Molecular correlates of response to eribulin and pembrolizumab in hormone receptor-positive metastatic breast cancer

Abstract

Immune checkpoint inhibitors (ICIs) have minimal therapeutic effect in hormone receptor-positive (HR+ ) breast cancer. We present final overall survival (OS) results (n = 88) from a randomized phase 2 trial of eribulin ± pembrolizumab for patients with metastatic HR+ breast cancer, computationally dissect genomic and/or transcriptomic data from pre-treatment tumors (n = 52) for molecular associations with efficacy, and identify cytokine changes differentiating response and ICI-related toxicity (n = 58). Despite no improvement in OS with combination therapy (hazard ratio 0.95, 95% CI 0.59-1.55, p = 0.84), immune infiltration and antigen presentation distinguished responding tumors, while tumor heterogeneity and estrogen signaling independently associated with resistance. Moreover, patients with ICI-related toxicity had lower levels of immunoregulatory cytokines. Broadly, we establish a framework for ICI response in HR+ breast cancer that warrants diagnostic and therapeutic validation. ClinicalTrials.gov Registration: NCT03051659.

Conflict of interest statement

R.B.-S. has served as an advisor/consultant to Eli Lilly and has received honoraria from Roche for participation in Speakers Bureau. J.L.G. is a consultant for GlaxoSmithKline (GSK), Array BioPharma, Codagenix, Verseau, and Kymera and receives sponsored research support from GSK, Eli Lilly, and Array BioPharma. I.V.L. has received institutional honoraria from Pfizer, AstraZeneca, and Amgen. M.X.H. has been a consultant to Amplify Medicines and Ikena Oncology. L.M.S. has been a consultant/advisor for Novartis and Avrobio. N.U.L. has received institutional research funding from Genentech, Cascadian Therapeutics, Array Biopharma, Seattle Genetics, Novartis, Merck, and Pfizer and has been a consultant/advisor to Seattle Genetics, Puma, and Daichii Sankyo. E.P.W. receives consulting fees from InfiniteMD and Leap Therapeutics, honoraria from Genentech, Roche, Tesaro, Lilly, and institutional research funding from Genentech. E.A.M. reports personal financial interests: research support for lab from GlaxoSmithKline; honoraria from Physician Education Resource; compensated service on Scientific Advisory Boards for AstraZeneca, Exact Sciences (formerly Genomic Health), Merck, Peregrine Pharmaceuticals, Roche/Genentech, Sellas Lifesciences, TapImmune Inc; uncompensated service on Steering Committees for BMS, Lilly, Roche/Genentech. E.A.M. reports institutional financial interests from MD Anderson: clinical trial funding from AstraZeneca, EMD Serono, Galena Biopharma, Roche/Genentech; and institutional financial interests from DFCI: clinical trial funding from Roche/Genentech (via SU2C grant). E.M.V.A. serves as a consultant/advisor to Tango Therapeutics, Invitae, Genome Medical, Dynamo, Foresite Capital, and Illumina; holds research support from Novartis and Bristol-Myers Squibb; and holds equity in Synapse, Genome Medical, Tango, and Microsoft Corp. S.M.T. receives institutional research funding from AstraZeneca, Lilly, Merck, Nektar, Novartis, Pfizer, Genentech/Roche, Immunomedics, Exelixis, Bristol-Myers Squibb, Eisai, Nanostring, Cyclacel, Odonate, and Seattle Genetics; has served as an advisor/consultant to AstraZeneca, Lilly, Merck, Nektar, Novartis, Pfizer, Genentech/Roche, Immunomedics, Bristol-Myers Squibb, Eisai, Nanostring, Puma, Sanofi, Celldex, Paxman, Puma, Silverback Therapeutics, G1 Therapeutics, AbbVie, Anthenex, OncoPep, Outcomes4Me, Kyowa Kirin Pharmaceuticals, Daiichi-Sankyo, and Samsung Bioepsis Inc. The remaining authors declare no competing interests.

© 2021. The Author(s).

Figures

Fig. 1. Cohort clinical and genomic characteristics and tumor mutation burden (TMB).

