Introduction

Circulating tumor cells (CTCs) were first described in 1869 [1]. Following the recent development of reproducible detection techniques, CTCs have been investigated as diagnostic, prognostic and/or predictive biomarkers in numerous types of cancer. In this review, after a brief reminder of detection techniques and the liquid biopsy concept, we present (1) an overview of the current clinical validity of CTCs in metastatic and non-metastatic disease, and (2) the main concepts and studies investigating the clinical utility of CTCs.

CTC detection techniques

CTCs are rare tumor cells; approximately 1 CTC per ml of blood released by primary tumors or metastases can be detected in peripheral blood. An ideal CTC detection platform must be able to isolate and detect all heterogeneous CTCs, while discarding the very large background of normal blood cells [2]. CTC selection is usually the first step of CTC detection, with positive or negative enrichment (Fig. 1). A detection step then further distinguishes (and possibly characterizes) CTCs from the remaining normal cells [2].

Fig. 1
figure 1

Circulating tumor cells (CTC): detection techniques

Full size image

The selection step is based on the biochemical or biophysical properties of CTCs. Differences in cell-surface antigen expression between CTCs and leukocytes allow positive selection of CTCs and negative selection of leukocytes. In carcinoma, the most commonly used antigen for positive immunoselection of CTCs is the epithelial cell adhesion molecule (EpCAM), which is expressed by most carcinoma cells. The Cellsearch® system, which is the only FDA-cleared detection technique, uses EpCAM-positive immunoselection as the CTC selection step. Negative immunoselection of leukocytes generally targets CD45, a pan-leukocyte antigen. Biophysics-based CTC selection is based on differences in size, deformability, bioelectric features or density between CTCs and leukocytes.

After CTC enrichment, verification of the identity of captured cells is generally based on high-resolution imaging combined with immunocytofluorescent staining. More sophisticated techniques include downstream high-throughput genomic, transcriptomic or proteomic characterization of isolated cells. A common definition of CTCs, used by CellSearch® for example, is an EpCAM+/cytokeratin+/DAPI+/CD45 cell detected in the blood [3]. Importantly, approaches based on detection of mRNA expression (by RT-PCR or probe-based assays) have been discontinued due to their lack of specificity and the results of these techniques will not be discussed in this review.

While different techniques allow the capture of CTCs with different phenotypes, their individual sensitivity and specificity are heterogeneous and vary according to the cancer type [4–6]. Because of its practical convenience (blood collection tubes allowing shipment of samples to remote laboratories) and FDA approval, most CTC clinical data have been obtained with the CellSearch® technique.

CTCs as liquid biopsy material: current status

Biopsies are invasive, costly, time-consuming, and potentially harmful and cannot be easily repeated during therapy. CTC detection is therefore very useful to avoid tissue biopsies, as CTCs can theoretically be characterized by all of the ‘omic’ assays that are usually performed on tissue biopsies. For instance, in breast cancer, point mutation [7], ERBB2 amplification [8], copy number variation [9], DNA methylation [10], transcriptomic [11] and proteomic [12] analyses have been shown to be feasible on CTCs.

The CTC count depends on the technique used and the volume of blood screened, but the CTC count ultimately impacts the validity of any CTC-based biomarker analysis. In early and metastatic breast cancer, we have demonstrated that assessment of HER2 levels by immunocytofluorescence on CTC was unreliable in patients with low CTC count (<5 CTCs/7.5 ml) [13]. Numerous other ‘omic’ techniques require more material (DNA, RNA, proteins) than what is available in a single cell, and consequently require either pooling of CTCs (e.g., mixing DNA or proteins from different CTCs) or pre-analytical amplification steps that may generate artefacts. In this context, it is therefore difficult to interpret the heterogeneous results obtained by single-cell approaches, e.g., PIK3CA mutation in CTCs from breast cancer patients [7, 14] and to demonstrate any clinical validity of these results. In addition to the above-mentioned intrinsic limitations of CTC-based biomarker analyses, the recent development of circulating tumor DNA [15] has limited the relevance of CTCs for DNA mutation analysis, especially for the detection of single nucleotide variations. The detection of complex chromosomal rearrangements, such as translocations, remains difficult on circulating tumor DNA, but is feasible on CTCs [16].

