Minimal residual disease and circulating tumor cells in breast cancer

Tumor cell dissemination in bone marrow or other organs is thought to represent an important step in the metastatic process. The detection of bone marrow disseminated tumor cells is associated with worse outcome in early breast cancer. Moreover, the detection of peripheral blood circulating tumor cells is an adverse prognostic factor in metastatic breast cancer, and emerging data suggest that this is also true for early disease. Beyond enumeration, the characterization of these cells has the potential to improve risk assessment, treatment selection and monitoring, and the development of novel therapeutic agents, and to advance our understanding of the biology of metastasis.


Introduction
Breast cancer (BC) is the most common cancer in women in Europe [1]. Despite surgery and adjuvant systemic therapy, many women with early BC still relapse and die of their disease. Minimal residual disease (MRD) after potentially curative surgery for BC is thought to contribute to disease relapse and to be the target of adjuvant treatment. MRD is defi ned as micrometastatic cells undetectable by conventional imaging and laboratory tests. Surrogates of MRD are tumor cells detected in the bone marrow (disseminated tumor cells (DTCs)) and peripheral blood (circulating tumor cells (CTCs)) [2]. Th e detection and characterization of DTCs/CTCs are expected to lead to personalized treatment strategies and accelerate the development of novel therapeutic agents for BC [2]. Furthermore, genotypic and phenotypic characterization of DTCs/CTCs at the single cell level may provide novel insights into the biology of tumor progression [3].

Detection methods
Th e detection of DTCs/CTCs in BC is challenging since these cells are rare, occurring at a frequency of one tumor cell per 10 6 to 10 7 mononuclear cells. To isolate DTCs/ CTCs, enrichment techniques are therefore typically applied. Th ese techniques are based either on the physical properties of the cells (for example, cell density by fi coll centrifugation or cell size by fi ltration) or on their immunological characteristics (for example, cell surface antigens of DTCs/CTCs by immuno enrichment or markers of hematopoietic cells by immunodepletion). Ficoll centrifugation was widely used in the initial clinical studies of bone marrow DTCs [4]. Currently, however, enrichment techniques incorporating immunomagnetically labeled monocolonal antibodies are more often used because they improve tumor cell recovery (recovery rates of >50% to 85%) [5,6] over fi coll enrichment (recovery rate of 40%) [7] in spiking experiments using cell lines. After the initial enrichment step, DTCs/CTCs have been detected using assays based on either antibodies (immuno cytochemistry, immuno fl uorescence) or nucleic acids (mRNA transcripts by reverse transcription PCR (RT-PCR)). Table 1 summar izes the main technologies for CTC detection in breast cancer. detection that uses microscopic ferrofl uids coated with an antibody against EpCAM to magnetically separate epithelial cells from whole blood [10]. Captured cells are stained with antibodies specifi c for cytokeratins 8, 18 and 19 (pan-CK) and CD45 (specifi c for leucocytes) and stained with 4'6-diamidino-2-phenylindole-2 (DAPI; to confi rm the presence of a cell nucleus). A CTC is defi ned as a cell staining for pan-CK and DAPI, but not for CD45. Currently, CellSearch® is the only technology that has received US Food and Drug Administration (FDA) approval for CTC detection as an aid in monitoring patients with metastatic breast, colorectal and prostate cancer [10][11][12]. Th e performance of CellSearch® for CTC detection in metastatic solid tumors has also been validated in ring studies [13,14].
Other technologies include the MagSweeper, which uses immunomagnetic separation and gently enriches target cells by 10 8 -fold from blood, eliminating cells that are not bound to magnetic particles [6]. Th is process has been shown to keep cell function intact and not to perturb rare cell gene expression [6]. CTCs have been also detected using multi-parameter fl ow cytometry, and their detection with this technology was associated with poor outcome in women with BC [15]. MAINTRAC®, another method, detects circulating epithelial tumor cells from whole unseparated blood, and uses a laser scanning cytometer after staining with anti-human epithelial and anti-CD45 fl uorescent antibodies [16]. Th is technology results in CTC counts up to 10 5 per milliliter of blood in all women with early BC, consequently raising concerns about the specifi city of the method to detect tumor cells [16].
Advances in optical technologies have also improved DTC/CTC detection. Several slide-based automated microscopic scanning devices, such as the Ikoniscope® [17] and the Ariol® system [18,19], have been applied for standardized micrometastatic cell detection and characteri zation. Another approach has been developed that uses fi ber-optic array scanning technology (FAST) for rare cell detection [20]. It has been demonstrated that FAST cytometry is capable of a 500-fold increase in speed over automated digital microscopy, with com para ble sensitivity and superior specifi city [20]. Th e combination of FAST and automated digital microscopy has allowed investigators to detect rare epithelial cells from whole unseparated blood after immunofl uorescence staining with a pan-CK antibody.

