The Genetic Landscape and Clonal Evolution of Breast Cancer Resistance to Palbociclib plus Fulvestrant in the PALOMA-3 Trial

Abstract

CDK4/6 inhibition with endocrine therapy is now a standard of care for advanced estrogen receptor-positive breast cancer. Mechanisms of CDK4/6 inhibitor resistance have been described preclinically, with limited evidence from clinical samples. We conducted paired baseline and end-of-treatment circulating tumor DNA sequencing from 195 patients in the PALOMA-3 randomized phase III trial of palbociclib plus fulvestrant versus placebo plus fulvestrant. We show that clonal evolution occurs frequently during treatment, reflecting substantial subclonal complexity in breast cancer that has progressed after prior endocrine therapy. RB1 mutations emerged only in the palbociclib plus fulvestrant arm and in a minority of patients (6/127, 4.7%, P = 0.041). New driver mutations emerged in PIK3CA (P = 0.00069) and ESR1 after treatment in both arms, in particular ESR1 Y537S (P = 0.0037). Evolution of driver gene mutations was uncommon in patients progressing early on palbociclib plus fulvestrant but common in patients progressing later on treatment. These findings inform future treatment strategies to address resistance to palbociclib plus fulvestrant.Significance: Acquired mutations from fulvestrant are a major driver of resistance to fulvestrant and palbociclib combination therapy. ESR1 Y537S mutation promotes resistance to fulvestrant. Clonal evolution results in frequent acquisition of driver mutations in patients progressing late on therapy, which suggests that early and late progression have distinct mechanisms of resistance. Cancer Discov; 8(11); 1390-403. ©2018 AACR. See related commentary by Schiff and Jeselsohn, p. 1352 This article is highlighted in the In This Issue feature, p. 1333.

Figure 1. Paired circulating tumor DNA (ctDNA) exome sequencing reveals frequent clonal selection on fulvestrant plus palbociclib

A – Day 1 and end of treatment plasma samples from the PALOMA-3 trial were screened using droplet digital PCR and a targeted SNP sequencing approach to identify patients from the palbociclib plus fulvestrant arm who had paired plasma samples of sufficient tumor purity (>10%) for plasma exome sequencing. B – Paired ctDNA exome sequencing in patient 390 analyzed for clonal composition. A newly emergent RB1-mutant clone is detected at EOT, harboring two inactivating RB1 mutations. C – Inferred phylogenetic tree of breast cancer from patient 390 derived from ctDNA. Yellow - truncal mutations present in all cancer cells; Purple -subclone present at day 1 that subsequently regressed on treatment: Gray - a newly emergent resistant clone characterized by two RB1 mutations, arising separately to the purple sub clone. D – Representation of mutational signatures identified in each individual sub clone for patient 390. The raw data are shown in Supplementary table 3. E – Digital PCR validation of the two RB1 mutations Q257X and N519fs identified in the treatment-resistant sub clone from patient 390. Results for day 1 and EOT are shown, Y axis mutant probe amplitude and X axis wild type probe amplitude. F – Paired ctDNA exome sequencing in patient 253 analyzed for clonal composition. A new FGFR2 mutant clone undetectable at day 1 is detected at EOT. G – Inferred phylogenetic tree of breast cancer from patient 253 derived from ctDNA. Yellow - truncal mutations present in all cancer cells; Purple - subclone characterized by an ESR1 D538G mutation present at day 1 that subsequently regressed on treatment; Orange - newly emergent resistant clone characterized by an FGFR2 kinase domain mutation, arising separately from the purple sub clone; Gray – a sub clone arising from the FGFR2-mutant sub clone characterized by a Q75E ESR1 mutation. H – Representation of mutational signatures identified in each individual clone from patient 253. The raw data are shown in Supplementary table 3. I – Digital PCR validation of the FGFR2 mutation from patient 253 showing results for plasma at day 1 and EOT. EOT – end of treatment, HR – homologous recombination, MMR – mismatch repair, ctDNA – circulating tumor DNA

Figure 2. Genetic landscape of breast cancer driver genes in paired plasma samples on PALOMA-3 with frequent selection of mutations on treatment

