Molecular characterization of circulating tumour cells identifies predictive markers for outcome in primary, triple‐negative breast cancer patients

Abstract mRNA profiles of circulating tumour cells (CTCs) were analysed in patients with triple‐negative breast cancer (TNBC) (pts) before (BT) and after therapy (AT) to identify additional treatment options. 2 × 5 mL blood of 51 TNBC pts and 24 non‐TNBC pts (HR+/HER2−; HR−/HER2+) was analysed for CTCs using the AdnaTest EMT‐2/Stem Cell Select™, followed by mRNA isolation and cDNA analysis for 17 genes by qPCR PIK3CA, AKT2, MTOR and the resistance marker AURKA and ERCC1 were predominantly expressed in all breast cancer subtypes, the latter ones especially AT. In TNBC pts, ERBB3, EGFR, SRC, NOTCH, ALK and AR were uniquely present and ERBB2+/ERBB3 + CTCs were found BT and AT in about 20% of cases. EGFR+/ERBB2+/ERBB3 + CTCs BT and ERBB2+/ERBB3 + CTCs AT significantly correlated with a shorter progression‐free survival (PFS; P = 0.01 and P = 0.02). Platinum‐based therapy resulted in a reduced PFS (P = 0.02) and an induction of PIK3CA expression in CTCs AT. In non‐TNBC pts, BT, the expression pattern in CTCs was similar. AURKA+/ERCC1 + CTCs were found in 40% of HR−/HER2 + pts BT and AT. In the latter group, NOTCH, PARP1 and SRC1 were only present AT and ERBB2 + CTCs completely disappeared AT. These findings might help to predict personalized therapy for TNBC pts in the future.

(pCR) rate 7 as well as the event-free survival and overall survival (OS) in some neoadjuvant trials. 8,9 Response to therapy in the neoadjuvant setting is assessed as pCR, which is achieved in about 27%-45% of the patients depending on tumour stage and which was shown to improve prognosis in some clinical studies. 2,3,[10][11][12][13][14] Those patients that do not achieve a pCR after NACT have a worse OS which might be improved by the addition of capecitabine after NACT, resulting in prolonged progression-free survival (PFS) and OS rates as shown in the CreateX trial. 15 However, despite achieving a pCR, about 5%-20% of patients will experience relapse. 14 Up to now, except for the pCR rate, the success or failure of anti-cancer therapies in BC is only assessed retrospectively by the absence or presence of overt metastases during the post-operative follow-up period. Therefore, other markers are needed to identify patients at high risk of recurrence to offer additional therapeutic options.
Circulating tumour cells (CTCs), the precursor of metastatic disease, would be an ideal surrogate marker to identify prognostic and predictive factors directly at primary diagnosis to guide optimal individual therapeutic strategies for metastasis prevention.
CTCs in primary BC have been extensively studied in large patient cohorts and hold strong potential to be translated into individual targeted therapy since their prognostic significance with regard to reduced PFS and OS has already been demonstrated. [16][17][18] Although their frequency in the primary setting is quite low, a variety of methods have been established to detect and characterize CTCs. 19,20 CTCs were shown to be very heterogeneous, even within the same patient, including the presence and persistence of resistant and stem cell like CTCs as well as CTCs in epithelial mesenchymal transition (EMT). [21][22][23][24] Furthermore, a discordant receptor status between ER, PR and especially HER2 on CTCs and the primary tumour has been demonstrated which explains why CTCs survive after targeted therapy based on markers detected on the primary tumour. [25][26][27] It might be assumed that CTCs may, therefore, also represent distinct metabolic profiles for survival, metastatic spread and therapy resistance since the prognosis in BC has been shown to be different in various BC subtypes. 3 For patients with TNBC, CTC counts were shown to have prognostic value in early as well as later stages of the disease, [28][29][30] but characteristics of these cells have rarely been shown. In this regard, our group demonstrated that CTCs found in primary, non-TNBC patients were frequently characterized as triple-negative phenotype regardless of the ER, PR and HER2 status of the primary tumour. 25 On the other hand, it was recently demonstrated that CTCs of early-stage TNBC patients frequently expressed ER, PR, HER2 and EGFR with a predomination of the latter one over the other phenotypes. Triple-staining experiments revealed that distinct subpopulations were identified in individual patients. 31 These data emphasize that a comprehensive characterization of these cells is essential to find targets for additional treatment options in this high-risk BC subgroup.
In the present study, therefore, we aimed to analyse mRNA profiles of CTCs from primary, non-metastatic TNBC patients before and after NACT using a multimarker gene panel to (a) identify additional individualized treatment options, (b) identify predictive markers for treatment outcome and (c) compare the results with CTCs of non-TNBC patients.

| Patient population and patient characteristics
The study was conducted in the Department of Gynecology and

| Eligibility criteria and response criteria
The eligibility criteria were as follows: histologically proven BC,

| Sampling of blood
2 × 5 mL EDTA blood was collected for CTC isolation in S-Monovettes ® (Sarstedt AG & Co., Nümbrecht, Germany) BT (n = 59 patients) and AT, before surgery (n = 58 patients). The samples were stored at 4°C and were processed not later than 4 hours after blood withdrawal.

| Enrichment of circulating tumour cells
CTCs were isolated from 2 × 5 mL blood by positive immunomagnetic selection targeting EpCAM, EGFR and HER2 (AdnaTest EMT-2/StemCell Select™, QIAGEN GmbH, Hilden, Germany). The method has been described in detail elsewhere. 33 Briefly, labelled CTCs were extracted using a magnetic particle concentrator and were lysed according to the manufacturer's instructions. Cell lysates were stored for a maximum of 2 weeks at -80°C until further processing.

| mRNA isolation and reverse transcription
mRNA was isolated from the cell lysates by oligo(dT) 25  The method has been recently published in detail. 33

| Data evaluation
CTC isolation was conducted in duplicate from 2 × 5 mL blood for each patient sample. cDNA was analysed separately from these duplicates. After binary evaluation of the qPCR data, described below, were excluded from the analysis.
Abbreviations: na, not applicable; nd, not done; nk, not known.

