Analysis of ESR 1 and PIK 3 CA mutations in plasma cell-free DNA from ER-positive breast cancer patients

Background: The measurement of ESR1 and PIK3CA mutations in plasma cellfree DNA (cfDNA) has been studied as a non-invasive method to quickly assess and monitor endocrine therapy (ET) resistant metastatic breast cancer (MBC) patients. Methods: The subjects of this retrospective study were a total of 185 plasma samples from 86 estrogen receptor-positive BC patients, of which 151 plasma samples were from 69 MBC patients and 34 plasma samples were from 17 primary BC (PBC) patients. We developed multiplex droplet digital PCR assays to verify the clinical significance of ESR1 and PIK3CA mutations both in a snapshot and serially in these patients. Results: cfDNA ESR1 and PIK3CA mutations were found in 28.9% and 24.6 % of MBC patients, respectively. The relation between ESR1 or PIK3CA mutations and clinical features showed that ESR1 mutations occurred mostly in patients previously treated by ET, which was not the case for PIK3CA mutations. The analysis of the clinical impact of those mutations on subsequent lines of treatment for the 69 MBC patients revealed that both ESR1 and PIK3CA mutations detection were related to a shorter duration of ET effectiveness in univariate analysis but only for ESR1 mutations in multivariate analysis. The monitoring of cfDNA in a subset of 52 patients showed that loss of ESR1 mutations was related to a longer duration of response, which was not the case for PIK3CA mutations. Conclusions: We have demonstrated the clinical significance of on-treatment ESR1 mutations both in a snapshot and serially in comparison with PIK3CA mutations.


INTRODUCTION
Approximately 80% of breast cancers (BCs) express the estrogen receptor alpha (ERα), encoded by the ESR1 gene, and endocrine therapy (ET) with selective ER modulators (SERMs) or aromatase inhibitors (AIs) is the mainstay of treatment for this group of patients because of their effectiveness balanced against their side effects. However, ET resistance occasionally occurs during the treatment of early BC and inevitably results in metastatic BC (MBC) [1]. ESR1 ligand binding domain (LBD) mutations constitutively activate the ER in a ligandindependent fashion [2][3][4] and they have attracted attention as a mechanism of ET resistance in MBC. These mutations were originally reported almost two decades ago [5][6][7][8], and recent large-scale next-generation sequencing (NGS) revealed that ESR1 mutations are present in approximately 20-50% of metastatic tissue samples treated

Research Paper
with endocrine agents while these variants are absent or only present at very low frequencies in primary tumor samples [2][3][4]. These features indicate that the presence of ESR1 mutations should be assessed in metastatic lesions. Circulating cell-free DNA (cfDNA) has been proposed to carry a comprehensive picture of metastatic tumor cells and genomic analysis of plasma cfDNA has been realized as a non-invasive method to quickly assess the mutational profiles and monitor molecular changes under treatment, using recent developments in digital genomic technologies [9]. Therefore, if ESR1 mutation status in cfDNA is predictive of response to ET, monitoring of this marker could be a useful method of informing treatment plans for subsequent metastatic disease.
PIK3CA is an oncogene that encodes the p110α component of phosphatidylinositol 3-kinase (PI3K) and PIK3CA is a representative frequently-mutated gene, whose frequencies are 20% to 40% of all BCs [10,11]. Recently, in phase III randomized trials, the clinical significance of ESR1 mutations have been reported in the comparison with PIK3CA mutations. In alteration frequency in metastatic versus primary tumors in the BOLERO-2 cohort, Hortobagyi et al. demonstrated that PIK3CA mutations had the highest frequency in PBCs and MBCs and that ESR1 mutations had higher frequency in MBCs than in PBCs [12]. More recently, in the BOLERO-2 study, Chandarlapaty and colleagues found that 28.8% (155/541) of ER-positive MBC patients had ESR1 mutations in plasma cfDNA [13] and 43.3% (238/550) of ER-positive MBC patients had PIK3CA mutations in plasma cfDNA [14]. They also demonstrated that the difference of clinical features between ESR1 and PIK3CA mutations, namely, progression free survival (PFS) benefit of mammalian target of rapamycin (mTOR) inhibitor everolimus was maintained irrespective of PIK3CA mutations, but that was decreased according to the presence of ESR1 mutations [13,14]. In another two phase III randomized trials, Fribbens and colleagues assessed ESR1 mutations in cfDNA using digital PCR (dPCR) [15]. ESR1 mutations were found in the plasma of 39.1% of patients (63/161) in the SoFEA study and 25.3% (91/360) in the PALOMA3 study. PIK3CA mutations were found in the plasma of 33% (129/395) of patients in the PALOMA3 study [16]. They also reported the effectiveness of the target drug by having the mutations or not. In the SoFEA study, patients with ESR1 mutations had improved PFS after taking fulvestrant compared with exemestane. In the PALOMA3 study, fulvestrant plus the CDK4/6-inhibitor palbociclib improved PFS regardless of the genomic status of ESR1 or PIK3CA [15,16].
In this retrospective study, we demonstrated the clinical significance of on-treatment hotspot ESR1 LBD mutations both in a snapshot and serially in 185 plasma samples from 86 patients in comparison with the hotspot mutation status of PIK3CA using multiplex droplet dPCR (ddPCR) assays. To our knowledge, this is the leading comparative study to identify the clinical significance of multiplex ddPCR detection of ESR1 mutations and PIK3CA mutations in plasma samples.

