Challenges and proposed solutions to the adoption of cell free DNA in screening, detecting and prognosticating colorectal cancer

Abstract

Detection and treatment of colorectal cancer (CRC) at an early stage is vital for long-term survival. Liquid biopsy has emerged as a promising new avenue for non-invasive screening of CRC as well as prognostication and surveillance of minimal residual disease. Cell free DNA (cfDNA) is a promising liquid biopsy analyte and has been approved for use in clinical practice. Here, we explore the current challenges of utilizing cfDNA in the screening and prognostication of CRC but also for detecting driver mutations in healthy, presymptomatic patients with normal colonic crypts. CfDNA for the detection of cancerous or premalignant colonic lesions has already been extensively explored, however few have considered utilizing cfDNA in the detection of driver mutations in healthy patients. Theoretically, this would allow us to detect patients who are at a higher risk of tumorigenesis decades in advance of established malignancy and stratify them into higher risk groups for early-intervention screening programs. We also explore the solutions necessary to overcome the challenges that prevent liquid biopsy from entering mainstream clinical use. The potential for liquid biopsy is immense if these challenges are successfully circumvented, and can dramatically reduce CRC rates as well as improve survival in patients.

Key Words: Cell free DNA; Circulating tumour deoxyribonucleic acid; Colorectal cancer; Liquid biopsy; Screening; Detection; Prognostication

Core Tip: Cell free DNA (cfDNA) carries information about colorectal cancer-specific genetic and epigenetic alterations which can aid in screening, detection and prognostication. Detection of genetic alterations is made difficult by low signal-to-noise ratio owing to an abundance of background non-tumorigenic mutations. Low amounts of cfDNA in healthy individuals negatively affects the sequencing performance and limit of detection of assays making screening non-feasible. One solution is to harvest cfDNA from peritoneal fluid or stool as this is more representative of the primary tumour compared to plasma-derived cfDNA. Alternatively, increasing the sensitivity of sequencing technologies would allow for the detection of low frequency mutations.

INTRODUCTION

Colorectal cancer (CRC) is the second leading cause of cancer death worldwide, with a lifetime risk in average individuals of approximately 4% to 5%[1]. Novel approaches in screening and detection of CRC are needed to reduce overall mortality. Current approaches in the detection of CRC include modalities such as endoscopy and faecal occult blood tests. Liquid biopsy-based approaches are a promising avenue for non-invasive cancer detection, prognostication and surveillance. Liquid biopsy is a molecular-biological diagnostic approach for detecting significant tumour-derived markers in bodily fluids without the need for invasive tissue biopsy. This includes circulating tumour cells (CTCs), cell-free DNA (cfDNA), mRNA, microRNA, exosomes, nucleosomes, and various glycoproteins and antigens[2]. Although each liquid biopsy analyte has unique advantages and disadvantages with regards to detection, this present review will focus on cfDNA, which are DNA fragments released into circulation during cellular apoptosis and necrosis[3].

This review focuses on cfDNA because it is an attractive candidate for non-invasive cancer detection. CfDNA carries information about cancer-specific genetic and epigenetic mutations[4]. CfDNA levels during treatment have been shown to correlate with oncologic outcome and provide direct evidence of minimal residual disease (MRD) in patients post-surgical resection[5]. Some studies report cfDNA to outperform imaging modalities like computer tomography in the detection of recurrent tumour[6]. Moreover, cancer patients’ blood has been shown to contain increased levels of cfDNA compared to that of healthy individuals[7]. The analysis of cfDNA methylation patterns has also shown to successfully detect a wide range of cancers with specificity and sensitivity performance approaching the standard for population-level screening[8]. However, few studies have explored the use of cfDNA analysis in healthy, presymptomatic individuals. This is a novel concept that is difficult to implement for a few reasons: Lack of knowledge regarding the molecular basis of tumour initiation, significantly lower cfDNA concentration in the plasma of healthy individuals compared to cancer patients[9], and lack of mutant ctDNA molecules present in the plasma of healthy individuals with zero tumour burden[10].

In this article, we will explore the role of cfDNA analysis across the oncogenic spectrum of CRC, beginning during tumour initiation at a point when the tissue is phenotypically normal, then proceeding into a precancerous polypoidal lesion, and finally at the point when cancer has been established. The clinical roles of cfDNA with regards to normal, precancerous and cancerous colonic epithelium are summarized in Figure 1.

