Complementary information provided by simultaneous sequencing of CTC, cfDNA and metastatic tissue in endocrine-resistant metastatic breast cancer

Background: Endocrine resistance is a major cause of therapeutic failure in metastatic breast cancer (MBC). The information provided by DNA sequencing of plasma cell free DNA (cfDNA) and by the analysis of circulating tumor cells (CTC) may be useful to determine the occurrence and type of endocrine resistance. However, different levels of concordance between cfDNA and CTC, and also between liquid biopsy and tissue biopsy, have been reported. Thus, the best strategy for performing DNA sequencing in the setting of therapeutic decisions for luminal MBC patients is not well dened. The purpose of this study was to determine the concordance between DNA sequencing data from CTC, cfDNA, metastasis, and primary tumor in endocrine-resistant luminal MBC. Methods: Using the same panel (10 genes) we performed simultaneous sequencing of cfDNA, CTC, metastases and primary tumor in 38 patients with luminal MBC. CTC isolation was performed with an Epcam-based immunomagnetic system. Concordance of DNA sequencing data of the different types of sample was analyzed and correlated with clinical data. Results: CTC were detected in 21% of the patients. The rate of detection of genetic alterations was 87.5% for CTC, 61.5% for cfDNA, 63.2% for metastasis and 67.7% for primary tumor, and involved mainly TP53, PIK3CA, ESR1 and AKT1. A higher number of mutations was found in cfDNA of MBC patients with progressive disease (p=0.02) and the same trend was observed for endocrine resistance (p=0.08), especially for PIK3CA, ESR1 and AKT1 mutations. Global concordance was low to intermediate between CTC and cfDNA, but signicantly higher between CTC and metastasis. Concordance also varied according to the specic genetic alteration, with no ESR1 mutations detected in CTC and with a low concordance for PIK3CA and AKT1 mutations between CTC and cfDNA samples. Conclusions: DNA sequencing of different types of tumor samples (cfDNA, CTC, metastasis and primary tumor) may provide different rates of genomic alterations detection in luminal endocrine-resistant MBC.


Background
Although new therapeutic options are now available, metastatic breast cancer (MBC) is still the rst or second cause of death by cancer in women. Luminal MBC must be treated with endocrine therapy except in those cases presenting with a visceral crisis [1]. However, endocrine resistance invariably occurs, leading to disease progression and making chemotherapy the only option for most patients after two or three lines of hormonal treatment [2]. A better knowledge of endocrine resistance causes and of genomic alterations in patients with luminal MBC may offer additional therapeutic options for some patients.
In previous publications, enrichment of metastatic disease with respect to the primary tumor has been observed in certain genomic alterations related to endocrine resistance, such as ESR1, RAS, AKT1 and RB1 [3,4]. The acquisition of at least part of these alterations, especially ESR1 and ERBB2, seems to be related to hormonal treatment [3]. Although better knowledge of these alterations could guide the treatment of luminal MBC, especially for guiding inclusion in clinical trials [5], until very recently most alterations have not modi ed the patient's treatment in clinical practice. However, the recent approval of alpelisib, a PIK3CA inhibitor, is linked to determination of PIK3CA mutations as predictors of response. There is also a growing conviction that the availability of new drugs, such as AKT inhibitors and new selective estrogen receptor degraders, among others, could increase the relevance of DNA sequencing in MBC.
Biopsies of metastatic tumor have been the main source of samples for next-generation DNA sequencing (NGS) in the metastatic disease setting. However, other alternatives, such as liquid biopsy, based either on cell-free circulating DNA (cfDNA) or on circulating tumor cells (CTC) obtention [6], may avoid invasive procedures and provide a dynamic strategy for analyzing resistance biomarkers in MBC [7,8]. Thus, it is relevant to understand the contribution of each of the potential sources of information on endocrine resistance mechanisms in patients with luminal MBC. While cfDNA is able to identify resistance markers and therapeutic targets, CTC are usually preferred for prognostic evaluation and tumor monitoring [9].
However, some recent works have shown that CTC evaluation might be potentially informative about endocrine resistance [10] and resistance mechanisms heterogeneity [11]. In studies with ctDNA in luminal CM, up to 78% of patients have shown GA [12], with an average of 2-2.5 GA, mainly affecting TP53, PKI3CA, ESR1 and, less frequently, AKT1 [12,13]. Differences in the degree of concordance of genomic alterations between CTC and cfDNA have been reported for ESR1 mutations [14,15] and for NGS panels [11,16]. These differences may be derived from technical sources of variation or from true differences in the cell compartments of origin. Since the level of concordance between cfDNA and CTC DNA is unclear, and discordant results have been communicated particularly for luminal MBC, the potential contribution of CTC DNA sequencing to therapeutic decisions in luminal MBC is also undetermined. A recent work has shown that CTC sequencing may provide additional information in certain patients [17], beyond the prognostic value of CTC levels [18]. Regarding primary tumor and metastatic tissue, the correlation of ndings between ctDNA and both types of tissue is highly variable among the different publications, with only 10-15% concordance rates in some of them [12].
Our objective was to determine the mutational concordance between CTC, cfDNA and metastatic tissue in luminal metastatic breast cancer (MBC) and to explore the potential mechanisms of endocrine resistance in the different tumor compartments. The nal aim was to determine which type of sample or sample combination might maximize the detection of targetable mutations in the setting of advanced disease.