ESR1 mutations in cfDNA baseline plasma samples
We developed a sensitive and quantitative multiplex ddPCR assay to screen for 3 hotspot mutations in the LBD of ESR1. Figure 1A and Supplementary Figure S1 show the comparative analysis of a dilution series of each indicated synthetic ESR1 mutation oligonucleotide by ddPCR. We used serial dilutions of three hotspot ESR1 LBD mutant recombinant DNAs: ESR1 Y537S, Y537N, and D538G, and analyzed them using a multiplex ESR1 mutant detection probe, which could simultaneously detect ESR1 Y537S, Y537N, and D538G, confirming that this assay was able to detect as few as three copies of the mutant allele in an abundance of wild-type DNA. There was no cross-reactivity for the detection of each ESR1 mutations.
Next, to validate the utility of the multiplex ddPCR assay, a subset of 26 women (62 blood samples), who were previously evaluated using a uniplex ESR1 mutant detection probe [17], were analyzed using a multiplex ESR1 mutant detection probe, and a statistically significant correlation between uniplex and multiplex ddPCR assays was found (Figure 1B, Supplementary Table S1: Y537S, κ = 1.0, r = 0.91, P < 0.0001; Y537N, κ = 1.0, r = 0.55, P < 0.0001; D538G, κ = 0.77, r = 0.71, P < 0.0001) 23.2 % (16/69) had not previously received any ET. The median duration of follow-up was 33 months (range, 13-101 months) in the PBC group and 49 months (range, 11-268 months) in the MBC group. There was no recurrence during the observation period in any of the PBC patients. Concerning the multiplex assay, it should be tested using negative control to help defining the background noise [18]. Therefore, we verified the multiplex ddPCR assay using PBC patients whose allele frequency (AF) of ESR1 mutations is very low [2][3][4]. We found 4 PIK3CA mutations in cfDNA, but we did not find any ESR1 mutations in the PBC group (Supplementary Table S3). Figure 2 shows the percentage of ESR1 mutations and PIK3CA mutations in plasma cfDNA of BC subsets. ESR1 mutations and PIK3CA mutations were evaluated using multiplex mutant detection probes. ESR1 mutations in cfDNA were detected in 30 Table S2). PIK3CA H1047L/R/Y, E545V/G/A/Q/K Q546L/R/P/E/K, E542K/V, and G1049R/S were found in 65.9%, 22.7%, 20.5% and 6.8% of PIK3CA mutant samples, respectively. The majority of cfDNA samples (88.7 %, 39/44) had only a single-site detectable PIK3CA mutation in cfDNA and exhibited markedly less heterogeneity than ESR1 mutations (Figure 2A, Supplementary Table S2). Nine patients had co-mutation (ESR1 and PIK3CA mutations) over the all of treatment. However, there was no statistically significant correlation of copies/μL between ESR1 mutations and PIK3CA mutations in plasma cfDNA (r = 0.28, P = 0.0001) ( Figure 2B, Supplementary Table  S2). Table 2 shows ESR1 and PIK3CA mutations as they relate to baseline clinical and pathological features. In the snapshot study, we analyzed the latest genetic state   Table S3).

Association of ESR1 and PIK3CA mutations with clinical outcome
We retrospectively analyzed whether ESR1 and PIK3CA mutations detected in cfDNA were associated with differential benefit in relation to the duration of ET effectiveness (Figure 3, Supplementary Figure S2). In that analysis, discontinuation of ETs caused by local recurrences, distant metastases, and disease progression at any site following the blood draw were considered as an event. These were tested by Kaplan-Meier analysis and verified by the log-rank test. Patients with detectable plasma ESR1 mutations (P < 0.0001) and PIK3CA mutations (P = 0.0034) showed statistically significant shorter duration of ET effectiveness ( Figure  3A, 3B). The Cox hazards model analysis of duration of ET effectiveness is shown in Table 3. The presence of ESR1 mutations in cfDNA was a significant prognostic parameter in univariate analysis (hazard ratio (HR): 3.2, 95% confidence interval (CI): 1.76-5.71, P = 0.0002) and      Table S4). Patients were grouped into those who had no ESR1 or PIK3CA mutations over the course of treatment (N = 28), those in whom ESR1 or PIK3CA mutations were acquired or maintained (N = 15), and those in whom ESR1 or PIK3CA mutations were disappeared after treatment in cfDNA (N = 9), and groups were compared by the patient response end-points of duration of ET effectiveness ( Figure 3C, 3D). Patients in the loss of cfDNA ESR1 mutations group had a longer duration of ET effectiveness than patients in the acquired or maintained numbers of cfDNA ESR1 mutations group, but had a shorter duration of ET effectiveness than patients without mutations over the course of treatment (P < 0.0001). On the other hand, there was no statistically significant differences in these three groups; no PIK3CA mutations during treatment group (N = 31), the loss of cfDNA PIK3CA mutations group (N = 3), and the acquired or maintained numbers of cfDNA PIK3CA mutations group (N = 18) (P = 0.10). We did not detect any trend towards particular ESR1 and PIK3CA mutation hotspots over the course of treatment, but indicated that polyclonal mutations appeared or disappeared more frequently in ESR1 than in PIK3CA (Supplementary Table S4; appeared polyclonal mutations, 40% for ESR1 vs 5.6% for PIK3CA: disappeared polyclonal mutations, 66.7% for ESR1 vs 0% for PIK3CA). Intrapatient changes of the number of copies/μL of ESR1 and PIK3CA mutations in each representative therapeutic drug group in 21 patients with longitudinal data are shown in Figure 4. Furthers, we listed the treatment just before the latest blood draw as "representative treatment between blood draw" in Table 4. All plasma samples were taken at the time of disease progression, so that it provided an evaluation of ESR1 and PIK3CA mutation status when each ET failed. In patients treated with AIs, 26.7% (4/15) had acquired ESR1 mutations, which was more frequent compared to the 13.3% (2/15) of patients lost ESR1 mutations. The two patients acquiring ESR1 mutations (Pt 33 and Pt 82) developed PIK3CA mutations over the course of treatment. In patients treated with selective ER down regulators (SERDs), 62.5% (5/8) showed acquired or maintained numbers of ESR1 mutations, but 25% (2/8) had decreases in the number of ESR1 mutations. Of the SERD-treated patients acquiring or maintaining ESR1 mutations, 60% (3/5) showed increases in PIK3CA mutations. In patients treated with SERMs, 50% (2/4) had acquired ESR1 mutations. In patients treated with ethinyl estradiol (EE2), 40% (2/5) of EE2-treated patients showed decreases in ESR1 mutations. Interestingly, among ETs, decreases in PIK3CA mutations were detected in one

DISCUSSION
dPCR is a highly sensitive technique for detecting rare mutations, but analysis is on the basis of a single mutation per assay. For screening multiple mutations in a limited amount of sample, an assay to detect multiple mutations in parallel has been demonstrated with dPCR [15,19,20]. Here, we used multiplex ddPCR to study a cohort of patients treated with multiple lines of hormonal therapy to determine whether on-treatment ESR1 and PIK3CA mutations can be detected noninvasively and to examine the potential clinical importance of these mutations both in a snapshot and serially.
We verified that up to three mutations in ESR1 and up to 12 and five mutations in PIK3CA could be multiplexed into a single assay, which could detect the three most common mutations in ESR1 (Y537S, Y537N, and D538G) and 17 mutations in the 5 most common mutation sites of PIK3CA (E542K/V, E545V/G/A/Q/K, Q546L/R/P/E/K, H1047L/R/Y, G1049R/S) (Figure 1, Supplementary Figure S4). Toy and colleagues revealed the spectrum of ESR1 mutations from more than 900 patients and their potential differential impact (e.g. tumors driven by Y537S, but not D538G, E380Q, or S463P, were less effectively inhibited by fulvestrant in comparison with more potent and bioavailable antagonists, including AZD9496 [21]. Therefore, we examined the less frequent ESR1 mutations, E380Q (LBx ® Probe ESR1 E380Q (A081), Riken Genesis, Tokyo, Japan) and Y537C (probe shown in [22]), but we could not detect them in our cohort. In validation assay, some of the mutations are found in the multiplex assay, but not using the uniplex assay (Y537N: 5 detection by multiplex, only 2 by uniplex) (Supplementary  Table S1). Because the cut-off level of the presence of the mutation was 3 positive droplets in both multiplex and uniplex assay, this study had the possibility that uniplex assay could not detect a mutation due to less than the cut-off level, which could be detected as one of target mutations using multiplex probe.
The subjects of this retrospective study were a total of 185 plasma samples from 86 ER-positive patients, of which 151 plasma samples were from 69 MBC patients and 34 plasma samples were from 17 advanced PBC patients. Plasma ESR1 mutations were found in 28.9% (20/69) of MBC patients while plasma PIK3CA mutations were found in 24.6% (17/69) of MBC patients, with a distribution of mutations that was highly similar to previously published data [2][3][4]19]. Interestingly, only 9 patients had both ESR1 and PIK3CA mutations over the all of treatment. These results were compatible with the recent report that PIK3CA was mutated in 37% (53/143) and ESR1 was mutated in 14% (20/143) of the ER-positive /HER2-negative MBC, in which only 9 had both mutations [23].
All patients with ESR1 mutations had resistance to prior AI (P = 0.0074), and the majority of patients with ESR1 mutations had resistance to prior SERM therapy (P = 0.0087), prior both AI and SERM therapy (P = 0.0014), and had received three or more prior endocrine regimens (P = 0.041) ( Table 2). Detailed information of three or more prior endocrine regimens was shown in Supplementary Table S5. On the other hand, PIK3CA mutations in cfDNA were not associated with previous endocrine exposure. Among all samples, 67.4% had polyclonal mutations in ESR1, which exhibited markedly more heterogeneity than PIK3CA mutations, 11.3% of which were polyclonal (Figure 2A). These findings are compatible with the report that polyclonal ESR1 mutations were present in 19.2-49.1% of ESR1 mutant patients [13,15,19,24], but PIK3CA mutations were often monoclonal [15,25].
In this study, we focused on "duration" of ET effectiveness. Because endocrine treatments can be administered repeatedly and consistently in ER-positive MBC patients without life-threating visceral metastases [26], the efficacy of endocrine treatments may contribute to the "duration" of ET effectiveness even if the endocrine treatments during the period varied. Furthers, this analysis had the problem that the time points and the number of samples analyzed were different among the patients. However, we overwhelmed it by comparing ESR1 mutations with PIK3CA mutations because the clinical role of PIK3CA mutations becomes clear as follow; PIK3CA mutations have the highest frequency in primary and metastatic breast tumors [12] and they are not statistically significant prognostic marker or predictor of ET effectiveness [14,16]. Patients with detectable plasma ESR1 mutations (P < 0.0001) and PIK3CA mutations (P = 0.0034) showed significantly shorter duration of ET effectiveness by log-rank test (Figure 3, Supplementary Figure S2). In the Cox hazards model, the presence of ESR1 mutations in cfDNA was a significant prognostic parameter in both univariate analysis (HR: 3.2, 95% CI: 1.76-5.71, P = 0.0002) and in multivariate analysis (HR: 2.04, 95% CI: 1.08-3.83, P = 0.029). However, the presence of PIK3CA mutations in cfDNA was a significant prognostic parameter in univariate analysis only (Table 3).
Potential interest of monitoring ESR1 mutations in the metastatic setting has been increasing. Recently, Clatot and colleagues reported that cfDNA ESR1 mutations are independent risk factors for poor outcome after AI failure, and they are frequently detectable before clinical progression [27]. On the other hand, Spoerkle and colleagues did not show clinical utility of ESR1 mutations as a monitoring tool [24]. In our tracking cfDNA ESR1 mutations and PIK3CA mutations study, patients in whom the number of cfDNA ESR1 mutations were lost had a longer duration of ET effectiveness than patients in whom the numbers of cfDNA ESR1 mutations were acquired or maintained, but had a shorter duration of ET effectiveness than patients without mutations over the course of treatment (P < 0.0001). On the other hand, there was no statistically significant differences in these three groups; no PIK3CA mutations during treatment group (N = 31), the loss of cfDNA PIK3CA mutations group (N = 3), and the acquired or maintained numbers of cfDNA PIK3CA mutations group (N = 18) (P = 0.10). These differences regarding the prevalence of ESR1 and PIK3CA mutations may be caused by polyclonal breast tumor evolution under the selective pressure of ET [24]. ESR1 mutations occur late in endocrine treatment and in a subclonal manner, so that these mutations are generally detected in metastatic lesions [2][3][4]22]. In contrast, most PIK3CA mutations occur early in the process of tumor development and its status does not change in the majority of patients who develop recurrent or progressive breast cancer [28] ( Figure  4).
The present study has limitations. This was a retrospective, single-institute study, and was prone to selection bias. The studied population was heterogeneously treated and all plasma samples were taken at the time of disease progression, so that we had insufficient data to examine whether or not ESR1 mutation detection is dependent on specific hormone therapies. The samples used in this study were obtained for biobanking. Therefore, the time from blood draw to spinning, freezing plasma and then thawing may affect the variability of the data.
In conclusion, our study demonstrates the clinical significance of the burden of on-treatment hotspot ESR1 LBD mutations, both in a snapshot and serially in MBC patients in comparison with PIK3CA hotspot mutation status, using multiplex ddPCR assays.

Patients and breast cancer samples
A total of 86 patients (185 plasma samples) with breast carcinoma, treated at Kumamoto University Hospital between 2003 and 2016, were enrolled in this study. Cases were selected if archival plasma samples were available. Informed consent was obtained from all patients before biopsy or surgery. The Ethics Committee of Kumamoto University Graduate School of Medicine (Kumamoto, Japan) approved the study protocol. Adjuvant and neoadjuvant treatment was administered in accordance with the recommendations of the St. Gallen international expert consensus on the primary therapy of early BC [29][30][31]. The treatment of MBC patients was performed in accordance with the National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology [26]. Recurrence was defined as the identification of positive spots by physical examination and/or by imaging diagnosis during the follow-up period. Patients were examined at the Kumamoto University Hospital or affiliated hospitals every 3 months for 5 years and every year thereafter and they were assessed monthly or at longer intervals depending on their disease status.

Sample preparation
Blood collected in EDTA K 2 tubes was processed as soon as possible and was centrifuged at 1,467 g for 10 min, with plasma stored frozen until DNA extraction. DNA was extracted from 500 μL aliquots of plasma using the ISOSPIN Blood & Plasma DNA kit (Nippon Gene, Tokyo, Japan) according to the manufacturer's instructions. All DNA extracts were quantified using a NanoDrop 2000 spectrometer (NanoDrop Technologies, Wilmington, DE, USA) and purity was determined from the A260/A280 absorbance ratios.

Analysis of ESR1 mutations by ddPCR
We performed duplicate ddPCR assay on a QX200 digital PCR system (Bio-Rad laboratories, Hercules, CA, USA) using the assays as described previously [17,32]. The PCR data were quantified as copies/μL and fractional abundance (allele frequency) using QuantaSoft ™ software (Bio-Rad laboratories). A mutation was considered positive with more than three ESR1 mutant or PIK3CA droplets. The uniplex ddPCR method had been optimized beforehand by comparative analysis of a dilution series of synthetic copies of each indicated mutant ESR1 oligonucleotide, as reported previously [17,22].

Immunohistochemistry
Immunohistochemical staining was carried out on 4-μm thick tumor sections. Serial sections were prepared from selected blocks and float-mounted on adhesivecoated glass slides for ERα, PgR, HER2, and Ki67