A conserved 8-gene epithelial barrier signature spatially quantifies immune exclusion across cancers and predicts immunotherapy outcome independent of HPV status

A conserved 8-gene epithelial barrier signature spatially quantifies immune exclusion across cancers and predicts immunotherapy outcome independent of HPV status | 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 A conserved 8-gene epithelial barrier signature spatially quantifies immune exclusion across cancers and predicts immunotherapy outcome independent of HPV status Tan Haosheng, Yu Dapeng, Jiao lianghe This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9472940/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 Tumor epithelial barriers physically impede immune cell infiltration, causing immune exclusion and resistance to immune checkpoint inhibitors (ICIs). However, a simple, spatially quantifiable biomarker of epithelial barrier strength is lacking. In head and neck squamous cell carcinoma (HNSCC), HPV status is used as a surrogate for ICI response, but its association with true immune exclusion remains unclear. Methods Using Xenium and Visium spatial transcriptomics data from nine cancer types (pancreatic, lung, breast, colorectal, cervical, ovarian, HNSCC, skin melanoma, kidney, and acute lymphoblastic leukemia bone marrow), we defined a compact 8‑gene epithelial barrier signature (EpiBarrier), together with a stromal signature and an immune activity signature. Spatial niche clustering, pan‑cancer comparisons, and external immunotherapy cohort validation were performed to test whether high EpiBarrier regions exhibit low stromal and low immune activity. In HNSCC, we compared the signature with HPV status. Results In six epithelial cancer types (pancreatic, lung, breast, colorectal, cervical, ovarian), high‑EpiBarrier regions consistently showed significantly lower stromal scores (17–87% reduction) and lower immune activity scores (5–73% reduction) (all P < 0.05). In skin melanoma, high‑EpiBarrier cells corresponded to normal epidermal structures and also exhibited reduced stromal and immune activity (P = 8e‑27), confirming the signature’s specificity to genuine epithelial cells. Non‑epithelial tumors (kidney, ALL bone marrow) did not support the model or showed opposite trends. In HNSCC, the model outcome was independent of HPV status: HPV‑high samples could be immune‑excluded (model supported) or immune‑hot (model not supported), while HPV‑negative samples were immune‑excluded. An external ICI‑treated HNSCC cohort (GSE159067, n = 102) validated that EpiBarrier predicted overall survival (HR = 1.744, P = 0.0205) and moderately predicted treatment response (AUC = 0.702). Conclusions The 8‑gene epithelial barrier signature is a pan‑epithelial‑cancer, spatially quantifiable biomarker of immune exclusion, independent of HPV status, and predicts immunotherapy outcome. It offers a clinically translatable tool for patient stratification in epithelial‑derived malignancies. Immunology Translational Medicine epithelial barrier signature spatial transcriptomics immune exclusion immunotherapy outcome Figures Figure 1 1. Background Immune checkpoint inhibitors (ICIs) have revolutionized cancer therapy, yet many patients do not respond due to primary resistance, often mediated by “immune exclusion” – the physical blockade of T cells by tumor epithelial barriers ( 1 , 2 ). Tight junctions, adherens junctions, and desmosomes create a barrier that prevents lymphocyte infiltration into tumor nests. However, no simple spatial biomarker exists to quantify epithelial barrier strength directly in tissue sections. In head and neck squamous cell carcinoma (HNSCC), HPV status (p16 IHC or HPV DNA) is widely used as a surrogate for ICI response. Nevertheless, up to 30% of HPV‑positive patients fail to respond, while some HPV‑negative patients derive benefit ( 3 ), indicating that HPV status poorly reflects functional immune exclusion. Thus, a direct, quantitative, and easily measurable epithelial barrier signature is urgently needed. Spatial transcriptomics technologies (Xenium, Visium) enable simultaneous gene expression profiling and spatial localization, allowing the construction of functional scores from defined gene sets. Here, we developed a compact 8‑gene epithelial barrier signature, validated it across six epithelial cancer types, compared it with HPV status in HNSCC, and tested its prognostic and predictive value in an external ICI‑treated HNSCC cohort. We also included non‑epithelial tumors (kidney cancer, melanoma, ALL bone marrow) to assess model specificity. 2. Methods Ongoing work and future updates To strengthen the generalizability of the epithelial barrier signature, we are actively collecting and analyzing larger, independent spatial transcriptomics cohorts across multiple cancer types. As our work progresses, more detailed figures, tables, and supplementary materials will be generated. Consequently, the results presented in this preprint version are preliminary and subject to refinement. We will continuously update the findings, and the latest versions of all figures, tables, and supplementary information will be made available in subsequent releases of this manuscript. Spatial transcriptomics datasets We collected publicly available Xenium and Visium datasets: pancreatic cancer (Xenium, 190,965 cells), lung cancer (Xenium, 67,763), breast cancer (Xenium Prime, 699,110), colorectal cancer (Xenium, 388,175), cervical cancer (Xenium, 205,082), ovarian cancer (Xenium, 388,175), head and neck