Nucleic acid-based assays
Nucleic acid-based assays have been initially hampered by false-positive results due to inability to assess tumor cell morphology, expression of target genes in normal cells, and the presence of pseudogenes (genes without protein-coding abilities) [21]. Newer quantitative assays have addressed some of these problems. To detect DTCs/ CTCs in breast cancer, nucleic acid-based assays, either as single genes or as part of multiplex assays [22][23][24][25][26][27], have mainly used CK19, mammaglobin-A (MGB1), HER2 and mucin 1 (MUC1) mRNA. Th e AdnaTest® BreastCancerSelect (AdnaGen AG, Langenhagen, Germany) is a commercially available molecular assay that utilizes immunomagnetic separation with antibodies against MUC1 and EpCAM followed by a multiplex RT-PCR for HER2, MUC1 and EpCAM [28].

Emerging detection technologies
Beyond immunomagnetic separators, microfl uidic devices have been developed for rare cell tumor capture, and these involve non-electrokinetic methods, such as immobili zation via antibody [29] and size-based sorting [30,31], or electrokinetic methods (for example, dielectro phoresis) [32]. An example of a microfl uidic platform is the 'CTC-chip' , which is capable of effi cient and selective separation of viable CTCs from peripheral whole blood samples, mediated by the interaction of target CTCs with EpCAM-coated microposts under precisely controlled laminar fl ow conditions [29]. A direct comparison between CellSearch® and two commercially available CTC-chips showed that these platforms provided similar sensitivity and yield in patient samples [33]. Stott and colleagues [34] recently reported improved sensitivity of the CTC-chip for CTC detection in patients with localized prostate cancer. Several other assays have also been developed. For example, a technique named EPISPOT (epithelial immuno spot) allows detection of viable DTCs and CTCs owing to their ability to secrete individual proteins after 48 hours of short-term culture [35]. A functional cell separation method called CAM, or the collagen adhesion matrix assay, was reported to detect CTCs with the invasive phenotype and to explore their molecular features [36]. Beyond these assays, new imaging procedures have been developed for the in vivo detection of CTCs [37].
Several investigators have also evaluated the potential utility of circulating cell-free DNA, either as a surrogate to monitor MRD [38], or as a 'liquid biopsy' for real-time monitoring of tumor mutations in cancer patients [39]. Moreover, some investigators have been able to identify patient-specifi c genomic rearrangements in plasmacirculating DNA as a way to monitor MRD [40]. Th ey employed next-generation sequencing to rapidly identify patient-specifi c genomic rearrangements in primary tumors and showed that PCR assays could reliably detect these rearrangements in plasma [40]. A recent review has summarized advances in cell-free nucleic acids (DNA, mRNA, microRNA) as potential biomarkers in cancer [41].

