A panorama of colon cancer in the era of liquid biopsy

Abstract

Colon cancer (CC) is one of the most frequent cancers worldwide being responsible for over 500 thousand deaths in 2022. Its financial and human burden is expected to increase in the next decades accompanying the growing and aging of the global population. Much of this burden could be alleviated considering that the lethality of CC is mostly due to its late diagnosis and failure in the individualized management of patients. Coordinated government actions and implementation of better diagnostic tools capable of detecting CC earlier and of tracking tumoral evolution are mandatory to achieve a reduction in CC's social impact. CtDNA-based liquid biopsy (LB) has great potential to contribute to patients' screening adhesion, CC earlier detection, and to longitudinal tumor follow-up. In this review, we will discuss the latest epidemiological data on CC disease, diagnostic, subtypes, genetics, and treatment management focusing on the advantages and limitations of ctDNA-based LB, including important bottlenecks and solutions necessary for its clinical translation. The latest ctDNA-directed CC clinical trials will also be examined.

Graphical abstract

Highlights

• •Better screening is required to counteract colon cancer increase under 50 years.
• •Many gene alterations occur in colon cancer but only some impact treatment.
• •Liquid biopsy is appealing for monitoring tumor heterogeneity, evolution and spread.
• •Digital PCR is attractive for detecting low-level drug resistance-associated alterations and MSI.
• •Numerous trials are evaluating ctDNA as markers of MRD, recurrence, treatment response and patients' stratification.

1. Introduction

Despite the discovery of several molecular mechanisms involved in the genesis/maintenance of colon cancer (CC) and the development of targeted drugs [1], the majority of metastatic patients still dye. This is mostly due to tumors' heterogeneity and evolution [2,3]. Such variability complicates classification and treatment, with tumors invariably adapting to medications by acquiring DNA alterations. Presently, diagnosis involves invasive methods, and longitudinal patient's follow-up relies on suboptimal image exams or biopsies [4]. The advent of liquid biopsy (LB), which consists in the analysis of tumoral analytes in body fluids, holds the promise to revolutionize the diagnosis and management of patients given its low invasiveness, speed, and heterogeneity-inclusive nature. In this review, updated information on epidemiology, subtypes, diagnosis, treatment, and genetics of sporadic CC will be summarized, followed by a discussion on LB types, ctDNA-LB's utility, advantages, disadvantages and impediments to robust clinical implementation, while exploring the latest works/trials involving ctDNA LB in CC.

2. Epidemiology

CC is the fourth most common cancer type, preceded only by breast, lung and prostate cancers (GLOBOCAN). The incidence of CC cases continues to rise due to factors such as increase in population, aging and behavioral changes [[5], [6], [7]]. It varies widely among countries, and, not surprisingly (i.e. aged population/industrialized lifestyle), those with high income accumulate the most important absolute incidence rates (Fig. 1A). In parallel, progresses achieved in the oncology field in the last decades have counteracted the contribution of populational aging, leading to a decrease in the proportion of elderly with CC [[8], [9], [10]] and of mortality (Fig. 1D). Countries such as Moldova, Zimbabwe, Somalia, Libya, Yemen, Paraguay and Suriname, still lag behind in terms of age-adjusted mortality rate (Fig. 1D) despite their intermediate level incidence rate (Fig. 1B), mirroring differences in screening and treatment accessibility. Developed countries, however, exhibit a clear rise in CC in individuals under 50 years old [8,[10], [11], [12], [13], [14]], which might reflect early-life contact with several agents [15] and longer exposure to an unhealthy/industrialized lifestyle [11,12]. The benefits of implementing regular screening programs is supported by studies showing reductions from 17 up to 70% in colorectal cancer (CRC) incidence and mortality rates [[16], [17], [18], [19], [20], [21], [22], [23]]. Indeed, in a 24 year follow-up randomized trial of tens of thousands of individuals screened by FOBT, a two-third decrease in the CC incidence was observed owing to early polypectomy [19], whereas other large studies reported incidence and mortality decreases ≈20–30% [[21], [22], [23]]. Noteworthy, the decrease in screening and CC treatment caused by the COVID-19 pandemic is expected to increase the incidence and mortality of this cancer [24].

Fig. 1.

Colon cancer incidence and mortality heat-maps in different countries in 2022. Darker colors represent higher incidence/rates. A. Estimated crude incidence rates. B. Estimated age-adjusted incidence rates. C. Estimated crude mortality rates. D. Estimated age-adjusted mortality rates. Asterisks indicate countries mentioned in the text. All maps were adapted from maps generated at the IEAC-Global Cancer Observatory site [25]. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3. Colon cancer subtypes and screening/diagnosis

3.1. Screening/diagnosis

CCs are slow-growing tumors which are highly controllable when detected at initial stages, with an overall 5-year survival of 90% [4,21]. However, only about 40% of patients are diagnosed at early stages [21]. Screening and diagnosis relies on colonoscopy, flexible sigmoidoscopy, computed tomography (CT) colonography and fecal-blood detecting tests [4]. While colonoscopy undertaken every 10 years is the gold standard method for individuals over 45–50 [20], it is a costly, unpleasant process (laxatives) which requires sedation/anesthesia and highly specialized personnel. Nevertheless, it is the only technique that allows lesions and polyps removal (curative), being considered a low-risk practice therefore constituting the best prevention method [4,26]. Sigmoidoscopy share the same disadvantages of colonoscopy but only covers the distal colon and demands higher testing periodicity (5 years) [4,26]. CT colonography every 5 years is an option for patients that are at high risk for invasive procedures. While dispensing anesthesia and allowing complete colon examination as well as extra-colonic masses, it involves radiologic exposure, the use of laxatives and insufflation, requires an expert radiologist and is expensive [4]. Fecal blood tests (FIT and FOBT) are low-cost, non-invasive, can be undertaken at home but demand more frequent testing, and usually display reduced specificity/sensitivity, performing better in advanced lesion [4,20,27]. They constitute the most practical first tier choice for screening owing to their low-cost/low-invasiveness/home-collection characteristics, and when positive, require the performance of colonoscopy to confirm/remove lesions. Alternatively to FIT/FOBT, commercially available stool DNA/multitarget tests (e.g. Cologuard, Colosafe, ColoSure, ColoTect, EarlyTect) can be undertaken annually, but high rates of false-positive [28] and aversion to stool-handling represent drawbacks. Other innovative tests/methods such as the plasma-based mSEPT9 (Epi proColon) LB and the capsule endoscopy have been developed, but the latter is not officially recommended for screening whereas the former indication for individuals who decline other screening modalities [20] was dismissed in the latest US Preventive Services Task Force guideline owing to sensitivity issues. Notwithstanding, driven by the abundancy of differentially methylated promoters early in CC [29,30], several alternative methylation-devoted LB panels capable of detecting CC years prior to its establishment have been developed [31,32] and await further validation. Diagnosis characteristics are summarized in Fig. 2.