Figure 1 The role of cell free DNA in patients with normal, precancerous or cancerous colon. Figure was created with Biorender.com. Cell free DNA can play different roles in different patient populations. In presymptomatic individuals, it can be used to screen for colorectal cancer (CRC) by detecting the presence of circulating tumour DNA in plasma or non-plasma samples. Even if patients are proven to be CRC-free, they can be evaluated for driver mutations which would predict their risk of developing CRC. In individuals with CRC however, cell free DNA can be used for the surveillance of minimal residual disease post-surgery, to prognosticate, or to inform therapeutic decision making. CRC: Colorectal cancer; cfDNA: Cell free DNA; ctDNA: Circulating tumour DNA.

CFDNA

CfDNA refers to extracellular DNA fragments of around 140-170 base pairs in length found in the plasma or non-plasma bodily fluids. CfDNA was first detected in the blood of individuals in 1948 by Mandel and Metais[11], but only in the last few decades have researchers started to explore the potential of cfDNA as a minimally invasive modality of obtaining biological information. CfDNA mainly originates from the apoptosis of hematopoietic cells[12], however, numerous other forms of cell release mechanisms have been hypothesised. A specific component of cfDNA is circulating tumour DNA (ctDNA), which represents DNA shed from tumour cells[13]. The tumour burden and its degree of metastasis have been positively correlated to ctDNA fraction (ctDNA%) in the plasma[14]. CtDNA constitutes 0.01%-90% of cfDNA[15] and has a half-life of 1-2.4 hours[16]. CtDNA in the plasma or bodily fluids can be detected through cfDNA-based assays that search for cancer-specific mutations with high sensitivity and specificity.

MECHANISMS OF CFDNA RELEASE

Existing literature suggests that cfDNA is released into the circulation via two different routes: Passive release mechanisms and active release mechanisms. Passive release mechanisms include processes like apoptosis, necrosis, and breakage of CTCs. Active release mechanisms involve cfDNA release from extracellular vesicles (EVs).

Passive release mechanisms

Apoptosis is a type of programmed cell death carried out by caspases occurring in both physiological and pathological conditions. This process is triggered by a complex signalling cascade and involving morphological changes like cell shrinkage, pyknosis, plasma and chromatin condensation. Membrane blebbing leads to the release of apoptotic bodies from cells. These apoptotic bodies are subsequently engulfed by phagocytic cells and their components recycled[17,18]. It is believed that the large majority of cfDNA originates from apoptotic processes.

Due to the multiple cancer hallmarks and cell-death mechanisms involved in tumorigenesis, an origin of cfDNA from apoptosis alone is unlikely[19]. Besides apoptosis, necrosis has been mentioned as a potential source of cfDNA in cancer patients. Cells undergoing necrosis exhibit organelle dysfunction and degradation of the plasma membrane, exposing its intracellular DNA to degradative agents such as nucleases and leading to the release of DNA into the extracellular space. CfDNA molecules are then fragmented by the various nucleases. Other passive mechanisms include cfDNA release from CTCs and chromosome fragments due to chromosomal instability. Other cell death mechanisms that have also been hypothesised to contribute to cfDNA load include necroptosis, oncosis, pyroptosis and ferroptosis[20-22].

Active release mechanisms

Despite the abundant literature suggesting that cfDNA is mainly associated with apoptotic and necrotic processes, one recent study showed that cfDNA concentration had no correlation with apoptosis and necrosis. Active release mechanisms via exosomes constitute an alternative mechanism in which cfDNA may be found in the blood. The study found that breast cancer cells in the G1 phase released cfDNA via exosomes, and that the majority of cfDNA from breast cancer cells was released via active cellular secretion processes[23]. Current literature mentions active release mechanisms via exosomes, apoptotic blebs, shedding vesicles, and microparticles[24,25], however, the majority of active secretion of cfDNA occurs via EVs, which are spherical lipid-bound particles acting as mediators of physiological and biological processes[26] such as homeostasis[27]. Tumour-derived EVs are known to promote tumour invasion, metastasis and tumour migration as they can transfer tumour traits by entering other cells[18].

DETECTION OF CFDNA

Genetic mutations such as single-nucleotide variations (SNVs) and copy number variations (CNVs) have been used as diagnostic biomarkers for cfDNA-based modalities[28]. Modalities which detect genetic mutations (i.e., SNVs and CNVs) include quantitative polymerase chain reaction (PCR), targeted sequencing, whole-genome sequencing (WGS), and whole-exome sequencing. However, these mutation-based diagnostic modalities might not be adequately sensitive for patients with precancerous lesions or early-stage cancer given the lower number of recurrent mutations. A possibly superior approach may be through the detection of large-scale epigenetic alterations instead as they are tissue and cancer-type specific, and therefore are not constrained by low cell numbers[29].

