Liquid Biopsy for Prognosis and Treatment in Metastatic Colorectal Cancer: Circulating Tumor Cells vs Circulating Tumor DNA

First discovered through LB in the late nineteenth century, CTCs are intact cancer cells originating from both the primary tumor and metastases, allowing an understanding of cancer spreading and metastasis [12]. Circulating tumor cells present with different proportions in various tumors, usually with a low concentration due to a short half-life (1–10 cells per 10 mL of blood as compared to 10⁷–10⁸ leukocytes) [13]. Heterogeneity hampers the definition of both sensitive and specific markers for collection. Circulating tumor cells are mainly separated by means of certain physical (i.e., filtration) and biological properties (i.e., cell surface expression), generally requiring an enrichment step (to maximize collection) and a detection step [14]. To date, the CellSearch system is the only US Food and Drug Administration-approved platform for the clinical detection of CTCs in patients with cancer, demonstrating a reproducible CTC count and prognostic value also in mCRC. It uses positive enrichment through epithelial cell adhesion molecule antibodies and then detection by immunostaining for cytokeratins and DAPI as markers for intact epithelial cells, together with absent CD45 as an exclusion marker labeling leukocytes [15]. In CRC, CTCs were detectable through CellSearch with a 36.2% sensitivity, correlating with clinical staging (up to 60.7% in mCRC) [16]. As an alternative, the AdnaTest integrates epithelial cell adhesion molecule-based enrichment with detection of cancer-specific genes through polymerase chain reaction (PCR), overcoming the expression loss of surface proteins due to epithelial-mesenchymal transition [17, 18]. Moreover, the combination of the CellSearch assay with the AdnaTest further improved the CTC detection rate in mCRC (CellSearch positivity rate 33%, AdnaTest 30%, combined 50%) [19]. The advent of microfluidic platforms such as the ‘CTC-chip’ further increased detection rates, efficiently collecting viable CRC-CTCs in microposts coated with a mixture of antibodies, under laminar flow [20]. When comparing CellSearch with IsoFlux, a microfluidic system with immunomagnetic implementation, the CRC-CTC yield was eight times higher with the latter [21]. Enrichment through leukapheresis could also substantially increase the CRC-CTC detection when combined with CellSearch [22]. The EPISPOT test and photoacoustic flow cytometry, the latter allowing in vivo CTC detection through the skin, are the most cutting-edge technologies in development [23, 24]. Overall, technological advances have led to dozens of new platforms with implemented capture efficiency, throughput rates, purity, and capability of molecular analysis [25, 26].

Circulating free DNA (cfDNA) was first detected in the mid-Twentieth century, consisting of fragments of non-encapsulated DNA in the bloodstream of healthy individuals [27]. In the late 1970s, comparable circulating tumor DNA (ctDNA) from malignant cells was discovered in patients with cancer [28]. As a result of increased cell turnover, patients with cancer usually have higher cfDNA levels than healthy subjects, with a variable ctDNA-to-cfDNA ratio of 0.01–90% depending on the tumor type, biological behavior, and cancer stage [29]. Among various malignancies, mCRC harbors the highest ctDNA amount [30]. Apart from genetic alterations, epigenetic modifications can be captured on ctDNA, most commonly in the form of DNA methylation. Difficulties in ctDNA detection include low-frequency aberrations and non-neoplastic age-dependent alterations in common driver genes. Hence, ctDNA detection requires highly sensitive and specific approaches [31]. Targeted methods such as quantitative and digital PCR (dPCR) mainly differ in their limit of ctDNA detection, reaching rates of 0.1% and 0.005%, respectively [32, 33]. Further high-resolution PCR-based technologies such as BEAMing, ARMS, and UltraSEEK further enhance the sensitivity to 0.001% and allele specificity up to a single base difference with the detection of multiple minor variants in a single reaction [34‐36]. While PCR remains confined to the analysis of few loci, NGS provides simultaneous targeted characterization of multiple genomic alterations focusing on regions of interest for mutations, but also copy number variations and chromosomal rearrangements [37, 38]. Untargeted whole genome sequencing further expands the genomic analysis, with potential applicability to extensive clonal tracking due to tumor progression or treatments, at the expense of higher costs and lower sensitivity especially for low-frequency ctDNA. However, fragment size analysis and selective sequencing of specific fragment sizes boosted ctDNA detection for whole genome sequencing in plasma samples from patients with mCRC [39]. Furthermore, a new whole genome sequencing method named Plasma-Seq overcame this issue by providing sequencing in less than 48 h [40, 41]. Finally, circulating methylated DNA (cmDNA) is similarly evaluated through targeted or untargeted interrogation. Apart from conventional direct bisulfite sequencing or pyrosequencing, novel methods such as methylation-specific PRC and methyl-BEAMing were found capable of sensitive detection of cmDNA [42, 43]. A summary of circulating DNA and CTCs with regard to advantages and drawbacks of both methods is available in Fig. 1, while the main features with regard to mCRC are listed in Table 1.