Fig. 2.

Characteristics of the available colon cancer diagnostic methods.

3.2. Subtypes

The following considerations will be focused on adenocarcinomas owing to their vast dominance (95% of all CC). They arise at the mucosa epithelium of the colon and their differentiation degree depends on the glandular/tubular composition (Fig. 3A–C). Tumors display high variability and can be classified according to several parameters. The TNM classification is widely adopted, helping to establish the extent of CC proliferation and spread. Combined to the degree of differentiation of the cancer cells, this classification helps to establish the tumor grade, treatment strategy and determine the patient's prognosis [33].

Fig. 3.

Examples of histological characteristics of colon tumors stained with hematoxilin-eosin. A. Moderately differentiated adenocarcinoma depicting glands with irregular and tortuous crypts, cells with moderate pleomorphism and anisokaryosis as well as stromal infiltrative pattern. Yellow arrows indicate the stroma between neoplastic nests, a pattern which characterizes infiltrative lesions (100x). B. Poorly differentiated adenocarcinoma presenting signet-ring cells with lateralized nuclei (yellow arrows) (200x). C. Moderately differentiated adenocarcinoma with the presence of areas with large amounts of extracellular mucus (mucin lakes – dashed areas) and glands with stromal infiltrative pattern having irregular/tortuous crypts as well as cells with moderate pleomorphism and anisokaryosis (40x). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Histologically, most adenocarcinomas are classified as “not otherwise specified” (NOS). Mucinous tumors account for 10% of CC, are defined as tumors with >50% area of extracellular mucus (mucin lakes – Fig. 3C) and have a controversial association with poor outcome [2]. Medullary tumors are rare (<0.5%), present poor differentiation and solid sheets of malignant cells with pushing borders associated with lymphocyte infiltration which accounts for its good prognostic [2]. The signet ring cell carcinoma (Fig. 3B) represents another rare subtype (< 1%) and is characterized by >50% of signet ring cells with peripheral nuclei [2]. When mucin-poor, it exhibits an aggressive behavior [2]. Finally, serrated adenocarcinomas (15%) are distinguished by the presence of glandular serration [34].

Colon tumors can be right-sized/proximal (arising at the cecum, ascending or transverse colon, 40%) indicating worse prognosis, or left-sized/distal (localized at the descending or sigmoid colon, 60%) [35,36], with differences attributable to distinct embryonic origin, microbiota, immune milieus and vascular supply [36,37]. Presently, classification by sublocations (cecum, ascending colon etc.) seems more informative [38].

From a molecular perspective, tumors can be considered microsatellite-instability-high (MSI-H), -low (MSI-L) or microsatellite-stable (MSS), with the latter being associated to a worse outcome and representing ≈75% of sporadic CCs [2]. Chromosomal unstable (CIN) tumors, in the other hand, indicate a worse prognostic [39], accumulate chromosomal aberrations/gross structural DNA changes and account for 70% of CC, whereas the CpG island methylator phenotype (CIMP) displaying disseminated promoter methylation represents 20% of cases [40,41]. A consensus molecular transcriptome-based classification comprising 4 subtypes has also been proposed [42], but fails to accommodate a proportion of sporadic CC that display mixed features from all subtypes [43]. Importantly, many of the aforementioned categories can overlap or be found in a single heterogenous tumor complicating classification and management [2].

4. Colon cancer treatment

Localized CCs are usually treated with surgical intervention, with or without adjuvant therapy with curative intent. However, 20% of patients will have metastatic disease at time of diagnosis, and up to 50% of those with localized primary tumors will progress to metastasis, resulting in only 10% overall survival (OS) after 5 years [44]. A minority of metastatic CC (mCC) patients has resectable tumors after initial systemic therapy. Systemic therapy consists of sequential or combined use of cytotoxic chemotherapy such as 5-fluorouracil (5-FU)/capecitabine (usually with leucovorin), irinotecan, oxaliplatin, and other effective fluoropyrimidines [45]. The choice of the drug combination will depend on availability, patient's preference, toxicity and history of prior therapy [46].

First- or second-line treatment for metastatic patients usually consists of a combination of two chemotherapeutic agents plus a biological agent directed against either EGFR (cetuximab, panitumumab) or VEGF (bevacizumab) [44,45]. VEGF-based biological agents are chosen in cases of patients with unknown RAS-status, in the absence of hematological/vascular system/atherosclerotic/proteinuria diseases/scheduled surgeries [44,45]. On the other hand, anti-EGFR is associated to better outcome, especially in KRAS wild-type individuals with left-sided tumors, and is the standard choice for treatment [45,47]. In maintenance therapy, a combination of both biological agents is also possible.

