Lymphovascular Invasion Is Associated With Doxorubicin Resistance in Breast Cancer

Lymphovascular invasion (LVI), the invasion of tumor cells into the lymphatic or vascular space, is an early indicator of potential metastasis, with its presence in breast cancer independently predicting poorer outcomes even after neoadjuvant chemotherapy. However, a major limitation is that LVI detection currently relies on postsurgical evaluation. To address this, we determined whether LVI+ breast tumors contain a unique gene signature that could facilitate earlier detection. Here, we conducted an integrative analysis of the gene profile between LVI+ and LVI- primary breast tumors from various sources, including published data and our own research, using both microarray and RNA-seq data. Our analysis revealed protein binding and vesicle-related genes to be the most enriched categories in LVI+ vs LVI- tumors. Furthermore, LVI+ tumors showed enrichment for xenobiotic metabolism genes, particularly drug metabolism enzymes, such as cytochrome P450 and uridine 5'-diphospho-glucuronosyltransferases. An elastic net regression model containing 13 of these uridine 5'-diphospho-glucuronosyltransferases and cytochrome P450 genes can predict LVI status with 92% accuracy. This suggests a potential link to drug resistance, which was further confirmed by the finding that patients with LVI+ tumors had a significantly lower clinical response rate than individuals with LVI- tumors. We also observed this resistance in patient-derived organoids, with LVI+ organoids exhibiting lower sensitivity to doxorubicin, implying that doxorubicin might be less effective for LVI+ breast cancer, potentially contributing to poorer outcomes. Overall, our study unlocked an exciting opportunity for personalized medicine, in that, therapy efficacy and patient outcomes can be improved by incorporating the LVI-associated gene signature into treatment plans. (c) 2025 United States & Canadian Academy of Pathology. Published by Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.