Genomic profile of advanced breast cancer in circulating tumour DNA

Abstract

The genomics of advanced breast cancer (ABC) has been described through tumour tissue biopsy sequencing, although these approaches are limited by geographical and temporal heterogeneity. Here we use plasma circulating tumour DNA sequencing to interrogate the genomic profile of ABC in 800 patients in the plasmaMATCH trial. We demonstrate diverse subclonal resistance mutations, including enrichment of HER2 mutations in HER2 positive disease, co-occurring ESR1 and MAP kinase pathway mutations in HR + HER2- disease that associate with poor overall survival (p = 0.0092), and multiple PIK3CA mutations in HR + disease that associate with short progression free survival on fulvestrant (p = 0.0036). The fraction of cancer with a mutation, the clonal dominance of a mutation, varied between genes, and within hotspot mutations of ESR1 and PIK3CA. In ER-positive breast cancer subclonal mutations were enriched in an APOBEC mutational signature, with second hit PIK3CA mutations acquired subclonally and at sites characteristic of APOBEC mutagenesis. This study utilises circulating tumour DNA analysis in a large clinical trial to demonstrate the subclonal diversification of pre-treated advanced breast cancer, identifying distinct mutational processes in advanced ER-positive breast cancer, and novel therapeutic opportunities.

Trial registration: ClinicalTrials.gov NCT03182634.

Conflict of interest statement

N.C.T., AR, JMB., L.S.K., C.S., L.M., S.K., K.W., S.M., H.B., M.H., B.K, I.G.M., M.B., G.W.-C., S.H. and R.C. report grants from Cancer Research UK, grants and non-financial support in the form of study drug provision from AstraZeneca and Puma Biotechnology and non-financial support in the form of ctDNA sequencing from Guardant Health and provision of reagents from BioRad during the conduct of the study. N.C.T. also reports grants and personal fees from AstraZeneca, Pfizer, and Roche/Genentech, personal fees from Bristol-Myers Squibb, Lilly, MSD, Novartis, Bicycle Theraputics, Taiiho, Zeno Pharmaceuticals and Repare Therapeutics and grants from BioRad, Clovis, Merck Sharpe and Dohme, and Guardant Health outside the submitted work. B.K. also reports personal fees from Guardant Health outside the submitted work. A.M.W. reports personal fees from Roche, personal fees and other support from Novartis, Pfizer, Lilly, Daiichi-Sankyo, MSD, AstraZeneca, Athenex and other support from Seattle Genetics, Andrew Wardley Ltd, Manchester Cancer Academy and Outreach Research and Innovation Group Limited outside the submitted work. I.R.M. reports personal fees and non-financial support from Roche Products UK Ltd, Eli Lilly and Eisai and personal fees from Novartis, Pfizer, Daichi Sankyo, Genomic Health, Pierre Fabre and MSD outside the submitted work. R.D.B. reports grants from AstraZeneca and Roche/Genentech outside the submitted work. R.R. reports personal fees from Novartis, Eli-Lilly and Pfizer, personal fees and non-financial support from Daiichi Sankyo and G1Therapeutics and non-financial support from Roche and AstraZeneca outside the submitted work. H.B. also reports personal fees from AstraZeneca outside of the submitted work. M.H. also reports also reports personal fees from Bristol Myers Squibb, Boehringer Ingelheim, Roche Diagnostics and Eli Lilly during the conduct of the study. J.M.B. also reports grants and non-financial support from AstraZeneca, Novartis, Janssen-Cilag, Merck Sharpe & Dohme, Pfizer, Roche, and Clovis Oncology and grants from Medivation outside the submitted work. A.R. also reports personal fees from Roche Products Limited, Pfizer, Novartis, Lilly and M.S.D. outside the submitted work. J.S.R.-F. is a consultant of Goldman Sachs and REPARE Therapeutics, a member of the scientific advisory board of Volition Rx and Paige.AI, and an ad hoc member of the scientific advisory board of Ventana Medical Systems, Roche Tissue Diagnostics, Genentech, Novartis and InVicro. I.F., K.S.B. and R.B.L. are employees with stock ownership in Guardant Health, Inc.