Patients and sample collection
We included 40 patients with luminal (ER positive and/or PgR positive; HER2 non ampli ed) MBC at the Department of Hematology and Medical Oncology of Hospital Universitario Morales Meseguer, Murcia, Spain. All patients had histologic con rmation of breast invasive carcinoma and clinical evidence of metastatic disease. Endocrine resistance was de ned for the rst recurrence or rst progression according to established clinical criteria [1] as primary, secondary or no resistance. Response evaluation for the current treatment of the patient was performed according to RECIST 1.1 criteria.
We collected two 10 ml tubes of whole blood (WB) for each patient, EDTA and PaXGene ccfDNA. Plasma was obtained from blood collected in PaxGene tube. Primary and metastatic tumors were obtained as FFPE blocks.
CTC enrichment and identi cation CTC isolation was performed in a whole blood (WB) sample of 7.5 mL by Epcam-based immunemagnetic enrichment with anti-Epcam antibody conjugated microbeads and a column system (MCAS Miltenyi Biotec, Madrid, Spain) following manufacturer's instruction.Parallel con rmation of the presence of CTC in each case was obtained in a simultaneous blood sample of 5mL: after immune-magnetic enrichment with ADNaTest Breast Cancer Select, (Qiagen, Hilden, Germany), multiplex RT-PCR was performed with AdnaTest Breast Cancer Detect (GA733-2, Muc1, HER2 and control actin) in an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). ADNaTest RT-PCR was further con rmed by qPCR using speci c primers for the three markers (GA733-2, Muc1, HER2); only con rmed results were considered as indicative of CTC presence. We used MCF-7 spike-n in human blood to determine the sensitivity of the test (Supplementary Figure S1). No CTC quanti cation was performed, and patients were classi ed either as CTC positive (CTC+) or CTC negative (CTC-).
DNA isolation from CTC, plasma and tumor tissue cfDNA was extracted from plasma (5 ml) derived from the PaXGene ccfDNA tubes with QIAmp Circulating Nucleic Acid Kit. After puri cation, we quanti ed DNA concentration with a QUBIT dsDNA HS Assay (Invitrogen, Thermo Fisher Scienti c, Waltham, MA, USA ). Median DNA concentration was 24 ng/mL (range: 4.7-84 ng/mL). ctcDNA from CTCs was isolated using the "Cells and Tissue Genomic DNA Isolation Microkit" (Norgen). The e ciency of the isolation was con rmed by a technical validation with digital PCR (Supplementary Figure S2).
Tumor DNA was extracted from para n-embedded tissues (breast primary tumor and/or metastatic biopsies) using the AllPrep DNA/RNA FFPE kit (Qiagen) after marking of those areas with at least a 50% of tumor cellularity by a pathologist (ACB, FMD).