Plasma and non-plasma sources of cfDNA

Though initially discovered in the blood, cfDNA fragments have now been found in all human body fluid types, such as pleural and peritoneal effusions, cerebrospinal fluids, urine, saliva, stool and seminal fluid. The advantages of using plasma as a source of cfDNA are as follows: It is easily obtainable hence allowing for longitudinal sampling at multiple timepoints, and tumour heterogeneity might be better captured than with tissue biopsy sampling. However, the low ctDNA to cfDNA ratio due to the predominance of clonal haematopoiesis makes using plasma cfDNA as a diagnostic modality difficult[30].

In contrast, the lower proportion of cfDNA originating from haematopoietic cells in non-plasma bodily fluids suggests a higher ctDNA fraction (ctDNA%) and higher variant allele frequencies (VAFs) in non-plasma sources compared to plasma[31]. This reduction in cfDNA levels can be beneficial when searching for low-frequency genetic alterations due to a reduction of the background noise created by clonal haematopoiesis. Furthermore, ctDNA from non-plasma sources might be more representative of the primary tumour, as shown in the pleural fluid of patients with advanced stage lung cancer[31] or in CSF from patients with leptomeningeal metastases[32]. Specifically in the context of CRC, stool-derived cfDNA can be particularly advantageous due to physical proximity to the colorectal neoplastic tissue. A multitarget stool DNA test (Cologuard^®) showed a sensitivity of 92.3% and specificity of 86.6% for the detection of CRC[33]. Hence, depending on the specific tumour type, non-plasma cfDNA can be a more viable option for cancer detection.

LIMITATIONS OF DETECTING CNVS AND SNVS FROM CFDNA

Mutation-based diagnostic modalities involve utilizing CNVs[34,35] and SNVs[36-38] as discriminative molecular features to reliably assess tumour-derived cfDNA. Current tumour fraction prediction methods based on CNVs rely on WGS with higher-coverage of more than 100-fold sequence coverage. Cutting edge algorithms, ichorCNA14[33] and ACE23[39], were initially developed from low-coverage WGS to generate an estimate of tumour fraction in cfDNA. However, both fail to provide an accurate estimate of tumour fractions due to high levels of aneuploidy and chromosomal instability[40,41]. Moreover, CNVs and SNVs are challenging to detect given the low number of mutated ctDNA fragments in early-stage cancer or certain tumour types[19]. While next-generation sequencing (NGS) technology enables high degrees of target multiplexing, the current depth of NGS sequencing is not deep enough to reliably search for mutations in a background of non-tumour-derived cfDNA[12,42]. The proportion of cfDNA fragments which harbour tumorigenic mutations is too low[43] which makes it difficult to search for bona fide variations amidst background signal from sequence changes introduced in library preparation. Extensive efforts have been made to improve the signal-to-noise ratio for more sensitive mutation detections, however these new methods rely on high-throughput sequencing and only analyze specific parts of the genome[37,42]. Such methods have limited efficacy for detecting cancer, especially at early stages, due to the low number of tumour genome equivalents in cfDNA[44,45]. The specificity of mutation detection is also hindered by the presence of somatic mutations in normal, non-malignant tissues[46,47]. Additionally, mutation-based screening modalities are largely incapable of localising the tissue of origin (TOO) of the tumour as the same driver mutation can be shared by many different cancers[37].

ADVANTAGES OF DETECTING DNA METHYLATION SIGNATURES AND OTHER EPIGENETIC ALTERATIONS FROM CFDNA

In contrast, DNA methylation signatures exhibit significant differences between healthy individuals and those with malignant tumours[19]. Compared to searching for point mutations, characterizing plasma epigenetic changes has shown to improve detection sensitivity by studies exploring the utility of cfDNA methylation for cancer detection[48]. Moreover, cfDNA methylation has shown utility in locating the cancer’s TOO[8], otherwise known as tissue deconvolution. This is possible as different tissues have different DNA methylation signatures[49,50] and even between different cell types within the same tissue[51]. Analyzing differentially methylated positions or regions can even allow detection of patients with malignant tumours[52-54]. Since cfDNA has various release sources, the methylation level measured at each cytosine-phosphate-guanine dinucleotide site is essentially a mixed signal originating from multiple tissues, including those harbouring tumorigenic driver mutations[55]. Hence, pinpointing the TOO is possible by deconvolution of blended methylation signatures.

