Gastric adenocarcinoma remains a leading cause of cancer-related mortality worldwide (1). The tumor microenvironment (TME) plays a critical role in tumor progression and response to immunotherapy (2). Recently, cellular senescence has emerged as a key modulator of the TME (3). The cyclin-dependent kinase inhibitor p16^INK4a, encoded by the CDKN2A gene, is a primary driver and marker of senescence (4).

In various malignancies, CDKN2A over-expression has been paradoxically linked to tumor progression, theoretically driven by the senescence-associated secretory phenotype (SASP). The SASP comprises a complex array of cytokines, chemokines, and immunomodulatory factors that can establish an immunoevasive contexture. Previous hypotheses suggest that senescent tumor cells utilize SASP factors – specifically transforming growth factor beta (TGF-beta), CD73 (NT5E), and indoleamine 2,3-dioxygenase 1 (IDO1) – to suppress or exclude cytotoxic immune infiltrates.

Recent literature highlights the complex role of cellular senescence in the gastric tumor microenvironment, particularly concerning the aberrant expression of the p16^INK4a protein. Wang et al. recently identified a distinct triple-classification staining pattern for p16^INK4a in gastric cancer, stratifying tumors into Loss, Wild-Type, and Over-Expression subgroups (5). Crucially, while p16^INK4a loss and wild-type tumors exhibit comparable clinical trajectories, p16^INK4a over-expression serves as a significant prognostic biomarker for aggressive disease. Clinically, this over-expression correlates with advanced pathologic T stage (pT stage), diminished overall survival, and marked resistance to both adjuvant chemotherapy and immune checkpoint inhibitors. This aggressive phenotype is biologically underpinned by an immunoevasive contexture, characterized by the exclusion of antitumor infiltrates like CD8+ T cells and elevated levels of immunosuppressive factors such as TGF-beta, CD73, and IDO (5, 6).

This study aimed to validate these mechanistic claims using transcriptomic data from The Cancer Genome Atlas Stomach Adenocarcinoma (TCGA-STAD) cohort, with computational deconvolution to assess whether CDKN2A over-expression correlates with CD8+ T cell exclusion and specific SASP factor up-regulation in a large patient population.

Data acquisition and processing. Bulk RNA-sequencing data (HTSeq-Counts) and matching clinical annotations for patients with gastric adenocarcinoma were obtained from the TCGA-STAD dataset via the Genomic Data Commons (GDC) portal. Gene expression values were normalized using the Trimmed Mean of M-values (TMM) method and log2-transformed. Duplicate gene entries were collapsed by calculating the mathematical mean using the limma framework (version 3.58.1).

Patient stratification. Patients were stratified into cohorts based on CDKN2A mRNA Z-scores calculated relative to the entire cohort (n=386): Loss (n=57): Z-score <−1.0; Wild-Type (WT, n=253): Z-score between −1.0 and +1.0; Over-expression (OE, n=76): Z-score >+1.0.

Computational deconvolution. Estimation of absolute immune cell infiltration fractions was performed using the immunedeconv R package (version 2.1.0) (7). The quanTIseq algorithm (8) was selected for its validated ability to provide absolute cell fractions, allowing for direct comparison of CD8+ T cells, M1 Macrophages, and Neutrophils across different patient samples.

Statistical analysis. All analyses were conducted in R (version 4.3.2). Differences in immune cell fractions and SASP gene expression (TGFB1, NT5E, IDO1) across cohorts were assessed using the Kruskal-Wallis test. Pairwise comparisons between the OE and WT groups were performed using the Wilcoxon rank-sum test. The relationship between SASP expression and CD8+ T cell infiltration was analyzed using Spearman rank correlation. Survival distributions were estimated using the Kaplan–Meier method and compared via the log-rank test using the survival (v3.5-7) and survminer (v0.4.9) packages. A p-value <0.05 was considered statistically significant.