Figures

Fig. 1. Mutation profile of advanced breast cancer determined by ctDNA sequencing.

a Mutational profile of advanced breast cancer determined by ctDNA targeted sequencing of 800 patients in the plasmaMATCH trial. Displayed are mutations and indels likely to be pathogenic (“Methods”), summarised by gene for each patient. Top bar refers to total counts of pathogenic mutations per patient. Right, variant classification of the alterations within each gene. FS, frameshift; IF, in-frame. b Breast cancer subtype association of the most frequently mutated genes within patients with known phenotype (HR + HER2- N = 515, HER2+ N = 72, TNBC N = 138). Comparison with false discovery corrected two-sided Fisher’s exact tests (TP53: HR + HER2- vs HER2+ q = 0.003, HR + HER2- vs TNBC q < 0.0001, HER2 + vs TNBC q = 0.04; PIK3CA: HR + HER2- vs TNBC q < 0.0001, HER2 + vs TNBC q = 0.0008; ESR1: HR + HER2- vs HER2 + q < 0.0001, HR + HER2- vs TNBC q < 0.0001, HER2 + vs TNBC q = 0.008; GATA3: HR + HER2- vs TNBC q < 0.0001; HER2: HR + HER2- vs HER2 + q = 0.05, HER2 + vs TNBC q = 0.005). ns, not significant. c Patient mutation frequency split by breast cancer subtype, overall and likely pathogenic mutations in patients with known phenotype (HR + HER2- N = 515, HER2 + N = 72, TNBC N = 138). Data are presented as a violin plot with inlayed boxplot, where the middle line is the median, the lower and upper hinges represent the 25th and 75th centiles respectively and the whiskers extend from the hinge to the smallest and largest value, respectively, no further than 1.5 x IQR (interquartile range) from the lower or upper hinge. Data outside of these ranges are plotted individually. Comparison of likely pathogenic mutations with false discovery corrected two-sided pairwise Kruskal-Wallis test, HR + HER2- vs HER2 + q = 0.03, HR + HER2- vs TNBC q < 0.0001. d Frequency of copy number increases in ctDNA split by breast cancer subtype in patients with known phenotype (HR + HER2- N = 515, HER2 + N = 72, TNBC N = 138). Comparison with false discovery corrected Chi-squared tests (MYCq = 0.01, PIK3CAq < 0.0001, FGFR1q = 0.01, EGFRq = 0.001, HER2q < 0.0001, CCNE1q < 0.0001, CCND1 q < 0.001, CDK6q = 0.0003, BRAFq = 0.0002, METq = 0.01). FS, frameshift; IF, in-frame.

Fig. 2. Polyclonal resistance with co-enrichment of MAPK pathway and ESR1 mutations in advanced breast cancer.