Critical interpretation of detection technologies
Th e diff erent technologies use diff erent enrichment and detection steps and therefore do not always detect the same CTC population (Table 1). In a study comparing two commercially available assays (CellSearch® and AdnaTest®) in the same metastatic BC patient samples, the concordance between the two assays was 64% for CTC detection and 50% for HER2-positive CTC detection [42]. Th erefore, it is important to study the clinical utility of the assay-dependent CTC detection and characterization. Moreover, most enrichment methods used by the diff erent assays are biased because they result in loss of a fraction of CTCs due to tumor cell heterogeneity. As an example, some available technologies detect only EpCAM+ CTCs (Table 1). However, it has been shown that BC cell lines with low EpCAM expression and high expression of mesenchymal markers cannot be effi ciently captured using a purely EpCAM-based mechanism [33,43,44]. Some other technologies are using enrichment based on red blood cell lysis or leukocyte depletion (CD45-negative depletion) aiming at a less biased CTC enrichment (Table 1). Another critical issue with all cell detection technologies is that blood cannot be stored and must be processed soon after it has been drawn, within up to 72 hours [13] depending on the technology used. Th ere fore, the clinical validation of CTCs depends on the availability of detection technologies in diff erent labs. Th is is a major diff erence between CTCs and biomarkers from paraffi n-embedded primary tumor blocks, for which real-time processing is not mandatory. Since all currently available platforms will continue to evolve rapidly, the challenge will be to prospectively evaluate the utility of each technology to address specifi c clinical questions.

Clinical relevance of DTCs/CTCs
CTCs and DTCs were cited for the fi rst time in the 2007 recommendations of the American Society of Clinical Oncology (ASCO) on tumor markers [45]. Recently, in the 7th edition of the American Joint Committee on Cancer Staging Manual (2010), a new M0(i+) category was proposed for TNM (tumor, node, metastasis) staging in BC [46]. Th is new category is defi ned as no clinical or radiographic evidence of distant metastases, but deposits of molecularly or microscopically detected tumor cells (no larger than 0.2 mm) in blood, bone marrow, or other non-regional nodal tissue in a patient without symptoms or signs of metastases.

Clinical relevance of DTCs
Th e inclusion of the M0(i+) category was driven at least in part by a pooled analysis of individual data from several studies, which showed that bone marrow CKpositive DTCs were detected at the time of surgery in 30.6% of 4,703 patients with invasive BC [4]. Bone marrow DTCs were signifi cantly more frequent in women with larger tumors, or tumors with higher histologic grade, hormone receptor negativity, and lymph node metastasis. In multivariate analysis, the presence of bone marrow DTCs predicted for signifi cantly higher risk of death from BC [4]. Recently, in the American College of Surgeons Oncology Group's (ACOSOG) Z0010 multicenter trial, bone marrow DTCs were identifi ed at surgery by immunocytochemistry in only 105 of 3,491 patients (3%) with clinical T1/T2 N0 M0 BC [47].
Although the DTC detection rate was very low, bone marrow DTCs still signifi cantly predicted decreased overall survival [47]. A pooled analysis of individual patient data from 676 women with stage I-III BC from three studies showed that bone marrow DTCs were detected in 15.5% of patients at a median 37-month follow-up after diagnosis [48]. Th e presence of DTCs was an independent indicator of poor prognosis and could be used to select patients for secondary adjuvant treatment strategies [48].

Clinical relevance of CTCs (CellSearch®)
Using CellSearch®, ≥5 CTCs/7.5 ml of blood were detected in 49% of 177 patients with measurable metastatic BC before a new treatment was started [10]. CTC detection was an independent predictor of progressionfree survival and overall survival [10]. Th is and other studies [49][50][51][52] have provided solid data about the adverse prognostic value when CTCs are detected by CellSearch® in metastatic BC.
Detecting CTCs in non-metastatic BC is more challeng ing because these cells occur at a very low frequency in this setting. Pierga and colleagues [53] found ≥1 CTC/7.5 ml in 23% of 97 patients before administrating neoadjuvant chemotherapy (NAC) and in 17% of 86 patients after NAC. Th e detection of ≥1 CTC/7.5 ml before NAC, after NAC, or at both time points in the above study was associated with worse distant metastasis-free survival and overall survival at a median follow-up of 36 months [54]. In another study ≥1 CTC/7.5 ml were detected in 21.6% of 213 patients before NAC and in 10.6% of 207 patients after NAC [55]. In both these studies, however, neither CTC detection before or after NAC, nor changes in CTC detection during treatment, were predictive of pathological complete response [53,55]. Rack and colleagues [56] detected ≥1 CTC/22.5 ml before the start of adjuvant treatment in 21.5% of 2,026 patients with early BC [56]. In this study, pretreatment detection of CTCs was confi rmed as an indepen dent predictor for both disease-free survival and overall survival [56]. Several other investigators have detected CTCs by CellSearch® in 9% to 38% of patients with early BC without reporting survival data [57][58][59].
Th ese diff erences in CTC detection rate in early BC could be attributed to the Poisson distribution of rare events [60], to diff erences in patient populations, sampling time points, blood volume analyzed, the use or not of fi coll enrichment before processing with CellSearch®, and diff erences in image interpretation between diff erent labs. Most women in this setting have only one detectable CTC/whole blood volume analyzed. Th erefore, in order to prospectively test potential clinical applications of CTCs in non-metastatic BC, it is important to standardize image interpretation across labs by taking into account cytomorphologic criteria.