Subsequent lines of treatment will depend on the patient's previous therapy and general clinical condition, but cytotoxic chemotherapeutic drugs and biological agents can be interchanged, therapeutic regimens can be altered, and combined-salvage-compounds can be necessary in refractory metastatic patients [48]. However, each subsequent line of treatment usually produces inferior rates of responses with shorter disease-free progression times. In the specific case of patients with advanced MSI tumors refractory to chemotherapy, immune checkpoint inhibitors (pembrolizumab, nivolumab, atezolizumab, ipilimumab) are usually adopted and can achieve durable responses [44,45,49]. Immune checkpoint inhibitors are also indicated for tumors with high mutational burden (>10 mut./mega base) [50].

5. Genetics of colon cancer with treatment implications

5.1. Microsatellites and polymerases

Microsatellites/short tandem repeats are ubiquitous genome DNA repeats of 1–6 nucleotides which are replication error-prone [51]. While in healthy individuals, repair mechanisms revert such errors, some cancer patients accumulate uncorrected microsatellites indels mainly through methylation or mutation of mismatch repair genes (MLH1, MSH2, MSH6, PMS2) [52]. Patients with indels in >30% of microsatellites are considered microsatellite unstable (MSI) and have better prognostic partly due to the frequent production of neoantigens [51]. Such patients, which account for 15% of CC, usually benefit from immunotherapy [51]. The same is true for patients with hypermutated tumors caused by mutations in polymerase genes (POLD1, POLE) [[53], [54], [55]].

5.2. TP53

TP53 is one of the most commonly affected genes in this disease, being altered by either point mutation, indels or gene deletion in over 50% of patients [[56], [57], [58], [59]]. Mutations of this gene usually occur between exons 5–8 and seem to contribute to the late stage of cancer development [59]. Its prognostic value is hard to establish [57,58,60] due to confounding effects such as differences in cohort and detection method. Notwithstanding, alterations in TP53 seem to negatively affect responses to 5-FU [60,61], constituting an important marker for the choice of chemotherapeutic backbone.

5.3. RAS and BRAF

Mutations in KRAS, a proto-oncogene encoding a kinase belonging to the MAPK/ERK pathway, are also frequent, being present in around half of CC cases [56,[62], [63], [64]]. Codons 12/13 concentrate 90% of KRAS activating mutations, while the remaining occur mainly at positions 61, 117 and 146 [62,64]. The homolog NRAS gene, which takes part in the same pathway, can equally be mutated in codons identical to KRAS, although at a much lower incidence (<10%) [63,65]. Though less frequently, amplification of both genes also occur [66]. BRAF constitutes an additional downstream MAPK/ERK kinase, whose gene presents relevant overactivation modifications (mostly p.V600E). Mutations, but not amplification, are observed in up to 15% of patients [62,67]. Importantly, alteration of all three genes predict poor prognostic [56,[68], [69], [70]] which is at least partly due to resistance to anti-EGFR treatment [45,69,[71], [72], [73], [74], [75]]. Despite KRAS mutation/amplification's association with poor outcome, patients with these alterations seem to be good responders to the FOLFOX combination [36]. Finally, some rare non-classical BRAF mutations (p.G469A, p.L485F, p.L525R, p.L597Q, p.V600R, p.K601E) that behave like p.V600E [[76], [77], [78]] should ideally be considered when stratifying for anti-EGFR treatment.

5.4. PIK3CA and PTEN

Another key kinase affected in CC is the PI3K, central to the PIK3/AKT pathway, and whose alpha catalytic subunit encoded by the gene PI3KCA appears mutated in approximately 15% of patients [57,58,62]. The mutations are concentrated in exons 9 (specially codons 542, 545) and 20 (codon 1047) and activate the pathway ultimately promoting oncogenesis [81]. In this same pathway, the PTEN gene, which encodes the phosphatase that opposes the phosphorylation of PIK3, displays inactivating mutations and deletion in around 4 and 20% of patients, respectively [62,79,80]. Once more, association between alterations of both genes and prognostic remains debatable [56,81,82] for the same reason mentioned previously for TP53. Although inconsistently, both genes have been associated to resistance to anti-EGFR [58,68,74,75,83,84]. Indeed, a definitive conclusion is hampered by their frequent combined occurrence with mutations of genes conferring anti-EGFR resistance [72,85]. For PTEN-altered tumors, the combination of anti-VEGF and mTOR inhibitors rather than anti-EGFR seems preferable [86,87].

5.5. APC

In the Wnt/β-catenin pathway, APC displays point mutation, insertions and deletion that arise early in up to 80% of CCs [88]. The inactivating mutations are enriched in exon 15 and commonly accompanied by loss of heterozygosity [88]. In this pathway, APC negatively regulates the levels of β-catenin, inhibiting proliferation while promoting differentiation [88,89]. It also binds to multiple partners outside this pathway, modulating adhesion, migration and chromosomal segregation [88]. Interestingly, truncated APC resulting from STOP-codons-generating mutations/indels exhibit a gain of function, fostering resistance to apoptosis, proliferation, migration and invasion [88]. Similar to TP53, PIK3CA and to PTEN, APC's prognostic value remains elusive [90,91], but its modifications are associated to poor response to 5-FU [92,93] and resistance to any chemotherapy that exploits base excision DNA repair (BER) [94] or pro-apoptotic drugs [95], justifying monitoring of APC's modification.

5.6. Other genes

Apart from classical genes, relevant modified genes have been uncovered in CC by whole genome, exome or targeted sequencing. Among these, some recurrently reported include CTNNB1 (β-catenin) [35,56,57,79], AMER1/FAM123B [57,79], SOX9 [57,79], AXINs [57,79], RNF43 [57], AKT1 [35,36,56], MAP2K1 (MEK1) [35,36,56,96,97], ERBBs [35,36,56,57,66,72,75,79,[96], [97], [98]], MET [75,[96], [97], [98], [99]], TGFBR2 [36,57,79], SMADs [36,56,57,79], ATM [79], FBXW7 [56,57,79] and ARID1A [57,79].