Patient characteristics and cohort stratification. The study analyzed a total of 386 patients from the TCGA-STAD cohort with complete transcriptomic and clinical annotations. Patients were stratified based on CDKN2A mRNA expression levels into three distinct groups: WT (n=253), Loss (n=57), and OE (n=76). The mean age at diagnosis for the entire cohort was 65.1 years (SD=10.5), and the population was predominantly male (65%). No statistically significant differences were observed across the three CDKN2A cohorts regarding age or sex, ensuring these demographic variables did not confound subsequent immune analyses. Notably, the CDKN2A OE subgroup was associated with more advanced disease, with 49% of patients presenting with Pathologic Stage III (IIIA, IIIB, or IIIC) compared to 41% in the WT group. This distribution aligns with reported phenotypes where p16^INK4a high expression correlates with advanced pT stage progression. A tumor pT stage refers to the pathological classification of the primary tumor based on direct examination of tissue removed during surgery. The “p” stands for pathologic, and “T” stands for tumor, evaluating the size and depth of invasion into nearby tissues. pT stage is more precise than clinical staging and guides post-surgery treatment (Table I).

CDKN2A over-expression predicts CD8+ T cell exclusion. To determine the impact of CDKN2A status on the tumor microenvironment (TME), quanTIseq estimated immune fractions were compared across the stratified cohorts. Tumors with CDKN2A over-expression exhibited a statistically significant depletion of infiltrating CD8+ T cells compared to the Wild-Type cohort. Conversely, no significant differences were observed in the infiltration of M1 macrophages (p=0.44) or neutrophils (p=0.57) across the cohorts. These findings transcriptomically validate p16^INK4a over-expression as a specific biomarker for a CD8+ T cell-depleted TME in gastric adenocarcinoma (Figure 1).

While CDKN2A over-expression was associated with statistically significant CD8+ T cell depletion (p=0.024), the magnitude of this effect was characterized as negligible to small [Cliff’s delta=-0.14; 95% confidence interval (CI)=-0.27, -0.01]. This modest transcriptomic effect size, contrasting with the prognostic disparities seen at the protein level, likely reflects the inherent decoupling between CDKN2A mRNA abundance and functional p16^INK4a protein activity in the gastric tumor microenvironment.

SASPfactor up-regulation and adaptive immune resistance. Expression levels of proposed SASP mediators were evaluated to assess their mechanistic role in immune evasion. IDO1 demonstrated significant up-regulation in the CDKN2A OE cohort compared to WT, whereas NT5E (CD73) and TGFB1 showed no significant expression differences (Figure 2).

Direct correlation analysis further revealed that IDO1 expression maintained a strong, highly significant positive correlation with CD8+ T cell infiltration (R=0.64, p=2.2×10^-16) (Figure 3).

Similarly, TGFB1 expression exhibited a significant positive correlation with CD8+ T cell fractions (R=0.36, p=4.1×10^-13). These positive correlations contradict primary exclusion models and instead suggest a mechanism of adaptive immune resistance, where these factors are reactively up-regulated in response to existing T cell presence rather than serving as the primary drivers of physical T cell exclusion (Figure 4).

Prognostic value of CDKN2A mRNA status survival outcomes were evaluated across the stratified CDKN2A cohorts (n=386) using Kaplan–Meier estimates and the log-rank test. While the OE group exhibited an early trend toward inferior overall survival, particularly between 24 and 48 months, the differences across the three groups did not reach statistical significance in the whole TCGA-STAD transcriptomic cohort (p=0.3). This contrasts with the highly significant survival disparities observed in the ZSHS IHC-based cohort (p=0.004) (5) and may reflect the inherent heterogeneity of bulk transcriptomic data compared to localized protein expression. Notably, the Loss cohort showed a trend toward superior survival outcomes in the first five years of follow-up compared to both the WT and OE groups (Figure 5).

Molecularsubtype survival.Subtype-stratified survival analysis was performed to determine if the prognostic significance of CDKN2A is context-dependent. In both the chromosomal instability (CIN n=124, p=0.76) and microsatellite instability (MSI n=53, p=0.73) subtypes, no statistically significant differences in overall survival were observed across the WT, Loss, and OE cohorts. These transcriptomic findings contrast with the protein-level immunohistochemistry (IHC) results reported by Wang et al. and suggest that CDKN2A mRNA levels may lack sufficient prognostic resolution compared to localized protein quantification.

