Pan-cancer atlas of cellular architecture reveals a nearest-neighbour distance-associated biomechanical-immune axis involving CD4+ memory T cells

Abstract

Background: The systematic link between cellular spatial organization and its biomechanical consequences remains a critical knowledge gap in human oncology.

Objective: To quantitatively map the cellular architecture of solid tumours and elucidate its mechanistic impact on mechanical signalling, immune infiltration, and patient survival.

Methods: We developed a scalable digital pathology framework, processing 7910 H&E whole-slide images across 21 solid tumours. A deep learning pipeline was employed to segment over 4.7 billion nuclei, enabling the calculation of cell density and nearest neighbour distance (NND) as key spatial metrics. To bridge morphology and function, we integrated these metrics with bulk RNA-seq data from 19 TCGA cohorts (n = 7401) using rigorous histological matching. Furthermore, to resolve microenvironmental heterogeneity, we performed unsupervised clustering of over 60 000 T-cell transcriptomes from independent cohorts. These findings were validated through high-resolution Visium-HD spatial transcriptomics to correlate physical proximity with localized gene expression.

Results: While tumours exhibited significant heterogeneity, NND, but not cell density, emerged as a primary determinant of biomechanical and immune signatures. A trend-level association was observed between lower NND and higher Hippo/YAP/TAZ pathway activity across nine matched cancer types (Spearman's ρ = -.65, p = .058, n = 9). In addition, lower NND was significantly correlated with increased CD4⁺ memory T-cell (CD4⁺ T_Mem cell) abundance (Spearman's ρ = -.86, p < .01). Single-cell analyses confirmed that CD4⁺ T_Mem cells intrinsically express mechanical stress signalling markers, which spatial transcriptomics localized to CD4⁺ T_Mem cell aggregation zones characterized by high pathway activity. Clinically, this spatial-mechanical-CD4⁺ T_Mem cell axis was associated with prognosis in breast, oesophageal, liver, and lung adenocarcinomas, where high mechanical signalling generally predicted poor outcomes but could be modulated by CD4⁺ T_Mem infiltration levels.

Conclusion: Our study identifies low NND as a spatial correlate of biomechanical crowding that is associated with CD4⁺ T_Mem cell programming and adverse clinical outcomes. By integrating deep learning-based spatial metrics with multi-omics, we highlight spatial mechanics as a critical, potentially targetable dimension of the tumour microenvironment for future immunotherapies.

Keywords

CD4 memory T-cell / digital pathology / Hippo pathway / mechanical stress / multi-omics integration / nearest neighbour distance / spatial analysis / survival analysis / tumour microenvironment

References

[1] Hunter MV, Joshi E, Bowker S, et al. Mechanical confinement governs phenotypic plasticity in melanoma. Nature. 2025; 647: 517-527.

[2] Jassim A, Rahrmann E, Simons B, Gilbertson R. Cancers make their own luck: theories of cancer origins. Nat Rev Cancer. 2023; 23: 710-724.

[3] Krishna C, DiNatale RG, Kuo F, et al. Single-cell sequencing links multiregional immune landscapes and tissue-resident T cells in ccRCC to tumor topology and therapy efficacy. Cancer Cell. 2021; 39: 662-677.e6.

[4] Lavin Y, Kobayashi S, Leader A, et al. Innate Immune landscape in early lung adenocarcinoma by paired single-cell analyses. Cell. 2017; 169: 750-765.e17.

[5] Bindea G, Mlecnik B, Tosolini M, et al. Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human cancer. Immunity. 2013; 39: 782-795.

[6] Guo X, Zhang Y, Zheng L, et al. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. Nat Med. 2018; 24: 978-985.

[7] Gupta P, Dang M, Oberai S, et al. Immune landscape of isocitrate dehydrogenase-stratified primary and recurrent human gliomas. Neuro Oncol. 2024; 26: 2239-2255.

[8] MacFawn IP, Magnon G, Gorecki G, et al. The activity of tertiary lymphoid structures in high grade serous ovarian cancer is governed by site, stroma, and cellular interactions. Cancer Cell. 2024; 42: 1864-1881.e5.

[9] Clark K, Vendt B, Smith K, et al. The Cancer Imaging Archive (TCIA): maintaining and operating a public information repository. J Digit Imaging. 2013; 26: 1045-1057.

[10] Akbar S, Peikari M, Salama S, Panah AY, Nofech-Mozes S, Martel AL. Automated and manual quantification of tumour cellularity in digital slides for tumour burden assessment. Sci Rep. 2019; 9:14099.

[11] The Clinical Proteomic Tumor Analysis Consortium Breast Invasive Carcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2020.

[12] Farahmand S, Fernandez AI, Ahmed FS, et al. Deep learning trained on hematoxylin and eosin tumor region of Interest predicts HER2 status and trastuzumab treatment response in HER2+ breast cancer. Mod Pathol. 2022; 35: 44-51.

[13] Campanella G, Hanna MG, Geneslaw L, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nat Med. 2019; 25: 1301-1309.

[14] The Clinical Proteomic Tumor Analysis Consortium Clear Cell Renal Cell Carcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2018.

[15] Cutaneous Melanoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2018.

[16] The Clinical Proteomic Tumor Analysis Consortium Colon Adenocarcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2020.

[17] Colorectal Cancer Collection. Cancer Moonshot Biobank (CMB). National Cancer Institute; 2022.

[18] Vrabac D, Smit A, Rojansky R, et al. DLBCL-Morph: morphological features computed using deep learning for an annotated digital DLBCL image set. Sci Data. 2021; 8: 135.