Exosomes are small extracellularly secreted vesicles (50–150 nm), containing evaluable biomarkers such as nucleic acids [44]. Given the preservation of RNA from RNAase enzymatic activity, exosomes may provide a source of messenger RNA and microRNA (miRNA), the latter being small non-coding RNA molecules implied in RNA silencing and carcinogenesis [45, 46]. Exosomes can be isolated from peripheral blood using biophysical properties such as centrifugation or precipitation, and immunoaffinity capture. After extraction procedures, targeted or untargeted methods are employed for RNA detection [47]. Despite limited research in this field for CRC, serum exosomal miRNA such as miRNA21, miRNA23, miRNA19a, and miRNA92a were identified as the most increased in this disease, some of them correlating with tumor prognosis, recurrence, and sensitivity to anti-cancer agents [48‐52]. However, an in-depth analysis of exosomes in mCRC is beyond the scope of this review.

Several studies demonstrated that CTC levels are associated with shorter survival in mCRC, as they positively correlate with tumor burden and aggressiveness. Cohen and colleagues proved feasibility of CTC isolation in 50 patients with mCRC, with 77% presenting with ≥ 1 CTC/7.5 mL of blood and a median yield of 2 CTCs/7.5 mL, dramatically less than other malignancies [53, 54]. In a following study on 430 patients with mCRC, ≥ 3 CTCs/7.5 mL was associated with shorter progression-free survival (PFS) and OS (PFS 4.5 vs 7.9 months, p = 0.0002; OS 9.4 vs 18.5 months, p < 0.0001). Conversion of baseline CTC count under cytotoxic chemotherapy from ≥ 3 to <3 within 5 weeks of treatment was associated with longer PFS and OS (PFS 6.2 vs 1.6 months; p = 0.02; OS 11.0 vs 3.7 months, p = 0.0002) [55]. The combination of CTCs and carcinoembryonic antigen (CEA) was capable of finer stratification in patients with baseline CEA ≥ 25 ng/mL, as those with < 3 baseline CTCs had longer OS (OS 20.8 vs 11.7 months, p = 0.001) and, among patients with a similar CEA reduction after treatment, the three CTC cut-off at 6–12 weeks distinguished two opposite prognostic groups (OS 19.8 vs 5 months, p = 0.0001) [56]. These results were corroborated by several studies, including ancillary studies of the CAIRO2 and the MACRO TTD trials, and a meta-analysis of 2388 patients with CRC [57‐59]. Association of the CTC count with poorer PFS and OS was also confirmed through a meta-analysis of 264 patients with resectable colorectal liver metastases (CRLM) [OS hazard ratio (HR) 2.47, p < 0.0001; PFS HR 2.07, p < 0.0001] [60]. High-CTC levels were related with non-resectability and impaired survival after metastasectomy, suggesting CTCs as a tool for decision making before liver resection. Furthermore, patients with ≥ 2 CTCs/7.5 mL experienced reduced PFS and OS when analyzing all patients and resectable patients only (p < 0.001) [61]. Given the proven prognostic value, some authors suggest that CTCs could help identify those patients benefiting the most from treatment intensification, avoiding high-toxicity regimens in low-CTC mCRC [62]. High-quality, well-designed, large-scale, multicenter prospective interventional studies are needed to assess if CTC-based therapeutic decisions could improve prognosis in mCRC. To date, the main effort to employ CTCs as a prognostic tool for treatment intensification or de-escalation in patients with mCRC comes from the VISNU-1 and VISNU-2 trials. In the phase III VISNU-1 trial, patients with mCRC with three or more baseline CTCs/7.5 mL (poor prognostic factor) were randomized to receive first-line bevacizumab with either FOLFOXIRI or FOLFOX, with superior PFS for the triplet arm [63]. Conversely, low-risk (< 3 CTCs/7.5 mL) patients with RAS wild-type (WT) mCRC were treated with doublet FOLFIRI with bevacizumab or cetuximab in the phase II VISNU-2 trial, to assess the influence of BRAF and PI3K mutational status on treatment efficacy [64]. Because a high-CTC count was established as a negative prognostic factor, this non-invasive biomarker may guide treatment selection in the mCRC first-line setting [65].