CFDNA TO DETECT EARLY ONCOGENESIS IN NORMAL AND PRECANCEROUS TISSUES

Driver mutations in normal/precancerous colonic epithelium

Carcinogenesis is classically considered to occur as a result of the gradual accumulation of multiple oncogenic cancer gene mutations, acquired from an early stage[56]. A study by Lee-Six et al[57] found that driver mutations were present in about 1% of normal colorectal crypts. The driver mutations found in the normal colonic epithelium included truncating mutations in the cancer genes STAG2 and AXIN2, and hotspot mutations in ERBB2 (R678Q, T862A, V842I), ERBB3 (R667 L, R475W), PIK3CA (E542K, R38H), and FBXW7 (R658Q, R505C)[57]. This list is unexpected as these genes rarely harbour mutations in cancerous colonic epithelium. Mutations in ERBB2 and ERBB3 were common in normal crypts with driver mutations (5 in 14), but rare in CRC (7 in 631). There is also a surprising lack of established CRC driver mutations such as APC, KRAS and TP53 which are common in colorectal neoplasms and account for 56% of base-substitution and indel driver mutations but were comparatively rare in normal epithelium (1 in 14).

Detecting mutations in normal tissue from cfDNA sources

Pre-symptomatic screening of such mutations in cancer-free individuals, however, is still largely unfeasible due to low input volume and poor signal-to-noise ratio. To date, only a few studies have reported the analysis of cfDNA in healthy, presymptomatic controls. One such study by Alborelli et al[10] detected genetic mutations in 7 out of 55 clinically healthy subjects from plasma-derived cfDNA using NGS, of which 6 were germline variants (APC, PDGFRA, TP53) and 4 were cancer hotspot driver mutations (GNAS, IHD2, PIK3CA, TP53)[10]. The 4 cancer hotspot driver mutations detected were PIK3CA, TP53, IDH2, and GNAS, which are clinically classified as pathogenic or likely pathogenic. This study proved that it was possible for cancer-associated driver mutations to be detected in healthy individuals by analysing the cfDNA. The development of error-corrected NGS methods in recent years has also allowed for the identification of low-frequency mutations in normal tissue by reducing sequencing errors or bias[58-62]. For instance, Duplex sequencing[59,63] was able to detect mutations clustered in cancer-associated TP53 hotspots at low frequencies of < 0.01% in peritoneal fluid of women without cancer[64]. Another similar study found considerable levels of driver mutations including KRAS, TP53, PTEN, PIK3CA and several others in women with endometriosis, a largely benign gynecological disease. Other studies have identified RAS and P53 mutations in the cfDNA from saliva of patients up to 2 years before lung cancer insurgence[65], or the presence of TP53 or KRAS2 mutations in the plasma of healthy individuals who subsequently developed bladder cancer[66]. However, there are still no studies which have evaluated mutations from the cfDNA of common mutations found in normal colonic epithelium, such as STAG2, AXIN2, PIK3CA, ERBB2, ERBB3 and FBXW7.

CHALLENGES AND PROPOSED SOLUTIONS IN THE DETECTION OF DRIVER MUTATIONS IN NORMAL TISSUE USING CFDNA

Challenge 1: Poor signal-to-noise ratio due to an abundance of background non-tumorigenic somatic mutations

The detection of driver mutations in the normal colon of presymptomatic individuals is hindered by several challenges, as summarized in Figure 2. To achieve sufficient sensitivity and specificity, a cfDNA-based modality must be able to distinguish between the high volume of background noise due to non-tumorigenic processes and the driver mutations of clinical interest. However, our current knowledge of how driver mutations impact on oncogenesis is unknown. Previous studies have found mutations in normal skin tissue which do indeed contribute to clonal hematopoiesis[37] or clonal expansion but not tumorigenesis[67,68]. However, the role of different mutations in healthy colonic crypts is still widely unknown, which hampers the identification of pro-tumorigenic driver mutations which are clinically significant amidst a background of non-tumorigenic somatic mutations.

Figure 2 The challenges of using cell free DNA for the detection of driver mutations in normal tissue and proposed solutions. Figure was created with Biorender.com. Challenges preventing the adoption of cell free DNA (cfDNA) for individuals with normal colonic epithelium include poor signal-to-noise ratio, the abundance of non-tumorigenic background mutations, the low amount of cfDNA found in healthy individuals, and that only 1% of healthy colonic crypts possess the driver mutations of interest. Potential solutions that can be explored are to harvest cfDNA from non-plasma sources like the stool and peritoneal fluid, increase the sequencing depth of current technologies or to analyze epigenetic changes instead. cfDNA: Cell free DNA.