a Association analysis for most frequent mutated genes with overall Fisher’s exact test two-sided p-values. Green genes showing mutual exclusivity, and purple showing co-occurrence with dark colours indicating significance following false discovery correction. b Frequency of MAPK pathway alterations comparing ESR1 mutant (77/265) vs ESR1 wild-type cancers overall (100/535) (left), ESR1 mutant (66/226) vs wild-type (59/289) in HR + HER2- cancers (middle), and within ESR1 mutant cancers between patients with single (27/138) and polyclonal ESR1 mutations (50/127) (right). p-values from two-sided Fisher’s exact test. c Example of polyclonal genomic resistance in a patient with multiple MAPK pathway and ESR1 mutations in ctDNA. Blue indicate dominant mutations with cancer fraction, and green subclonal mutations with cancer fraction. d Overall survival (OS) in patients with HR + HER2- disease who entered a treatment cohort in plasmaMATCH divided by combined ESR1 and MAPK pathway mutation status. ESR1 WT and MAPK WT, median 18.5 months, hazard ratio (HR) -. ESR1 mt and MAPK WT, median 17.7 months, HR 0.82, 95% confidence interval (CI) 0.40 to 1.69. ESR1 WT and MAPK mt, median 10.1 months, HR 1.65, 95% CI 0.56 to 4.88. ESR1 mt and MAPK mt, median 7.9 months, HR 1.65, 95% CI 0.84 to 3.23. p-value from log-rank test. HR > 1 indicate worse OS for that group. WT, wild-type; mt, mutant. e Mutational profile of ctDNA in plasmaMATCH (N = 725 patients with known breast cancer subtype) compared to published large metastatic breast cancer tissue sequencing dataset (MSKCC, N = 715 patients with known breast cancer subtype). Red dots indicate significant change in frequency after false discovery adjusted two-sided Fisher’s exact test (HR + HER2-: ESR1q < 0.0001, TP53q = 0.0009). Included are genes with an incidence 1.5% in both data sets.

Fig. 3. Clinical and pathological associations of breast cancer mutation profile.

a Association of number of mutations (SNVs/indels, left) and the maximum variant allele frequency (mVAF, right, as a proxy of ctDNA purity) with indicated clinical and pathological features. p values from pairwise two-sided Kruskal–Wallis test with correction for multiple testing (number of mutations: HR + HER2- vs TNBC q = 0.008; 0 vs ≥5 lines of treatment q = 0.0005, 1–2 vs ≥5 lines of treatment q = 0.0005; 0 vs ≥3 lines of chemotherapy q = 0.003. mVAF: 0 vs ≥5 lines of treatment q = 0.03, 1–2 vs ≥5 lines of treatment q = 0.006; 0 vs 1–2 lines of chemotherapy q = 0.003, 0 vs ≥3 lines of chemotherapy q = 0.0003; soft tissue/nodal vs visceral disease q = 0.002). MBC, metastatic breast cancer; CTx, chemotherapy; IDC, invasive ductal carcinoma; ILC, invasive lobular carcinoma. b Association of clinical and pathological features with pathogenic alterations in the four targetable genes in plasmaMATCH: PIK3CA, ESR1, HER2 and AKT1. p-values from Chi-squared test (PIK3CA: histological subtype p < 0.0001, lines of treatment p = 0.006; ESR1: histological subtype p < 0.0001, lines of treatment p < 0.0001, disease site p = 0.003; HER2: histological subtype p = 0.004, primary breast cancer subtype p < 0.0001). cHER2 mutation incidence in patients with HER2 + cancer, by line of therapy. 0–1 lines of therapy mutation incidence 7.3% (3/41) and 2–3 lines of therapy mutation incidence 25% (8/32) HER2 mutations, p = 0.04, Chi-squared test. mt, mutant; wt wild-type. d) Adjusted HER2 copy number (CN) in targeted sequencing, in patients with tissue assessed HER2 + (amplified, N = 72) and HER2- (non-amplified, N = 605) cancers. (left) receiver operator curve of adjusted HER2 plasma copy number, (right) HER2 plasma copy number adjusted for purity. Data are presented as mean + SD. The p-value indicated is derived from a two-sided Mann–Whitney U-test.

Fig. 4. Organotropism of mutations in advanced breast cancer.

a Association of mutations in indicated genes with sites of metastasis. p-values from false discovery corrected two-sided Fisher’s exact test (ESR1: bone q < 0.0001, liver q < 0.0001, lymph node q = 0.0001; GATA3: bone q = 0.0009; PIK3CA: lymph node q = 0.005; TP53: lymph node q = 0.002, liver q = 0.002, bone q = 0.02). b Association of mutations in indicated genes with sites of metastasis in left HR + HER2- and right TNBC. p-values from false discovery corrected two-sided Fisher’s exact test (HR + HER2- ESR1: liver q = 0.004, bone q = 0.02).