The key advantage of CTCs, especially in relation to circulating tumor DNA, is to allow all other ‘omic’ analyses in addition to DNA sequence analysis, and especially to allow the study of mRNA and proteins expressed by tumor cells. Numerous proteins are the actual targets of antitumor drugs and can be used to tailor therapy. In prostate cancer, resistance to abiraterone and enzalutamide has recently been related to a splicing variant of the androgen receptor mRNA (AR-V7 [17, 18]). PD-L1 [19] and estrogen receptor [20] are two examples of proteins whose expression levels by the tumor may influence response to immune checkpoint blockers and hormone therapy, respectively.

As of January 2017, there is currently no validated use of CTCs as a liquid biopsy technique in clinical practice, but some of these tests may become approved and reimbursed in the near future, pending further demonstration of their clinical validity and favorable economic evaluation profile [21].

Clinical validity of CTCs

Cancer screening

Studies using the Cellsearch® technique in early non-metastatic cancer have usually reported very low CTC detection rates (5–30% [22–30], depending on the cancer type). Moreover, the specificity of this technique is somewhat limited, as some circulating epithelial cells can be found in individuals with inflammatory disease or even in some healthy individuals [31]. The use of this system for cancer screening is therefore clearly impaired by its limited sensitivity and specificity. Although most of the other CTC detection techniques have not been shown to present better performances, striking results have been reported for early lung cancer detection using a filtration-based detection technique, ISET. In a preliminary report, Fiorelli et al. found that a CTC count >25 could help distinguish lung cancer from benign lesions in patients with abnormal lung imaging [32]. Using the same technology, Ilie et al. showed, in high-risk patients diagnosed with tobacco-induced chronic obstructive pulmonary disease, that only patients with detectable CTCs developed lung cancer during follow-up [33]. These results are now being investigated in a confirmatory study (NCT02500693), while our team has initiated a study on BRCA1 carriers who are at high risk of breast and ovarian cancers (NCT02608346).

Prognostic value of CTCs in metastatic cancers

The highest level of clinical evidence for the prognostic value of elevated CTC count (CellSearch®) was obtained in a pooled analysis including data from 1944 individual metastatic breast cancer patients from 20 European studies [34]. In this study, coordinated by our group, 46.9% of patients had ≥5 CTCs/7.5 ml of blood at baseline, before the start of a new line of therapy. These patients had decreased progression-free survival (PFS) (hazard ratio [HR] 1.92, p < 0.0001] and overall survival (OS) (HR 2.78, p < 0.0001) compared to patients with a low CTC count (<5 CTCs/7.5 ml). Moreover, the addition of baseline CTC count to optimized clinicopathological models (including all available significant prognostic factors) significantly improved survival predictions.

Publication-based meta-analyses confirmed that an elevated baseline CTC count in metastatic colon cancer is an independent prognostic factor for PFS and OS with CellSearch® [35] or RT-PCR/immunostaining techniques [36]. The prognostic value of CTCa has also been demonstrated during chemotherapy [37–39]. Detection of CTCs undergoing epithelial−mesenchymal transition (EMT) also appeared to predict outcome in a study with plastin-3, an EMT marker [40]. In a phase II study, Krebs et al. found that patients with elevated baseline CTC counts (≥3 CTCs/7.5 ml) could benefit from an intensive chemotherapy regimen, unlike patients with low CTC counts [41]. This result suggests that CTC count may help to tailor treatment regimens in metastatic colorectal cancer patients; a confirmatory randomized clinical trial is currently underway (Table 1, NCT01640405).

Table 1 Clinical utility of CTCs: published or ongoing phase II or III trials
Full size table

In metastatic prostate cancer, CTCs have repeatedly been reported to be a strong independent prognostic factor, particularly in castration-resistant prostate cancer (CRPC) patients. Most of these studies have been reviewed elsewhere [42]. In a large prospective study on 711 CRPC patients previously treated by docetaxel, Scher et al. demonstrated that elevated CTC counts (≥5 CTCs/7.5 ml, CellSearch®) after 12 weeks of therapy was a strong prognostic factor [43]. Interestingly, the combination of CTC count and lactate dehydrogenase (LDH) level at 12 weeks was investigated as a surrogate marker of survival at the individual-patient level and fulfilled some of the surrogacy criteria; the 2-year OS was 46% in the low-risk group (low CTC and LDH) versus 2% in the high-risk group (≥5 CTCs/7.5 mL of blood and LDH >250 U/L) and 10% in the intermediate-risk group.