CTCs versus DTCs
Since blood is more easily obtained than bone marrow, an important question is whether peripheral blood CTCs can be used as surrogate markers for bone marrow DTCs. In one study, peripheral blood and bone marrow were collected from 341 patients at a median follow-up of 40 months after initial surgery [61]. In this study, 8 patients were CTC+/DTC+, 26 were CTC+/DTC-, and 40 were CTC-/DTC+. Although both CTCs (10% of the patients) and DTCs (14% of the patients) were significantly associated with worse clinical outcome, DTCs were more informative than CTCs [61]. Th is and other studies [62] showed that there was no good correlation between CTC and DTC detection. However, it is not clear whether this is because CTCs and DTCs represent diff erent tumor cell populations or whether this is also related to limitations of the detection technologies used. At present there are no data to support that CTCs can replace DTCs.

Clinical relevance of nucleic acid-based assays
In early BC, initial single-center studies have reported that the detection of peripheral blood CK19 mRNA by RT-PCR after fi coll enrichment of mononuclear cells was an independent prognostic factor for reduced diseasefree survival and overall survival [63,64]. In another study, 13% of 431 early BC patients were CTC-positive according to the AdnaTest®; however, no correlation with clinical outcome was reported [65]. In metastatic BC, CTC detection by AdnaTest® was reported in 52% of 42 women and predicted therapy response in 78% of cases [66]. Finally using immunomagnetic tumor cell enrichment and a multi-marker quantitative PCR based assay, CTCs were detected in 7.9% of 733 stage I/II breast cancer patients with a median follow-up time of 7.6 years and their detection was an independent predictor of metastasis-free survival and breast cancer specifi c survival [67]. However, despite these initial results, no nucleic acidbased assay has received FDA approval nor has demonstrated utility in treating patients with BC.

Clinical trials with DTCs/CTCs
Interestingly, CTC or DTC clearance after systemic treatment has been used as an endpoint in BC clinical trials. In one single-center study, it was shown that a short course of trastuzumab (3 cycles every 3 weeks) eliminated chemotherapy-resistant CK19 mRNA-positive cells in peripheral blood or bone marrow in 20 of 30 women with stage I-IV BC [68]. Another study randomized women with stage II and III BC to NAC with or without zolendronic acid [69]. Th e primary endpoint of the trial was the number of patients with detectable DTCs at 3 months' post-treatment. At 3 months, DTCs were detected in 17 of 56 patients receiving zoledronic acid versus 25 of 53 patients who did not. Although fewer women had detectable DTCs after NAC with concurrent zoledronic acid than with chemotherapy alone, this was not the case when only women who tested DTC-positive at baseline were analyzed. A critical question is if CTC clearance can be used as a 'surrogate' for survival for regulatory purposes. Such an eff ort is ongoing and investigators are studying CTC detection by CellSearch® before and after treatment in the phase 3 registration trials of abiraterone acetate in prostate cancer [70].
Although data on the adverse prognostic value of CTC d etection by CellSearch® in metastatic BC are solid, evidence from prospective trials is needed that CTC detection can lead to changes in treatment decision and thus improve clinical outcome in metastatic BC. Such an eff ort is ongoing in a phase III trial run by the Southwest Oncology Group, which is testing the strategy of changing chemotherapy versus continuing the same chemotherapy for patients with metastatic BC who have elevated CTC levels at their fi rst follow-up assessment (ClinicalTrials.gov NCT00382018).