As with CTCs, circulating DNA was found to be prognostic in mCRC. Spindler and colleagues measured cfDNA in 55 patients with mCRC, achieving a wide range distribution (22–3922 ng/mL blood) and a mean value of 1157 ng/mL, with cfDNA levels > 1000 ng/mL correlating with shorter survival (p = 0.02) [29]. Similarly, a retrospective post hoc study of the NORDIC-VII trial found that high cfDNA levels before chemotherapy were associated with an impaired outcome (OS HR 1.83, p < 0.001), then confirmed by a meta-analysis of 1076 patients with mCRC (OS HR = 2.39, p < 0.0001) [66, 67]. A retrospective analysis of the CORRECT trial showed that clinical benefit from regorafenib was associated with KRAS and PIK3CA WT ctDNA status (KRAS WT vs mutant-pinteraction 0.74; PIK3CA WT vs mutant-pinteraction 0.85) as well as low cfDNA concentration (low vs high cfDNA-pinteraction 0.601) [68]. Furthermore, combined analysis of cfDNA and ctDNA added a supplementary prognostic impact [69]. Early changes in ctDNA levels were suggested as a biomarker for treatment response and outcomes in patients with mCRC. In a study of 29 patients with mCRC assessing ctDNA levels at baseline and 8 weeks after chemotherapy, ctDNA halved the predicted radiological response and resulted in longer PFS and OS (PFS HR 0.33, p = 0.04; OS HR 0.30, p = 0.08) [70]. Similarly, serial ctDNA analysis was able to identify a subsequent increase in ctDNA in initial responders to cytotoxic agents, effectively anticipating tumor progression as proven by shorter PFS and OS [71]. These results are consistent with those from the 82 patients with mCRC in the PLACOL study, monitoring ctDNA or hypermethylated alleles by dPCR [72]. The prognostic role of ctDNA was confirmed also in 41 patients with CRLM, where a decreased ctDNA/b-globin ratio after liver surgery predicted longer disease-free survival (366 vs 102 days, p < 0.001) [73]. In another cohort of 18 patients with CRLM, ctDNA was more sensitive than CEA in detecting minimal residual disease (ctDNA 100% vs CEA 56%, p = 0.008) and predicting postoperative recurrence (p = 0.003) [74]. In the PREDATOR trial, sensitivity for minimal residual disease was improved by personalized tumor-informed ctDNA detection, assessing plasma mutations according to erstwhile tumor tissue profiling in 113 oligometastatic patients with CRC (HR 4.59, p < 0.001) [75]. Finally, cmDNA due to epigenetic changes was employed to circumvent the absence of cancer-specific mutations in cfDNA. In a study assessing five methylated loci, 87% of 182 patients with mCRC showed positivity in one or more loci, dynamically correlating with overall response rate and PFS in those receiving anticancer therapy [76]. Amatu et al. confirmed baseline and dynamic cmDNA as prognostic and predictive in patients with mCRC treated with regorafenib [77]. Validation of circulating DNA in prospective clinical trials is warranted. The ongoing trial NCT03844620 is currently evaluating ctDNA capability to serve as a guide for precocious interruption of regorafenib or TAS-102 independently from radiological progression [78].

Cayrefourcq and colleagues were the first to report the successful establishment of a stable mCRC-derived CTC line. However, only one long-term CTC line was achieved from 71 unique blood samples (1.4%). Noteworthy, this required a CTC count over 300/10 mL, available only for two samples. Despite the low success rate of CTC line generation, culturing allowed NGS-based genomic analysis, followed by single-cell transcriptome, proteome, and secretome analysis, and xenotransplantation. Importantly, KRAS and BRAF concordance was proven between primary tumor, metastases, cell line, and xenografts [79]. Moreover, CTC genotyping allowed the identification of tissue-undetected mutations and secondary mechanisms of resistance to anti-epidermal growth factor receptor (EGFR) drugs [80, 81]. Altogether, the rarity of CTCs and strain from CTC line establishment are the major drawbacks in this setting [82].

The tumor mutational landscape is reflected by ctDNA, and high concordance with tissue analysis (> 80%) has been reported regarding main genomic alterations in mCRC [10, 30, 70]. The ongoing COLOMATE (NCT03765736) and TARGET trials are screening ctDNA genetic alterations in mCRC for enrolment onto targeted therapy-based/early-phase clinical trials [83]. Here, follows the main fields of interest for ctDNA profiling in mCRC.