Challenge 2: Low amount of cfDNA in presymptomatic healthy individuals

As first reported in 1977 by Leon et al[9], the total plasma cfDNA concentration is significantly lower in healthy subjects compared to cancer patients. This makes the recovery and characterization of driver mutations from cfDNA in healthy individuals challenging due to the negative effect on NGS library concentration. In a study by Alborelli et al[10], NGS library concentration was found to be significantly affected by the lower plasma cfDNA concentration in healthy individuals as limited DNA input was used for library preparation, often below the minimal manufacturers’ recommended volume. The low level of input was shown to have a negative impact on sequencing performance and limit of detection (LOD) of certain assays, whereas generally higher molecular coverage was found in plasma cfDNA samples with higher amount of input cfDNA[10]. Hence, the low level of signal as well as the heterogeneity of mutations present in normal tissue still needs to be overcome to achieve efficacy. Specific to the colon, driver mutations are present in only 1% of colorectal crypts in healthy middle-aged individuals[57], making the cfDNA concentration shed by this small proportion of crypts incredibly low. This further exacerbates the poor signal-to-noise ratio. Despite the development of cutting-edge biotechnology, such as quantitative PCR or fluorescent dyes, whether the sequencing is deep enough to detect the ultralow cfDNA concentration is still unknown.

Solution 1: Harvesting cfDNA from non-plasma sources

One possible solution is to analyze the cfDNA derived from non-plasma sources. CfDNA from non-blood sources has been previously shown to be more representative of the primary tumour[31,32]. Detection of stool-derived cfDNA for CRC detection is one of the areas where non-plasma cfDNA testing is already being utilized clinically - the United States Food and Drug Association-approved screening test, Cologuard^®, is able to analyze stool DNA samples using sequencing panels comprising mutations in KRAS, TP53, APC and BAT-26[33]. Similarly, stool cfDNA might be more sensitive in detecting mutations in the normal colonic epithelium compared to plasma cfDNA. This might be due to stool’s physical proximity to the colonic epithelium, as human DNA is hypothesised to directly enter the stool from the colonic crypts via a combination of cellular shedding and colonocyte apoptosis[69]. However, human DNA accounts for only 0.01% of the total DNA content in stool, with the majority of DNA derived from the diverse consortium of microorganisms which inhabit the gastrointestinal tract[69,70]. The low cfDNA component poses a challenge to the detection of mutations from the normal colon. Another possible non-plasma source would be peritoneal fluid, which has previously been shown to have high detectability in CRC patients with peritoneal metastasis[71].

Solution 2: Increasing sequencing depth

Another solution would be to improve the depth of current sequencing technologies. However, even with the recent advent of various NGS technologies and methods where the LOD has decreased to < 0.01%, the LODs are unlikely to be low enough to detect low-frequency driver mutations amidst background noise from the abundance of somatic mutations in normal tissues. Another possible avenue would be utilizing DNA methylation signatures of normal colonic cfDNA instead as epigenetic changes have been shown to improve detection sensitivity[48]. If this is possible, cfDNA methylation might even allow location of the driver mutation’s TOO[8].

CFDNA FOR THE DETECTION OF ESTABLISHED CRC

Despite the efficacy of colonoscopy as a screening test for CRC, its invasive nature, interobserver variability leading to interpretive error, and need for bowel preparation, significantly limit its potential for population-wide screening. In contrast, liquid biopsy is a potential non-invasive modality for screening CRC through the detection of ctDNA which possess CRC-related genetic and epigenetic mutations. To date, many liquid biopsy kits have already been commercialised and approved for clinical use. Table 1 summarizes the current commercial liquid biopsy kits which analyze cfDNA, while Table 2 summarizes the liquid biopsy kits which analyze epigenetic changes.

Table 1 Liquid biopsy kits analyzing cell free DNA.

Name	Company	CfDNA source	Function	Detection of CRC sensitivity	Main limitation	Price per kit (dollar)
CancerSEEK	Thrive Earlier Detection Corp (Cambridge, Massachusetts, United States)	Plasma	Detection of 8 different cancers	64.0%	Low sensitivity for CRC detection	500
ColoAlert®	Pharm Genomics (Mainz, Germany)	Stool	Detection of CRC	84.6%	Low to very low certainty of reliable evidence available	151
Guardant360® CDx	Guardant Health (Palo Alto, California, United States)	Plasma	Genome profiling	NIL	Only some tumours shed detectable ctDNA into circulation	5000
FoundationOne® Liquid CDx	Foundation Medicine (Ontario, Canada)	Plasma	Genome profiling	NIL	Majority of the mutations detected are non-actionable	5800
Oncomine™ Colon cfDNA Assay	Thermo Fisher (Yokohama, Japan)	Plasma	Prognostication	78.6%	Low feasibility in clinical practise	12000
Signatera™	Natera (Austin, Texas, United States)	Plasma	Surveillance of MRD	99.9%	Not as sensitive as standard-of-care imaging surveillance	1750

cfDNA: Cell free DNA; CRC: Colorectal cancer; MRD: Minimal disease recurrence; ctDNA: Circulating tumour DNA.

