Phase 2 study of neoadjuvant enzalutamide and paclitaxel for luminal androgen receptor-enriched TNBC: Trial results and insights into "ARness"

Abstract

Luminal androgen receptor (LAR)-enriched triple-negative breast cancer (TNBC) is a distinct subtype. The efficacy of AR inhibitors and the relevant biomarkers in neoadjuvant therapy (NAT) are yet to be determined. We tested the combination of the AR inhibitor enzalutamide (120 mg daily by mouth) and paclitaxel (80 mg/m² weekly intravenously) (ZT) for 12 weeks as NAT for LAR-enriched TNBC. Eligibility criteria included a percentage of cells expressing nuclear AR by immunohistochemistry (iAR) of at least 10% and a reduction in sonographic volume of less than 70% after four cycles of doxorubicin and cyclophosphamide. Twenty-four patients were enrolled. Ten achieved a pathologic complete response or residual cancer burden-I. ZT was safe, with no unexpected side effects. An iAR of at least 70% had a positive predictive value of 0.92 and a negative predictive value of 0.97 in predicting LAR-enriched TNBC according to RNA-based assays. Our data support future trials of AR blockade in early-stage LAR-enriched TNBC.

Keywords: ARness; LAR subtype; TNBC; androgen receptor; biomarker of response; enzalutamide; neoadjuvant systemic therapy; triple-negative breast cancer.

Graphical abstract

Figure 1

Summary of clinical characteristics (A) A total of 311 patients were enrolled under ARTEMIS, the principal biomarker assessment protocol, by data closure. Of these, 119 patients, who were determined to have a suboptimal response to AC, were evaluated for compatibility with the ZT protocol. Following the exclusion of 5 patients, 24 patients proceeded to undergo treatment as per the ZT protocol. (B) The trial met its primary endpoint target by achieving the goal rate of greater than 20% of patients having a pCR or RCB-I. Four patients had pCR, and six had RCB-I, resulting in 42% of all patients accrued to this trial having a pCR or RCB-1. (C) The 48-month EFS rate was 33.4% (95% CI, 12.6%–88.2%) for all patients, 16.9% for patients with RCB-II/III, and 67% for patients with pCR or RCB-I. The difference between these groups was statistically significant (p = 0.0043). (D) The 48-month OS rate was 63.0% (95% CI, 40.2%–98.4%) for all patients, 36.4% for patients with RCB-II/III, and 100% for patients with pCR or RCB-I. The difference between these groups was statistically significant (p = 0.032).

Figure 2

LAR-enriched TNBC and correlation with IHC and clinical response to enzalutamide-based treatment (A) We assessed two different RNA expression measuring assays (RNA microarray, RNA-seq) and IHC measurement of AR nuclear staining (iAR) as ARness biomarkers and correlated them with ZT response. The LAR-enriched TNBC subtype label is based on the RNA Affymetrix microarray, and RNA-seq had 139 overlapping genes included in each assay. (B) The subtype calling showed 99% matching between microarray and sequencing methods, and the high iAR values were concordant with high RNA-based LAR calling (positive correlation with R=0.87). (C) The LAR-enrichment TNBC subtype calls according to the RNA Affymetrix vs. iAR and RNA-seq vs. iAR showed similar concordance. In both correlation analyses, most LAR-enriched TNBCs fell into the iAR ≥ 70% category. (D) Among patients with LAR-enriched TNBC subtype treated with ZT, neither LAR enrichment score nor iAR differed significantly between patients with pCR or RCB-I and those with RCB-II/III. p values were not significant (0.96 in RNAseq, 0.81 in IHC comparison. (E) Among 14 patients who had profiling of pretreatment biopsies and complete molecular profiling, Hallmark gene set analysis showed the AR response gene pathway as the only marker that correlated with GRG. The AR response gene set expression between GRG vs. PRG showed a difference (p = 0.05). (F) The EFS of iAR <30% vs. >30% and iAR <70% vs. >70% showed a trend favoring superior EFS in higher iAR. However, the difference was not statistically significant.

Figure 3

Evolution and stability of AR-enriched TNBCs and TME over treatment with AC (A) Unlike other TNBC subtypes, the LAR signature remained stable after exposure to AC in LAR TNBCs profiled longitudinally throughout T0 (baseline) and T1 (after four cycles of AC) except for one case. (B) Among investigated TME cells, B cells, macrophages, natural killer cells, and T cells were reduced regardless of the response of tumors to ZT or AC. However, neutrophil and myeloid dendritic cell marker expressions increased throughout the treatment. Error bar in the box plot indicates standard deviation of each value (ranging from −1 to 1).

Figure 4

The potential therapeutic target of LAR TNBC (A) 220 ARTEMIS tumor cohorts with available microarray profiles were assessed for the overlap between overexpressed genes and the available drug library. LAR-enriched TNBCs show overlap with 52 genes that had available targeted inhibitors. AR was the highest expressed gene, while others like FOXA1, SPDEF, AGR2, FGFR4, PIP, and ZNF552 also showed elevated expression. Among these genes, FOXA1, SPDEF, and FGFR4 showed overlap with the Dependency Map (DepMap) screening results in genes. (B) DepMap screening using the LAR-enriched cohort revealed FOXA1, FGFR4, and SPDEF as the top three genes that had the highest expression with high dependency among LAR-enriched tumors. (C) Baylor College of Medicine patient-derived xenograft (PDX) breast cancer cohort showed various expression levels of AR and SPDEF, FOXA1, and FGFR4. Pearson pairwise correlation assay showed that these four genes correlate with each other with a high coefficient. (D) Pearson pairwise correlation assay using 4 genes (AR, SPDEF, FOXA1, FGFR4) + a 150 AR response gene signature showed clustering of these genes with others (FOXA1, SPDEF, and FGFR4 clustered with FASN, AGR2, SLC44A4, TSPAN13, and XBP1 genes with strong correlation, while AR showed stronger correlation with the genes RBM25 and BRD2.