is biologically meaningful. Tumor stage reflects macroscopic growth and invasion, whereas ICOS reflects the immune microenvironment state [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. This suggests that once established, ICOS-associated immune architecture remains stable throughout tumor evolution.\u003c/p\u003e \u003cp\u003eDespite this stage independence, ICOS consistently correlated with immune exhaustion across etiologies and stage strata. Multivariable modeling confirmed that ICOS independently predicts exhaustion architecture even after adjusting for tumor stage. Together, these data show that the ICOS\u0026ndash;exhaustion axis is embedded within the tumor microenvironment and maintained irrespective of tumor burden, consistent with recognized models of exhaustion stability in chronic disease [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eICOS Shows a T-Cell–Inflamed State Without Preferential Regulatory Skewing\u003c/h3\u003e\n\u003cp\u003eA key question is whether ICOS enrichment reflects preferential expansion of regulatory T cells. Prior studies have shown that ICOS can be expressed by both effector and regulatory T-cell subsets, depending on context [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. However, our data does not support selective regulatory dominance. ICOS correlated strongly with both Treg and CD8 cytotoxic signatures, and no association was observed with the Treg/CD8 balance metric.\u003c/p\u003e \u003cp\u003eUnsupervised clustering identified two immune ecosystems: an immune-low subtype and a T-cell\u0026ndash;inflamed subtype characterized by coordinated elevation of ICOS, exhaustion, Treg, and CD8 programs. ICOS was enriched within the inflamed subtype but did not skew immune balance toward suppressive predominance. This is consistent with reports that ICOS functions as a context-dependent modulator of immune activation rather than an inherently suppressive signal [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThus, ICOS expression at the bulk tumor level likely reflects global T-cell engagement instead of selective regulatory expansion, consistent with tumor-infiltrating lymphocyte heterogeneity described in HCC [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. These findings suggest that ICOS-defined immune states may represent biologically relevant contexts for future ICOS-targeted immunotherapies, where modulation of the ICOS pathway could potentially amplify pre-existing T-cell\u0026ndash;inflamed tumor microenvironments.\u003c/p\u003e \u003cdiv id=\"Sec37\" class=\"Section2\"\u003e \u003ch2\u003eImmune Remodeling Does Not Independently Predict Survival\u003c/h2\u003e \u003cp\u003eDespite clear immune stratification, ICOS expression and immune cluster membership did not independently predict overall survival in TCGA. Tumor stage remained the major determinant of prognosis, consistent with established epidemiological and clinical observations in HCC [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis highlights a biological separation between immune architecture and anatomical tumor progression. In largely untreated cohorts such as TCGA, tumor burden is expected to dominate survival outcomes [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Immune polarization may instead influence therapeutic responsiveness rather than baseline prognosis, notably in the context of immune checkpoint blockade [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Therefore, ICOS-defined immune states should be interpreted as markers of the immune context rather than as indicators of intrinsic tumor aggressiveness.\u003c/p\u003e \u003cdiv id=\"Sec38\" class=\"Section3\"\u003e \u003ch2\u003eA Hierarchical Model of Immune Remodeling\u003c/h2\u003e \u003cp\u003eThe results support a hierarchical framework:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eICOS is not induced during early metabolic dysfunction.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eIt emerges during advanced inflammatory remodeling and malignant transformation.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eWithin the established HCC, ICOS defines a stable T-cell\u0026ndash;inflamed immune ecosystem.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eThis immune architecture is independent of tumor stage and etiology.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eImmune ecosystem stratification does not independently determine survival in untreated cohorts.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThis model clarifies the apparent paradox between disease-stage emergence and tumor-stage independence. ICOS is induced during inflammatory transformation instead of anatomical tumor expansion, consistent with progressive immune adaptation described in chronic liver disease and cancer immunology [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec39\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThis study is based on bulk transcriptomic data and therefore cannot resolve cell-type-specific ICOS expression or spatial immune organization. As a direct next step, single-cell and spatial profiling approaches\u0026mdash;such as single-cell RNA sequencing and multiplex imaging\u0026mdash;should be employed to dissect the cellular sources of ICOS and map its distribution within the liver tumor microenvironment [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Additionally, trajectory inference methods could unravel the dynamics of immune cell state transitions during disease progression [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Another limitation is that the TCGA dataset predominantly reflects untreated tumors. Future studies should specifically examine immunotherapy-treated patient cohorts to determine whether ICOS-defined immune ecosystems can predict treatment responsiveness, especially in the context of immune checkpoint blockade, where T-cell inflamed states regularly forecast response [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. By prioritizing these directions, future research can address the current gaps and directly test the translational relevance of ICOS-associated immune states.