Development and validation of multivariate predictors of primary endocrine resistance to tamoxifen and aromatase inhibitors in luminal breast cancer reveal drug-specific differences

Abstract

Background : Endocrine therapy is highly effective in blocking the estrogen receptor pathway in HR+/HER2- early breast cancer (EBC). However, up to 40% of patients experience relapse during or after adjuvant endocrine therapy. Here, we investigate molecular mechanisms associated with primary resistance to endocrine therapy and develop predictive models. Patients and Methods : In the WSG-ADAPT trial ( NCT01779206 ), HR+/HER2- EBC patients underwent pre-operative short-term endocrine therapy (pET). Treatment response was determined by immunohistochemical in-situ labeling of cycling cells (G1 to M-phase) with Ki67 before and after pET. We performed targeted next generation sequencing and Infinium MethylationEPIC-based DNA methylation analysis post-pET in a discovery cohort (n=364, responder (R) and non-responder (NR) pairs matched for clinicopathologic features) and a validation cohort (n=270, unmatched). Predictive indices of endocrine resistance under both treatments were constructed using lasso penalized logistic regression. A subset of breast cancers from The Cancer Genome Atlas project (TCGA-BRCA) was used for external validation. Results : TP53 mutations were prominently associated with primary resistance to both tamoxifen (TAM) and aromatase inhibitors (AI), with AI non-responders exhibiting resistance in up to 32% of cases. Additionally, we identified distinct DNA methylation patterns in TAM and AI non-responders, with TAM non-responders showing global DNA methylation loss, associated with KRAS signaling, apical junctions and epithelial-mesenchymal transition (EMT). Conversely, we observed methylation gain in AI non-responders affecting developmental transcription factors, hypoxia and estrogen signaling. TAM or AI resistance was associated with increased methylation-inferred proportions of immune cells and decreased proportions of endothelial cells. Based on these findings and patient age, we developed the Predictive Endocrine ResistanCe Index (PERCI). PERCI stratified NR and R cases in both treatment groups and cohorts with high accuracy (ROC AUC TAM discovery 93.9%, validation 83%; AI discovery 98.6%, validation 76.9%). A simplified PERCI efficiently predicted progression-free survival in the TCGA-BRCA sub-cohort (Kaplan-Meier log-rank p-value = 0.03 between low and high PERCI groups). Conclusions : We identified genomic and epigenomic features associated with primary resistance to TMA and AI. By combining information on genomic alterations, patient age, differential methylation and tumor microenvironment (TME) composition, we developed PERCI TAM and PERCI AI as novel predictors of primary resistance, with potential additional prognostic value. Applying PERCI in a clinical setting may allow patient-specific drug selection to overcome resistance. WSG-ADAPT, NCT01779206 , Registered 2013-01-25, retrospectively registered.