In the United States, CRC is the third most common cancer diagnosed among men and women and the second leading cause of cancer death with the estimation of 142280 new cases and 50830 deaths in 2013[6]. The five-year survival rate is 90% for cancer found localized or confined to the bowel wall, 70% for cancer with lymph node involvement, and 10% for cancer that has metastasized. Clearly, these numbers demonstrate that screening and early detection would lead to better survival, prognosis, treatment options, and hence quality of life. In 1980, the American Cancer Society (ACS) issued a formal guideline for CRC screening in average-risk adults, including an annual digital rectal exam and stool guaiac slide test in addition to the performance of a sigmoidoscopy every three to five years[7]. Since the guideline was issued, the cancer morbidity and mortality rates, which peaked around 1985 in the United States, have been in steady decline[6]. It is conceivable that the decline of CRC rates is at least partially attributable to the implementation of early screening and surveillance programs[8].

As of 2008, the basic screening modalities remain remarkably similar to those used in 1980 when the original guideline was issued, even when taking into account the development of newer technology in subsequent years[8]. In general, ACS, American College of Radiology (ACR), and the United States Preventive Services Task Force (USPSTF)[9] all agree on and emphasize the importance of CRC screening[8,10-12]. The recommended CRC screening modalities can be roughly divided into two different categories: fecal tests and direct structural exams.

The fecal tests are essentially “blood in the stool” tests. They can be performed using either a hemoglobin test [the guaiac-based Fecal Occult Blood Test (gFOBT)] or a newer and more sensitive version of an antibody-based globin test, known as the immunochemical FOBT or Fecal Immunochemical Test (FIT)[13]. In general, the gFOBT test is a non-invasive, inexpensive and easily applicable screening test which patients can readily perform in the comfort of their own home. Specimens from a FIT must be submitted to a laboratory for testing. The fecal tests help to reduce the risk of CRC death but has no effect on all-cause mortality[14]. They are not specific tests for CRC markers, and if found positive, the presence of CRC must still be confirmed by a direct structural exam such as colonoscopy or imaging procedures[15]. The fecal tests have high false positive rate for detecting CRC as gastrointestinal bleeding may occur in other conditions like colitis and hemorrhoids[16-18]. This, therefore, increases the burden of unnecessary colonoscopies and anxiety among patients[19]. It also may not detect precancerous lesions or early stage adenomas as bleeding may not be readily detectable in the presence of these conditions[20,21]. Regarding the fecal tests in general, the opportunity for CRC prevention is both limited and incidental and they are therefore not recommended as the solo screening test for CRC[8].

Direct structural exams include endoscopic procedures, such as flexible sigmoidoscopy and colonoscopy, and imaging procedures, such as double-contrast barium enema and computed tomographic colonography. In general, both flexible sigmoidoscopy and colonoscopy are invasive procedures using a colonoscope. Sigmoidoscopy is a small-scale colonoscopy which can be performed with a simple preparation without sedation, and is used to examine the lower half of the colon lumen as opposed to the entire colon. The complete colonoscopy allows direct mucosal inspection of the entire colon from the appendiceal orifice to the dentate line. Same-session biopsy sampling or definitive treatment by polypectomy in the case of precancerous polyps and some early-stage cancers can also be performed. The double-contrast barium enema and computed tomographic colonography are both imaging examinations of the colon in its entirety and are either noninvasive or minimally invasive. However, although they allow for complete examination of the colon, there is no opportunity for biopsy or polypectomy and must therefore be followed up by therapeutic colonoscopy when polyps are found.

In the United States, colonoscopy has become the gold standard of CRC screening. It is one of the critical screening procedures recommended by ACS, ACR, and USPSTF, and it is also recommended by the American College of Gastroenterology as the preferred screening test[22]. The principal benefit of colonoscopy is that it allows for a full structural examination of the colon and rectum in a single session and for the detection of colorectal polyps and cancers accompanied by biopsy or polypectomy. Therefore, it has been performed with much higher frequency than all other procedures[23].

However, even in the United States where the technology and procedure are widely available, the colorectal screening participation is still low among average-risk adults in the range of 29.8% to 55.2%[24]. The participation rate is also surprisingly low at 40% for people at increased risk of CRC[25,26]. The majority of United States adults are not receiving regular age- and risk-appropriate screening due to concerns of cost, risk, and the discomfort and cumbersome preparation associated with the procedure[27-29]. The same is true in other European and Asian nations[2,30-32].

Although colonoscopy is the most effective screening method for CRC, there are various reported risks associated with the procedure, including bleeding (1.64 per 1000 patients), perforation (0.85 per 1000), death (0.074 per 1000), missed adenoma (6%-12%), and missed cancer (5%)[33]. The observed rate of missed polyps and/or cancer are largely due to variations in polyp size and other factors such as sub-optimal bowel preparation, experience of the endoscopists, and patient anatomical variations[34]. When it is taken into consideration that the guideline for the average-risk adult is to undergo colonoscopy every 10 years beginning at age 50[8,22] coupled with the rate of missed polyps being between 6% and 12%, there is still risk of developing CRC even when regular colonoscopy screening guidelines are followed.

The goals of any test are to detect disease early, improve duration and quality of life, reduce mortality and/or morbidity, and augment patient participation for that disease process-all at a very low risk and cost. To this end, the current CRC screening modality based on colonoscopy is severely inadequate. Despite all of the benefits that colonoscopy can offer as a screening procedure for CRC, concerns about its cost, risks, cumbersome preparatory procedure, and willingness of the general public to participate seriously compromise its effect in undermining the global CRC burden[35-37].