SMAD4-altered tumors show poor response to 5-FU [100,101] which should be avoided in mutated-individuals. As expected for effectors downstream of the EGFR receptor, mutations (p.K57T, p.L115P and p.V211D) in MAP2K1 (MEK1) were associated with anti-EGFR acquired resistance [71,102]. Moreover, in patients undergoing anti-EGFR therapy, appearance of ERBB1 (EGFR) mutations that prevent binding of anti-EGFR antibodies to EGFR extracellular domain are also indicative of secondary resistance [75,96,103], and should be monitored for therapy switch. Interestingly, tumors with ERBB1 (EGFR) mutations p.S492R, p.K467T and p.R451C still respond to panitumumab [72] rendering it an alternative for these EGFR-mutated patients. Other genes whose modifications induce resistance to anti-EGFR include ERBB2 (HER2) [75,83,96,103,104] and MET [75,96,103], for which dual HER blockade [97,105,106] and dual MET inhibition [99,107] are indicated, respectively. Importantly, the accumulation of up to 13 different DNA alterations linked to anti-EGFR resistance has been reported in a single patient under treatment, with high heterogeneity in the modifications’ profiles among individuals [75,103]. Such heterogeneity and co-occurrence of resistance-associated DNA alterations (polyclonal rather than single-mutation-resistant clone) represent a tremendous challenge for implementing successful therapeutic strategies capable of overcoming resistance. Importantly, the observation that particular tumors from a patient continue to respond to therapy even when others acquire resistance [97] imply that suspension of drugs should be carefully evaluated. Likewise, failure in identifying resistance-associated mutations in different tumors while considering solely the primary tumor molecular landscape allows disease to progress [102], indicating that multiple drugs targeted to the mutation ensemble would be beneficial whenever toxicity permits. Genetic alterations relevant for patient management are summarized in Fig. 4.

Fig. 4.

Main localized genomic alterations with impact in colon cancer first-line and anti-EGFR-resistant patients' treatment. All gene alterations relevant for treatment are depicted in the central boxes, and their relationships with poor and good responses to distinct compounds are displayed at the right and left boxes, respectively. Brown-shaded boxes relate to genes associated to poor response to anti-EGFR, green-shaded boxes relate to those linked to poor response to 5-FU, whereas blue-shaded boxes relate to genes linked to good response to immunotherapy. APC: Adenomatous polyposis coli; BRAF: V-Raf murine sarcoma viral oncogene homolog B; EGFR: Epithelial growth factor receptor; ERBB: Erythroblastic leukemia viral oncogene homolog receptor; HER2: Human epidermal growth factor receptor 2; MAP2K1: Mitogen-activated protein kinase kinase 1; MET: MNNG HOS transforming gene; MLH1: MutL homolog 1; MSH2/6: MutS homolog 2 or 6; PIK3CA: Phosphatidylinositol-4,5-bisphosphate 3-kinase Catalytic subunit alpha; PMS2: Post meiotic segregation increased 1 homolog 2; POLD1: DNA polymerase delta 1; POLE: DNA polymerase epsilon; PTEN: Phosphatase and tensin homolog; RAS: Rat sarcoma 2 viral oncogene homolog; SMAD4: Mothers against decapentaplegic homolog 4; TP53: Tumor protein 53; VEGF: Vascular endothelium growth factor; HER2i: HER2 inhibitor; METi: MET inhibitor; mTORi: mTOR inhibitor; 5-FU: 5-Fluorouracil; FOLFOX: Folinic acid + 5- Fluorouracil ​+ ​Oxaliplatin. ∗Conflicting reports - no consensus reached. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

6. Liquid biopsy

6.1. Definition, advantages/limitations and types

Liquid biopsy (LB) is broadly defined as any detection method capable of identifying/characterizing tumor presence using liquid human samples such as blood, saliva, urine, nasal secretion, seminal/lymphatic/cerebrospinal/pleural/peritoneal/lymphatic fluid and maternal breast milk. Blood is the most exploited substrate, probably due to its low invasiveness and extensive perfusion reach (contact with cancer in any organ). Analytes which are exclusive to cancers are considered superior to markers produced in different amounts (e.g. CAE), owing to the difficulty in establishing a reliable threshold distinguishing healthy from diseased individuals [[108], [109], [110]].

Solid biopsies are considered the gold-standard method for tumor malignancy confirmation and staging, and treatment decision is often based on data from this single biopsy. Nevertheless, tissue removal is a painful, risky, expensive, and time-consuming procedure that cannot be performed in either inaccessible or minute tumors, fragile patients or longitudinally. LB provides the opportunity to repeatedly screen a readily accessible biological source for cancer-derived markers in a rapid, simple, safe, and less expensive way. Indeed, most body fluids are collected through minimally invasive procedures allowing monitoring tests to be performed more often, with wide implications for tracking tumor's evolution, relapse, metastasis, and treatment response. Furthermore, while solid biopsies only account for part of the often-heterogeneous tumor, LB (originating from all tumors in their totality) provides a more comprehensive picture of the patient's disease. Finally, LB eliminates the need for surgical facilities, highly specialized personnel and, most importantly, accelerates access to test results allowing treatment decisions to be made earlier. Therefore, LB represents an extremely powerful complimentary tool to tissue biopsy.

Among the most popular LB analytes, circulating tumor cells (CTCs) and circulating tumor exosomes (ctExos) represent complex multitarget entities [111]. CTCs are cancerous cells believed to colonize metastatic niches to produce new extracolonic tumors [112], whereas ctExos consist of small vesicles of endosomal origin secreted by cancerous cells for communication, implicated in several tumoral processes [113]. While CTCs are cultivable and exploitable for personalized drug screening [112,114], they arise mostly at later stages [115] at minute proportions and display challenging isolation [116]. Likewise, low yields, high variability and impurities inherent to the isolation of ctExos still limit their clinical application [117].