Fig. 5. Clonal dominance and mutational signatures in dominant and subclonal mutations.

a Cancer fractions of mutations in indicated genes, ordered by mean cancer fraction (N = 1974 mutations with assessable cancer fractions). The mean value is indicated with a blue line. *indicates significant difference in cancer fraction compared to remaining cases, false discovery corrected two-sided Wilcoxon signed-rank test (AKT1q < 0.0001, PIK3CAq < 0.0001, GATA3q = 0.0003, ESR1q < 0.0001, SMAD4q = 0.02, KRASq < 0.0001). Cancer fraction—allele fraction of the mutation relative to the maximum somatic allele fraction in the sample. b Proportion of mutations that occur as a single versus multiple mutations per patient in indicated genes. *indicates significant difference in proportion to single to multiple mutations in the gene compared to remaining cases, false discovery corrected two-sided Fisher’s exact test (AKT1q = 0.0009, CDH1q = 0.05, GATA3q < 0.0001, ESR1q < 0.0001). c Bootstrap mutational signature analysis on aggregated mutations from all HR + HER2- (left, clonally dominant mutations N = 328, subclonal mutations N = 968) and TNBC (right, clonally dominant mutations N = 121, subclonal mutations N = 190) breast cancers, for dominant and subclonal mutations. Signature contributions for clonal versus subclonal alterations were ascertained using deconstructSigs and compared using a two-sided Mann–Whitney U-test. Signatures with significant difference in signature contribution and no overlap in interquartile range are identified with the p-value (HR + HER2-: signature 3 p < 0.0001, signature 13 p < 0.0001; TNBC: signature 1 p < 0.0001, signature 5 p < 0.0001, signature p < 0.0001).

Fig. 6. Subclonal multiple PIK3CA mutations and resistance to fulvestrant.

a Cancer fractions of individual pathogenic hotspot mutations in indicated gene, including hotspots with at least 3 mutations in the overall data set or for any indel. p-value from two-sided Kruskal–Wallis test for variation in cancer fraction across mutations in gene (ESR1p < 0.0001, PIK3CAp < 0.0001). b Analysis of patients with H1047R (left, N = 39) and E726K (right, N = 26) dual pathogenic PIK3CA mutations, with linkage of cancer fraction of indicated PIK3CA mutation with the other PIK3CA mutation present in the same patient. The p-values indicated are derived from two-sided Mann–Whitney U-tests (H1047R p < 0.0001, E726K p = 0.02). c Analysis of individual recurrent hotspot mutations in HR + HER2- PIK3CA mutant disease (N = 202) with mean cancer fraction, proportion of mutations detected as a single PIK3CA mutation, and indication of whether the mutation occurs at an APOBEC consensus site. Mutations occurring at least 3 times in the HR + HER2- disease dataset included. Cancer fraction and proportion single mutations is lower at APOBEC sites, P < 0.001, two-sided Fisher’s exact test both comparisons. d Proportion of PIK3CA mutations that occur at APOBEC consensus sites, by cancer subtype (HR + HER2- N = 197, TNBC N = 21), and clonally dominant (N = 194 mutations within HR + HER2- breast cancers and N = 16 within TNBC breast cancers) versus subclonal PIK3CA mutation (N = 82 mutations within HR + HER2- breast cancers and N = 9 within TNBC breast cancers). P-value from two-sided Fisher’s exact test. ns, not significant. e Progression free survival (PFS) in patients on fulvestrant in treatment cohort A in plasmaMATCH divided by PIK3CA mutation status. Cohort A patients with available sequencing data included (78/84). PIK3CA WT, median 2.4 months, HR -. PIK3CA single mt, median 2.4 months, HR 0.71, 95% CI 0.41 to 1.22. PIK3CA multiple mt, median 1.6 months, HR 3.15, 95% CI 0.88 to 11.33. p-value from log-rank test. HR > 1 indicates worse PFS for that group. WT, wild-type; mt, mutant.