The complex interplay between cellular senescence and tumor immunity represents a critical frontier in gastric cancer research. This computational analysis of the TCGA-STAD cohort (n=386) confirms that CDKN2A mRNA over-expression is a statistically significant transcriptomic biomarker for a CD8+ T cell-depleted tumor microenvironment. These results provide independent validation of the clinical observations reported by Wang et al., who established that the p16^INK4a high phenotype correlates with advanced pathologic stage and an “immunoevasive contexture”.

However, our findings reveal a notable divergence regarding the specific mediators of this evasion. While Wang et al. utilized IHC to demonstrate significantly higher protein expression of TGF-beta and CD73 in p16^INK4a high tumors, our transcriptomic analysis found no significant differences in TGFB1 or NT5E mRNA levels across the CDKN2A cohorts.

Our extensive subtype-stratified analysis (CIN and MSI) failed to replicate the significant survival disparities reported by Wang et al. While we confirmed the biological immune contexture (CD8+ T cell depletion and IDO1 up-regulation), the lack of survival significance in CIN and MSI subtypes (p=0.76 and p=0.73, respectively) underscores a profound decoupling between CDKN2A transcript abundance and clinical outcomes. This divergence suggests that the aggressive clinical trajectory of p16^INK4a high tumors is likely driven by post-translational protein stability, spatial immune exclusion, or enzymatic SASP activity that is not captured by bulk RNA-sequencing. Consequently, while CDKN2A mRNA is a marker for immune infiltration status, protein-level IHC remains the gold standard for clinical prognosis in gastric adenocarcinoma.

This discordance likely reflects the inherent decoupling between mRNA transcripts and functional protein levels. Factors such as TGF-beta undergo extensive post-translational regulation, and the enzymatic activity of CD73 (encoded by NT5E) is not always mirrored by transcript abundance. Furthermore, the IHC evidence provided by Wang et al. captures spatial localization that bulk RNA-sequencing averages across the tumor biopsy. These results suggest that while CDKN2A over-expression remains an indicator of a “cold” tumor, the specific SASP exclusionary mechanism may be driven by proteomic or metabolic shifts rather than a generalized up-regulation of SASP-related transcripts.

IDO1and adaptive immune resistance. A critical finding of this study is the significant up-regulation of IDO1 in the CDKN2A over-expressed cohort (p=0.015), which aligns with the protein-level increases observed by Wang et al. (p=0.026). However, the strong, highly significant positive correlation between IDO1 expression and CD8+ T cell infiltration (R=0.64) challenges the model of IDO1 as a primary exclusionary factor.

Instead, this dynamic is indicative of adaptive immune resistance. In this paradigm, initial infiltration by CD8+ T cells leads to the secretion of interferon-gamma (IFN-gamma), which in turn drives the reactive up-regulation of IDO1 as a feedback inhibitory mechanism by the tumor and surrounding stroma.

Therefore, in CDKN2A-over-expressing gastric cancers, IDO1 acts as a potent reactive immune checkpoint rather than a factor that physically bars T cells from the tumor microenvironment. This distinction is clinically important, as it suggests that the poor immunotherapy response observed by Wang et al. in p16^INK4a high patients may be partially reversible through the strategic use of IDO1 inhibitors in combination with existing checkpoint blockade.

This mechanistic refinement has therapeutic implications. Wang et al. demonstrated that p16^INK4a OE tumors respond poorly to immune checkpoint inhibitors, particularly in PD-L1 CPS ≥1 and CIN subsets. Our data suggest that IDO1-mediated adaptive resistance may contribute to this phenomenon. Consequently, CDKN2A-over-expressing gastric adenocarcinomas might benefit from combination immunotherapy strategies incorporating IDO1 inhibitors (e.g., epacadostat, linrodostat) alongside PD-1/PD-L1 blockade (9). While early-phase IDO1 inhibitor trials have yielded mixed results (10), patient selection based on CDKN2A status and IDO1 expression may identify a more responsive subset.