[19] Glioblastoma Multiforme Collection (CPTAC-GBM) (Version 16). The Cancer Imaging: The Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2018.

[20] Bakas S, Sako C, Akbari H, et al. The University of Pennsylvania glioblastoma (UPenn-GBM) cohort: advanced MRI, clinical, genomics, & radiomics. Sci Data. 2022; 9: 453.

[21] Amin SB, Anderson KJ, Boudreau CE, et al. Comparative Molecular Life History of Spontaneous Canine and Human Gliomas. Cancer Cell. 2020; 37: 243-257.e7.

[22] Gastroesophageal Cancer Collection. Cancer Moonshot Biobank (CMB). National Cancer Institute; 2022.

[23] Chowdhury S, Kennedy JJ, Ivey RG, et al. Proteogenomic analysis of chemo-refractory high-grade serous ovarian cancer. Cell. 2023; 186: 3476-3498.e35.

[24] Lung Cancer Collection. Cancer Moonshot Biobank (CMB). National Cancer Institute; 2022.

[25] Lung Cancer Screening. National Lung Screening Trial (NLST). National Cancer Institute; 2013.

[26] Lung Squamous Cell Carcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2018.

[27] The Clinical Proteomic Tumor Analysis Consortium Lung Adenocarcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2018.

[28] Wilkinson S, Ye H, Karzai F, et al. Nascent prostate cancer heterogeneity drives evolution and resistance to intense hormonal therapy. Eur Urol. 2021; 80: 746-757.

[29] Pancreatic Ductal Adenocarcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2018.

[30] Sarcomas Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2019.

[31] Uterine Corpus Endometrial Carcinoma Collection. National Cancer Institute Clinical Proteomic Tumor Analysis Consortium (CPTAC). National Cancer Institute; 2019.

[32] Sturm G, Finotello F, Petitprez F, et al. Comprehensive evaluation of transcriptome-based cell-type quantification methods for immuno-oncology. Bioinformatics. 2019; 35: i436-i445.

[33] Aran D, Hu Z, Butte A. xCell: digitally portraying the tissue cellular heterogeneity landscape. Genome Biol. 2017; 18: 220.

[34] Racle J, Gfeller D. EPIC: a tool to estimate the proportions of different cell types from bulk gene expression data. Methods Mol Biol. 2020; 2120: 233-248.

[35] Becht E, Giraldo NA, Lacroix L, et al. Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol. 2016; 17: 218.

[36] Zhang M, Yang H, Wan L, et al. Single-cell transcriptomic architecture and intercellular crosstalk of human intrahepatic cholangiocarcinoma. J Hepatol. 2020; 73: 1118-1130.

[37] Lu Y, Yang A, Quan C, et al. A single-cell atlas of the multicellular ecosystem of primary and metastatic hepatocellular carcinoma. Nat Commun. 2022; 13: 4594.

[38] Jiang S, Lu H, Pan Y, et al. Characterization of the distinct immune microenvironments between hepatocellular carcinoma and intrahepatic cholangiocarcinoma. Cancer Lett. 2024; 588:216799.

[39] Pelka K, Hofree M, Chen JH, et al. Spatially organized multicellular immune hubs in human colorectal cancer. Cell. 2021; 184: 4734-4752.e20.

[40] Oliveira MFD, Romero JP, Chung M, et al. High-definition spatial transcriptomic profiling of immune cell populations in colorectal cancer. Nat Genet. 2025; 57: 1512-1523.

[41] Bankhead P, Loughrey MB, Fernández JA, et al. QuPath: open source software for digital pathology image analysis. Sci Rep. 2017; 7:16878.

[42] Stuart T, Butler A, Hoffman P, et al. Comprehensive integration of single-cell data. Cell. 2019; 177: 1888-1902.e21.

[43] Chang K, et al. The cancer genome atlas pan-cancer analysis project. Nature Genetics. 2013; 45: 1113-1120.

[44] Yost KE, Satpathy AT, Wells DK, et al. Clonal replacement of tumor-specific T cells following PD-1 blockade. Nat Med. 2019; 25: 1251-1259.

[45] Zhang P, Yang M, Zhang Y, et al. Dissecting the single-cell transcriptome network underlying gastric premalignant lesions and early gastric cancer. Cell Rep. 2019; 27: 1934-1947.e5.

[46] Schmidt U, Weigert M, Broaddus C, Myers G. Cell Detection with Star-Convex Polygons. Springer International Publishing; 2018: 265-273.

[47] QuPath Documentation, StarDist in QuPath. https://qupath.readthedocs.io/en/0.4/docs/deep/stardist.html

[48] Bolstad B, Irizarry R, Åstrand M, Speed T. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics. 2003; 19: 185-193.

[49] Hafemeister C, Satija R. Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression. Genome Biol. 2019; 20: 296.

[50] Zheng L, Qin S, Si W, et al. Pan-cancer single-cell landscape of tumor-infiltrating T cells. Science. 2021; 374:abe6474.

[51] Cheng S, Li Z, Gao R, et al. A pan-cancer single-cell transcriptional atlas of tumor infiltrating myeloid cells. Cell. 2021; 184: 792-809. e23.

[52] Luecken MD, Büttner M, Chaichoompu K, et al. Benchmarking atlas-level data integration in single-cell genomics. Nature methods. 2022; 19: 41-50.

Rights & permissions

2026 The Author(s). Clinical and Translational Medicine published by John Wiley & Sons Australia, Ltd on behalf of Shanghai Institute of Clinical Bioinformatics.