TA B L E 1 (Continued)
The quantification cycle was defined to be reached at the threshold 0.5 in all cases. CTC expression data were normalized to matched expression data of healthy donor controls (n = 20), described in detail by Keup et al. 33 Transcripts not exclusively expressed in CTCs but also in the 100-200 contaminating leucocytes were normalized to the leucocyte-specific transcript PTPRC (∆∆C q = [Cut-off(gene)-

| RE SULTS
The clinical characteristics of all patients at the time of first diagnosis are shown in Table 1

| Gene expression profiles in CTCs of TNBC and non-TNBC patients before and after therapy
In total, 51 patients with TNBC were analysed for CTCs, 39  ERBB2 + CTCs, initially detected in 50% of the cases, completely disappeared AT, most likely due to anti-HER2 targeted treatment which seemed to also markedly reduce initial PI3K/AKT/MTOR overexpression. Notably, AT, AURKA + CTCs and ERCC1 + CTCs were found in 38% and 50% of HR−/HER2 + cases and in 15% and 38% of the HR+/ HER2− patients.

| Gene expression profiles in CTCs of TNBC and non-TNBC patients before and after therapy (pairwise)
Pairwise expression of the different genes BT and AT is detailed in  were correlated with PFS before therapy in COX univariate proportional hazard analysis (detailed in Figure S1). As shown in Figure 3A and B, the presence of EGFR+ or ERBB2+ or ERBB3 + CTCs in TNBC patients BT and ERBB2+/ERBB3 + CTCs AT significantly correlated with a reduced PFS (0.01 and P = 0.02). Interestingly, the presence of ERBB3 + CTCs alone BT was sufficient to significantly (P = 0.04) indicate a shorter PFS (data not shown). OS analysis did not reach statistical significance.

| Survival analysis
The relationship between PFS and the given therapy is shown in Figure 4A  (data not shown).
Using COX multivariate proportional hazard analysis with standard staging parameters like nodal stage and tumour size BT and AT as well as menopausal status and grading, the CTC 'all ERBB family status' turned out as a significant, independent unfavourable predictor for PFS ( Figure 5).

| D ISCUSS I ON
We here demonstrated that markers representing the PIK3CA signalling pathway as well as the resistance marker ERCC1 and In non-TNBC patients, we demonstrated that in CTCs from HR−/ HER2+ patients most of the 17 genes were down-regulated AT, most likely because of anti-HER2 therapy which was the common treatment choice in that group. ERBB2 + CTCs were no longer detected after treatment but also genes involved in the PIK3CA signalling pathway as well as the resistance markers were strongly affected.
In the HR+/HER2− group, the PIK3CA signalling pathway also got down-regulated but some of the genes appeared de novo (SRC, ERBB2) and ERCC1 was strongly up-regulated which might be considered as an onset of resistance.
Although the heterogeneity of CTCs has widely been con- For TNBC, the role of EGFR but also HER3 has often been discussed. 44 Although the expression of EGFR could frequently be observed in TNBC tumour tissue, the success of EGFR-targeted regimens was reported, so far, to be insufficient in terms of sensitivity.
It is discussed that molecular mechanisms other than EGFR alone must be considered to develop successful combination therapies.  48 The role of BRCA1, correlated with PIK3CA expression in this study, has generally been described to be involved in DNA repair mechanisms and cell cycle regulation which explain its role in counteracting DNA-damaging chemotherapy. 49 In general, in all BC subtypes, genes associated with resistance, especially AURKA and ERCC1, were frequently expressed, mostly AT, perhaps representing one of the major problems in cancer treatment. We recently demonstrated that CTCs detected after NACT were associated with tumour stem cell characteristics as well as ERCC1 expression which may also suggest a potential selection of this CTC subset by chemotherapy. 23

| CON CLUS I ON S AND LIMITATI ON S OF THE S TUDY
Taking all these considerations into account, a comprehensive characterization of CTCs, probably on the single cell level, might help to identify patients for further targeted therapy. However, we would like to emphasize that this is a small 'proof of principle study' which has to be confirmed in a larger study group. In this context, also statistical analysis has an exploratory character; thus, multiple testing correction (eg a Bonferroni correction) was not applied but will be considered in further confirmatory settings. 50

ACK N OWLED G EM ENTS
We thank Ute Kirsch for her excellent technical assistance.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
AKB contributed to protocol/ project development, patient recruitment, data collection, data analysis and manuscript writing; CK contributed to method evaluation and visualization of the results; OH contributed to protocol/ project development, patient recruitment, data collection, data analysis and manuscript writing; SH contributed to data analysis and establishment of multimarker panel; RK contributed to overall responsibility; SKB contributed to protocol/ project development, data collection, data analysis, manuscript writing.

E TH I C A L A PPROVA L
All specimens were obtained and collected after written informed consent from all subjects using protocols approved by the clinical ethic committee of the University Hospital Essen (05/2856).

CO N S E NT FO R PU B LI C ATI O N
Not applicable.

DATA AVA I L A B I L I T Y S TAT E M E N T
C q values of the patient samples (in duplicate) and healthy donors are listed in Table S1. Further information is available from the corresponding author on reasonable request.