Study limitations. While this study provides computational validation of CDKN2A as a transcriptomic biomarker, several limitations must be acknowledged:

Transcriptomic vs. proteomic divergence: This analysis relied exclusively on bulk RNA-sequencing data from the TCGA-STAD cohort. In contrast, Wang et al. primarily utilized IHC to define p16^INK4a status and immune cell densities. The lack of correlation observed here for M1 macrophages and neutrophils may reflect the inherent “decoupling” between mRNA expression and functional protein levels, or differences in spatial localization that bulk sequencing cannot capture.

Deconvolution constraints: Although the quanTIseq algorithm is highly validated for quantifying absolute immune fractions, it remains a mathematical estimation based on standardized gene signatures. These estimates may be influenced by the “purity” of the tumor samples or the presence of non-malignant stromal cells that also express SASP-related transcripts.

Retrospective nature and cohort bias: The TCGA-STAD dataset is a retrospective repository representing a diverse global population. The specific clinical outcomes and immune phenotypes reported by Wang et al. were derived from localized cohorts in Shanghai and Seoul. Variations in patient ethnicity, dietary factors, and H. pylori prevalence across these cohorts may contribute to the differing immune landscapes.

Bulk vs. single-cell resolution: Bulk RNA-seq averages the signal across the entire tumor biopsy. Consequently, we cannot definitively distinguish whether IDO1 or TGFB1 transcripts are originating from the senescent tumor cells themselves (as a true SASP) or from infiltrating myeloid and stromal cells in the surrounding niche.

Lack of functional validation: This study is purely observational and computational. While the strong positive correlation between IDO1 and CD8+ T cells suggests a mechanism of adaptive immune resistance, in vitro functional assays or multiplexed spatial profiling would be required to confirm this interaction.

The computational deconvolution and transcriptomic analysis of the TCGA-STAD cohort (n=386) provides validation of CDKN2A mRNA over-expression as a significant biomarker for a CD8+ T cell-depleted tumor microenvironment in gastric adenocarcinoma. These findings align with the clinical and pathological observations reported by Wang et al., confirming that the p16^INK4a high phenotype identifies a subset of patients with aggressive disease and an immunoevasive contexture.

However, our data suggests a more nuanced mechanistic landscape than a generalized, secretory exclusion model. The lack of transcript-level up-regulation for TGFB1 and NT5E suggests that the immunoevasive properties of these specific factors in CDKN2A-over-expressing tumors may be driven by post-translational regulation or spatial sequestration rather than a simple increase in SASP-related mRNA.

Furthermore, the strong positive correlation between IDO1 and CD8+ T cell infiltration indicates that IDO1 serves as a marker of adaptive immune resistance. In this paradigm, IDO1 is reactively up-regulated in response to cytotoxic T cell activity, functioning as a localized immune checkpoint rather than a primary barrier to T cell entry. These results refine the biological understanding of senescence in the gastric microenvironment and highlight IDO1 inhibition as a potential therapeutic strategy to overcome immunotherapy resistance in CDKN2A-over-expressing gastric cancers.

The Authors declare that they have no competing interests in relation to this study.

Steven Lehrer conceptualized the study, performed data acquisition and statistical analyses, interpreted the results, and drafted the manuscript. Peter Rheinstein contributed to study design, provided critical revisions of the manuscript for important intellectual content, and assisted in the interpretation of findings. Both Authors reviewed and approved the final version of the manuscript and agree to be accountable for all aspects of the work. This work was supported in part through the computational and data resources and staff expertise provided by Scientific Computing and Data at the Icahn School of Medicine at Mount Sinai.

None.

During the preparation of this manuscript, a large language model was used solely for language editing and stylistic improvements in select paragraphs. No sections involving the generation, analysis, or interpretation of research data were produced by generative AI. All scientific content was created and verified by the authors. Furthermore, no figures or visual data were generated or modified using generative AI or machine learning–based image enhancement tools.