squamous cell carcinoma (5 samples, 611,962 cells), kidney cancer (Xenium, 465,334), melanoma (Xenium, 87,499), and acute lymphoblastic leukemia bone marrow (Xenium, 225,906, negative control). All data were log‑normalized (scale factor 10,000). Samples with > 50,000 cells were randomly down‑sampled to 50,000. Gene set definitions Three functional gene sets were defined based on literature and pan‑cancer expression stability: Epithelial barrier (EpiBarrier): 8 genes involved in cell adhesion, desmosomes, keratins, and pan‑epithelial markers. Stromal barrier: 8 cancer‑associated fibroblast and extracellular matrix genes (FAP, COL1A1, ACTA2, VIM, PDGFRA, PDGFRB, FN1, POSTN). Immune activity: 8 CD8 + T‑cell effector molecules, chemokines, and antigen‑presenting genes (CD8A, IFNG, GZMB, PRF1, CXCL9, CXCL10, HLA‑A, HLA‑B). For each cell/spot, the module score was the mean log‑normalized expression of the respective genes (Xenium data were first Z‑scored across cells). Cell type annotation (for neighborhood features) Epithelial cells were defined by expression of EPCAM, KRT8, KRT18, or KRT19 (> 0); immune cells by PTPRC, CD8A, CD4, CD68, or HLA‑A (> 0); stromal cells by FAP, COL1A1, or VIM (> 0). Priority: epithelial > immune > stromal, others as “Other”. Spatial niche analysis For each cell, neighbors within 50 µm were identified (RANN::nn2). Neighborhood averages of EpiBarrier, stromal, and immune scores were calculated. PCA (6 features: self scores + neighbor averages) followed by Louvain clustering (resolution 0.4) identified spatial niches. The top 3 niches by mean EpiBarrier were defined as high‑EpiBarrier niches. Statistical validation Non‑spatial: Cells were split by median EpiBarrier into high/low groups; Wilcoxon test compared stromal and immune scores. Correlation: Spearman correlations among the three scores. Linear regression: Immune activity ~ EpiBarrier + stromal score. Spatial comparison: Wilcoxon test between high and low EpiBarrier niches. HPV status: Detected by HPV16‑E6/E7 expression (> 0 cells defined as HPV‑positive). External immunotherapy cohort GSE159067 (102 HNSCC patients treated with anti‑PD‑1/PD‑L1) was downloaded. Log2CPM expression and clinical data (overall survival, event, response) were extracted. EpiBarrier score was calculated (Z‑scored mean). Optimal cutoff was determined by `surv_cutpoint`. Kaplan‑Meier and Cox regression assessed overall survival; ROC evaluated response prediction (CR/PR vs SD/PD). Statistics Two‑sided P < 0.05 was considered significant. 3. Results 3.1 The 8‑gene epithelial barrier signature identifies immune‑excluded niches in epithelial cancers We first validated the model in pancreatic cancer Xenium data. Cells were split into high‑ and low‑EpiBarrier groups by the median score. The high‑EpiBarrier group exhibited significantly lower stromal scores (mean 0.485 vs 1.044, P < 1e‑300) and lower immune activity scores (mean 0.075 vs 0.158, P < 1e‑130). Spatial niche analysis confirmed that high‑EpiBarrier niches had significantly lower stromal and immune scores (P < 0.001). Linear regression showed a negative coefficient for EpiBarrier (‑0.025, P < 1e‑112), indicating independent negative prediction of immune activity. The same pipeline was applied to lung, breast (Xenium Prime), colorectal, cervical, ovarian, head and neck (HNSCC), skin melanoma, kidney, and acute lymphoblastic leukemia (ALL) bone marrow datasets. Results are summarized in Table 1 . Table 1 Pan‑cancer validation of the 8‑gene EpiBarrier signature Cancer type Platform Model supported Immune activity change (high vs low EpiBarrier) Stromal change Remarks Pancreatic Xenium Yes ↓ 53% (0.075→0.158, P < 1e-130) ↓ 54% Epithelial Lung Xenium Yes ↓ 73% (0.021→0.078, P < 1e-146) ↓ 66% Epithelial Breast (Xenium Prime) Xenium Yes ↓ 54% (0.036→0.078, P = 3.2e-12) ↓ 59% Only CDH1 and EPCAM detected Colorectal Xenium Yes ↓ 13% (0.402→0.461, P = 4.9e-22) ↓ 18% Epithelial Cervical Xenium Yes ↓ 31% (0.035→0.051, P = 0.049) ↓ 87% Epithelial Ovarian Xenium Yes ↓ 5% (0.091→0.096, P = 3.3e-9) ↓ 73% Small but significant HNSCC (HPV-negative) Xenium Yes ↓ 52% (0.264→0.555, P < 0.001) ↓ 55% Laryngeal primary/metastasis HNSCC (HPV-positive, immune-hot) Xenium No (reversed) ↑ (0.228→0.175, P < 1e-26) ↑ Low HPV viral load Skin melanoma (normal epidermis) Xenium Yes ↓ 15% (0.996→1.172, P = 8e-27) ↓ 27% Corresponds to normal epidermal cells Kidney (clear cell RCC) Xenium No ↑ (P < 1e-180) ↑ Non-epithelial ALL bone marrow Xenium No (reversed) ↑ (0.261→0.203, P = 1.4e-7) ↑ Non-epithelial (negative control) In all epithelial cancers, high‑EpiBarrier regions consistently showed reduced stromal and immune activity, with reductions ranging from 5% to 73% (all P < 0.05). Notably, in breast Xenium Prime only two of the eight genes (CDH1 and EPCAM) were detected, yet the model remained highly significant, demonstrating robustness. Importantly, in skin melanoma – a non‑epithelial tumor – the model did identify high‑EpiBarrier cells that corresponded to normal epidermal structures (e.g., keratinocytes). Those cells exhibited significantly lower stromal and immune activity (P = 8e‑27), confirming that the signature faithfully captures epithelial barrier function wherever true epithelial cells are present. In contrast, in kidney cancer and ALL bone marrow, where no epithelial cells exist, the model either failed or showed the opposite trend (immune‑hot phenotype), underscoring its specificity to epithelial tissues. 