ESR1 mutations are frequent in newly diagnosed metastatic and loco-regional recurrence of endocrine-treated breast cancer and carry worse prognosis

A retrospective cohort of hormone receptor-positive (HR+) breast cancer patients experiencing either local or metastatic recurrence was assembled based on available archival samples. All clinical data were obtained from the clinical records of the patients by an expert breast oncologist. This included age, TNM staging, grade, immunohistochemistry scores for estrogen receptor (ER), progesterone receptor (PR), human epidermal growth receptor 2 (HER2), and treatment lines. Outcome information included the date of next local/regional/metastatic recurrence, date of death, and date of the last follow-up. ER and PR positivity were determined based on local pathology practice (> 1% of positively stained cells). Archival formalin-fixed and paraffin-embedded (FFPE) tissue blocks were obtained from the Sheba Medical Center Pathology Institute. A total of 130 archival tumor samples were collected from 103 breast cancer patients who were treated with endocrine therapy prior to their recurrence and provided consent to sample and data collection. For 11 patients, matched samples from the primary tumor were available. In addition, for 14 patients, multiple samples from several recurrence events were collected. The cohort consisted of 41 patients having at least 1 sample from local (ipsilateral or contralateral breast) or regional (axillary nodes) recurrence and 62 patients with metastatic samples. The metastatic samples included 41 samples from newly diagnosed metastases and 28 samples from advanced metastatic disease. Samples from 1 normal breast reduction and 4 normal breast tissue adjacent to fibroadenomas served as controls. This project and tissue collection were approved by the institutional review board.

Tumor samples were sectioned and stained for hematoxylin and eosin to evaluate tumor cell percentage in the slide. Slides were reviewed by a pathologist to enrich for 70% cancer cellularity. One to 10 10 μM sections of the FFPE tumor samples were macro-dissected to enrich for tumor cellularity and deparaffinized at 90 °C for 5 min. Tumor DNA was isolated using the All Prep DNA/RNA FFPE Kit (QIAGEN) following the manufacturer’s instructions and stored at − 20 °C. DNA concentration was measured using the Qubit HS dsDNA kit (Invitrogen).

Hotspot mutations in the ESR1 LBD region (527aa to 557aa) were identified by deep sequencing of a 129-bp amplicon at a coverage of about X10,000-X30,0000 using an Illumina platform. Ten nanograms of genomic DNA purified from FFPE samples was PCR amplified using the following primer: ESR1 F: AACAAAGGCATGGAGCATCTG, ESR1 R: CTCCACGGATGCCCCTC. The amplified PCR product was separated on and purified from a 2% agarose gel using a QIAGEN gel purification kit. For each sample, DNA concentration was measured by the Qubit HS dsDNA kit (Invitrogen). Library preparation steps were as follows: 50 ng of the PCR product was used for end repair and A-tail using NEB End prep enzyme mix (E7422s/E6090) for 20 min at 25 °C and 20 min at 72 °C. The DNA was purified using AMPure beads (Beckman Coulter), and ligation reaction was performed by T4 DNA quick ligase and index-oligo adaptor for Illumina platform for 15 min at 25 °C. The ligation product was cleaned again with AMPure beads and amplified by Illumina primers and hotstart DNA polymerase (Promega). The final library size was verified by TapeStation (Agilent). Paired-end sequencing (2 × 100 bp) was performed using Illumina HiSeq2500 or MiSeq (reagent micro kit, v2). Mutations were validated by either repeating library preparation and sequencing, or Droplet Digital™ PCR (ddPCR) (Bio-Rad).

Bio-Rad QX100 Droplet Digital PCR (ddPCR) System was used for testing mutation abundance. Validation of the sequencing data was performed with primers and probes designed for D538G and wild type (WT). WT probe: CCCCTCTATGaCCTGCT-HEX, D538G mutant: CTCTATGgCCTGCTGC-FAM 900 nM for ESR1 F primer: TACAGCATGAAGTGCAAG, and ESR1 R primer: TGGGCGTCCAGCA were also used. The local recurrence cohort was examined using Bio-Rad-specific kits for D538G, L536R, and Y537S/N/C mutations. Digital PCR conditions were optimized with a temperature gradient to identify the optimal annealing/extension temperature on a QX200 ddPCR system (Bio-Rad) using TaqMan chemistry. Ten to 50 ng DNA from FFPE samples were used in each reaction, with the addition of 10 μl of ddPCR Supermix for probes (no dUTP) (Bio-Rad) and 250 nM TaqMan WT probe and 250 nM TaqMan D538G probe in a volume of 20 μl for each reaction. The emulsified PCR reaction was partitioned into ~ 14,000 droplets per sample in a QX100 droplet generator according to the manufacturer’s instructions.

