Primary tumor transcriptomic signature is associated with locoregional dissemination in canine mammary carcinoma

Abstract

Canine mammary carcinoma (CMC) exhibits heterogeneous metastatic behavior, while accurate assessment of locoregional dissemination may be limited by incomplete sampling, occult metastasis, and variable sentinel lymph node anatomy. We investigated whether primary-tumor transcriptomic programs are associated with locoregional dissemination. The training cohort comprised 113 CMC RNA sequencing samples classified by peritumoral lymphatic vessel invasion, and the independent evaluation cohort comprised 27 microarray samples classified by histopathologically confirmed lymph node (LN) metastasis. Expression profiles were mapped to human ortholog gene symbols, and single-sample gene set enrichment analysis (ssGSEA) was performed using Molecular Signatures Database (MSigDB) gene sets. Seven candidate gene sets selected through univariable screening and literature-guided curation were entered into least absolute shrinkage and selection operator (LASSO) logistic regression, yielding a four-gene-set transcriptomic risk score (RS). The RS showed an apparent receiver operating characteristic area under the curve (ROC-AUC) of 0.949 (95% confidence interval, 0.906-0.993) and precision-recall area under the curve (PR-AUC) of 0.725 in the training cohort; bootstrap internal validation yielded an optimism-corrected ROC-AUC of 0.920. In the evaluation cohort, the ROC-AUC was 0.890 (95% confidence interval, 0.766-1.000) and PR-AUC was 0.869. Discrimination was attenuated but retained in histologically restricted sensitivity analyses, and higher RS remained associated with locoregional dissemination after clinicopathologic adjustment. These findings support gene-set-level transcriptomic scoring as a potential complementary approach for characterizing dissemination-associated primary-tumor states in CMC.