All cells release DNA in body fluids in either a passive or active way, giving rise to cell-free DNA (cfDNA). In addition to CfDNA from healthy cells, cancer patients possess a variable proportion (0.003–95% [118]) of modified cfDNA originating from tumor cells, named circulating tumor DNA (ctDNA). Such ctDNA constitutes another LB substrate whose attractiveness relies on its simplicity of isolation, high specificity, and stability after purification. Interestingly, sensitivity of detection of cancer-specific mutations has been reported to be higher when using ctDNA than CTCs [115,119] although this is debatable. Among solid tumors, CC displays one of the highest ctDNA levels [120], which renders it particularly suited for ctDNA-based LB. CtDNA has been successfully used for the detection of minimal residual disease (MRD) and recurrence following colonic tumor resection surgery [121,122], as well as for predicting treatment response, disease-free survival and OS [107,[123], [124], [125], [126], [127]], even correlating tumor size variation with concordant fluctuations of ctDNA levels [57,97,128,129] weeks to months earlier and with superior accuracy than radiological images [75,121,122,[129], [130], [131]]. Also, several groups have demonstrated the detection of mutated ctDNA in patients with tumors tested negative for the same mutations [75,96,102,107,115,122,123,132,133], evidencing ctDNA's superiority in the cases of heterogeneous tumors with single limited solid biopsy sample availability. Low amounts of ctDNA at the initial stages of CC can, however, limit its broad application in early-stage patients, with only 34% detection rate for stage I CRCs and differences in ctDNA levels of up to 500 fold [123]. Notwithstanding, novel ctDNA methylation-based assays indicate that early CC detection by LB can eventually become a reality [32,134], with exciting implications for CC incidence and mortality. Differently from mutations/CNVs which arise in many genes or only occur in a subset of CC patients, disease-related promoter methylation arises in a limited set of genes in a fairly consistent way among patients, holding the promise for a universal CC test. Also, methylated ctDNA levels seem to be higher than their mutated counterparts, which would facilitate their adoption in patients with small/early tumors [135]. Lastly, ctDNA has been effectively explored for MSI-status characterization, showing similar or superior performance compared to traditional methods [[136], [137], [138]], which is significant for patients with unavailable/insufficient/bad quality tissues or with etiologies other than MMR gene silencing that could not have their MSI status defined otherwise.

Other promising blood markers for CC detection and management have been described. Most of them are aimed at diagnosis as expected if considering the low adherence to available tests, while some are devoted to predicting prognostic and treatment response. Not surprisingly owing to the involvement of immunity in CC, several immune-associated molecules have been proposed as CC markers, such as GDF15 [139] and IL-8 [140], but display ≤70% sensitivity. More studies are needed to confirm GDF15 as an indicator of liver metastasis. The best performing protein markers include TFF3, CA11-19 and the panels IGFBP2/DKK3/PKM2 and DC-SIGN/DC-SIGNR, most of which were associated to CC establishment/dissemination [140]. Although displaying superior sensitivity than immune-related markers, their capacity for discriminating polyps was poor and the number of subjects analyzed modest. Among miRNAs, mi-1290 showed 70% sensitivity/91% specificity in detecting CRC but failed in adequately distinguishing adenomas from healthy subjects, whereas the panel consisting of mi-24/mi-320a/mi-423-5p exhibited suboptimal specificity [140]. Long non-coding RNAs NEAT_v1 and 2 also exhibited sensitivity ≈70% although with high specificity but were unable to distinguish early from late CRC [140]. Additionally, the transcript SALL4 demonstrated over 90% sensitivity/specificity but lacked comparison to polyps and was probed in a small cohort [140]. Noteworthy, a limitation of all these markers if compared to mutated ctDNA is their reliance on quantitative thresholds influenced by several additional conditions. Moreover, larger confirmation studies with a better design must be undertaken to validate the most promising indicators.

6.2. Liquid biopsy using ctDNA for CC management

Next generation sequencing (NGS), digital PCR (dPCR) and real-time PCR (qPCR) constitute the foundation of ctDNA LB. Contrarily to qPCR, dPCR permits a heterogeneous sample to be partitioned in a tremendous number of sub-samples analyzed separately [141], abruptly improving rare events' detection sensitivity even in inhibitory backgrounds, while allowing its quantification in the absence of standard curves, based on the Poisson's distribution. Several NGS methodologies exist, all based on the parallel sequencing of millions of small DNA fragments followed by in silico analysis to identify targets while eliminating noise.

Even enhanced qPCR-based methods (e.g. E-ice-COLD, Idylla, COBAS) tend to perform worse if compared to dPCR [[142], [143], [144]]. If contrasted to NGS, dPCR has an equal or superior sensitivity, also produces quantitative data, is cheaper, generates faster results easier to analyze, but is limited to the evaluation of only a few known simultaneous “hot-spot” alterations [123,133,145,146]. Importantly, NGS is the only method capable of estimating tumor mutational burden, predict primary tumor origin [147], and to analyze thousands of alterations of distinct nature in a single assay, although it usually requires higher DNA input. Differently from dPCR, it can also comprehensively access alterations in tumor-suppressor genes which can be modified in many positions. Moreover, since metastatic masses not always preserve alterations from the primary tumor [97], dPCR limited target capacity can preclude longitudinal patient's evaluation. A comparison of dPCR and NGS is available in Table 1.

Table 1.

Comparison of NGS and dPCR characteristics. CNV: copy number variation; MRD: minimal residual disease; MSI: microsatellite Instability; VAF: variant allele frequency.