Plates were read on a Bio-Rad QX100 droplet reader using QuantaSoft v1.6.6.0320 software (Bio-Rad). At least two negative control wells with no DNA were included in every run and one positive sample for the mutant (using gBLocks from IDT of purified mutated FFPE DNA sample). A sample was considered mutation positive if it contained three or more mutant droplets. The detection limit was 0.1%. The results are reported as a fractional abundance of mutant DNA alleles to total (mutant plus WT) DNA alleles.

ESR1 mutation status was defined as a binary parameter (mutated at any AF vs. WT or AF > 1% vs. AF < 1% that includes WT). Two-tailed p values were considered significant if p < 0.05. The dissimilarity between the percentages of censored cases was tested by chi-square tests. The normal distribution of the clinical parameters was tested for by Kolmogorov-Smirnov test. To determine if there were differences between the groups (ESR1 mutated vs. WT) in terms of the clinical parameters, we used the appropriate tests depending on the covariate type. An independent-samples t test was run for the normally distributed parameter (age). The non-parametric independent-samples Mann-Whitney test was used for clinical parameters that were not normally distributed (duration of endocrine treatment post-recurrence). The Fisher’s exact test was used for categorical covariates (stage, lymph node status, local recurrence type—local or regional). A chi-square test for the association was conducted between the mutation group and clinical parameters. Survival analyses were conducted by Kaplan-Meier survival analyses using the log rank (Mantel-Cox) test. Progression-free survival (PFS) was calculated as the time from AI treatment to any recurrence or death event. Disease-free survival (DFS) was calculated as the time from the tested local recurrence to any recurrence or death event. Distant recurrence-free survival (DRFS) was calculated as the time from either diagnosis or from the tested local recurrence to any distant recurrence or death event. Univariate Cox regression was used to calculate the hazard ratio for each parameter. Data analysis was performed using Statistical Package for the Social Sciences (SPSS), version 25 (IBM SPSS Statistics, Armonk, NY, USA) or with MATLAB® and Statistics Toolbox Release 2016b, The MathWorks, Inc., using the LogRank and KMPLOT functions by G. Cardillo.

In total, we tested 41 FFPE samples from newly diagnosed metastases and 28 samples from advanced metastatic lesions (Fig. 1a) obtained from 62 patients. Prior endocrine treatments at each stage (either adjuvant or metastatic) are noted in the lower bars for each patient (Fig. 1a). For 4 patients, an additional sample from the primary tumor was available and tested. Patients’ characteristics are presented in Table 1.

ESR1 hotspot mutations were found in 10/62 (16%) of the tested patients. Nine of 10 mutations were in D538G, and 1/10 mutation was in L536R. Mutations were detected in liver metastasis (4/10), bone metastasis (3/10), skin (1/10), mediastinum (1/10), and brain (1/10). To the best of our knowledge, ESR1 mutations were not described before in brain metastasis [15]. To validate the sequencing results and allele frequencies, ddPCR analyses were performed on the mutated samples, resulting in a high concordance of allele frequencies (AF) of the mutated clones between both methods (Additional file 1: Table S1).

In line with previous findings [11, 18], ESR1 mutation prevalence was lower among newly diagnosed metastases (5/41; 12%) and higher in advanced disease (5/28; 17%) (not significant by Fisher’s exact test) (Fig. 1b). Notably, 4/5 patients harboring ESR1 mutations in newly diagnosed metastases (Met3, Met5, Met73, Met76) have been treated with TAM only as an adjuvant therapy without AI exposure. Mutated samples at advanced metastatic stages were detected after both TAM and AI treatment (Fig. 1b).

Matched primary tumors were available for four patients (Met5, Met9, Met24, Met34), two from ESR1-WT metastatic patients, and two from ESR1-mutated metastatic patients. The primary tumor of patient Met5 was positive for D538G mutation (AF of 22%, detected by ddPCR). The frequency of ESR1 mutation in the metastatic sample was at AF of 39%, higher than in the primary tumor. Recurrence of metastatic disease for patient Met5 occurred early at 1.5 years post-surgery, while on TAM treatment.

To examine the prognostic significance of ESR1 mutations, we performed a survival analysis and tested the interaction with other clinical parameters. No dissimilarity between the percentages of censored cases between the ESR1 mutant and WT groups was found. No clinical parameters were significantly different between the groups (unpaired t test for age and chi-square test for other categorical parameters). DFS from the primary tumor to metastatic recurrence was similar for both mutant and WT cohorts. There was a trend for worse overall survival from the time of metastatic recurrence in the ESR1 mutant group, which was not statistically different (Additional file 1: Figure S1), possibly due to the small sample size. However, PFS on AI treatment was significantly shorter in the ESR1-mutated patients, four of which were never previously treated with aromatase inhibitors (log rank p value = 0.017; median PFS 3 months for mutation carriers vs. 15 months for no mutation Cox hazard ratio 3.1 [95% confidence interval 1.1–8.3]) (Fig. 1c). These results are in line with previous observations that patients harboring ESR1 mutations demonstrate lower progression survival rates under AI [34].