CTC count has also been shown to be an independent prognostic factor in non-small cell lung cancer [44, 45], and small cell lung cancer [46].

Prognostic value in non-metastatic cancers

CTCs have been extensively studied in breast cancer, in both neoadjuvant and adjuvant settings [22], mostly with the CellSearch® detection technique. In the neoadjuvant setting, a meta-analysis based on >2000 patients, reported by our group, showed that CTC detection was a strong independent prognostic factor [47], as each CTC detected added a further ‘quantum’ of poor prognosis in terms of distant disease-free survival (DFS) and OS as well as locoregional relapse-free survival. Inflammatory breast cancers have a higher CTC detection rate (almost 40%) [48], whereas CTC counts are generally lower in non-inflammatory breast cancers (approximately 20% before chemotherapy and 15% before surgery [22]). In the postoperative setting, Rack et al. showed that CTC detection before or after adjuvant chemotherapy in 21% of patients was also a prognostic factor [49]. In these studies, CTC detection appeared to be independent of other prognostic factors, such as lymph node involvement, breast cancer size, grade or subgroup.

In locally advanced pancreatic cancer, the prospective LAP07 study also showed CTC detection to be an independent prognostic factor for OS, although based on a low detection rate (9% with CellSearch®) [30]. Similar findings have been reported in non-metastatic colon cancer using CellSearch® [24, 25, 50] or RT-PCR [51] before surgery. In contrast, CTC detection by CellSearch® after surgery was not associated with poorer outcome in a large prospective study (N = 472 patients) [52]. In that study, Van Dalum et al. detected CTCs both before and after surgery. CTC detection rates were similar (24% before and 20% after surgery), but only preoperative CTC detection was a prognostic factor. Using Cellsearch®, only studies that analyzed CTCs before surgery found CTCs to be a prognostic factor [24, 25, 50], unlike postoperative CTC detection studies [24, 52]. However, there is no explanation for this clinical observation, which is apparently restricted to colorectal cancer. In patients with high-risk colorectal cancer, higher relapse rates were described for patients in whom CTCs were detected after completion of adjuvant chemotherapy [52, 53].

The prognostic value of CTCs has been less clearly demonstrated in other non-metastatic cancers, due to small-scale studies, as in lung cancer, for which few studies have found that patients with positive CTCs after surgery had a higher recurrence rate [54–56]. A review has discussed these studies in the non-metastatic setting [57].

In conclusion, CTC detection is an independent prognostic factor in the localized and metastatic settings in several cancer types (Fig. 2). However, CTCs are rare in the non-metastatic setting, at the limit of detection of current detection methods. For this reason, the cut-off in breast cancer is 5 CTCs per 7.5 ml in the metastatic setting, while it is generally 1 CTC per 7.5 ml in localized disease. Moreover, it is noteworthy that the CTC cut-off demonstrated to be a prognostic factor by one CTC detection technology cannot be extrapolated to another technique due to the sensitivity difference between techniques. Dichotomized CTC count (high or low) has generally been used. However, the absolute value of CTCs also has a prognostic value [3, 58]. Moreover, patients with very high CTC counts in blood generally have an extremely poor prognosis [59]. The timing of CTC detection is also crucial and positive CTCs before, during or after surgery may have different biological and clinical significance. The sensitivity and specificity of CTC detection are currently too low to allow the detection of minimal residual disease, unlike ctDNA [60].