3.2 HNSCC: model independent of HPV status We analyzed five HNSCC/laryngeal samples (Table 2 ). HPV status was determined by HPV16‑E6/E7 expression. HPV‑negative samples (GSM8193670/3671) were immune‑excluded (model supported). A high‑HPV sample (82.4% HPV+) was also immune‑excluded, whereas low‑HPV samples (< 2.2% HPV+) were immune‑hot (model not supported). Thus, HPV status cannot predict immune exclusion; the EpiBarrier score directly quantifies the functional phenotype. Table 2 HNSCC samples: HPV status and model outcome Sample HPV+ cells Model outcome Immune phenotype GSM9054474 82.40% Supported Excluded GSM9054471 0.80% Not supported Hot GSM9054487 2.20% Not supported Hot GSM8193670 0% Supported Excluded GSM8193671 0% Supported Excluded 3.3 Matched primary and metastatic lymph node analysis A paired analysis of primary laryngeal cancer (GSM8193670) and its regional lymph node metastasis (GSM8193671) showed that the model was significant in both sites, although the effect size was smaller in the metastasis (31% vs 52% immune reduction, both P < 0.001). This suggests that epithelial barrier function and its association with immune exclusion are maintained during tumor progression. 3.4 External immunotherapy cohort validation In the GSE159067 cohort (102 anti‑PD‑1/PD‑L1‑treated HNSCC patients), the EpiBarrier score as a continuous variable predicted overall survival (HR = 1.744, 95% CI [1.09–2.78], P = 0.0205). Using the optimal cutoff (‑0.246), high‑score patients had significantly worse survival (log‑rank P = 0.012, Fig. 1 A). The score moderately discriminated responders from non‑responders (AUC = 0.702, Fig. 1 B). Multivariable Cox regression adjusting for HPV status confirmed the independent predictive value of the EpiBarrier score (P = 0.026, Fig. 1 C). 4. Discussion We developed and validated a compact 8‑gene epithelial barrier signature that spatially quantifies immune exclusion in 6 epithelial cancer types (pancreatic, lung, breast, colorectal, cervical, ovarian). The signature is robust even when only two of the eight genes are detected (as in breast Xenium Prime). In skin melanoma, high‑EpiBarrier cells corresponded to normal epidermal structures and exhibited the expected immune‑excluded phenotype, further confirming that the signature faithfully captures epithelial barrier function wherever genuine epithelial cells are present. Conversely, in non‑epithelial tumors (kidney, ALL bone marrow) the model failed or showed opposite trends, underscoring its cancer‑type specificity. In HNSCC, the signature outperforms HPV status: HPV‑positive tumors can be either immune‑excluded or immune‑hot, and HPV‑negative tumors can also be immune‑excluded. Thus, HPV status should not be used as a surrogate for functional immune exclusion; direct measurement of epithelial barrier strength is required. The signature predicted overall survival and treatment response in an external ICI‑treated HNSCC cohort, suggesting clinical utility. The 8‑gene composition is attractive for translation into an immunohistochemistry (IHC) or qPCR panel. Limitations include the modest size of the external cohort (n = 102) and the lack of prospective validation. The model does not apply to immune‑hot tumors (e.g., some HPV‑positive HNSCC, kidney cancer) where active immunity may disrupt epithelial barriers. Future work will develop an IHC assay and test the signature in larger immunotherapy trials across multiple epithelial cancer types. Conclusions The 8‑gene epithelial barrier signature is a pan‑epithelial‑cancer, spatially quantifiable biomarker of immune exclusion. It is independent of HPV status and predicts immunotherapy outcome. This signature offers a clinically translatable tool for patient stratification in epithelial‑derived malignancies. Declarations Author contributions: T.H. conceived the study, performed data analysis, wrote the manuscript, and acted as corresponding author. Y.D. contributed to data interpretation and critical revision. J.L. supervised the project and contributed to funding acquisition. All authors approved the final version. This manuscript is prepared for preprint submission. All code and data are available upon request. References Joyce JA, Fearon DT (2015) T cell exclusion, immune privilege, and the tumor microenvironment. Science 348(6230):74–80 Chen DS, Mellman I (2017) Elements of cancer immunity and the cancer-immune set point. Nature 541(7637):321–330 Cillo AR et al (2020) Immune landscape of viral- and carcinogen-driven head and neck cancer. Immunity 52(1):183–199 Saintigny P et al (2022) Gene expression profiling of HNSCC patients treated with anti-PD-1/PD-L1 (GSE159067). GEO 10x Genomics Xenium In Situ and Visium Spatial Gene Expression documentation Additional Declarations The authors declare no competing interests. 