PARAMETERS	TECHNOLOGY
	cfDNA NGS	cfDNA dPCR
Costs	Medium to high	Medium
Labor-intensive protocol	Yes	No
Analysis complexity	High	Low
Time to results	Weeks	Days
Sensitivity	Lower to equal (from 0.02% VAF, most from 0.1% VAF)	Higher to equal (from 0.001VAF, most from 0.01% VAF)
VAF Output	Yes	Yes
Target breadth/multiplexability	Hundreds to thousands simultaneous	A few to tens simultaneous
Compatible with identification of targets of distinct nature in a single assay (e.g.mutations, indels, CNV, translocations, methylations)	Yes	No
Need for prior alteration knowledge	No	Yes
Comprehensive tumor characterization performance	Good	Poor
Performance for tumor suppressors	Good	Poor
Metastasis identification/monitoring performance	Good	Medium to good (depends on alteration's choice)
Diagnostic/Screening performance	Medium	Medium
Tumor Mutational Buden performance	Good	Unsuited
MSI performance	Good	Good
Treatment response performance (VAF fluctuations)	Good (for >1% VAF, but improving)	Good
Classical acquired resistance monitoring performance	Good	Good
Novel resistance-associated alterations discovery performance	Good	Unsuited
Patients' triage performance (e.g. therapy non-responders, benefiting from adjuvant therapy)	Good	Good
MRD performance	Variable (assay' sensitivity-dependent)	Good
Identification of best up-to-date medication	Supported (report consensus)	Not supported (no report consensus)
Commercial/accreditedtests options	Several assays/platforms available	Only BEAMing coupled to few assays

In CC, NGS-based LB has been pervasively used. KRAS mutations were uncovered by NGS in CC patients considered as wild-type by tissue analysis, illustrating its advantages for stratification of patients with heterogeneous tumors [102,148]. Incongruencies between plasma and tissue NGS were also observed by others [102,107,115,123,133]. The utility of NGS for determining initial tumors' genomic profile has also been established [57,76,121,123], with NGS combination to dPCR enhancing LB’ sensitivity [123]. Plasma NGS has been successful for treatment monitoring/identification of CC MRD/early relapse/therapy response [57,97,130,149] ahead of image exams [97], prediction of PFS and OS [125,126], perceiving unknown acquired resistance alterations [57,75,76,96,102,107] and unknown actionable mutations [57,150]. Satisfactory evaluation of MSI [136,138] and use of ctDNA fragmentomics for early CC detection [151] were also achieved.

Continuous measurement of the frequency of tumor-associated alleles can predict poor outcome when reflecting overgrowth of an aggressive tumoral sub-clone/tumor growth, spread or indicate allele amplification. Coupled to tumor characterization by NGS at baseline and resistance acquisition, monitoring of an aggressive metastatic disease linked to the appearance of a rare anti-EGFR resistance allele (BRAF p.L597Q) was achieved by plasma dPCR [76]. The emergence and steady rise in the allele frequency of KRAS p.Q61H in a distant progressing metastasis was accessed by plasma dPCR and reflected resistance to the dual panitumumab/trametinib therapy [102], while in BRAF p.V600E mutated patients receiving BRAF inhibitor combined to anti-EGFR, mutant allele frequency (MAF) > 2 was linked to worse progression-free survial (PFS) and OS and established as an indicator of patients more likely to benefit from intensification therapy with MEK inhibitor [152]. Follow-up of patients undergoing intermittent cycles of anti-EGFR treatment using dPCR also demonstrated KRAS mutation appearance in plasma at disease progression and MAF decrease over therapy modification, validating the positive impact of re-challenging patients that were previously resistant to anti-EGRF [75,96], which was confirmed by clinical trials [153,154]. Conversely, the identification of “EFGR-holiday” patients unresponsive to re-challenging in clinical trials was also possible [155]. Cetuximab resistance was also anticipated by 2 months when utilizing plasma dPCR targeted against MET and ERBB2 (HER2) amplifications, whose levels reflected tumor growth [75,129]. DPCR was equally employed for identification of FGFR1 amplification in patients’ plasma, which was associated to a poor outcome [156]; for detection of EGFR ectodomain mutations, ERBB2 and MET amplifications [96,98] and to monitor the response to panitumumab/trametinib by analyzing the MAF of MAP2K1 (MEK1) p.K57T [102]. Takahashi and colleagues have evidenced by dPCR that the threshold of mutations conferring resistance to anti-EGFR therapy might be relevant once half of the patients displaying <0.5% MAF still responded to treatment [157]. Silveira and colleagues developed drop-off dPCR assays against three microsatellites (BAT-26, ACVR2A and DEFB105 A/B) and applied them to blood-derived cfDNA from 14 MSI-high, 28 MSS and 48 healthy individuals, achieving impressive 100% specificity and 100% concordance with microsatellite status determined by other methods [137]. Other than that, the sizes of the cfDNA fragments have also been successfully used for recurrence prediction, with short fragments being indicative of relapse [133]. Finally, several additional groups demonstrated the applicability of plasma dPCR for prognosis, treatment response and MRD [121,132,158], including works using ctDNA methylation [131].

Regarding the limitations of LB, it has been suggested that CC patients with lung and peritoneal metastasis might be more susceptible to false-negative results and display lower VAF [159,160]. Moreover, false positive results are difficult to distinguish from true positive when blood alterations cannot be confirmed in the tumor tissue or excluded to originate from clonal hematopoiesis.

6.3. Considerations for ctDNA clinical implementation

6.3.1. Experimental considerations

One of the main limitations to the adoption of LB in the clinics is methodological standardization and adequation to hospitals’ routine. For sample collection, blood mononuclear cells’rupture/ctDNA degradation must be prevented and prolonged collection-to-processing timeframes must be allowed. Efforts have been made to establish standards, concluding that for clinical routine, generous sample volumes should be collected in “Streck” tubes to obtain plasma by 2-step centrifugation before 7 days, while preservation of buffy coat for eliminating clonal hematopoietic alterations should be undertaken [[161], [162], [163]]. Storage of aliquoted cfDNA at −80 °C is recommended for long periods, and quantification should ideally be performed with qPCR, capillary electrophoresis or fluorimetry [163]. Also, sample collection time must be carefully chosen taking into consideration measurements goals, once surgery, radio and chemotherapy modify ctDNA levels [163].