Table 2 Liquid biopsy kits analyzing epigenetic changes.

Name	Company	cfDNA source	Function	Detection of CRC sensitivity	Main limitation	Price per kit (dollar)
Cologuard®	EXACT Sciences Corporation (Madison, Wisconsin, United States)	Stool	Detection of CRC	92.3%	Poor compliance rate due to the inconvenience of collecting stool sample	649
Epi proColon®	Epigenomics AG (Berlin, Germany)	Plasma	Detection of CRC	74.0%	Large amount of blood plasma (> 3.5 mL) is required	192
ColoProbe	NIL	Plasma	Detection of CRC	82.7%	Performance could exhibit variability since age was shown to be a confounding factor	NIL
ColonSecure	NIL	Plasma	Detection of CRC	85.3%	Performance could exhibit variability since age was shown to be a confounding factor	NIL
Guardant Reveal™	Guardant Health (Palo Alto, California, United States)	Plasma	Surveillance of MRD	91.0%	High false-negative and high false-positive rates	5000

cfDNA: Cell free DNA; CRC: Colorectal cancer; MRD: Minimal disease recurrence.

CancerSEEK evaluates the presence of mutations in 16 cancer genes and 8 tumour-associated protein biomarkers in the plasma. CancerSEEK can detect the presence of multiple cancers through assessment of cfDNA, and its specificity for detection of CRC was over 99% but the sensitivity was only around 60%[37]. However, multi-analyte tests like CancerSEEK are not meant to replace stool-based cfDNA assays for CRC detection, but to provide additional information that could help identify patients most likely to harbour a malignancy.

Cologuard^® on the other hand is a stool-based cfDNA assay specifically designed to detect CRC, involving NGS of the methylation of NDRG4, bone morphogenic protein, 7 mutation sites of KRAS and an immunohistochemical assay for haemoglobin. It was previously evaluated in a large prospective study by Imperiale et al[72] including 9989 participants with an average risk for CRC. Its evaluated sensitivity is higher than fecal immunohistochemistry test (FIT) (92.3% vs 73.8%) however with lower specificity (89.8% vs 96.4%)[72]. However, the stool-based tests usually have suboptimal compliance rates due to inconvenience[73], hence, blood-based detection assays still need to be explored.

To date, Epi proColon^® is the only FDA-approved blood-based detection assay for CRC. However, its usage is currently limited for individuals aged 50 and above who are noncompliant to traditional screening modalities like colonoscopy or FIT tests[74]. It utilizes a PCR with a hydrolysis probe for the detection of methylated Septin-9 DNA (mSEPT9), which has been associated with the occurrence of CRC[75,76]. Multiple studies have demonstrated Epi proColon^®’s high sensitivity and specificity of mSEPT9 for the detection of CRC[77-79]. A recent study evaluating the performance of mSEPT9 showed an aggregate sensitivity and specificity of 74% and 84% respectively when comparing CRC patients to healthy individuals[80]. Another recent study by Loomans-Kropp et al[81] evaluated the accuracy of Epi proColon^® V2.0 as a screening tool for early-onset CRC, which are CRC cases diagnosed at an age of 50 years or younger[81]. They showed that the mSEPT9 assay’s detection of early-onset CRC had a sensitivity of 90.8% and specificity of 88.9%.

Other upcoming cfDNA assays for CRC detection include ColoAlert^®, ColonSecure and ColoProbe. For example, ColoProbe is a multitarget plasma-based assay which detects three methylation markers, mSEPT9, SDC2, and ALX4. ColoProbe exhibited sensitivity of 82.7% for detecting CRC and 55.0% for detecting precancer, along with a specificity of 90.1%. Hence, compared to Epi proColon^®, ColoProbe has a significant advantage in detecting precancerous lesions. Another upcoming blood-based assay known as ColonSecure demonstrated excellent sensitivity (85.3% and 87.0% respectively) in discriminating CRC patients from healthy controls in control and test groups[82], surpassing even conventional biomarkers such as carcinoembryonic antigen (CEA), C-reactive protein and carbohydrate antigen 19-9. Compared to the mSEPT9 assay used in Epi proColon^® and ColoProbe, ColonSecure exhibited increased sensitivity and comparable specificity[83]. ColonSecure has shown similar detection sensitivity compared to the stool-based Cologuard^® test and clear superiority over the FIT test, in addition to its likely advantage of having a significantly improved compliance rate[72,73].