DNA sequencing
Sequencing of cfDNA and ctcDNA was performed in an Ion Torrent PGM System (Thermo Fisher Scienti c, Waltham, MA, USA) using the Oncomine Breast cfDNA Assay, a panel of 152 hotspots and indels targeting 26 amplicons of 10 genes (AKT1, EGFR, ERBB2, ERBB3, ESR1, FBXW7, KRAS, PIK3CA, SF3B1, TP53). Libraries were constructed using multiplex PCR according with the manufacturer's instructions and were puri ed with Agentcourt AMPurez XP (Beckman Coulter, Barcelon, Spain. Limit of detection range was 0.01-0.15%, with 92.1% cases equal or less than 0.1%. Sequencing of tissue samples was performed with an Ampliseq custom panel designed to target just our sequences of interest (the same 152 hotspot mutations that the commercial panel Oncomine Breast cfDNA). This design was made using the algorithm Ion Ampliseq Designer 6.1.3.
For bioinformatic analysis, we used Torrent Suite software 5.8 (reference: hg19) and annotation with Ion Reporter 5.6

Statistical analysis
All statistical analyses were performed with SPSS 21.0 (IBM España, Madrid, Spain). The comparison of the mutational pro le of CTC and cfDNA and its correlation with endocrine resistance and patient characteristics was performed with Chi-squared test for comparison of proportions and with Kendall's kappa for concordance. Overall survival was calculated from the time of inclusion in the study; Kaplan-Meier curves were obtained, and univariate comparisons were done with the log-rank test. A p value of < 0.05 was considered as signi cant. The study report is in accordance to the REMARK guidelines [19].