In the clinical setting, automated devices for both DNA extraction, test running, and analysis would be preferable (decreased error/faster results). Notwithstanding, the consensus is that the manual QIAamp Circulating Nucleic Acid kit delivers the highest cfDNA yields [[164], [165], [166], [167]]. Among automatic extraction devices, The Qiagen EZ1&EZ2 ccfDNA Kit coupled with the EZ2 instrument showed good results [168], which awaits confirmation.

Performant “plug-&-play” execution/analysis platforms compatible with high-throughput assays at an affordable cost are ideal. Presently, Food and Drug Administration (FDA)/Conformité Européenne (CE)-endorsed qPCR platforms for CC include the Epi proColon (mSEPT9), Idylla (BRAF/RAS, MSI) and COBAS (KRAS, PIK3CA, BRAF p. V600W), while many others are Laboratory Developed Test (LDT)/CE-marked only. COBAS and Idylla were only approved for tissue but have been used for ctDNA assessment showing contradictory (Idylla) [142,169] and worse (COBAS) [142] performance compared to dPCR. Three NGS assays aimed at advanced solid cancers' treatment-guidance (Guardant360 CDx, Foundation One Liquid CDx, Tempus xT CDx) have been approved by the FDA and CE-marked while the Guardant Reveal (blind) and Signatera (tumor-informed), directed at CC MRD detection, received FDA Breakthrough Device Designations and a CE mark (Signatera). Moreover, the Shield assay is under premarket approval evaluation as a CC screening test, whereas the pan-cancer MSK-IMPACT assay has been approved by New York State. Other CE-marked or LDTs assays commercially available include Galleri (pan-tumor characterization/tumor origin), AlphaLiquid100, Follow it (pan-tumor characterization), FoundationOne Monitor, Cancertrack, Caris Assure, CellMax-LBx (monitoring), Foresight Solid Tumor Recurrence, FoundationOne Tracker (MRD), CancerSEEK (early detection) and ColoNext (CC risk), illustrating the importance of NGS for LB. Regarding dPCR, only the BEAMing (magnetic partitioning) OncoBEAM RAS or RainSure RAS/BRAF assays have been certified (CE and/or CLIA). In one CC study, BEAMing dPCR was considered superior to droplet dPCR, probably due to the higher cfDNA volume used (123 x 8uL) [170]. The scarcity of cfDNA recovered from each plasma sample and the multitude of alterations with therapeutic implications for CC necessitate the development of dPCR multiplex assays and equipments with an increased number of detectors for clinical validity. The recent launching of dPCR devices (e.g. Neo, QX600, LightCycler2.0, QIAcuity, AbsoluteQ) with relevant clinical-compatible features should facilitate dPCR-based LB transition to the clinics. Innovations include up to 7 detection channels, automated workflows, superior input capacities (i.e. >ctDNA) and reaction speed, which should increase assays' multiplexing, reproducibility, sensitivity, partitions' integrity, and time-to-result. Using platforms with distinct input capacities, Crucitta et al. observed that QIAcuity outperformed QX200 in CC [171], but since an equal input (4uL) has been used for both equipments, differences might rather reflect partitions’ properties.

Another limitation hampering LB’ s clinical adoption consists of its low sensitivity reflecting the scarcity of plasma ctDNA. CtDNA enrichment steps (e.g. pre-amplification steps, enriching for smaller/altered fragments, inhibiting wild-type alleles amplification, enzymatic) represent a wet-lab-possibility to increase sensitivity [172]. Several groups have demonstrated the advantages of such approaches resulting in increased VAF/detectability of alterations [[172], [173], [174], [175]]. Higher ctDNA recovery through in vivo blockage of DNA clearance has also been achieved [176]. Increasing coverage depth represent an alternative way of enhancing sensitivity for NGS assays [177]. For dPCR, combining probes at different concentrations [178], using modified probes [179] or designing drop-off assays could expand multiplexability [180]. Indeed, successful drop-off dPCR assays have been described for the detection of 99% of pathogenic RAS/BRAF/PIK3CA alterations from CC [181], as well as mediator probe assays [182] and a probe concentration-gradient-based test [183] for KRAS/BRAF and KRAS identification, respectively.

Standardization and improvement of the analysis step are also essential before LB’s clinical use, by which solid bioinformatic and artificial intelligence tools must be validated in order to establish optimal calling thresholds. The adoption of cut-offs not too close to the limit of detection and in silico size-filtering of NGS reads would help to avoid false positive, preventing overdiagnosis/patient distress whereas improvements in allele calling algorithms would contribute to minimizing false negative results. Noteworthy, the limits, advantages and more indicated application of each assay/platform should be clearly determined, considering its regulatory and sensitivity implications (e.g. screening, early-stage and MDR detection require higher sensitivity).

Finally, efficient communication between test providers/oncologists, creation of working groups with important players (e.g. BloodPAC), educational programs to form broadly skilled professionals are needed as well. Much of the afore mentioned initiatives are underway, such as the European Liquid Biopsy Academy training program, guidelines publications from precision medicine working groups [163,184] and in silico solution development [[185], [186], [187], [188], [189]].

6.3.2. Regulatory considerations

Regulatory environments are also paramount for the adoption of LB into the clinics. In the European Union, LB tests’ regulation follows the requirements of the in-vitro diagnostics (IVD) directive, which exempt them from a Notified Body certificate for marketing, but this might change [162]. While In the United States, LB assays/platforms also follow under the IVD umbrella, most LB platforms/assays must face more tightly regulated sub-pathways, namely LDT and PMA. However, owing to their potential to improve treatment/diagnosis of a life-threatening disease, some LB products might be considered under the expedited breakthrough device designation (BDD) pathway [190]. Faster pathways should undeniably accelerate clinical use of LB.

