In triple-negative breast cancer (TNBC), platinum-based chemotherapy increases pathological complete response rates relative to non-platinum neoadjuvant chemotherapy, but its potential toxicity underscores the need for predictive biomarkers. While homologous recombination deficiency (HRD) has been proposed as a platinum sensitivity marker, FDA-approved HRD testing has shown only modest predictive power in TNBC clinical trials. Nucleotide excision repair (NER) deficiency is an alternative mechanism of platinum sensitivity identified in bladder cancer using next-generation sequencing-based mutational signatures. Here, we apply the ERCC2mut score, derived from bladder cancers with inactivating ERCC2 mutations, to identify potential NER-deficient breast cancers and evaluate their association with platinum response. In the Cancer Genome Atlas (TCGA) breast cancer cohort (n = 943), 4.7% of cases showed indications of NER deficiency without concurrent HRD. In a cisplatin-treated TNBC cohort (n = 50), both ERCC2mut (p = 0.0077) and HRDetect (p = 0.036) were associated with pathological response, but the two biomarkers identified largely non-overlapping patients. In the TBCRC030 randomized trial (n = 54 cisplatin, n = 45 paclitaxel), multiple signatures associated with NER deficiency were significantly associated with response to cisplatin, but not to paclitaxel. These findings demonstrate that NER deficiency-associated mutational signatures can identify a subset of platinum-sensitive TNBC patients missed by HRD testing alone, supporting a combined biomarker strategy.
Pathogenic BRCA1 and BRCA2 germline variants confer an increased risk of hereditary breast and ovarian cancer (HBOC). In this high-risk cohort, novel biomarkers are needed to support patient management. While circulating cell-free DNA-based conventional CpG-level averaging is promising, it dilutes tumor-specific signals in samples with low circulating tumor DNA (ctDNA) fractions. Here, we developed a targeted ultra-deep NGS assay for detecting tumor-derived ctDNA based on fragment-level hypermethylation of CpG-rich cfDNA. A panel of 360 cancer-specific CpG-rich regions derived from 450k methylation array datasets from TCGA breast cancer and GEO ovarian cancer, was applied to plasma from 31 breast cancer, 16 ovarian cancer patients, and 26 healthy controls using enzymatic methylation sequencing. Hypermethylated ctDNA (hm-ctDNA) fragments, were significantly enriched in cancer cases, identifying 62.5% of ovarian and 38.7% of breast cancers at 100% specificity. Elevated hm-ctDNA was observed across different tumor stages, with higher burden in BRCA1-associated cases. This approach enables highly specific, non-invasive hypermethylated ctDNA detection in HBOC populations.
Pivotal CDK4/6 inhibitor trials in HR+/HER2− metastatic breast cancer defined renal eligibility by serum creatinine, excluding all patients with chronic kidney disease (CKD) stage 4–5; no head-to-head data exist. Within the TriNetX Global Collaborative Network, we emulated three target trials as pairwise propensity-score-matched comparisons (1134/545/520 per arm) of women with CKD stage 3b–5 initiating first-line therapy. Co-primary outcomes were 2-year composite treatment failure and all-cause mortality. Compared with palbociclib, abemaciclib and ribociclib were each associated with lower treatment failure (hazard ratio: 0.79, 95% CI: 0.68–0.90; and 0.75, 0.60–0.94) and lower mortality (0.84, 0.71–0.99; and 0.69, 0.52–0.92); ribociclib and abemaciclib did not differ. Treatment-failure associations held in all sensitivity cohorts; mortality associations did not (E-value for the confidence-interval bound, 1.11). Drug-specific safety signatures from pivotal trials were preserved. Acute kidney injury hazard was stable for abemaciclib (interaction p = 0.081) but fell for ribociclib versus palbociclib (hazard ratio from 1.29 to 0.75; interaction p = 0.012). Mapping onto a published single-cell kidney atlas places the renal transporters these agents inhibit at the site of creatinine secretion. These exploratory, unregistered analyses lack familywise error control, but the toxicity profiles may inform drug selection in advanced CKD; an isolated creatinine rise on abemaciclib may warrant cystatin C measurement before discontinuation.
Estrogen targets multiple cell types that impact tumor progression. However, effects on metastatic processes and interplay of signals to cancer cells and stromal cells are poorly understood. Using an immunocompetent preclinical model of estrogen receptor alpha (ER + ) breast cancer and two matched CRISPR-Cas9 Esr1-ablated (ER-) cell lines, we demonstrate that supplemental 17β-estradiol fosters a pro-tumor immune environment for both ER+ and ER- mammary tumors, but exerts disparate effects on early pulmonary colonization following intravenous injection: promoting ER + , but inhibiting ER- tumor lesions. Interrogation of mammary tumor environments and splenic and pulmonary immune cells revealed marked effects of 17β-estradiol on immune subpopulations and activity. RNASeq analyses illuminated the potential of ER+ tumor cells for immune evasion, revealing a new facet of estrogen action in ER+ cancer. This context-specific interplay among estrogen responsive cells in the mammary and premetastatic pulmonary environments has implications beyond breast cancers, including ER- cancers of varying origins.
Accurate assessment of breast cancer risk is essential for risk-stratified screening strategies. We developed BREST-AI, a deep-learning model that predicts three-year breast cancer risk from negative screening mammograms. Model development used a curriculum-learning strategy and data from 65,419 women; the model was trained on 58,591 women (9020 cancer, 1633 benign, and 47,938 three-year cancer-free). In the external UK case–control validation (1899 cases and 5697 matched controls), BREST-AI achieved an AUC of 0.751 (95% CI 0.738–0.764), outperforming Mirai (0.700, 95% CI 0.687–0.714; p < 0.001). In a US external cohort (97 cases and 291 matched controls), AUCs were 0.670 (95% CI 0.609–0.731) for BREST-AI and 0.633 (95% CI 0.570–0.696) for Mirai (p = 0.199); this analysis is underpowered to detect modest differences and should be interpreted as exploratory. In a UK cross-vendor external cohort acquired on Siemens systems (400 cases and 1200 matched controls), cross-vendor domain shift modestly reduced discrimination, but BREST-AI still exceeded Mirai (AUC 0.719 versus 0.685; p = 0.019). When projecting case–control results to the English screening prevalence (8 cancers per 1000 examinations), BREST-AI yielded an estimated positive predictive value of 8.0% and a relative risk of 10.1 at a high-risk operating point defined by 99% specificity. These prevalence-adjusted estimates suggest that BREST-AI could help to identify a small subgroup of women at high three-year breast cancer risk within the NHS Breast Screening Programme.
