Multifocal clonal evolution characterized using circulating tumour DNA in a case of metastatic breast cancer

Abstract

Circulating tumour DNA analysis can be used to track tumour burden and analyse cancer genomes non-invasively but the extent to which it represents metastatic heterogeneity is unknown. Here we follow a patient with metastatic ER-positive and HER2-positive breast cancer receiving two lines of targeted therapy over 3 years. We characterize genomic architecture and infer clonal evolution in eight tumour biopsies and nine plasma samples collected over 1,193 days of clinical follow-up using exome and targeted amplicon sequencing. Mutation levels in the plasma samples reflect the clonal hierarchy inferred from sequencing of tumour biopsies. Serial changes in circulating levels of sub-clonal private mutations correlate with different treatment responses between metastatic sites. This comparison of biopsy and plasma samples in a single patient with metastatic breast cancer shows that circulating tumour DNA can allow real-time sampling of multifocal clonal evolution.

Figure 1. Inference of clonal structure from multiregional tumour biopsies.

(a) Tumour samples collected from the patient, labelled P (primary) and M (metastasis). Numbers preceding dot (1,2 and 3) correspond to time of collection: 1, collected at diagnosis; 2, collected at the time of resection of brain metastasis; 3, collected at autopsy. (b) Timeline describing clinical course, samples collected, treatments administered and selected imaging assessments. Plasma DNA samples are labelled 1 through 9. Imaging assessments were performed using computed tomography scans. Histopathological and imaging findings are summarized in Supplementary Tables 1 and 2 and Supplementary Figs 1–6. (c) Distribution of 207 validated functional mutations in tumour and plasma samples, ordered by mutation clusters inferred using PyClone from mutant allele fractions in all tumours. Red rectangles indicate high-confidence mutations with AF >2%. Blue rectangles indicated mutations detected significantly above background but with AF of 2% or lower. Stem mutations (observed ubiquitously in all tumour samples and comprising tumour cluster 1) and metastatic-clade mutations (high confidence in metastatic tumours and comprising tumour cluster 2) are readily identifiable/detectable in plasma samples. Detailed values of allele fractions are documented in Supplementary Data 1–3. (d) Tumour phylogenetic tree, inferred by clonal ordering given distribution of high-confidence mutations in tumour samples shown in a. Length of each branch of the tree correlates with the number of mutations on the branch as indicated. Exome-sequencing results for samples P1.2 and M3.4 were not available and therefore private mutations for these branches cannot be identified. Assignment of mutations to each branch is documented in Supplementary Data 1. (e) Mean predicted cellular frequency of each cluster identified by PyClone across the tumour samples.

Figure 2. Serial plasma analysis during systemic treatment.

(a) Circulating levels of stem, metastatic-clade and private mutations during treatment. Mean allele fractions at each time point are presented. Mean AF for private mutations is multiplied by 10 to highlight trend. Shaded areas represent treatment lines. (b) Mean predicted cellular frequency of each cluster identified by PyClone across the plasma samples T2–T9. PyClone identified five mutation clusters from variation of circulating allele fractions (without reliance on tumour data). Clusters 1 and 2 are largely comprised of stem and metastatic-clade mutations. Cluster 5 is comprised of 11 mutations, 10 of which are private M3.1 mutations. (c) Plasma abundance calculated as the product of AF in a tumour sample (normalized for mean of stem mutations) and the corresponding AF in a plasma sample, summed across all private mutations for each tumour. To normalize for different number of private mutations in each tumour (3–70), we calculated plasma abundance relative to T1. (d) Dynamics of ERBB4 mutation (p.H809G) in deep sequencing data. (e) Allele fractions measured by deep amplicon sequencing for the PIK3CA mutation (p.E542K) identified in exome sequencing of plasma sample T2. Mutation was significantly detectable (>3 s.d.'s above the mean allele fraction in control samples) on days 727, 762 and 937 (yellow diamonds).