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-9472940","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":626351754,"identity":"0c7634a7-eebb-4057-81e6-e14840a4c515","order_by":0,"name":"Tan Haosheng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2UlEQVRIiWNgGAWjYHACNhCRAMSMDxIqagir50HSwmzw4Mwx0rSwST5sYSasxZ797LEHPyrq8vil269VJDawMfC3dyfgt4UnL92w58zhYsk5Z8puJO6QYZA4c3YDAYflmEnwth1I3HAjJ+1G4hk2BgOJXAJa+N+YSf5tq0vcD9RSkNjGTIQWiRwzad425sQNEunHGIjTcuONmbTMmcOJM27kMEsknDnGQ9Av7P05ZpJvKuoS+2ekP/z4o6JGjr+9F78WZAsNwCSxysEWPiBF9SgYBaNgFIwgAAA160j88ULQjQAAAABJRU5ErkJggg==","orcid":"","institution":"Department of General Surgery, The Affiliated Taizhou People's Hospital of Nanjing Medical University, Taizhou School of Clinical Medicine, Nanjing Medical University, Taizhou, China","correspondingAuthor":true,"prefix":"","firstName":"Tan","middleName":"","lastName":"Haosheng","suffix":""},{"id":626351938,"identity":"373a7041-d69e-4718-9ba8-b4308a988bcf","order_by":1,"name":"Yu Dapeng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAs0lEQVRIiWNgGAWjYHACA+YfFTZybOztB0jQwnAmzZiP50wCCVoY2w4nzpNwMCBS/Y3kbY8LzqSlt0kwJDD8qNhGjJa0cuMZFTa5bdKNBxh7ztwmRkuOmQTPmbTcNpkDCUAXEquFt+1wOptEggHxWqSBWhKI1yJ55lm54YwzaYZtwEA+SJRf+I4nb3vwocJGXr69/eCDHxVEaFE4wMAG5xwgrB4I5BuQtIyCUTAKRsEowAoAb2NAFce6tnQAAAAASUVORK5CYII=","orcid":"","institution":"Department of General Surgery, The Affiliated Taizhou People's Hospital of Nanjing Medical University, Taizhou School of Clinical Medicine, Nanjing Medical University, Taizhou, China","correspondingAuthor":true,"prefix":"","firstName":"Yu","middleName":"","lastName":"Dapeng","suffix":""},{"id":626352116,"identity":"a4a7f9a5-86c7-476c-8c71-f3934697c0ad","order_by":2,"name":"Jiao lianghe","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsUlEQVRIiWNgGAWjYBACxmYg8bDBRo6Nvf0ACVoSG9KM+XjOJJBgVWLD4cR5Eg4GxKlmbmd++CBxR1p6mwRDAsOPim3EOIzN2CDxjE1um3TjAcaeM7eJ0cJgJpHYlpbbJnMggZmxjSgt7N+AWg6ns0kkGBCrhQdky+EEkrQUGwAdZtgGDOSDRPnFsP/4xgcf22zk5dvbDz74UUGMlgYkzgHC6oFAnihVo2AUjIJRMLIBALMxO5VKceZwAAAAAElFTkSuQmCC","orcid":"","institution":"Department of General Surgery, The Affiliated Taizhou People's Hospital of Nanjing Medical University, Taizhou School of Clinical Medicine, Nanjing Medical University, Taizhou, China","correspondingAuthor":true,"prefix":"","firstName":"Jiao","middleName":"","lastName":"lianghe","suffix":""}],"badges":[],"createdAt":"2026-04-20 13:34:38","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9472940/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9472940/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107453115,"identity":"322636ac-3763-46af-b475-4203c3655973","added_by":"auto","created_at":"2026-04-21 15:31:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":770571,"visible":true,"origin":"","legend":"\u003cp\u003eThe EpiBarrier score predicts overall survival and treatment response in anti‑PD‑1/PD‑L1‑treated HNSCC patients (GSE159067 cohort).\u003c/p\u003e\n\u003cp\u003e(A) Kaplan‑Meier curves for overall survival stratified by the optimal cutoff of the EpiBarrier score (‑0.246). High‑score patients (red) had significantly worse survival compared with low‑score patients (blue) (log‑rank P=0.012). The number of patients at risk at each time point is shown below the graph.\u003c/p\u003e\n\u003cp\u003e(B) Receiver operating characteristic (ROC) curve for the ability of the EpiBarrier score to discriminate responders (CR/PR) from non‑responders (SD/PD). The area under the curve (AUC) was 0.702, indicating moderate predictive performance.\u003c/p\u003e\n\u003cp\u003e(C) Forest plot showing the hazard ratios (HRs) from multivariable Cox regression analysis. After adjusting for HPV status, the EpiBarrier score remained an independent predictor of overall survival (HR=1.744, 95% CI: 1.09–2.78, P=0.026). HPV status itself was not significantly associated with survival (HR=1.086, P=0.697). Horizontal lines represent 95% confidence intervals.\u003c/p\u003e","description":"","filename":"Figurecombinedvertical.png","url":"https://assets-eu.researchsquare.com/files/rs-9472940/v1/c6525e2b10946533367c85e4.png"},{"id":107868320,"identity":"ddc2d4c5-d109-4c33-ac72-6656b7899413","added_by":"auto","created_at":"2026-04-27 07:10:10","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":835243,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9472940/v1/0eb8d1be-fcfc-4d18-ab42-6296072f764b.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eA conserved 8-gene epithelial barrier signature spatially quantifies immune exclusion across cancers and predicts immunotherapy outcome independent of HPV status\u003c/p\u003e","fulltext":[{"header":"1. Background","content":"\u003cp\u003eImmune checkpoint inhibitors (ICIs) have revolutionized cancer therapy, yet many patients do not respond due to primary resistance, often mediated by \u0026ldquo;immune exclusion\u0026rdquo; \u0026ndash; the physical blockade of T cells by tumor epithelial barriers (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Tight junctions, adherens junctions, and desmosomes create a barrier that prevents lymphocyte infiltration into tumor nests. However, no simple spatial biomarker exists to quantify epithelial barrier strength directly in tissue sections.\u003c/p\u003e \u003cp\u003eIn head and neck squamous cell carcinoma (HNSCC), HPV status (p16 IHC or HPV DNA) is widely used as a surrogate for ICI response. Nevertheless, up to 30% of HPV‑positive patients fail to respond, while some HPV‑negative patients derive benefit (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e), indicating that HPV status poorly reflects functional immune exclusion. Thus, a direct, quantitative, and easily measurable epithelial barrier signature is urgently needed.\u003c/p\u003e \u003cp\u003eSpatial transcriptomics technologies (Xenium, Visium) enable simultaneous gene expression profiling and spatial localization, allowing the construction of functional scores from defined gene sets. Here, we developed a compact 8‑gene epithelial barrier signature, validated it across six epithelial cancer types, compared it with HPV status in HNSCC, and tested its prognostic and predictive value in an external ICI‑treated HNSCC cohort. We also included non‑epithelial tumors (kidney cancer, melanoma, ALL bone marrow) to assess model specificity.