2026-10-14T14:23:06Z https://www.scidb.cn/oai

10.57760/sciencedb.41856 2026-07-06T10:54:34Z

2026-07-06 The LTB–LTBR–COL1A1 axis orchestrates a tolerogenic peritumoral tertiary lymphoid niche and enables multimodal prediction of metastasis in clear cell renal cell carcinoma doi:10.57760/sciencedb.41856 en Distant metastasis is the principal cause of death in clear cell renal cell carcinoma (ccRCC), yet the intercellular programs that license dissemination—and their spatial organization—remain incompletely defined. Integrating single-cell RNA sequencing (scRNA-seq) of primary, lung-metastatic and adjacent-normal ccRCC, spatial transcriptomics, validation in The Cancer Genome Atlas (TCGA-KIRC), and a multimodal machine-learning framework, we dissect the regulatory logic of metastatic competence. Pseudotime trajectory analysis of malignant cells, intersected with ligand–receptor regulatory modeling, nominated a lymphotoxin-driven LTB–LTBR–COL1A1 axis as the program most consistently divergent between primary and metastatic disease, with LTB elevated in metastasis-associated immune cells, LTBR in metastatic tumor cells, and LTBR–COL1A1 co-expression across the microenvironment. In spatial data the axis was concentrated in a peritumoral, tertiary lymphoid structure (TLS)-rich niche—Niche 1, the Tolerogenic Coronal Niche (TCN)—that was M2-skewed and immunosuppressive (PDCD1, CTLA4, FOXP3 programs) and dominated by COL1A1–CD44/collagen signaling, in contrast to an intratumoral, immunoreactive niche (Niche 4, the Immunoreactive Medullary Niche, IMN). The TLS chemokine CXCL13 marked the TCN, correlated with immunosuppressive checkpoints, and behaved as a risk factor. In TCGA-KIRC, a TCN-positive/IMN-negative signature—concordant between ssGSEA and NNLS deconvolution—stratified markedly worse overall survival (log-rank P < 0.0001) and higher metastasis frequency (Fisher P < 0.001). Finally, a weakly-supervised histopathology model (CLAM) predicted metastasis from primary-tumor whole-slide images (mean out-of-fold AUC ≈ 0.78); a systematic fusion search adding niche/axis-gene panels yielded only marginal, non-significant gains over histology alone—confirmed by repeated cross-validation and bootstrapping—indicating that routine histology captures most of the metastasis-predictive signal while the niche genes supply concordant, interpretable risk that resolves discordant cases. Multiplex immunofluorescence is planned to corroborate the spatial co-localization of the axis [Fig. 5, pending]. These data define an LTB–LTBR–COL1A1–driven tolerogenic peritumoral niche as a determinant of ccRCC metastasis and provide a deployable, image-anchored risk model. Keywords: clear cell renal cell carcinoma; metastasis; lymphotoxin-β receptor; tertiary lymphoid structures; COL1A1; CXCL13; single-cell RNA sequencing; spatial transcriptomics; digital pathology; machine learning Xiaochen Qi Guangzhen Wu RESTRICTED dataset Science Data Bank