Results
Patient characteristics and availability of CTC, cfDNA and primary and metastatic tissue.
We performed DNA sequencing of CTC and cfDNA samples in a cohort of 38 women with metastatic breast cancer patients treated in our center. A valid sample was available in 34 of them. In 20 of them there was available tissue of metastasis, and 32 also had tissue from primary tumor. Median age of the patients was 58 (range, 37-85). As shown in Table 1, most patients (81.6%) had endocrine resistant disease, either primary (21.1%) or secondary (60.5%). The plasma samples were taken in different response settings corresponding to different moments of the clinical evolution: 2 at time of metastatic recurrence or after metastasis surgery (5.3%), 21 with stable disease (55.3%), 5 with partial or complete response (13.2%), and 9 of them at time of disease progression (23.7%). Treatment at time of sample obtention consisted in endocrine therapy in 60.5% of the patients and was chemotherapy for the remaining 39.5%. Median overall survival for the whole group has not been reached, and 2-years survival is 60% (95%CI, 45-75%).
Among the 38 patients, only 8 (21%) had detectable CTC. All CTC + patients showed endocrine resistance (37.5% primary and 62.5% secondary), although no signi cant association was found between endocrine resistance and CTC detection (Fisher's exact test, p = 0.31). Detection of CTC was associated with progressive disease (45.5% vs. 10.3% of patients with SD/OR, p = 0.03) and showed a trend to worse overall survival (median: 13.2 months vs. not reached; p = 0.13). The speci c type of genetic alterations also differed among the different types of samples (  and only one case with an ESR1 mutation (3.1%). Interestingly, a few mutations were only found in some type of samples. In particular, FBXW7 and ERBB3 mutations were detectable only in CTC, while ERBB2 mutations were only present in the primary tumor sample of one patient.
The number of different mutations of the same gene for a particular patient was also higher for TP53 and PIK3CA, especially in the primary tumor and cfDNA, thus yielding a larger total number of genomic alterations (Fig. 2).
Matched asssesment and concordance of somatic mutations on liquid biopsies (CTC, ctDNA) and metastasis tissue Globally, we found a low to intermediate concordance between CTC and cfDNA, for the detection of any genomic alterations (Table 3). Concordance was also low to moderate between primary and metastatic tissue. A higher overall concordance was found between CTC and primary tumor sequencing (k = 0.57; p = 0.12), which was signi cantly better between cfDNA and metastasis (k = 0.51, p = 0.03).
We performed a concordance analysis speci cally focused on the three main groups of mutations related to endocrine resistance. Also, in order to determine the clinical utility of each type of sample to maximize the identi cation of targetable mutations, we calculated the percentage of positivity provided for each type of sample over the total number of cases with mutations identi ed by sequencing every pair of matched samples. As shown in Table 3, our ndings are consistent with different degrees of overall concordance according to the type of mutation.
For PIK3CA, the concordance between CTC and cfDNA was low, and the same occurred when CTC results were compared with tissue biopsy of either the primary tumor or metastases, which suggest some degree of complementarity between CTC and other type of samples. Our results also showed a low concordance of cfDNA positive results with those provided by sequencing of tumor (25%) or metastasis (50%) biopsy, which might provide a higher sensitivity for detection of mutations (70-87.5% rate of positive results in comparison with 25-40% for cfDNA).
Regarding AKT1, complete concordance was found for detection of mutations between CTC and primary tumor or metastasis biopsy, while only 33% of mutations identi ed in cfDNA sequencing were found in the matched CTC samples. The concordance between cfDNA and primary or metastatic biopsies was also low, with the identi cation of only 16-25% of the mutations found in cfDNA.
ESR1 mutations were not found in any of the CTC samples, which has limited the concordance analysis with othder samples. However, the nding of ESR1 mutations in cfDNA again showed a low concordance (25%) with metastases biopsy detection, and null concordance with primary tumor, as expected. Finally, although not shown in Table 3, the concordance for TP53 mutations was very low in every pair of matched samples, with kappa values below 0.20 in all cases.
Association of genomic alterations in metastatic samples (CTC, cfDNA, metastatic tissue) with progressive disease and endocrine resistance The average number of mutations found in cfDNA was signi cantly higher in those patients with disease progression (2.27 vs. 0.81; Mann-Whitney U, p = 0.002), but we did not nd the same differences for CTC (p = 0.39) or metastatic tissue (p = 0.79) samples. Similarly, the number of patients with any mutation present in cfDNA showed a non-signi cant association with the response status (progression disease: 90.9%; stable disease or objective response: 60.9%) (Fisher's exact test, p = 0.11). Endocrine resistant patients also showed a trend to a higher number of genetic alterations in cfDNA when compared with patients without resistance (1.42 vs. 0.43; p = 0.08), again without differences for the rest of samples (Fig. 1).
We also analyzed the association of nding speci c genetic alterations in any of the metastatic samples (CTC, cfDNA, metastasis biopsy) with endocrine resistance. The association differed according to the speci c gene mutation: the frequency of PIK3CA mutations trended to be higher (p = 0. 22   The nding of relevant genomic alterations may have implications for clinical trial inclusion and for decision-making concerning next treatments. Our analysis focused on concordance, which was low to moderate between the different types of samples. However, in the clinical setting, the maximization of mutation detection is the main goal to improve therapeutic options for patients. Consequently, positive concordance is the true measure of potential clinical utility, and our evaluation was directed to determine whether sequencing other type of samples might increase the detection rate. While CTC sequencing results were similar to primary tumor, a higher concordance was found between cfDNA and metastatic tumor, in agreement with previous whole-exome sequencing data [20]. The low identi cation of ESR1 mutations in CTC is also interesting, and previous data have shown similar results concerning the low concordance of ESR1 variants between CTC and cfDNA [17]. Although these data are limited by the small number of cases and by technical factors such as the extent of the gene panel, there are also plausible biological explanations related to space and time tumor heterogeneity. In particular, the low overall concordance between both forms of liquid biopsy, CTC and cfDNA, suggest relevant biological differences among the different tumor compartments.
The use of pooled CTC samples for sequencing deserves a comment. While isolated CTC sequencing may provide additional information on tumor heterogeneity of resistance mechanisms [21], pooled CTC sequencing may be a more accessible and practical clinical approach for liquid biopsy. Additionally, recent data on CTC clusters suggest a prominent role of CTC aggregates in metastasis generation and resistance to cancer therapy [22], thus supporting a less selective approach for studying CTCs in the clinical setting. In any case, CTC sequencing provided additional data to cfDNA in order to identify genomic alterations in luminal MBC patients. When only potentially targetable mutations were analyzed, 25% patients with CTC + had additional ndings for PIK3CA mutations. Consequently, partly due to the low yield of CTC in this group of patients, less than 5% of the total group of luminal MBC patients obtained therapeutically relevant results after CTC isolation and sequencing. Together with the absence of ESR1 mutations in CTC pooled samples, our results suggest that, in most cases, cfDNA alone might provide su cient information from a clinical perspective, and that a metastasis biopsy should be the second preferred option in order to maximize the identi cation of targetable genomic alterations. Alternatively, a sequential approach with CTC sequencing in those cases in which cfDNA or tissue sequencing do not yield any targetable genomic alteration or in which a metastasis biopsy is not feasible might be useful in some patients. CTC sequencing is still a valid research tool and may be a relevant source of information about tumor heterogeneity; prognostic complementarity of ctDNA and CTC dynamic evaluation has also been suggested [18,23,24] and new approaches for liquid biopsy are being developed that include both types of information [25].
A higher number of genomic alterations was found in patients with endocrine resistance and progressive disease. This nding is in agreement with previous reports [3,26], and poses the question of the best moment for obtaining a liquid or tissue biopsy for DNA sequencing. Although our work is based on a onetime sample collection, our data support a dynamic evaluation of tumor genomics and a preference for obtention of tissue or plasma at time of progression or when endocrine resistance occurs [4,8].
Our work has limitations, the rst of them being the different moments in the evolution of the disease and the different response settings in which samples were obtained. However, it should be noted that blood obtention for cfDNA and CTC samples was simultaneous, making sequencing results directly comparable. Comparability was further assured with the utilization of the same gene panel for all samples of the same patient and our results of around 2 genomic alterations in each patient is in accordance with previous publications [13]. A second potential limitation is the utilization of an Epcambased method for CTC isolation, with the potential loss of undifferentiated or mesenchymal CTC [27].
However, Epcam-positive CTC seem to be the most relevant population in terms of metastasis generation [28] and prognosis [9] of MBC. Also, although our CTC isolation method was validated with spike-in experiments, its sensitivity is probably lower than that of other marketed systems such as CellSearch; in fact, the 21% detection rate suggest that positivity for CTC with this approach might be equivalent to nding > 5 CTC/10 mL with other methods. Finally, the utilization of a panel with a limited number of hotspots, although including the main genomic alterations previously linked to resistance, might have provided a partial estimation of the concordance between samples.