Having said so, the lack of concordance observed among commercial LB assays has prompted regulatory agencies to plead for increased supervision to preserve patients' safety. If approved, such measures would limit patients’ access to LB results (especially in-house tests developed by healthcare laboratories that rapidly adapt to evolving biomarkers) and slowdown LB clinical integration.

6.4. Clinical trials involving the use of ctDNA-based liquid biopsy

Several clinical trials have been evaluating the utility of ctDNA for diverse purposes. The NCT01983098 has validated the prognostic role of ctDNA in mCRC patients treated with regorafenib, concluding that plasma baseline tumor fractions >5% were strong predictors of reduced OS [191]. Reinforcing the prognostic value of ctDNA, the VALENTINO study observed a poorer OS in RAS wild-type mCRC patients with high VAF of patient-specific genes previously detected by NGS [192]. The GALAXY trial also observed poor prognostic in resected stage I-IV CRC patients with post-surgical ctDNA (MRD), evidencing the value of ctDNA as a stratification marker for adjuvant therapy [193], while DYNAMIC demonstrated the power of ctDNA in selecting stage II CRC patients who could be spared from additional chemotherapy in hundreds of individuals recruited from 23 hospitals [194]. In the specific case of KRAS-mutated ctDNA, its negative prognostic value has also been demonstrated in mCRC patients under anti-EGFR-based therapy in several clinical trials [[195], [196], [197], [198]].

Regarding the application of ctDNA as a marker of patient selection, NCT01442935 showed that patients positive for ctDNA before liver metastasis removal lived at best for 24 months, while 30% of ctDNA-negative were still alive after 48 months [199], indicating that the latter group benefits from surgery. Sartore-Bianchi and collaborators have recently published the results of the trial NCT03227926/CHRONOS in which they reported the positive impact of using ctDNA to guide rechallenge of RAS/BRAF/EGFR wild-type mCRC patients with panitumumab [200]. Indeed, 30% and 63% of patients selected based on ctDNA profile displayed partial response and disease control, respectively.

A 21% response and 54% disease control rate was observed in mCRC RAS/BRAF wild-type individuals who became resistant to cetuximab and underwent 3rd line cetuximab plus irinotecan rechallenge in the NCT02296203 trial [153]. In this work, evaluation of ctDNA status during treatment demonstrated that patients with KRAS-mutated ctDNA had shorter PFS and that mutation was absent in cfDNA from responders, identifying sequential cfDNA interrogation for KRAS status as a measure of rechallenge effectiveness. Results of the CAIRO5 trial established that following therapy, a decrease of at least 98% in the frequency of the highest-mutant-ctDNA-allele-before-treatment correlated with a longer OS in mCRC patients, and validated LB for the detection of resistance-related ctDNA alterations [201].

Impressively, the PROSPECT-C trial has evidenced how integrating mathematical modeling to frequent cfDNA sampling can result in accurate prediction of treatment failure time [75]. Taken together, these results, although from a still limited number of clinical trials, endorse the power of ctDNA as a LB marker for prognostication, therapy response, resistance detection and patient selection.

Regarding new or ongoing trials, a selection of the most recent protocols registered at https://clinicaltrials.gov/ that intend to evaluate the impact of ctDNA-based LB for CC management were summarized in Supplementary Table S1. Importantly, there is an impressive number (around 100) of clinical trials directed at CC which use ctDNA-based LB as outcome measure, highlighting the enormous potential of this analysis. Most measure ctDNA using a panel assay to predict relapse/MRD, response to a certain procedure/pharmacological regimen, de-escalation therapy (ctDNA-) or earlier intervention (ctDNA+) suitability. Of note, several trials aiming to validate ctDNA methylation and fragmentation patterns for early diagnosis are underway (NCT05508503; NCT05485077; NCT05633342/CADENCE; NCT05587452/RECOMMEND; NCT05431621), as well as for LB's validation for the determination of CC tumor MSI status (NCT03594448; NCT03561350).

7. Conclusion and future perspectives

Available data indicate that the recurrent screening of alterations in genes proven to influence treatment response would positively impact CC patients' management and should be adopted in the clinical practice worldwide as soon as methodological standardization, clinical trial-proven benefits and regulation compliance are achieved. Despite insipient levels of ctDNA in early-stage CC, blood-based novel ctDNA-methylation-directed assays hold great potential as screening tools. DPCR-based LB represents a simple, rapid, and affordable way to determine MSI status and to monitor disease progression, therapy response and MRD that outperforms image-based follow-up and is particularly suited for large and metastatic tumors, whereas NGS-founded LB has the capacity to simultaneously determine a multitude of alterations independently of prior knowledge. In the future, more efficient therapies might be tailored to each individual's molecular landscape and modified according to the tumor's molecular evolution on the basis of LB. Although dPCR-based LB is limited to a few known alterations, it can still positively affect decision making while its NGS counterpart has a unique role in determining the tumor genetic landscape and identifying new targets, including non-classical resistance-associated alterations, constituting complementary approaches. Importantly, methodological standardization must be consolidated, and efforts to increase dPCR sensitivity and multiplexability must be undertaken for dPCR to achieve its full potential. For NGS, cost-effective panels directed to mixed nature-alterations and reliable thresholds based on AI-guided algorithms must be standardized. While ongoing clinical trials will further assess ctDNA value for CC patients' management, existing expedited regulatory pathways should promote a faster adoption of LB in the clinics, a scenario that might change soon.

Funding

This study was supported by the Brazilian agencies Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq Universal 427912/2018-0) and Fundação de Amparo à Pesquisa do Rio de Janeiro (FAPERJ 25191).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: