Abstract PD6-03: High prevalence and clonal heterogeneity of ESR1 mutations (mt) in circulating tumor DNA (ctDNA) from patients (pts) enrolled in FERGI, a randomized phase II study testing pictilisib (GDC-0941) in combination with fulvestrant (F) in pts that failed a prior aromatase inhibitor (AI)

Background: Mutations in the ligand binding domain of the estrogen receptor gene (ESR1) have been associated with resistance to AI therapy in pts with ER+ breast cancer. To assess if ESR1 status has prognostic or predictive significance in the post-AI metastatic setting ESR1 mutation status was analyzed in circulating tumor DNA (ctDNA) from 168 pts enrolled on the FERGI study (NCT01437566; Krop et al., SABCS 2014).

Methods: Baseline and longitudinal mutational analysis for hotspot mutations in ESR1 (E380Q, S463P, V534E, P535H, L536R/H/P, L536Q, Y537N/S/C, D538G) and PIK3CA (C420R, E542K, E545K/G, Q546K, M1043I, H1047Y/R/L) was performed using droplet digital PCR (ddPCR) on ctDNA derived from plasma. Archival tissue was analyzed via RT-PCR and ddPCR.

Results: Baseline ctDNA analysis demonstrated a total of 62/156 (40%) and 57/153 (37%) pts with PIK3CA and ESR1 mutations, respectively. The most common ESR1 mutations are D538G, Y537S, and E380Q, representing 54%, 33% and 26% of the pts with a detectable ESR1 mutation at baseline, respectively. There was a numeric increase of ESR1 mutations in patients with LumA (41/99, 41%) vs LumB disease (14/44, 31%). PIK3CA mutations in asynchronously collected archival tissue were 85% concordant with plasma ctDNA mutations (sensitivity 78%, specificity 91%). PIK3CA mutations in baseline ctDNA showed a higher median allele frequency (AF) than ESR1 mutations (3.6% vs 0.46%), consistent with PIK3CA being an early event and ESR1 mutations occurring later in pts with recurrent disease. Of the pts with a detectable ESR1 mutation at baseline (n=57), 23 (40%) pts had multiple ESR1 mutations and 10 (18%) had ≥3 ESR1 mutations. The PFS outcomes for patients with and without ESR1 mutations detected at baseline are summarized below, indicating no obvious prognostic or predictive effect for combination of F with pictilisib compared with F in these underpowered subsets.

ArmESR1 MT - mPFS (mo)ESR1 WT - mPFS (mo)HR (95% CI)F + placebo5.4 (30 pts, 24 events)3.7 (40 pts, 31 events)1.06 (0.62, 1.81)F+pictilisib5.8 (27 pts, 20 events)6.7 (56 pts, 34 events)1.36 (0.78, 2.38)

PIK3CA and ESR1 ctDNA analysis on serial plasma samples from 40 pts and the assessment of ESR1 mutation status in the patient's tumor sample by ddPCR is currently in progress and will be reported.

Conclusions: Mutations in ESR1 detected by ddPCR in patient plasma samples occur in nearly 40% of pts that failed a prior AI. The polyclonal nature of ESR1 mutations is consistent with the convergent evolution of multiple AI resistant subclones. While these conclusions should be interpreted with caution due to the relatively small sample size and post hoc nature of the analysis, this data does not support a prognostic or predictive PFS hypothesis for ESR1 mutations with F or in combination with pictilisib.

Citation Format: Spoerke J, Gendreau S, Johnston S, Schmid P, Krop I, Qui J, Derynck M, Chan I, Walter K, Amler L, Hampton G, Lackner M. High prevalence and clonal heterogeneity of ESR1 mutations (mt) in circulating tumor DNA (ctDNA) from patients (pts) enrolled in FERGI, a randomized phase II study testing pictilisib (GDC-0941) in combination with fulvestrant (F) in pts that failed a prior aromatase inhibitor (AI). [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr PD6-03.

Abstract P6-05-09: Development of a predictive biomarker gene expression signature for the PIK3CA inhibitor, GDC-0032, in breast cancer cells

Introduction

The PI3-Kinase pathway is one of the most commonly mutated pathways in cancer and plays a major role in cell proliferation and survival. Mutations in PIK3CA, the gene encoding the p110 subunit of PI3K, are among the most common alterations in breast cancer, occurring in approximately 45% of luminal A, 30% of luminal B, 30% of HER2 positive and 8% of triple negative breast cancers. Additional pathway activating alterations include loss of PTEN, AKT mutations and overexpression of PIK3CA and HER2. Development of a pharmacodynamic biomarker is challenging with the more isoform specific PI3K inhibitors as multiple upstream pathways can funnel into common downstream immunohistochemical evaluable endpoints. In addition, phosphorylated epitopes are often labile and do not always lend themselves to immunohistochemical evaluation in the clinical setting. GDC-0032, which is currently under clinical investigation, is a class I PI3K inhibitor with 30-fold less inhibition on PI3K beta relative to PI3K alpha, and the development of a predictive and on-study pharmacodynamic signature may prove informative as compared to traditional IHC endpoints.

Methods

We screened a panel of 53 breast cancer cell lines, incorporating all subtypes, to GDC-0032 using the cell proliferation assay cell titer glo. To determine if there was a relationship between pathway activation and sensitivity to GDC-0032, we correlated response to PIK3CA mutations, loss of PTEN and HER2 overexpression. Using RNA sequencing, we compared the baseline gene expression between the sensitive and refractory cell lines. Next, to identify an on-study pharmacodynamic gene expression signature, we treated both sensitive and refractory cell lines with GDC-0032 and ran an in-house custom designed 800 gene NanoString breast cancer gene set that incorporated published PI3K pathway signatures, intrinsic subtyping genes and immunological related genes. Finally, the GDC-0032 signature was applied to a set of 160 FFPE breast cancer samples and overlaid with relevant biomarkers.

Results and Conclusions

Sensitivity to GDC-0032 correlated strongly with PI3K pathway activation including PIK3CA mutations and HER2 overexpression in breast cancer cells. Comparing baseline whole genome RNA expression of GDC-0032 sensitive and refractory cell lines, we identified 293 genes that were differentially expressed. Applying a more stringent statistical cutoff (greater than 2 fold difference and t-test less than 0.01) refined the gene list to 51 genes, which defined the baseline GDC-0032 sensitivity signature. Applying the 800 gene breast cancer NanoString panel to a set of 160 FFPE breast cancer samples, the GDC-0032 sensitivity signature correlated with luminal status and was enriched in PIK3CA mutant tumors. In conclusion, our in-house designed GDC-0032 sensitivity signature correlated strongly with PIK3CA mutations in clinical specimens. However the lack of complete correlation may identify tumors that have an activated PI3K pathway outside of PIK3CA mutations and/or HER2 amplification that may derive clinical benefit from GDC-0032.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P6-05-09.