a, b Updated overall survival curves for eribulin ± pembrolizumab in metastatic TNBC show no difference in all patients (a) and in patients with PD-L1 + tumors (b). c Comutation plot shows no association between TMB and response. Each column represents a tumor. Tumors are ordered by RECIST response, and within each response subgroup by decreasing nonsynonymous (Nonsyn) mutational load (top row). Nonsynonymous mutational burden is further subdivided into clonal (dark blue) and subclonal (light blue) mutational load. Tumor purity is the inferred proportion of the tumor sample that is from cancer cells compared to other cell types (Methods). The protocol therapy and biopsy timing (archival primary or metastatic tumor vs. baseline metastatic tumor collected during trial screening) are indicated. Mutational signatures (sig) indicate the dominant signature present at the highest relative proportion and the presence or absence of APOBEC signatures consisting of COSMIC signatures 2 or 13,. Mutations and copy number alterations in genes commonly mutated in breast cancer are shown for each tumor. d, e In tumors treated with eribulin and pembrolizumab, nonsynonymous mutational burden was not different in patients with clinical benefit (green) vs. those without clinical benefit (yellow) (d) and did not differ by RECIST response (e). Unadjusted two-sided Mann–Whitney–Wilcoxon p-values are shown. Boxplot limits indicate the interquartile range (IQR; 25th to 75th percentile), with a center line indicating the median. Whiskers show the value ranges up to 1.5 × IQR above the 75th or below the 25th percentile with outliers beyond those ranges shown as individual points. Mb, megabase; NE, not evaluable; PD, progressive disease; PR, partial response; SD, stable disease; TNBC, triple-negative breast cancer.

Fig. 2. Tumor heterogeneity and purity correlate with resistance to eribulin and pembrolizumab.

a–c Focused analysis of eribulin pembrolizumab WES cohort (n = 27): a Tumor ploidy, defined as the overall genomic copy number (a normal diploid cell has a copy number of 2; Methods) was not different in patients with clinical benefit (green) versus those without clinical benefit (yellow) or by RECIST response. b Tumor heterogeneity, defined as the proportion of subclonal mutations in each tumor (Methods), was lower in patients with clinical benefit (green) versus those without clinical benefit (yellow) and was lower in patients with partial response (PR, green) vs. progressive disease (PD, yellow). c Tumor purity, defined as the proportion of DNA from tumor versus other cells in the sample (Methods), was lower in patients with clinical benefit (green) versus those without clinical benefit (yellow) and was lower in patients with partial response (PR, green) vs. progressive disease (PD, yellow). Unadjusted two-sided Mann-Whitney-Wilcoxon p values are shown. Boxplot limits indicate the interquartile range (IQR; 25th to 75th percentile), with a center line indicating the median. Whiskers show the value ranges up to 1.5 × IQR above the 75th or below the 25th percentile with outliers beyond those ranges shown as individual points. d, e In the overall WES cohort treated with eribulin ± pembrolizumab (n = 50), patients with tumors that had top quartile heterogeneity (d) and purity (e) had longer progression-free survival. Unadjusted two-sided log rank p-values are shown. NE, not evaluable; SD, stable disease; WES, whole exome sequencing.

Fig. 3. Immune gene set enrichment correlates with response to eribulin and pembrolizumab.

a For tumors treated with eribulin and pembrolizumab (pembro), the top 5 cancer hallmark gene sets (GSEA) enriched in patients with clinical benefit (n = 6) versus those without clinical benefit (n = 8) consisted of immune-related pathways with normalized enrichment scores (NES) > 2.25 and false discovery rate (FDR) q values < 0.001. b For tumors treated with eribulin ± pembrolizumab, the top 5 cancer hallmark gene sets (GSEA) enriched in patients with clinical benefit (n = 11) versus those without clinical benefit (n = 19) also consisted of immune-related pathways with NES > 2.25 and FDR q values < 0.001. c For tumors treated with eribulin alone, the same immune-related gene sets (GSEA) were less enriched in patients with clinical benefit (n = 5) versus those without clinical benefit (n = 11), as these gene sets were not all ranked in the top 5 enriched gene sets, had lower NES down to 1.52, and larger FDR q values up to 0.023. d A heatmap of single-sample gene set enrichment analysis (ssGSEA) score, where each column is a tumor arranged first by clinical benefit (green) versus no clinical benefit (yellow) and then by eribulin pembrolizumab treatment arm (purple) versus eribulin treatment arm (gray), shows that tumors with clinical benefit had enrichment of these immune gene sets. Color indicates the ssGSEA score from least enriched (blue) to most enriched (red). e, f In the overall RNAseq cohort treated with eribulin ± pembrolizumab (n = 30), patients with tumors that had high versus low antigen presentation (AP) gene set enrichment scores (ssGSEA; divided by the median) had longer progression-free (e) and overall survival (f) in unadjusted analyses. All Hallmark pathways and their GSEA enrichment scores are included in Supplementary Data File 1: Supplementary Table 3, and the antigen presentation gene set is shown in Supplementary Data File 1: Supplementary Table 4. Unadjusted two-sided log rank p-values are shown.