Identifi cation of therapeutic targets
Beyond enumeration, further characterization of DTCs and CTCs holds the promise to improve treatment outcome in women with BC. Because of the availability of anti-HER2 agents, HER2 expression was studied on DTCs [71][72][73] and CTCs [33,42,55,59,66,[74][75][76][77][78] (Table 2) and was correlated with HER2 expression on the primary tumor. In most studies, HER2 expression on DTCs/CTCs is more prevalent in women with HER2-positive BC than in women with HER2-negative BC in both non-metastatic and metastatic settings. Interestingly, among women with HER2-negative primary tumors defi ned by standard pathology and detectable CTCs, between 14% and 50% may have at least one HER2-positive CTC. However, it is not known whether the discordant cases can be attri buted to technical causes or whether there is any underlying biological explanation. Clinical testing for HER2 in the primary tumor is known to result in false-negative and false-positive results [79]. Furthermore, in most cases diff erent technologies are used to evaluate HER2 in the primary tumor and the CTCs. Beyond technical issues, functional HER2 protein up-regulation on CTCs cannot be excluded, and the acquisition of HER2 amplifi cation during the course of the disease has been suggested [74]. It was shown that four out of nine patients with metastatic BC whose primary tumors were HER2-negative and who had CTCs showing HER2 gene amplifi cation derived benefi t from trastuzumab-containing therapy [74].
Beyond HER2, several other markers have been studied on DTCs/CTCs. Markers related to angiogenesis, such as vascular endothelial growth factor (VEGF), VEGF2, and hypoxia inducible factor (HIF)-1α, were observed in CTC-positive samples from metastatic BC patients [80]. Using the CTC-chip technology to purify CTCs, epidermal growth factor receptor mutations conferring drug resistance were detected in CTCs from non-small-cell lung cancer patients who had received tyrosine kinase inhibitors [81]. Androgen receptor mutations were also identifi ed in CTC-enriched peripheral blood samples from castration-resistant prostate cancer patients [82]. In most of these studies, DTC/CTC characterization was performed in few patients in the metastatic setting, and therefore validation in independent larger patient series is required. Characterizing DTCs/CTCs in nonmetastatic tumors poses additional challenges since such cells are only rarely detected in this setting. Finally, clinical trials are needed to demonstrate that CTC characterization is important for patient management.

Identifi cation of DTCs/CTCs with 'tumor-initiating cell' phenotype
Beyond the potential for improving patient outcome, the study of DTCs/CTCs aims to lead to a better under standing of the metastatic process. Research has shown a signifi cant proportion of DTCs to be resistant to conventional chemotherapy [48]. Furthermore, using Ki67 immuno staining, most micrometastatic cells have been found to be in a non-proliferative state [83]. Interestingly, the CD44 + CD24 -/low tumor-initiating cell phenotype [84] was observed in a signifi cant number of bone marrow DTCs using triple-staining by immunocytochemistry [85]. Moreover, the CK19+/MUC1 stem cell-like phenotype was demonstrated in a signifi cant number of DTCs in BC by the EPISPOT assay [35]. Epithelial mesencymal transition markers and aldehyde dehydrogenase 1 (ALDH1) were also identifi ed in a major proportion of CTCs from patients with metastatic BC [86]. However, clinical studies are needed to associate the presence of CTCs/DTCs with tumor-initiating cell phenotype with clinical outcome in women with BC.