To test the prevalence of ESR1 mutations, in the context of local recurrence, we assembled a cohort of 41 patients who experienced at least 1 local or loco-regional HR+ breast cancer recurrence (Fig. 2a). Patient characteristics are detailed in Table 2.

All patients were treated in a curative intent for their local recurrence (surgery and if appropriate—radiation therapy). Additional primary tumors were available for 7 patients, and samples from a later metastatic recurrence were available for 4 patients. Two patients had samples from more than 1 loco-regional recurrence. One patient had undergone neoadjuvant endocrine treatment, and samples from pre- and post-treatment were available. Endocrine treatment before the first tested local recurrence biopsy is noted in the lower bars for each patient (Fig. 2a).

ESR1 mutations were examined using hotspot-specific ddPCR assays performed on the FFPE samples. This method is more sensitive than NGS, with lower limits of detection of 0.05% [25, 26]. The detection limit in our analysis was set to 0.1%.

ESR1 mutations were detected at any AF above the detection limit, in 15/41 (36%) patients with the D538G mutation being the most prevalent occurring in 14/15 patients. Y537C was detected in 1 patient while Y537S was detected together with D538G in 3 patients (Fig. 2a). ESR1 mutations at local recurrence were mostly found in the first local recurrence. For 2 patients with ESR1 mutation, only the second local recurrence was available. Interestingly, 1 patient had 4 available local recurrence samples, in which the first 3 were ESR1-WT and only the fourth recurrence harbored ESR1 mutation.

ESR1 mutated with AF greater than 1% were observed only in 4/41 (9.7%) patients. This cutoff was used for the sequencing data in the metastatic cohort. In line with the findings in the newly diagnosed metastatic cohort, mutations in the local recurrence samples were detected under TAM treatment alone, with no previous AI exposure. Mutations in TAM only-treated patients (23 patients) were found at any AF in 9/23 patients and at AF > 1% in 3/23 patients (Fig. 2b). There was no significant difference between patients developing mutations on TAM vs. AI/AI+TAM (Fisher’s exact test p = 0.75).

Matched primary tumors were available for seven patients in this cohort. One patient (LR32) harbored D538G mutation in the primary tumor, which reappeared in two local recurrences after 7 years and 16 years. Interestingly, the mutation AF in the primary tumor was 1.8% and remained at a similar level of AF (around 1% or less) in both local recurrences. This patient did not develop a metastatic disease, and her disease is stable for over 20 years.

To explore the prognostic and clinical implications of ESR1 mutations in loco-regional recurrences, we performed survival analyses and tested the interaction with other clinical parameters. First, survival events were compared between patients with a mutation at any AF and without a mutation. In this case, the differences in survival distributions between the groups were not statistically significant (log rank test, p value > 0.05) (Additional file 1: Figure S2). There was no statistically significant difference (p value > 0.05) between the mutation groups and other clinical parameters (age at diagnosis, type of endocrine treatment, duration of endocrine treatment post-local recurrence, stage, LN, HER2 status, and type of local recurrence—local or regional). The differences between the groups were not significant also when adjusting for age and for the duration of AI post-local recurrence.

We speculated that ESR1 mutation at low AF may not have a significant survival effect and therefore decided to test the effect on survival by setting a cutoff of 1% AF, similar to the cutoff used in the metastatic cohort. Kaplan-Meier survival analysis was conducted to compare the effect on survival between patients with mutations at first tested local recurrence with AF > 1% (n = 4 patients) vs. AF < 1% or no mutation (n = 37 patients). DFS from the first tested local recurrence was significantly shorter in the > 1% mutant-positive ESR1 patients (log rank p value = 0.041, median time 0.63 [0–1.96] vs. 4.79 [2.59–6.77]) (Fig. 3a). Additionally, DRFS from tested local recurrence (Fig. 3b) as well as from primary tumor (Fig. 3c) was also significantly shorter in the AF > 1% ESR1 mutant-positive patients (log rank p value = 0.017 and 0.011, respectively). A univariate Cox regression analysis that included other relevant clinicopathological features showed that besides ESR1 mutations (AF > 1%), AI treatment post-local recurrence is significantly correlated with longer DFS and DRFS from local recurrence and from primary tumor diagnosis, respectively (Fig. 3d–f), while other parameters such as lymph node involvement at initial diagnosis trended towards worse outcomes but were not significant. Due to the small cohort size, we were unable to perform a multivariate analysis.