CFDNA TO PROGNOSTICATE AND AID IN THERAPEUTIC DECISION MAKING IN ESTABLISHED CRC

Comprehensive genome profiling cfDNA tests for CRC - Guardant360^® CDx and FoundationOne^® Liquid CDx - were recently approved by the FDA for the detection of genomic changes in cancer-associated genes. Both assays are recommended as companion diagnostics for guidance of therapeutic decision making in multiple cancers. In particular, Guardant360^® CDx has been shown to accurately identify 28 of 29 (96%) of pre-treatment plasma of CRC patients as bearing an amplification of ERBB2, which as previously stated in this review, is a common driver mutation in normal colonic epithelium. This might potentially allow clinicians to identify patients at a higher risk of malignant colonic transformation. This would also allow the prediction of the patients’ response to HER2-targeted therapy[84].

Another test that could have prognostication utility in CRC patients is the commercial Oncomine^™ cfDNA assay which constructs 48 amplicons covering key hotspot mutations of 14 genes. In a pre-planned analysis of the VALENTINO trial using Oncomine^™, Manca et al[85] studied the ctDNA VAF as a prognostic marker in patients with wild-type RAS metastatic CRC treated with an anti-EGFR-based treatment (folinic acid, fluorouracil, and oxaliplatin). They noted that higher VAF was found in patients with liver metastases, and that patients with high VAF had shorter overall survival (OS) compared to those with low VAF (21.8 months vs 36.5 months). The prognostication value of VAF hence exceeded that of baseline CEA by being significantly correlated with OS (P = 0.003), hence confirming its reliability for this purpose.

CfDNA based assays are also emerging as promising noninvasive approaches for detection of CRC recurrence post-resection and evaluating treatment response. The DYNAMIC trial highlighted that the ctDNA-based detection approach significantly reduced the use of adjuvant chemotherapy without increasing the risk of CRC recurrence in patients with stage II CRC[5]. Various NGS assays for surveillance of MRD post CRC resection have been developed over the past decade. One such example is Signatera^™ (Natera), a personalized, tumour-informed, multiplex PCR-based NGS assay for ctDNA detection currently commercially available in the United States. A large observational surveillance trial by Reinert et al[86] evaluated CEA levels, computed tomography imaging, and ctDNA in patients with stage I to III CRC. Patients underwent post-resection surveillance by Signatera^™ after adjuvant chemotherapy as well as by radiographic imaging. They found that Signatera^™ identified disease recurrence at a median of 8.7 months before radiographic imaging. More notably, while patients were awaiting radiologic detection, their ctDNA levels increased 5-fold, indicating that tumour burden increased markedly during the 8.7 months of lead time.

Guardant Reveal^™ is another NGS assay that utilizes ctDNA detection as a surveillance strategy. Unlike Signatera^™, it is a tumour-uninformed assay which analyses epigenetic signatures related to anomalous DNA methylation in addition to the detection of CRC-specific somatic alterations employed by most MRD assays. A prospective cohort study by Parikh et al[87] found that the augmentation of MRD detection by the integration of epigenomic signatures increased sensitivity by 36% compared to somatic alterations alone. In most CRC patients, ctDNA was detected from both genetic and epigenetic alterations, however a significant proportion were detected as positive by either genetic or epigenetic alterations, demonstrating that combining these two modalities may improve sensitivity for MRD detection. Other than its utility as a postoperative surveillance tool, the Guardant Reveal™ assay is also being tested in prospective clinical trials to assess its efficacy as a guide for adjuvant therapy.

CHALLENGES AND PROPOSED SOLUTIONS IN THE CLINICAL USE OF CFDNA FOR CRC DETECTION AND PROGNOSTICATION

Despite the major strides of progress in the last decade regarding cfDNA assay for CRC detection, surveillance and prognostication, only a handful of tests have been approved for widespread clinical use due to multiple limitations, as summarized in Figure 3.