Fig. 2
figure 2

Clinical validity of circulating tumor cells (CTC): level of evidence according to clinical settings

Full size image

Monitoring treatment response

Numerous studies in metastatic breast cancer have shown that women with high baseline CTC counts but low CTC counts after one cycle of chemotherapy have a significantly better prognosis than women with persistently elevated CTC counts [PFS median: 9.6 months, 95% CI 8.2–11.1 vs 4.8 (3.7–6.5), p < 0.0001 and OS median: 27 months 95% CI 21.7–31.5 vs 13.1, 95% CI 9.4–16.4, p < 0.0001)]. Moreover, this prognosis was almost the same as that of patients with low baseline CTC counts [61]. We showed that the combination of CTC count and variation of CTC count during treatment significantly increased the accuracy of multivariate prognostic models based on standard clinical and pathological characteristics, unlike serum tumor markers (CEA or CA15.3) [34]. Similar findings were reported in metastatic colon cancer, in which changes in CTC levels during chemotherapy were correlated with outcome [37–39]. In the largest study (N = 319 patients) [38], Cohen et al. showed that PFS and OS were significantly better in patients with high baseline CTC counts and low CTC counts at 3–5 weeks than in patients with high CTC counts at both time points (PFS 6.2 vs 1.6 months, p = 0.02; OS 11.0 vs 3.7 months, p = 0.0002). As observed in metastatic breast cancer, the outcome of patients with decreasing CTC counts was similar to that of patients with persistently low CTC counts. Variations in CTC levels have also been shown to be a prognostic factor in CRPC treated by docetaxel [62, 63] and in small cell lung cancer [46].

Due to the low rate of CTC detection in non-metastatic cancers, the clinical validity of changes in CTC counts during therapy is intrinsically limited; as reported by Poisson’s law of rare events, a patient with 1 CTC detected before therapy and 0 CTC after therapy cannot be considered to present a true decrease in the CTC count. Keeping this statistical issue in mind, CTC ‘decrease’ was not significantly associated with pathological complete response of breast cancer treated by neoadjuvant chemotherapy [22]. During adjuvant therapy, persistently detectable CTCs after chemotherapy were associated with a negative impact on survival, while patients with ‘increasing’ CTC counts (i.e., 0 CTC before therapy and 1 CTC after therapy) did not have a poorer prognosis [49]. In ovarian cancer, CTC monitoring by means of a sensitive assay is more accurate than serum marker CA125 to predict response to chemotherapy and cancer relapse [64]. A phase 2 study conducted in 80 patients with non-metastatic gastric cancer showed that CTC decrease during neoadjuvant chemotherapy was an independent prognostic factor of survival [65], using a technique claiming an unusually high CTC detection rate (>90% of patients). In rectal cancer, a reduction in CTC counts (CellSearch®) was also observed in patients treated by neoadjuvant chemoradiotherapy with complete or partial remission, confirming the findings of two older studies using RT-PCR techniques [66, 67], but CTC detection was not a prognostic factor for post-treatment survival [27].

Altogether, these reports suggest that CTC monitoring is able to predict the efficacy of treatments, particularly in metastatic breast, colon and prostate cancers. However, this approach is limited by the low CTC detection rate before therapy, particularly in patients with non-metastatic cancer.

CTCs as liquid biopsy material

Several studies have correlated the phenotypic or genotypic characterization of CTCs with that of the matched tumor tissue, with a number of discordant results [7, 68]. A critical aspect of any CTC-based liquid biopsy approach is the number of CTCs examined. By comparing HER2 expression on CTCs with that of matched primary tumors in patients with early or advanced breast cancer, we demonstrated that CTC-based status may be unreliable in patients with low CTC counts. On the basis of these results, the validity of any predictive test based on CTC characterization can be expected to be limited to patients with high CTC counts, i.e., patients with a poorer prognosis. Building a clinical scenario taking into account both the intrinsic prognostic value of CTC counts and any predictive value of CTC characterization is feasible, but would require separately investigating patients with low and high CTC counts [69].

Clinical utility of CTCs

Although the clinical validity of CTC detection has been demonstrated, as described above, the clinical utility of CTC detection (i.e, does it improve patient outcome) has yet to be demonstrated before it can be implemented in routine clinical practice. Clinical trials assessing the clinical utility of CTCs have investigated three main concepts—(1) using CTCs as surrogate tumor material for liquid biopsy (e.g., AR-V7 detection predicting resistance to androgen-deprivation therapy), (2) CTC counts and CTC changes during therapy, and (3) targeting biological features that are specific for CTCs or related to metastatic spread. This last objective should be distinguished from the liquid biopsy concept, as the target is a molecular process or biomarker whose expression is specific to CTCs and not shared in common by all tumor cells. As CTCs reflect tumor heterogeneity, lead to metastasis and correlate with poorer prognosis, targeting molecular events responsible for CTC spread may help to improve patient outcome. Interventional phase 2 and 3 studies based on CTCs are summarized in Table 1 and several of them are described below.