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cp\u003eOngoing work and future updates\u003c/p\u003e \u003cp\u003eTo strengthen the generalizability of the epithelial barrier signature, we are actively collecting and analyzing larger, independent spatial transcriptomics cohorts across multiple cancer types. As our work progresses, more detailed figures, tables, and supplementary materials will be generated. Consequently, the results presented in this preprint version are preliminary and subject to refinement. We will continuously update the findings, and the latest versions of all figures, tables, and supplementary information will be made available in subsequent releases of this manuscript.\u003c/p\u003e \u003cp\u003eSpatial transcriptomics datasets\u003c/p\u003e \u003cp\u003eWe collected publicly available Xenium and Visium datasets: pancreatic cancer (Xenium, 190,965 cells), lung cancer (Xenium, 67,763), breast cancer (Xenium Prime, 699,110), colorectal cancer (Xenium, 388,175), cervical cancer (Xenium, 205,082), ovarian cancer (Xenium, 388,175), head and neck squamous cell carcinoma (5 samples, 611,962 cells), kidney cancer (Xenium, 465,334), melanoma (Xenium, 87,499), and acute lymphoblastic leukemia bone marrow (Xenium, 225,906, negative control). All data were log‑normalized (scale factor 10,000). Samples with \u0026gt;\u0026thinsp;50,000 cells were randomly down‑sampled to 50,000.\u003c/p\u003e \u003cp\u003eGene set definitions\u003c/p\u003e \u003cp\u003eThree functional gene sets were defined based on literature and pan‑cancer expression stability:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eEpithelial barrier (EpiBarrier): 8 genes involved in cell adhesion, desmosomes, keratins, and pan‑epithelial markers.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eStromal barrier: 8 cancer‑associated fibroblast and extracellular matrix genes (FAP, COL1A1, ACTA2, VIM, PDGFRA, PDGFRB, FN1, POSTN).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eImmune activity: 8 CD8\u0026thinsp;+\u0026thinsp;T‑cell effector molecules, chemokines, and antigen‑presenting genes (CD8A, IFNG, GZMB, PRF1, CXCL9, CXCL10, HLA‑A, HLA‑B).\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eFor each cell/spot, the module score was the mean log‑normalized expression of the respective genes (Xenium data were first Z‑scored across cells).\u003c/p\u003e \u003cp\u003eCell type annotation (for neighborhood features)\u003c/p\u003e \u003cp\u003eEpithelial cells were defined by expression of EPCAM, KRT8, KRT18, or KRT19 (\u0026gt;\u0026thinsp;0); immune cells by PTPRC, CD8A, CD4, CD68, or HLA‑A (\u0026gt;\u0026thinsp;0); stromal cells by FAP, COL1A1, or VIM (\u0026gt;\u0026thinsp;0). Priority: epithelial\u0026thinsp;\u0026gt;\u0026thinsp;immune\u0026thinsp;\u0026gt;\u0026thinsp;stromal, others as \u0026ldquo;Other\u0026rdquo;.\u003c/p\u003e \u003cp\u003eSpatial niche analysis\u003c/p\u003e \u003cp\u003eFor each cell, neighbors within 50 \u0026micro;m were identified (RANN::nn2). Neighborhood averages of EpiBarrier, stromal, and immune scores were calculated. PCA (6 features: self scores\u0026thinsp;+\u0026thinsp;neighbor averages) followed by Louvain clustering (resolution 0.4) identified spatial niches. The top 3 niches by mean EpiBarrier were defined as high‑EpiBarrier niches.\u003c/p\u003e \u003cp\u003eStatistical validation\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eNon‑spatial: Cells were split by median EpiBarrier into high/low groups; Wilcoxon test compared stromal and immune scores.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eCorrelation: Spearman correlations among the three scores.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLinear regression: Immune activity\u0026thinsp;~\u0026thinsp;EpiBarrier\u0026thinsp;+\u0026thinsp;stromal score.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eSpatial comparison: Wilcoxon test between high and low EpiBarrier niches.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHPV status: Detected by HPV16‑E6/E7 expression (\u0026gt;\u0026thinsp;0 cells defined as HPV‑positive).\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eExternal immunotherapy cohort\u003c/p\u003e \u003cp\u003eGSE159067 (102 HNSCC patients treated with anti‑PD‑1/PD‑L1) was downloaded. Log2CPM expression and clinical data (overall survival, event, response) were extracted. EpiBarrier score was calculated (Z‑scored mean). Optimal cutoff was determined by `surv_cutpoint`. Kaplan‑Meier and Cox regression assessed overall survival; ROC evaluated response prediction (CR/PR vs SD/PD).\u003c/p\u003e \u003cp\u003eStatistics\u003c/p\u003e \u003cp\u003eTwo‑sided P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered significant.\u003c/p\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e3.1 The 8‑gene epithelial barrier signature identifies immune‑excluded niches in epithelial cancers\u003c/h2\u003e \u003cp\u003eWe first validated the model in pancreatic cancer Xenium data. Cells were split into high‑ and low‑EpiBarrier groups by the median score. The high‑EpiBarrier group exhibited significantly lower stromal scores (mean 0.485 vs 1.044, P\u0026thinsp;\u0026lt;\u0026thinsp;1e‑300) and lower immune activity scores (mean 0.075 vs 0.158, P\u0026thinsp;\u0026lt;\u0026thinsp;1e‑130). Spatial niche analysis confirmed that high‑EpiBarrier niches had significantly lower stromal and immune scores (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Linear regression showed a negative coefficient for EpiBarrier (‑0.025, P\u0026thinsp;\u0026lt;\u0026thinsp;1e‑112), indicating independent negative prediction of immune activity.