Fig. 4. Tumor immune infiltration correlates with response to eribulin ± pembrolizumab.

a Absolute immune cell infiltrate inferred by CIBERSORTx was higher in patients with clinical benefit (green) versus those without clinical benefit (yellow) to eribulin + /- pembrolizumab. b Tumor infiltrating lymphocytes, calculated as the sum of all lymphocytes inferred by CIBERSORTx (Methods), were higher in patients with clinical benefit (green) versus those without clinical benefit (yellow) to eribulin ± pembrolizumab. c, d Resting memory CD4 + T cells (c) and follicular helper T cells (d) were higher in patients with clinical benefit (green) versus those without clinical benefit (yellow) to eribulin ± pembrolizumab. e, f M2 macrophages (e) were higher in patients with clinical benefit (green) versus those without clinical benefit (yellow) to eribulin ± pembrolizumab, while M1 macrophages (f) trended towards being higher in patients with clinical benefit (green) with borderline significance (p = 0.05). Unadjusted two-sided Mann–Whitney–Wilcoxon p-values are shown. Boxplot limits indicate the interquartile range (IQR; 25th to 75th percentile), with a center line indicating the median. Whiskers show the value ranges up to 1.5 × IQR above the 75th or below the 25th percentile with outliers beyond those ranges shown as individual points.

Fig. 5. Estrogen signaling correlates with resistance to eribulin and pembrolizumab.

a For tumors treated with eribulin and pembrolizumab (pembro), the top 2 cancer hallmark gene sets (GSEA) enriched in patients without clinical benefit (n = 8) versus those with clinical benefit (n = 6) consisted of estrogen response pathways with normalized enrichment scores (NES) < −1.75 and false discovery rate (FDR) q values < 0.01. b For tumors treated with eribulin ± pembrolizumab, the top 2 cancer hallmark gene sets (GSEA) enriched in patients without clinical benefit (n = 19) versus those with clinical benefit (n = 11) consisted of protein secretion and early estrogen response pathways with NES < −1.75 and FDR q values < 0.001. c For tumors treated with eribulin alone, the top cancer hallmark gene sets (GSEA) enriched in patients without clinical benefit (n = 11) versus those with clinical benefit (n = 5) was the protein secretion pathway. d A heatmap of single-sample GSEA (ssGSEA) score, where each column is a tumor arranged first by clinical benefit (green) versus no clinical benefit (yellow) and then by eribulin pembrolizumab treatment arm (purple) versus eribulin treatment arm (gray), shows that tumors with no clinical benefit had enrichment of these gene sets, particularly estrogen response signaling. Color indicates the ssGSEA score from least enriched (blue) to most enriched (red). e, f In the overall RNAseq cohort treated with eribulin ± pembrolizumab (n = 30), patients with tumors that had top quartile early estrogen response (ER) gene set enrichment scores (ssGSEA) had shorter progression-free (e) and overall survival (f) in unadjusted analyses. Unadjusted two-sided log rank p values are shown. All Hallmark pathways and their GSEA enrichment scores are shown in Supplementary Data File 1: Supplementary Table 3.

Fig. 6. Response correlate interplay and cytokine analyses.

a Spearman’s correlation coefficients between genomic and transcriptomic features associated with response and b hierarchical clustering of these correlation coefficients. Color indicates the Spearman’s correlation between features, from perfect negative correlation (Spearman’s r = −1, blue) to perfect positive correlation (Spearman’s r = 1, red). Tumor infiltrating lymphocytes (TILs), immune infiltrate, and antigen presentation single-sample gene set enrichment analysis (ssGSEA) score all clustered together, while tumor purity negatively correlated with this cluster. Tumor heterogeneity and estrogen response ssGSEA score were independent from the immune-infiltrated cluster and tumor purity. The sample size for each correlation depended on the number of available data points: correlations involving exclusively genomic or transcriptomic data had n = 50 or 30 tumor samples, respectively, whereas correlations involving genomic and transcriptomic features had n = 28 tumor samples with data available. c Schematic representation of potential interplay of genomic and transcriptomic features. Antigen presentation may increase responses in non-immune infiltrated tumors, while heterogeneity and estrogen signaling may reduce responses in immune-infiltrated tumors. d Heatmap of median pre- to on-treatment fold changes in cytokines across three immune-related toxicity groups: patients treated with (1) eribulin or (2) eribulin and pembrolizumab with no immune-related toxicity versus patients treated with (3) eribulin and pembrolizumab with immune-related toxicity. Shown are unadjusted two-sided Mann–Whitney–Wilcoxon p-values for significant fold-change (FC) differences between adjacent groups. The sample size for each group depended on the number of available data points and ranged from 15–19 patients as indicated in Supplementary Data File 1: Supplementary Tables 10–11. AP, antigen presentation; ER, estrogen response; TMB, tumor mutational burden; tox, toxicity.