Figure 3 The challenges of using circulating tumour DNA for the detection and prognostication of colorectal cancer and proposed solutions. Figure was created with Biorender.com. Challenges preventing the adoption of cell free DNA to detect and prognosticate colorectal cancer include the exorbitant cost of deep sequencing which makes it infeasible for population-wide screening, inefficient circulating tumour DNA extraction from plasma or non-plasma sources, and the lack of technologies with sufficient sequencing depth. Potential solutions that can be explored are to channel more resources into the development of sequencing technologies with sufficient depth and sensitivity, and to formulate an effective and standardized method of circulating tumour DNA extraction. CRC: Colorectal cancer; ctDNA: Circulating tumour DNA.

Challenge 1: High cost and inefficient ctDNA extraction

One of the main limitations is that ctDNA is not readily available in patients with early-stage tumours due to the low input volume and poor signal-to-noise ratio. Since ctDNA is not very different from normal cfDNA, specific extraction is challenging with no current standard for extraction. Extraction efficiency is pivotol for a successful ctDNA analysis, especially in the early stages of CRC when ctDNA load is low to begin with. However, there is currently no standardized protocol for ctDNA isolation. Most methods require centrifugation of plasma which is time-consuming, inefficient and drains resources. Most methods also lack extraction efficiency as they are unable to detect for low molecular weight DNA which is typical of most cfDNA fragments[88]. Hence, the entire process of ctDNA isolation is not only labour-intensive but also costly, calling for the need of a standardized and efficient purification protocol.

Challenge 2: Poor signal-to-noise ratio

Additionally, somatic mosaicism in blood plasma remains an immense challenge for accurate cfDNA analysis[89]. Clonal hematopoiesis is a common age-related process involving the expansion of a clonal population of hematopoietic stem cells[90]. However, the detection of these non-tumorigenic clonal hematopoietic mutations is a common source of background signal for cfDNA-based assays[89]. Hence, the detection of CRC-specific mutations from plasma-derived cfDNA remains largely infeasible.

Solution 1: Detection technologies with higher sensitivity and specificity

In order to address the issue of poor signal-to-noise ratio, we need detection technologies with higher sensitivity and specificity to detect the < 1.0% of ctDNA in total cfDNA to allow for earlier intervention. However, the cost of high-sensitivity sequencing is generally more costly and is hence infeasible for population-wide screening. It might be advantageous to explore the use of new cutting edge sequencing technologies such as Beads, Emulsion, Amplification, and Magnetics, clustered regularly interspaced short palindromic repeats-mediated, Ultrasensitive detection of Target DNA-PCR-PCR, or CAncer Personalised Profiling by Deep Sequencing for the augmentation of cfDNA analysis to increase sensitivity and specificity[91].

As the consistency of mutation profiles between paired plasma and tumour tissue samples increases, it is becoming clear that cfDNA-based liquid biopsy for CRC detection, surveillance and prognostication has immense potential for clinical and precision medicine. However, two main challenges still stand in the way. The first challenge is the poor signal-to-noise ratio amidst a background of somatic non-tumorigenic mutations, which is further compounded by a lack of technologies with sufficient sequencing depth. The only way to circumvent this issue would be to develop technologies capable of deep sequencing to pick up mutations more sensitively. However, even if such technologies are developed, the cost of high-sensitivity detection is likely to be expensive and infeasible for population-wide screening. The second challenge is the low quantity of plasma ctDNA found in healthy, precancerous and even early-stage CRC patients. This significantly affects the sequencing performance of some, if not all, ctDNA assays. It might be possible to circumvent this issue by harvesting ctDNA from non-plasma sources like the stool or peritoneal fluid, which might be more representative of the primary tumour. However, obtaining peritoneal fluid is unrealistic for population-wide screening due to the invasive nature of a paracentesis. On the other hand, obtaining stool as a non-plasma source of ctDNA is more feasible and is already being used clinically. However, due to the gastrointestinal microbiome, only 0.01% of the total DNA content of stool is human-derived. This leaves us to face the first challenge again: The issue of poor signal-to-noise ratio and the lack of technologies with sufficiently deep sequencing. Hence, there is no conceivable way to utilize ctDNA assay as a mainstream modality of CRC detection and prognostication if the above two challenges are not addressed. If they are circumvented, however, the potential of liquid biopsy for the detection, surveillance and prognostication of CRC would be limitless.

CONCLUSION

The potential for cfDNA to revolutionize the screening, detection and prognostication of patients with CRC rests on overcoming the challenges detailed in this manuscript. Given the potential for improved screening, as well as improved OS in patients with CRC, efforts need to be focused on overcoming these challenges with expediency.