\u003c/p\u003e \u003cp\u003eThe same pipeline was applied to lung, breast (Xenium Prime), colorectal, cervical, ovarian, head and neck (HNSCC), skin melanoma, kidney, and acute lymphoblastic leukemia (ALL) bone marrow datasets. Results are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePan‑cancer validation of the 8‑gene EpiBarrier signature\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCancer type\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePlatform\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModel supported\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eImmune activity change (high vs low EpiBarrier)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStromal change\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eRemarks\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePancreatic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 53% (0.075\u0026rarr;0.158, P\u0026thinsp;\u0026lt;\u0026thinsp;1e-130)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 54%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEpithelial\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLung\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 73% (0.021\u0026rarr;0.078, P\u0026thinsp;\u0026lt;\u0026thinsp;1e-146)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 66%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEpithelial\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBreast\u003c/p\u003e \u003cp\u003e(Xenium Prime)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 54% (0.036\u0026rarr;0.078, P\u0026thinsp;=\u0026thinsp;3.2e-12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 59%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eOnly CDH1 and EPCAM detected\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eColorectal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 13% (0.402\u0026rarr;0.461, P\u0026thinsp;=\u0026thinsp;4.9e-22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 18%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEpithelial\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCervical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 31% (0.035\u0026rarr;0.051, P\u0026thinsp;=\u0026thinsp;0.049)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 87%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEpithelial\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOvarian\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 5% (0.091\u0026rarr;0.096, P\u0026thinsp;=\u0026thinsp;3.3e-9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 73%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSmall but significant\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHNSCC (HPV-negative)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 52% (0.264\u0026rarr;0.555, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 55%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eLaryngeal primary/metastasis\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHNSCC (HPV-positive, immune-hot)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo (reversed)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026uarr; (0.228\u0026rarr;0.175, P\u0026thinsp;\u0026lt;\u0026thinsp;1e-26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026uarr;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eLow HPV viral load\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSkin melanoma (normal epidermis)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026darr; 15% (0.996\u0026rarr;1.172, P\u0026thinsp;=\u0026thinsp;8e-27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026darr; 27%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eCorresponds to normal epidermal cells\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKidney\u003c/p\u003e \u003cp\u003e(clear cell RCC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026uarr; (P\u0026thinsp;\u0026lt;\u0026thinsp;1e-180)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026uarr;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNon-epithelial\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALL bone marrow\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXenium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo (reversed)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026uarr; (0.261\u0026rarr;0.203, P\u0026thinsp;=\u0026thinsp;1.4e-7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026uarr;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNon-epithelial (negative control)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn all epithelial cancers, high‑EpiBarrier regions consistently showed reduced stromal and immune activity, with reductions ranging from 5% to 73% (all P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Notably, in breast Xenium Prime only two of the eight genes (CDH1 and EPCAM) were detected, yet the model remained highly significant, demonstrating robustness.\u003c/p\u003e \u003cp\u003eImportantly, in skin melanoma \u0026ndash; a non‑epithelial tumor \u0026ndash; the model did identify high‑EpiBarrier cells that corresponded to normal epidermal structures (e.g., keratinocytes). Those cells exhibited significantly lower stromal and immune activity (P\u0026thinsp;=\u0026thinsp;8e‑27), confirming that the signature faithfully captures epithelial barrier function wherever true epithelial cells are present. In contrast, in kidney cancer and ALL bone marrow, where no epithelial cells exist, the model either failed or showed the opposite trend (immune‑hot phenotype), underscoring its specificity to epithelial tissues.