Detection of Activating Estrogen Receptor Gene (ESR1) Mutations in Single Circulating Tumor Cells

This article is featured in Highlights of This Issue, p. 5997

Breast cancer is the most frequent cancer and the leading cause of cancer-related death among women. Despite advances in prevention, diagnosis, and adjuvant treatment, about 30% of breast cancer patients develop metastatic disease (1). Recent advances suggest that the presence of different tumor cell clones plays an important role in metastatic progression and resistance to chemotherapy (2). According to the clonal theory of tumor evolution, cancer is an evolving process (3), and the selective pressure exerted by multiple lines of treatment may lead to selection of much more aggressive subclone populations or even those with an acquired drug resistance (4).

About 75% of breast cancers express the estrogen receptor (ER); and, acting on this signaling pathway is a key treatment strategy. The main endocrine therapeutic approaches are: (i) selective ER modulators (SERM); (ii) inhibitor of aromatase (AI); and (iii) selective ER downregulators (SERD; ref. 5). However, in 20% to 25% of metastatic breast cancer (MBC) patients, endocrine therapy failure has been reported after several lines of treatment, and new targeted therapies have been approved to be combined with hormone therapy (6–8). Several molecular mechanisms of resistance may be involved, including downregulation and posttranslational modification of the ER encoded by the ESR1 gene (9). Because ESR1 mutations are rare, occurring in only 1% of primary breast cancer, their ability to confer endocrine resistance has been speculated for many years (10). However, in metastatic tissues the incidence of such mutations is estimated at 20% (11). In the last several years, 14 ESR1 point mutations have been reported, mainly localized in the ligand-binding domain (LBD) including 3 hotspot mutations in codons 380, 537, and 538. Functional study of these mutations showed an ER ligand-independent activity, highlighting their role in acquired endocrine resistance (12, 13). Therefore, genomic characterization of distant metastasis may provide clinically useful information for the selection of specific therapeutic treatments (14, 15). Even though genetic testing on repeated metastatic biopsies may not be representative of the whole tumor mass and leads to an underestimation of tumor heterogeneity (16).

Liquid biopsy using either circulating tumor cells (CTC) or cell-free tumor DNA (ctDNA) has become one of the most sensitive approaches to monitor tumor molecular evolution (17, 18). CTCs can be isolated noninvasively over time (19) and over the past decade, the prognostic value of CTCs has been shown in metastatic breast, colorectal, and prostate cancer (20–26). Significant advances in cancer diagnosis and in the evaluation of disease progression and treatment can be reached with single CTC analysis because of the improvements made in single-cell genomics analysis (27). Currently, there are few techniques available for single cell isolation, including micromanipulation, laser microdissection, and high-throughput fluorescence-activated cell sorting (FACS). These approaches have several disadvantages, including inadequate detection sensitivity for CTC population, a required high number of cells as a starting population and are manual and laborious methods (28–30).

We decided to study the molecular features of CTCs in patients with hormone-receptor–positive (HR⁺) MBC receiving endocrine therapy. We planned to validate a laboratory workflow for single CTC detection, isolation and molecular analysis by combining the sequential use of CellSearch (20–26), DEPArray systems (31, 32) and MALBAC techniques (33). The purpose of this study was to investigate the incidence and heterogeneity of ER expression and to evaluate the detection of ESR1 mutations in individual CTCs. We also planned to compare our single CTC data with matched cfDNA.

Thirty MBC patients were enrolled at the Department of Medical Oncology, Thomas Jefferson University in Philadelphia, between February and September 2015. Only patients with a primary ER-positive MBC were included. Clinical parameters included sex, age at surgery, differentiation grade, lymph node metastasis, distant metastasis, TNM stage, and histology. Progesterone receptor (PR) and ER status of the primary tumor and of available metastases were recorded. All subjects gave informed consent, and the study was approved by the Institutional Review Board. For each patient, 10 mL of blood was collected in two CellSave tubes (Veridex, LLC) for enrichment, enumeration, and molecular characterization of CTCs using the FDA-approved CellSearch System. All samples were taken at least 5 days after the last treatment. Matching primary tumor tissues were tested for the presence of mutations in ER receptor before starting hormonal therapy.

Two human breast cancer cell lines (MCF-7 and FC-IBC-02) and a prostate cancer cell line (C4-2) were used to validate whole genome amplification experiments. MCF-7 cells were maintained in DMEM containing 10% (v/v) FBS, 100 units/mL penicillin and 100 mg/mL streptomycin. FC-IBC-02 primary cells were isolated from pleural effusion of IBC patients and cultured in Ham's F12 with 10% (v/v) FBS, 5 mL insulin and 100 μg/μL of hydrocortisone with antibiotic antimicotic. The C4-2 cell line was cultured in RPMI with 2.5% to 10% (v/v) FBS as previously described (34). All cell lines were maintained in T-25 or T-75 flasks using prescribed cell culture conditions [5% (v/v) CO₂, 37°C].

Single tumor cells were obtained from MCF-7 and FC-IBC-02 cell lines through micromanipulation and from C4-2 cell line by serial dilution. All collected cells were processed for MALBAC. Meanwhile, 30 pg of genomic DNA carrying the V600E mutation in the BRAF gene was used as MALBAC-positive control for all WGA products. To validate the CellSearch ability to capture CTC, 100 C4-2 cells were spiked into a healthy donor blood sample. Captured C4-2 cells together with WBC from the healthy donor were loaded on the DepArray cartridge to achieve single cell isolation and capture.

DEPArray system is a semiautomated system that allows the isolation of rare fluorescently labeled cells. An electric field is generated on the surface of a silicon chip directly interfaced to a microfluidic chamber containing the cell suspension and an array of electrodes. Each electrode can be programmed to achieve a cage of dielectrophoresis, inside of which single CTC can be trapped and then analyzed individually.

Individual C4-2 and WBC cells were subsequently MALBAC amplified and screened for 7 known mutations in the AR, CDH1, PIK3C3, NCOR2, ERBB2, CDK4, and ETV1 genes by Sanger sequencing. C4-2 and WBC genomic DNA from a healthy donor also were used to confirm the 7 variants by Sanger sequencing.

Standard CellSearch protocol for CTCs enrichment and enumeration was used according to the manufacturer's instructions. Briefly, CTCs were enriched on the CellTracks Autoprep using ferrofluid conjugated with EpCAM antibody. Cells were stained with fluorescently labeled monoclonal antibody for cytokeratin CK8-, CK18-, CK19-FITC as well as for leukocyte common antigen CD45-APC and nuclear-stained with DAPI. Moreover, ER expression on MCF-7 cells and CTCs was assessed by staining the cells with a PE-conjugated anti-ER nuclear antibody.

Because the DEPArray system (Silicon Biosystem) provides the analysis of only 66% of the loaded volume, to optimize single CTC recovery rate, only patients exhibiting >20 CTCs (ER⁺ and ER⁻) were processed.

Briefly, for each CellSearch enriched sample, 13 μL were loaded with 325 μL manipulation buffer (SB115, Silicon Biosystem) into an A300K cartridge. Approximately 8.6 μL of the sample is de facto dielectrophoretically processed in which cells are individually trapped in cage. The cartridge is then scanned by an automated fluorescence microscope and cells detected by DAPI staining. Three different populations of cells were isolated: (i) ER⁺ CTCs, defined as ER-positive, CK-8, CK-18, CK-19–positive, CD45-negative; (ii) ER⁻ CTCs, defined as ER-negative, CK-positive, CD45-negative; and (iii) WBCs, defined as CD45-positive, ER⁻, and CK-negative. Each cell was collected individually, washed two times in PBS and stored at −80°C or immediately lysed in accordance with MALBAC protocol (35). To minimize DNA contamination in the same isolation cage containing the individual cell, an aliquot of the elution buffer from the single cage was MALBAC-amplified and subjected to DNA Sanger sequencing. No mutation was detected on all elution buffer reactions. Moreover, for each sequencing run, a no template control was tested.

Cell lysis and genome amplification was performed using the MALBAC kit (Yikon Genomics YK001A/B version 1302.1; ref. 33), following the manufacturer's instructions. A negative no template control (NTC), a blank control (SB115), and a MALBAC-positive control were used for each MALBAC reaction. WGA products were then purified according to the Agencourt AMPure XP bead kit (Beckman Coulter) manufacturer's protocol (36) and QC using Qubit dsDNA High Sensitivity Assay kit (ThermoFisher). WGA products were run on 0.8% (v/v) agarose gel and checked for expected distribution in size (300–2,000 bp).

Sanger sequencing was performed to genotype all WGA products as well as 14 hotspot mutations in the ESR1 gene found in MBC tissues and related controls on an AB 3730 following the manufacturer's protocol. Detailed PCR conditions (Tm) and primer sequences are available in Supplementary Methods (Table 1). Sequences were analyzed and genotyped by SeqScape v3.0 analysis software (ThermoFisher).

To investigate whether detection of ESR1 mutations in individual CTCs in MBC patients could be used as a tool to enable monitoring of the metastatic burden for clinical decision-making, a 4-step protocol was implemented with the following workflow: (i) CTC enrichment, (ii) single cell isolation, (iii) whole genome amplification, and (iv) Sanger sequencing (Fig. 1).

A cohort of 30 MBC patients was characterized by a median age of 56 years. Patients (36.6%) showed evidence of one single metastatic lesion, while 64% showed more than one at the time of the first draw. Clinical and pathological features of primary and metastatic tumor tissues are summarized in Table 1. Among the total number of patients, 28 had histologically confirmed ER positivity even at metastatic sites. At the time of surgery, primary tumor tissues were investigated for presence of ER mutations and all harbored a wild-type genotype.

Enrichment and enumeration of CTCs performed on CellSearch involved a total of 50 blood samples taken from the 30 patients enrolled. The number of CTCs based on ER expression (Fig. 2) for each patient (ID) are shown in Table 2. Overall, the average of total CTCs enumerated was 80, with a maximum of 1,375 cells. Samples (22%) were negative (no CTCs) for the presence of CTCs, defining a group of patients currently responsive to current treatment. The remaining 39 samples were divided into two groups depending on the established cutoff of 5 CTCs (37) used to identify patients with high risk of disease progression (12 samples <5 CTCs vs. 27 samples ≥5 CTCs). Only 4 of the 50 samples analyzed showed a number of CTCs greater than or equal to 20 and were processed on the DEPArray. Seventy-one single CTCs and 12 white blood cells (WBC) were retrieved. Forty of these CTCs and all the WBCs were subjected to WGA. The number of CTCs (subdivided in ER⁺ or ER⁻) isolated for each individual patient and the corresponding number of selected cells for WGA are summarized in Table 3. Samples processed on DEPArray showed between 21% and 30% of CTCs recovered.

Validation of single-cell genome amplification was conducted on 30 individual single cells. DNA-positive controls, after MALBAC amplification showed heterozygosity for the BRAF V600E (c.1860T>a) mutation (data not shown), as expected. No other BRAF mutations were found in all the wild-type single cells analyzed, demonstrating feasibility of the protocol.

Following spiking, enrichment and immune labeling, 91 positive CTCs were detected by CellSearch. Sixty-five percent of these CTCs were identified on the DEPArray and finally 10 individual cells were recovered and subjected to whole genome amplification. In addition, 10 WBCs were recovered and subjected to WGA, as negative control. The sequences obtained from all the C4-2 cells (single or pooled) revealed all 7 carried known mutations (AR/T878A, CDH1/P94T, CDK4/P110L, ErbB2/E930D, ETV1/G207E, NCOR2/L167P, PIK3C3/F524C; data not shown). Sequences obtained from the 10 WBCs from healthy donor's buffy coat showed wild-type genotypes for all variants tested.

ESR1 mutation analysis was successfully performed on all single cells isolated. All mutations were located within the ligand-binding domain of the ESR1 gene in exons 4, 5, 6, 7, and 8 (Fig. 3). All 12 WBCs showed a wild-type genotype confirming the absence of mutations in the germline. Overall, we found ESR1 mutations in a total of 8 CTCs belonging to 2 MBC patients. High levels of intra- and intertumor genetic heterogeneity in ER-positive CTCs populations was also revealed. The remaining 32 CTCs analyzed showed a wild-type ESR1 genotype. For patient ID20, a total of 5 single CTCs were analyzed and all showed wild-type phenotype. Patient ID19 exhibited a heterogeneous ESR1 genotype in their CTC populations. Among the ER⁺ CTC population, we detected 3 different genotypes: (i) one single wild-type CTC; (ii) 3 CTCs heterozygous for a single mutation (Y537S) in exon 8 (Fig. 3a); (iii) one single CTC homozygous for the same Y537S (LOH; Fig. 3b). Matching cfDNA was tested, confirming the Y537S mutation at 0.25% allele frequency. Patient ID10 was the only one who had 2 serial blood samples taken, 3 months apart. The first sample showed an ESR1 wild-type genotype in all the 12 CTCs recovered. The second showed a wild-type genotype in all 8 ER⁻ CTCs, and in 4 of the ER⁺ CTCs recovered. Three CTCs were heterozygous for the Y537S mutation, whereas the remaining ER⁺ CTC harbored 2 different mutations in exon 8. Other than the Y537S we found a new mutation, not reported until now, the T570I (Fig. 3c). Also, in this case, data were compared with those obtained on matching cfDNA. No mutation was detected at the first draw, while Y537S was detected at the second sampling at a percentage of 0.18%.

Patient ID19 with the Y537S mutation had only one sampling for CTC enumeration and circulating free tumor DNA (cfDNA) analysis. The first diagnosis of inflammatory ductal breast cancer was made in 2010, and the patient opted for holistic remedies. In 2011, the patient developed ascites and pleural effusions and started a chemotherapeutic treatment (docetaxel–cytoxan) for 6 months, followed by endocrine therapy with aromasin. In 2014, the patient was subjected to 6 cycles of Doxil and Faslodex, but in January 2015, a liver metastasis was found. This progression to metastatic disease during treatment indicated a failure in the therapeutic approach. The analysis of CTCs and cfDNA confirmed this suspicion, with the finding of the Y537S-activating mutation.

Patient ID10 reported two mutations, the Y537S and the newly reported T570I. This case report shows how the monitoring of ESR1 mutations can be crucial to monitor and predict disease evolution. First diagnosis was made in 2011, which was followed by a mastectomy. In 2012, she was irradiated and subsequently treated with tamoxifen. Due to poor tolerability of the drug and to the occurrence of bone metastasis, the patient switched endocrine therapy, examestane followed by fulvestrant. At the time of the first draw, she was negative for ESR1 mutations, but a high increase in the number of CTCs was found compared with baseline CTCs count (80 vs. 54). The second sampling was taken only 1 month apart and was positive for the Y537S-activating mutation. Furthermore, a third CTC count was made after a further therapy switch to combination with palbociclib, showing a relative decrease in total number of CTCs (43 vs. 80), but an almost unchanged number of the ER⁺ population (23 vs. 28). The prospective clinical evidence clearly indicate that the patient did not benefit from fulvestrant at time of ESR1 mutation detection. Instead, she benefited from the prompt switch to a combined therapy of palbociclib and fulvestrant, a highly effective regimen evaluated in the prospective, randomized, phase III study PALOMA-3, whose benefit appears irrespective of common genomic abnormalities such as ESR1 and PI3KCA (38, 39).

The frequency of ESR1 mutations in breast cancer is a matter of intense debate for the potential clinical utility of this information. Primary ESR1 mutations are relatively rare in primary tissue, up to 7% of specimens analyzed with very low allele frequencies (0.07%–0.2%), when compared with a much higher detection in patients with metastatic disease (11). The first study on the detection of ESR1 mutations in patients that were exposed to endocrine therapy was conducted on metastatic biopsies and matched cfDNA samples (37). Two hotspot mutations in codons 537 and 538 of the ESR1 gene were investigated by digital PCR (dPCR). Those findings showed monitoring ESR1 mutations by dPCR was feasible, but not all mutations found in the metastatic biopsies were detected also in matched cfDNA (40). In our study, we used the analysis on cfDNA as a validation of the results found at the CTC level. All mutations found in cfDNA were confirmed in the ER⁺ CTC population. In CTCs, we detected a new mutation in codon 8 (T570I) that was not detected in cfDNA. On the other hand, ER expression in CTCs showed a wide heterogeneous status. Most samples positive for CTCs showed a mixed population (ER⁺ and ER⁻), but 5 samples positive for CTCs were negative for ER expression. These results suggest that analysis of both CTCs and cfDNA can be a useful guide in clinical practice.

Our study is the first in which the combined systems (CellSearch and DEPArray) were applied to assess both ER expression and all 14 ESR1 hotspot mutations by the MALBAC single-cell amplification method. The combined approach, CellSearch and DEPArray, was tested already on cancer patients' samples (41–43). In a previous study, 510 CTCs were isolated from 66 MBCs. Thirty-seven CTCs were subjected to adaptor–ligation-mediated whole-genome amplification and subsequently analyzed for the expression of the ErbB2 gene and for analysis of two hotspots in PIK3CA (exons 20 and 9; ref. 41). They demonstrated applicability of that workflow and also found some heterogeneity between the analyzed CTCs and primary tumor. Another group studied the entire population of CTCs and white blood cells enriched from the CellSearch system to genotype ESR1, PIK3CA, TP53, FGFR1, and FGFR2 genes, by a next-generation sequencing (NGS) panel. Analysis of such markers was also done on cfDNA and 4 repeated samples over time during patient therapy monitoring. Two patients showed changes at the level of ESR1 gene mutations detected in cfDNA, only one in cfDNA and CTCs. This discordance can be explained by the fact that CTCs were analyzed as a pool with WBCs, where the predominant component is wild-type (42). Only recently the same group published a new study where the NGS panel was also performed on single CTC isolated by DEPArray. The purpose of the study was to determine whether cfDNA can be compared with single CTC analysis to detect tumor mutation heterogeneity (44).

Our work is the first to evaluate detection of all activating ESR1 mutations among the LBD, at the single circulating tumor cell level in MBC patients. We also monitored the acquisition of endocrine resistance and validated data confirming mutations in matched cfDNA. Of the 4 samples processed and corresponding to 3 different patients, 2 had at least one mutation that was also confirmed in cfDNA, but not in primary tissue. In both samples, the mutation detected was Y537S, positioned in exon 8 of the LBD domain of ER, as well as one of the most common mutations found in metastatic lesions. This mutation was present only in some of the CTCs from the same patient, highlighting the importance of single cell analysis instead of the previously pooling strategy (40, 42).

The role of wild-type tyrosine 537 and the effects of a number of possible amino acid substitutions have been thoroughly investigated (13). This site is located in domain E, ligand binding, and recognition region, containing the functional transcription activating domain ligand dipendene-2 and involved in the regulation of ER transcriptional activity. Among all the substitutions tested, the Y537S was the only one that showed 100% activity of the receptor in the absence of ligand (45). A recent study showed that such activity can be partially reduced by increasing the tamoxifen or fulvestrant doses, a possible strategy to avoid endocrine resistance (46). This mutation could definitely be one of the main causes of poor or inadequate response to hormone therapy.

Finally, we also demonstrate how our workflow allows investigation of intratumor heterogeneity. A loss of heterozygosity (LOH) was detected in one single CTC. We cannot totally exclude that this may be due to a technical error deriving from WGA. Analyzing a higher number of CTCs for each patient has in part solved this issue. For the first time, in our work, MALBAC has been used in combination with the CellSearch and the DEPArray systems. None among the various techniques can ensure the absence of errors like false positive/negative results or allelic dropout (ADO). Several comparison studies have shown that each WGA technique has its own benefits and drawbacks. The main advantages of MALBAC are associated with reduced ADO and PCR bias, high amplification efficiency even of GC-rich regions and high genome coverage (up to 90%; ref. 47). Moreover, an improved ability of MALBAC in SNP variant identification has been recently reported, with better performance in uniformity and reproducibility (48). Other methods, like SurePlex, allow for a better copy-number alterations (CNA) detection, with a more uniformity of amplification across the genome (49). These observations led us to endorse MALBAC as the technique to be used in genotyping the ESR1 gene. MALBAC is therefore confirmed by us to be a reliable WGA method to address single CTC molecular profiling.

In conclusion, this study demonstrated the feasibility of our protocol to detect and monitor ESR1 gene mutations at the single CTC level in MBCs. Early detection is essential to set the correct treatment plan for patients, before onset of metastatic disease. In addition, analysis of individual CTCs could allow identification of new potentially driving mutations or even new genes involved in resistance. Further studies with larger numbers of patients are required to make this approach of use in the clinic.

M. Cristofanilli is a consultant/advisory board member for Vortex. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Paolillo, K. Knudsen, M. Cristofanilli, P. Fortina

Development of methodology: C. Paolillo, Z. Mu, M. Cristofanilli

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Paolillo, M.J. Schiewer, T. Nguyen, L. Austin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Paolillo, T. Nguyen, M. Cristofanilli

Writing, review, and/or revision of the manuscript: C. Paolillo, Z. Mu, M.J. Schiewer, E. Capoluongo, M. Cristofanilli, P. Fortina

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Paolillo, Z. Mu, E. Capoluongo

Study supervision: M. Cristofanilli, P. Fortina

Other (collection and organization of the data in a database): G. Rossi

We wish to thank Dr. Saul Surrey for helpful and valuable suggestions.

This work was supported in part by NIH–National Cancer Institute Core Grant P30CA56036 (P. Fortina), Thomas Jefferson University Institutional funds (P. Fortina), by the Catholic University of the Sacred Heart Institutional funds (C. Paolillo).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.