A – Paired ctDNA sequencing results with frequency of observed variants in genes included in the targeted driver gene panel (SNVs and indels). Results are shown for both day 1 and end of treatment in 195 patients with matched data from day 1 and EOT. P values McNemar’s test with continuity correction. B – RB1 mutations identified at EOT in patients treated with palbociclib plus fulvestrant (6/127). No RB1 mutations were identified in the EOT plasma samples from patients treated with placebo and fulvestrant (n = 68). The digital PCR plot shows orthogonal validation of Q257X in the EOT sample from patient 418. C – End of treatment (EOT) ctDNA sequencing results from 195 patients with paired samples, split by treatment, and whether the mutation status changed on treatment between day 1 and EOT. The cohort of 195 patients is formed from both the targeted sequencing cohort (n = 184) and exome sequencing cohort (n = 14, with n = 3 in both sets, see Supplementary figure 1). The pattern of mutation acquisition is similar across both treatment arms. SNV – single nucleotide variant, indel – insertion or deletion, EOT – end of treatment, ctDNA – circulating tumor DNA. Mixed – patients with different variants in the same gene at day 1 and EOT.

Figure 3. Positive selection of ESR1 Y537S on fulvestrant based treatment

A – Percentages of patients at baseline and end of treatment with specific ESR1 mutations observed in paired ctDNA sequencing data (n = 195, treatment groups combined). Data includes bothfulvestrant plus placebo and fulvestrant plus palbociclib groups together. P values calculated from McNemar’s test with continuity correction, and q value after Bonferroni correction for multiple comparisons. B – Validation of acquired ESR1 Y537S end of treatment mutations with digital PCR (17/17). Allele fraction is plotted for each technique. C – Concordance between sequencing and digital PCR for allele fraction in patients with an ESR1 Y537S mutation in ctDNA sequencing at either day 1 or end of treatment. Blue - concordant calls present in both sequencing and digital PCR. Orange - present only in digital PCR. ctDNA – circulating tumor DNA

Figure 4. Clonal evolution of breast cancer driver genes through treatment

A – Individual variants for each gene in the treatment groups combined (n=195), split by variants maintained between day 1 and end of treatment, lost over the course of treatment, or acquired during treatment. A number of patients had polyclonal variants in a single gene, particularly ESR1. The majority of TP53 acquired variations are accounted for by a single patient acquiring 8 separate new variants at EOT (see also Supplementary figure 9). B – Cartoon with data from patient 237 illustrating subclonal selection on treatment. A clonal PIK3CA mutation and an ESR1-mutant sub clone are detectable at day 1. Over the course of treatment the ESR1-mutant subclone present at day 1 is lost, with acquisition of a new ERBB2-mutant sub clone. C – Sankey diagram to illustrate changes in individual PIK3CA mutations through treatment in both treatment groups combined. Polyclonal mutations from a single patient are displayed separately. Only two PIK3CA mutations detected at day 1 are undetectable at EOT, one from a patient with the other polyclonal mutation detected at EOT. D – Clonal state diagram to illustrate changes in PIK3CA polyclonality through treatment, with each individual patient represented once at day1 and EOT. Inner track demonstrates clonal states, representing different combinations of PIK3CA mutations indicated by segments of the circle. The middle tracks show individual mutations in the clonal states. The outer track shows the number of patients with that specific combination of mutations at day 1 (green bar) and end of treatment (purple bar). The central arrows show changes between day1 and EOT. The plot incorporates data from both treatment arms (n = 195). E – Sankey diagram to illustrate changes in individual ESR1 mutations through treatment in both treatment groups combined. F – Clonal state diagram to illustrate changes in ESR1 polyclonality through treatment, with each individual patient represented once at day1 and EOT, legend see D. EOT – end of treatment

Figure 5. New driver mutations are selected late on treatment with palbociclib plus fulvestrant

A – Swimmers plot of patients with paired sequencing data who received palbociclib plus fulvestrant (n = 127) comparing PFS between patients with any acquired driver mutation at end of treatment (n = 35) versus patients that did not acquire a new driver mutation (n = 92). P value calculated using log rank. B – Genetic landscape of breast cancer progressing after palbociclib plus fulvestrant treatment. Each box shows the percentage of patients with a mutation identified at EOT (gray) and the subset of these patients who have newly acquired a mutation by EOT (yellow) for each gene. Patients with the same gene detected mutant and day 1 and EOT, but with a different pattern of mutations in the gene, are not counted as being acquired. PFS – progression free survival, HR – hazard ratio, CI – Confidence interval, EOT – end of treatment