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e3.2 HNSCC: model independent of HPV status\u003c/h2\u003e \u003cp\u003eWe analyzed five HNSCC/laryngeal samples (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). HPV status was determined by HPV16‑E6/E7 expression. HPV‑negative samples (GSM8193670/3671) were immune‑excluded (model supported). A high‑HPV sample (82.4% HPV+) was also immune‑excluded, whereas low‑HPV samples (\u0026lt;\u0026thinsp;2.2% HPV+) were immune‑hot (model not supported). Thus, HPV status cannot predict immune exclusion; the EpiBarrier score directly quantifies the functional phenotype.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eHNSCC samples: HPV status and model outcome\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSample\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHPV+ cells\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModel outcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eImmune phenotype\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGSM9054474\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e82.40%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSupported\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eExcluded\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGSM9054471\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.80%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNot supported\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHot\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGSM9054487\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.20%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNot supported\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHot\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGSM8193670\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSupported\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eExcluded\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGSM8193671\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSupported\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eExcluded\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Matched primary and metastatic lymph node analysis\u003c/h2\u003e \u003cp\u003eA paired analysis of primary laryngeal cancer (GSM8193670) and its regional lymph node metastasis (GSM8193671) showed that the model was significant in both sites, although the effect size was smaller in the metastasis (31% vs 52% immune reduction, both P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This suggests that epithelial barrier function and its association with immune exclusion are maintained during tumor progression.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.4 External immunotherapy cohort validation\u003c/h2\u003e \u003cp\u003eIn the GSE159067 cohort (102 anti‑PD‑1/PD‑L1‑treated HNSCC patients), the EpiBarrier score as a continuous variable predicted overall survival (HR\u0026thinsp;=\u0026thinsp;1.744, 95% CI [1.09\u0026ndash;2.78], P\u0026thinsp;=\u0026thinsp;0.0205). Using the optimal cutoff (‑0.246), high‑score patients had significantly worse survival (log‑rank P\u0026thinsp;=\u0026thinsp;0.012, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). The score moderately discriminated responders from non‑responders (AUC\u0026thinsp;=\u0026thinsp;0.702, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Multivariable Cox regression adjusting for HPV status confirmed the independent predictive value of the EpiBarrier score (P\u0026thinsp;=\u0026thinsp;0.026, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC).\u003c/p\u003e "},{"header":"4. Discussion","content":"\u003cp\u003eWe developed and validated a compact 8‑gene epithelial barrier signature that spatially quantifies immune exclusion in 6 epithelial cancer types (pancreatic, lung, breast, colorectal, cervical, ovarian). The signature is robust even when only two of the eight genes are detected (as in breast Xenium Prime). In skin melanoma, high‑EpiBarrier cells corresponded to normal epidermal structures and exhibited the expected immune‑excluded phenotype, further confirming that the signature faithfully captures epithelial barrier function wherever genuine epithelial cells are present. Conversely, in non‑epithelial tumors (kidney, ALL bone marrow) the model failed or showed opposite trends, underscoring its cancer‑type specificity.\u003c/p\u003e \u003cp\u003eIn HNSCC, the signature outperforms HPV status: HPV‑positive tumors can be either immune‑excluded or immune‑hot, and HPV‑negative tumors can also be immune‑excluded. Thus, HPV status should not be used as a surrogate for functional immune exclusion; direct measurement of epithelial barrier strength is required. The signature predicted overall survival and treatment response in an external ICI‑treated HNSCC cohort, suggesting clinical utility.\u003c/p\u003e \u003cp\u003eThe 8‑gene composition is attractive for translation into an immunohistochemistry (IHC) or qPCR panel. Limitations include the modest size of the external cohort (n\u0026thinsp;=\u0026thinsp;102) and the lack of prospective validation. The model does not apply to immune‑hot tumors (e.g., some HPV‑positive HNSCC, kidney cancer) where active immunity may disrupt epithelial barriers. Future work will develop an IHC assay and test the signature in larger immunotherapy trials across multiple epithelial cancer types.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThe 8‑gene epithelial barrier signature is a pan‑epithelial‑cancer, spatially quantifiable biomarker of immune exclusion. It is independent of HPV status and predicts immunotherapy outcome. This signature offers a clinically translatable tool for patient stratification in epithelial‑derived malignancies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor contributions:\u003c/h2\u003e \u003cp\u003eT.H. conceived the study, performed data analysis, wrote the manuscript, and acted as corresponding author. Y.D. contributed to data interpretation and critical revision. J.L. supervised the project and contributed to funding acquisition. All authors approved the final version.\u003c/p\u003e\u003cp\u003eThis manuscript is prepared for preprint submission. All code and data are available upon request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eJoyce JA, Fearon DT (2015) T cell exclusion, immune privilege, and the tumor microenvironment. Science 348(6230):74\u0026ndash;80\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen DS, Mellman I (2017) Elements of cancer immunity and the cancer-immune set point. Nature 541(7637):321\u0026ndash;330\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCillo AR et al (2020) Immune landscape of viral- and carcinogen-driven head and neck cancer. Immunity 52(1):183\u0026ndash;199\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaintigny P et al (2022) Gene expression profiling of HNSCC patients treated with anti-PD-1/PD-L1 (GSE159067). GEO\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e10x Genomics Xenium In Situ and Visium Spatial Gene Expression documentation\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":"Taizhou People's Hospital","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"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":"epithelial barrier signature, spatial transcriptomics, immune exclusion, immunotherapy outcome","lastPublishedDoi":"10.21203/rs.3.rs-9472940/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9472940/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eTumor epithelial barriers physically impede immune cell infiltration, causing immune exclusion and resistance to immune checkpoint inhibitors (ICIs). However, a simple, spatially quantifiable biomarker of epithelial barrier strength is lacking. In head and neck squamous cell carcinoma (HNSCC), HPV status is used as a surrogate for ICI response, but its association with true immune exclusion remains unclear.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eUsing Xenium and Visium spatial transcriptomics data from nine cancer types (pancreatic, lung, breast, colorectal, cervical, ovarian, HNSCC, skin melanoma, kidney, and acute lymphoblastic leukemia bone marrow), we defined a compact 8‑gene epithelial barrier signature (EpiBarrier), together with a stromal signature and an immune activity signature. Spatial niche clustering, pan‑cancer comparisons, and external immunotherapy cohort validation were performed to test whether high EpiBarrier regions exhibit low stromal and low immune activity. In HNSCC, we compared the signature with HPV status.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eIn six epithelial cancer types (pancreatic, lung, breast, colorectal, cervical, ovarian), high‑EpiBarrier regions consistently showed significantly lower stromal scores (17\u0026ndash;87% reduction) and lower immune activity scores (5\u0026ndash;73% reduction) (all P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In skin melanoma, high‑EpiBarrier cells corresponded to normal epidermal structures and also exhibited reduced stromal and immune activity (P\u0026thinsp;=\u0026thinsp;8e‑27), confirming the signature\u0026rsquo;s specificity to genuine epithelial cells. Non‑epithelial tumors (kidney, ALL bone marrow) did not support the model or showed opposite trends. In HNSCC, the model outcome was independent of HPV status: HPV‑high samples could be immune‑excluded (model supported) or immune‑hot (model not supported), while HPV‑negative samples were immune‑excluded. An external ICI‑treated HNSCC cohort (GSE159067, n\u0026thinsp;=\u0026thinsp;102) validated that EpiBarrier predicted overall survival (HR\u0026thinsp;=\u0026thinsp;1.744, P\u0026thinsp;=\u0026thinsp;0.0205) and moderately predicted treatment response (AUC\u0026thinsp;=\u0026thinsp;0.702).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe 8‑gene epithelial barrier signature is a pan‑epithelial‑cancer, spatially quantifiable biomarker of immune exclusion, independent of HPV status, and predicts immunotherapy outcome. It offers a clinically translatable tool for patient stratification in epithelial‑derived malignancies.\u003c/p\u003e","manuscriptTitle":"A conserved 8-gene epithelial barrier signature spatially quantifies immune exclusion across cancers and predicts immunotherapy outcome independent of HPV status","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-21 15:31:18","doi":"10.21203/rs.3.rs-9472940/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":"cb8befb9-878d-4b35-a899-1df9b58e87b5","owner":[],"postedDate":"April 21st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":66663935,"name":"Immunology"},{"id":66663936,"name":"Translational Medicine"}],"tags":[],"updatedAt":"2026-05-16T00:31:18+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-21 15:31:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9472940","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9472940","identity":"rs-9472940","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}