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eTo sum up, ICOS functions as a stage-emergent architect of immune remodeling in hepatocellular carcinoma. It is induced during advanced inflammatory progression, persists following malignant transformation, and defines a conserved T-cell\u0026ndash;inflamed tumor ecosystem independent of tumor stage and etiology. By reframing ICOS as a structural regulator of immune ecology rather than a simple checkpoint marker or prognostic factor, this study provides a mechanistic model linking chronic inflammatory liver disease to stable immune architecture in HCC.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cem\u003eConflict of Interest\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eClinical Trial Number\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eClinical Trial Number not applicable\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eFunding Declaration\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eNo funding was received for this project\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAvailability of Data and Materials\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAll data used in the project is publicly available on GEO repository. All supplementary data and figs that support this project are uploaded on Zenedo and can be accessed via the following link.\u0026nbsp;\u003cbr\u003e\u0026nbsp; https://doi.org/10.5281/zenodo.18903688\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAuthor Contributions\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contributed equally to the \u0026nbsp;conception, design, and execution of the study. The overall manuscript was drafted and critically revised by all authors for important intellectual content. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eInformed Consent\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eInformed consent was not required for this study, as it involved secondary analysis of publicly available, anonymised data.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, Jemal A (2024) Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Cancer J Clin 74(3):229\u0026ndash;263. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3322/caac.21834\u003c/span\u003e\u003cspan address=\"10.3322/caac.21834\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGlynn KA, Petrick JL, El-Serag HB (2021) Epidemiology of hepatocellular carcinoma. Hepatology 73(S1):4\u0026ndash;13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/hep.31288\u003c/span\u003e\u003cspan address=\"10.1002/hep.31288\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYounossi ZM, Kalligeros M, Henry L (2024) Epidemiology of metabolic dysfunction-associated steatotic liver disease. Clin Mol Hepatol 31(Suppl):S32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3350/cmh.2024.0431\u003c/span\u003e\u003cspan address=\"10.3350/cmh.2024.0431\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEstes C, Razavi H, Loomba R, Younossi Z, Sanyal AJ (2018) Modeling the epidemic of nonalcoholic fatty liver disease demonstrates an exponential increase in burden of disease. Hepatology 67(1):123\u0026ndash;133. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/hep.29466\u003c/span\u003e\u003cspan address=\"10.1002/hep.29466\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePfister D, N\u0026uacute;\u0026ntilde;ez NG, Pinyol R, Govaere O, Pinter M, Szydlowska M, Gupta R, Qiu M, Deczkowska A, Weiner A et al (2021) NASH limits anti-tumour surveillance in immunotherapy-treated hepatocellular carcinoma. Nature 592(7854):450\u0026ndash;456. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-021-03362-0\u003c/span\u003e\u003cspan address=\"10.1038/s41586-021-03362-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLlovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, Lencioni R, Koike K, Zucman-Rossi J, Finn RS (2021) Hepatocellular carcinoma. Nat Reviews Disease Primers 7:6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41572-020-00240-3\u003c/span\u003e\u003cspan address=\"10.1038/s41572-020-00240-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Q, He Y, Luo N, Patel SJ, Gao R, Modak M, Carotta S, Haslinger C, Kind D, Van den Brink MRM et al (2019) Landscape and dynamics of single immune cells in hepatocellular carcinoma. Cell 179(4):829\u0026ndash;845. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2019.10.003\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2019.10.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun Y, Wu L, Zhong Y, Zhou K, Hou Y, Wang Z, Zhang Y, Wang Y, Zhang Y, Yuan J et al (2021) Single-cell landscape of the ecosystem in early-relapse hepatocellular carcinoma. Cell 184(2):404\u0026ndash;421. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2020.11.041\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2020.11.041\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeltra JC, Manne S, Abdel-Hakeem MS, Kurachi M, Giles JR, Chen Z, Casella V, Ngiow SF, Khan O, Huang Y et al (2020) Developmental relationships of four exhausted CD8 T cell subsets reveal underlying transcriptional and epigenetic control mechanisms. Immunity 52(5):825\u0026ndash;841. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.immuni.2020.04.014\u003c/span\u003e\u003cspan address=\"10.1016/j.immuni.2020.04.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFinn RS, Qin S, Ikeda M, Galle PR, Ducreux M, Kim TY, Kudo M, Breder V, Merle P, Kaseb AO et al (2020) Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma. N Engl J Med 382(20):1894\u0026ndash;1905. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1056/NEJMoa1915745\u003c/span\u003e\u003cspan address=\"10.1056/NEJMoa1915745\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWikenheiser DJ, Stumhofer JS (2016) ICOS co-stimulation: Friend or foe? Front Immunol 7:304. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fimmu.2016.00304\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2016.00304\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrotty S (2019) T follicular helper cell biology: a decade of discovery and diseases. Immunity 50(5):1132\u0026ndash;1148. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.immuni.2019.04.011\u003c/span\u003e\u003cspan address=\"10.1016/j.immuni.2019.04.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmatore F, Gorvel L, Olive D (2020) Role of inducible co-stimulator (ICOS) in cancer immunotherapy. Expert Opin Biol Ther 20(2):141\u0026ndash;150. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14712598.2020.1693540\u003c/span\u003e\u003cspan address=\"10.1080/14712598.2020.1693540\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlank CU, Haining WN, Held W, Hogan PG, Kallies A, Lugli E et al (2019) Defining \u0026lsquo;T cell exhaustion\u0026rsquo;. Nat Rev Immunol 19(11):665\u0026ndash;674. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41577-019-0221-9\u003c/span\u003e\u003cspan address=\"10.1038/s41577-019-0221-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSade-Feldman M, Yizhak K, Bjorgaard SL, Ray JP, de Boer CG, Jenkins RW et al (2019) Defining T cell states associated with response to checkpoint immunotherapy in melanoma. Cell 176(1\u0026ndash;2):404\u0026ndash;419. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2018.12.034\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2018.12.034\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNikanjam M, Kato S, Nishizaki D et al (2026) Inducible T-cell co-stimulator (ICOS) and ICOS ligand: Dealing with a two-faced cancer immunoregulatory system. Cancer Med 15(1):e71467. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/cam4.71467\u003c/span\u003e\u003cspan address=\"10.1002/cam4.71467\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen KY, Zhu Y, Xie SZ et al (2024) Immunosuppressive tumor microenvironment and immunotherapy of hepatocellular carcinoma: Current status and perspectives. J Hematol Oncol 17:25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13045-024-01549-2\u003c/span\u003e\u003cspan address=\"10.1186/s13045-024-01549-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhan O, Giles JR, McDonald S et al (2019) TOX transcriptionally and epigenetically programs CD8 T cell exhaustion. Nature 571(7764):211\u0026ndash;218. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-019-1325-x\u003c/span\u003e\u003cspan address=\"10.1038/s41586-019-1325-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa L, Hernandez MO, Zhao Y, Mehta M, Tran B, Kelly M et al (2019) Tumor cell biodiversity drives microenvironmental reprogramming in liver cancer. Cancer Cell 36(4):418\u0026ndash;430. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ccell.2019.08.007\u003c/span\u003e\u003cspan address=\"10.1016/j.ccell.2019.08.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"HCC, ICOS, immune remodeling, TCGA, transcriptomics","lastPublishedDoi":"10.21203/rs.3.rs-9060848/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9060848/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eHepatocellular carcinoma (HCC) arises within chronic liver disease and is molded by complex immune remodeling. Inducible T-cell costimulator (ICOS) is an immune checkpoint associated with T-cell activation, but its timing and role across disease progression remain unclear. This study aimed to clarify the stage-specific regulation of ICOS and determine its role in molding immune architecture during hepatocellular carcinoma development.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe conducted staged transcriptomic profiling integrating early metabolic liver dysfunction (GSE89632), advanced inflammatory remodeling and HCC (GSE164760), and external validation in The Cancer Genome Atlas Liver Hepatocellular Carcinoma cohort (TCGA-LIHC). ICOS dynamics were evaluated across disease stages, followed by pathway enrichment, immune signature profiling, multivariable modeling, and unsupervised immune ecosystem stratification.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eICOS expression was stable in early disease but increased significantly during advanced inflammatory remodeling and malignant transformation. ICOS-high tumors exhibited coordinated enrichment of interferon, costimulatory, inhibitory, and effector T-cell programs, showing a stable T-cell\u0026ndash;inflamed immune architecture. In TCGA-LIHC, ICOS expression was independent of viral etiology and tumor stage but remained tightly associated with immune exhaustion signatures after multivariable adjustment. Unsupervised clustering defined an ICOS-enriched immune subtype, distinct from tumor burden and not independently predictive of survival.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eICOS emerges during advanced liver disease and remains as a structural regulator of immune architecture in HCC. These outcomes position ICOS as a marker of immune ecosystem state rather than tumor progression per se. Looking ahead, ICOS-defined immune states could serve as a practical basis for stratifying patients in future interventional studies, enabling personalized selection for ICOS agonist or antagonist therapies.\u003c/p\u003e","manuscriptTitle":"Inducible T-Cell Co-Stimulator Emerges During Chronic Liver Disease Progression and Defines a T-Cell–Inflamed Immune State in Hepatocellular Carcinoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-24 15:01:01","doi":"10.21203/rs.3.rs-9060848/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1bc829b3-daa3-46aa-88d1-f76afe0ea9af","owner":[],"postedDate":"March 24th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-24T15:01:02+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-24 15:01:01","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9060848","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9060848","identity":"rs-9060848","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}