Tumor dormancy
An issue related to the role of DTCs in the metastatic process is determining which of them will grow into overt metastases and which will not. According to clinical studies on bone marrow DTCs, 50% to 70% of patients with detectable DTCs will not develop metastases, although even patients without DTCs may relapse and die of BC [4]. For patients who relapse without such cells detectable in their bone marrow, it is possible that the DTCs have actually settled into other organs; alternatively, lack of DTCs could be the result of sampling error or refl ect the suboptimal sensitivity of CKs as a marker for DTC detection. Indeed, tumor cell dissemination has been linked to epithelial mesencymal transition and the down-regulation of epithelial cell markers [87]. Conversely, it is possible that non-relapse in the case of patients with DTCs/CTCs can be attributed to the detection of apoptotic cells or to tumor dormancy. Interestingly, CTCs have been detected in one-third of women without clinical evidence of disease up to 22 years after mastectomy for BC [88].
Th ere is evidence that several mechanisms of dormancy exist, including cellular dormancy, in which DTCs enter a state of quiescence (G0-G1 arrest), and tumor mass dormancy, in which DTCs divide but the lesion does not grow beyond a certain size [89]. Th ere is also evidence that mechanisms regulating the switch between cellular dormancy and escape from it are related to the cross-talk between DTCs and the microenvironment [89,90]. For example, loss or absence of a surface receptor like HER2, urokinase-type plasminogen activator receptor (uPAR) or integrins that transduce growth signals from the microenvironment may result in a dormant DTC, whereas the presence of such a receptor and a permissive microenvironment may result in a proliferating DTC. Interestingly, the overexpression of HER2 [71] or uPAR [91] on DTCs was associated with poor prognosis in patients with breast and gastric cancer, respectively.
Th e mechanisms that regulate the switch between tumor mass dormancy and expansion have been suggested to be related to angiogenesis [92] and the immune system [93]. When there are limitations in blood supply and/or when there is an active immune system, for example, the micrometastasis cannot grow into overt metastasis. By contrast, a shift in favor of pro-angiogenic factors and activation of transcriptional programs that allow the recruitment of new blood vessels (angiogenic switch) or an escape of immune surveillance (immunosupression) may cause the expansion of the micrometastatic cells into macrometastasis. It is not clear how all these mechanisms operate in a given patient, or how they are infl uenced by exogenous factors like stress and diet or by host genetic factors [94].

Genomic characterization of DTCs/CTCs
Th us far, only limited information is available about the global gene expression programs that determine the fate of DTCs and CTCs. Some studies have performed molecular characterization of CTC-enriched samples and reported mRNA or microRNA expression of CTCspecifi c genes in metastatic BC [95,96]. Using single cell comparative genomic hybridization, it has been shown that bone marrow DTCs are genetically heterogeneous and display fewer genetic aberrations than primary tumor cells [97][98][99]. In addition, the most prevalent chromo somal aberrations of primary breast tumors (including 8q gain, 13q loss, 16q loss and 17p loss) have rarely been found in DTCs with abnormal karyograms isolated at the time of curative surgery [3]. Husemann and colleagues [100] provided evidence that systemic spread occurs early in BC by showing that tumor cells can disseminate from earliest breast epithelial alterations in transgenic mice and from breast ductal carcinoma in situ in women. Th ese results have led to the proposal of a parallel progression model in which tumor cells disseminate early at ectopic sites and evolve in parallel with tumor cells in the primary site [3]. Finally, beyond gene expression profi ling and comparative genomic hybridization, the characterization of DTCs/CTCs using next generation sequencing may provide new insights into the cellular programs that regulate tumor dormancy and metastasis.

Future directions
Th e characterization of DTCs/CTCs might lead to the identifi cation of targets for the design of new drugs. CTCs might also be used to accelerate drug development if ongoing or future trials demonstrate that CTC clearance is a 'surrogate' for drug effi cacy. In order to move DTCs/CTCs into clinical practice, prospective trials with innovative designs and endpoints are needed to demonstrate both clinical utility and cost-eff ectiveness. Such eff orts are currently ongoing. Because the techno logies used to detect and characterize tumor cells in peripheral blood are rapidly evolving, issues like easy access to newer technologies and standardization across laboratories will be critical for prospective validation. CTC detection and characterization have the potential to improve risk assessment and provide a 'liquid biopsy' for real-time monitoring of tumor genotype/phenotype in metastatic BC. In early BC, the presence of MRD after patients have completed standard adjuvant treatment may contribute to a better selection of patients to evaluate secondary adjuvant treatment strategies. Overall, the integration of information from both the primary tumor and MRD may eventually lead to personalized treatment strategies.

Competing interests
MI has received consultancy fees and research support from Veridex. MR has no fi nancial competing interests.