Conclusions
In conclusion, in luminal MBC the performance of sequencing different types of tumor samples (CTC, cfDNA, metastasis, primary tumor) is different, and may be variable for the different genetic mutations.
Sequencing of CTC does not seem to be the best approach due to the low yield of CTC and the low rate of identi cation of targetable mutations. The obtention of liquid biopsy at time of progression or when endocrine resistance develops is associated with a higher rate of detection genomic alterations. This information may be valuable for developing sequential algorithms of sample sequencing to maximize the rate of detection of potentially targetable genetic mutations in luminal metastatic breast cancer.

Declarations
Ethics approval and consent to participate The study was performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants and the protocol was approved by the institutional Ethics and Clinical Research Committee (CEIC Hospital Universitario Morales Meseguer; EST24/15).

Consent for publication
Not applicable Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was funded by Instituto de Salud Carlos III (Project PI15/1499), including FEDER funding by the European Union. ENM was funded by the Ministerio de Educación, Cultura y Deporte (Spain) (FPU16/06537). This work was also partially funded by "Calasparra se mueve", a research-funding initiative inspired by women from Calasparra (Murcia), Spain. The funding bodies did not have any role in the design, analysis and interpretation of the study.
Authors' contributions ENM participated in the conception and design of the study, performed laboratory procedures, participated in data retrieval and analysis, performed statistical analysis and drafted the manuscript; MPFP participated in the design of the study and performed laboratory procedures and data retrieval; EGM, PMB, AIR and EGT participated in the obtention of clinical data, data analysis and critically revised the manuscript; ACB and FMD performed the pathologic studes, selected the samples for DNA sequencing, participated in data retrieval and critically revised the manuscript; RTM participated in the design of the study, performed laboratory procedures and critically revised the manuscript; FAP conceived of the study, participated in its design, obtained clinical data, participated in data analysis and revised the manuscript. All authors read and approved the nal manuscript. Sequencing results for each type of sample.

Figure 2
Total number of genomic alterations found in each type of sample for each gene.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.