Treatment based on CTCs used as liquid biopsy

The NCT02621190 phase II trial was proposed to study the splice variant of AR-V7, which has been shown to be a biomarker of resistance to anti-androgen therapy such as enzalutamide or abiraterone [17]. In the NCT02621190 trial, the investigators hypothesized that patients with AR-V7-positive CTCs would have a meaningful response to cabazitaxel, but this trial was withdrawn prior to enrollment.

Treatment based on CTC counts or CTC variations

The STIC CTC phase III randomized trial investigates the clinical utility of baseline CTC counts in first-line estrogen receptor-positive metastatic breast cancer patients (NCT01710605). In this trial, the type of first-line treatment (hormone therapy or chemotherapy) is determined by clinicians (standard arm) or by baseline CTC level (CTC arm). In the CTC arm, patients with low baseline CTC counts are given hormone therapy, while patients with high CTC counts are treated by first-line chemotherapy. The rationale of this trial is that CTC count is a stronger prognostic factor than other clinical prognostic factors currently used by clinicians to choose between the two treatment modalities. In this trial, >700 patients have been randomized and the results are expected in 2017.

The utility of monitoring changes in CTC counts during therapy was investigated by the SWOG S0500 trial in first-line metastatic breast cancer [70]. Patients with CTC counts remaining >5 CTCs/7.5 ml (poorer prognosis) after one cycle of chemotherapy were randomized and were possibly switched earlier to second-line chemotherapy. After having included 595 patients and randomized 120 patients, this study was reported as negative; an early switch of chemotherapy did not improve OS in this subgroup of patients. Various explanations for this negative result have been proposed [71]; it is likely that this study selected patients displaying early chemoresistance and who are generally not responsive to further chemotherapy. Other clinical trials are currently ongoing, such as the CirCe01 trial, addressing the utility of a similar strategy in the third-line setting of metastatic breast cancer (NCT01349842).

Treatment based on CTC biomarker expression

Only limited proof of concept studies are available for treatment based on CTC biomarker expression. In the HER2-negative breast cancer setting, lapatinib, which binds both EGFR and HER2, was administered to patients with HER2-positive CTCs [72] or EGFR-positive CTCs [73]. No objective tumor responses were observed in these two studies, but these results could possibly be attributed to the limited efficacy of lapatinib administered as monotherapy in HER2-positive breast cancer. In the neoadjuvant setting, another study randomized 75 women with non-metastatic HER2-negative breast cancer and HER2-positive CTCs (by a RT-PCR based technique), to receive either trastuzumab or observation [74]. Twenty-seven of 36 (75%) HER2-negative women treated with trastuzumab became CK19 mRNA-negative compared to seven of 39 (17.9%) women in the observation arm (p = 0.001) and trastuzumab also decreased the risk of disease recurrence and prolonged DFS.

Conclusions and future perspectives

Although CTC count has been shown to be a strong prognostic factor and can be used to monitor the efficacy of treatment, CTC detection has not been approved in everyday clinical practice due to the absence of clinical utility.

Large-scale ongoing studies, especially in breast cancer, should determine whether CTC detection can change clinical practice. However, CellSearch® and other techniques using epithelial biomarkers may not capture all CTCs, as mesenchymal CTCs or stem cell-like CTCs present limited expression of epithelial markers, while they probably play a crucial role in cancer metastasis and drug resistance [75–77]. The clinical validity of new CTC detection techniques and the role of mesenchymal CTCs and stem cell-like CTCs should be investigated in larger studies, which may lead to the development of new anticancer strategy in the future. Improvement in the sensitivity of CTC detection could also improve monitoring of treatment response, cancer screening [32], and the use of CTCs as liquid biopsy. Lastly, another potential future application of CTCs would be to develop CTC-derived xenografts that could lead to drug screening [78].