In an ideal world, the first line screening should be performed to identify a high risk segment of the population and then use a more extensive test (colonoscopy) on this sub group to reduce incidence of advanced diseases. In other words, it is crucial for the first line screening program to separate the following three entities: the general population (average risk), high risk group, and cancer group. Despite its non-specific nature, the simple FIT, when coupled with colonoscopy, has helped to dramatically reduce cancer incidence and number of deaths - In 100000 average risk patients, this screening has helped to reduce the number of cancer cases from 4875 to 1393, and number of cancer deaths from 1782 to 457[38]. Therefore, a more effective and sensitive blood-based biomarker test, supported by evidence from larger studies with solid results, can readily replace the stool-based test.

In order to establish a screening test, it must be evaluated for the following elements: frequency of performance, risk of complications, limitations, and false positive and negative rates. A blood-based test could be ideally used as a first line screening if all these elements were reliably determined and optimized. Colonoscopy would then become the secondary test, not the primary one. There will be greater willingness, by physicians and patients alike, to perform a blood test every several years than to justify the bowel preparation and complications of colonoscopy every 5-10 years.

RNA was originally thought to be highly labile, easily degradable, and therefore not likely to be stable or detectable outside of the protective cellular environment. However, numerous recent studies have shown that RNA are actually stable outside of cells[45,46], and all species of RNA, including both coding messenger RNA (mRNA)[47] and non-coding RNA, which includes microRNA (miRNA) and long non-coding RNA (lncRNA)[48,49], can be extracted and detected in the circulating blood plasma, serum, and other bodily fluids[50-52]. Furthermore, RNA expression is highly regulated in normal state but becomes increasingly dysregulated in a pathological state such as cancer[48,53]. Therefore, numerous studies have focused on profiling RNA expression, which may correspond to cancer state, and finding the indicator biomarkers for cancers[54-57].

Various research groups have investigated the potential use of circulating mRNA as markers for cancer. The general experimental strategy is to employ microarray technology for mRNA expression profiling, which is then followed by validation using real time quantitative reverse transcription polymerase chain reaction (RT-qPCR). The specimens used are either mRNA extracted directly from blood serum/plasma or from peripheral blood cells[58]. Kopreski et al[47] demonstrated the possibility of detecting tumor mRNA, tyrosinase, in the serum of malignant melanoma patients although the result remains controversial[59]. Tsouma et al[60] extracted RNA from peripheral blood cells and used the multiplex RT-qPCR technology to determine the expression of three transcripts (carcinoembryonic antigen, cytokeratin 20 and epidermal growth factor receptor) to determine the disease stage and overall survival of CRC patients. DePrimo et al[61] and Twine et al[62] performed microarray-based mRNA expression profiling in peripheral blood mononuclear cells in 2003 and proposed some potential markers. However, this research generally remained at a proof-of-concept or pilot study stage, and further follow-up study has been sparse as the strategy they originally employed is now gradually being replaced by the new technology of Next Generation Sequencing (NGS), which will be discussed in more detail later.

Marshall et al[63] from GeneNews Ltd. developed a blood-based test using a seven-gene biomarker panel (ANXA3, CLEC4D, LMNB1, PRRG4, TNFAIP6, VNN1 and IL2RB) testing RNA extracted from peripheral blood cells. This seven-gene panel was derived from a 196-gene expression profile using 112 CRC patients (including those with stage I, II, III, and IV disease) and 120 controls. The panel was confirmed using 202 CRC patients (from all stages) and 208 controls, all from the Canadian population. They reported a sensitivity of 72% and specificity of 70% for this initial study. Then, they validated the seven-gene profile using 99 CRC patients (presumably from all stages) and 111 controls from the Malaysian population and reported 61% sensitivity and 77% specificity[64]. The researchers further validated their panel with an even larger population of 314 CRC patients (from all stages) and 328 controls from Canada and the United States, and they reported a sensitivity of 78% and specificity of 66%[65]. GeneNews now offers the ColonSentry test, presumably based on this seven-gene profile, as the world’s first commercially available blood test for colon cancer screening, which is licensed to Enzo Clinical Labs of Enzo Biochem for the United States market. The test has recently been approved by the New York State Department of Health as a test to determine a person’s risk of having CRC[66].

The ColonSentry molecular diagnostic test is marketed as a risk assessment tool rather than a cancer detection test. Although the experimental design for this seven-gene profile appeared to focus on identifying a pan-CRC marker panel when it profiled and validated a total of 727 CRC patients from all stages (estimated to be 30% stage I, 30% stage II, 30% stage III, and 10% stage IV), there is no mention of any study on high risk individuals, advanced adenomas (AA), or patients with colon polyps that ultimately turned cancerous. It is therefore unclear how a set of pan-CRC markers for all CRC stages can be marketed as a risk assessment test. In any case, the test is considered experimental and investigational with many independent experts still questioning its effectiveness.

MiRNA are small non-coding RNA about 18-25 nucleotides in size[67]. A large body of publications indicates that miRNA regulate gene expression at the post-translational level in almost every biological event and play important roles in tumorigenesis, cancer development, migration and metastasis[68]. The differential expression of miRNA has been related to various cancers[69], and efforts have been made to profile the global and circulating miRNA expression patterns associated with various cancers, including breast cancer[70], lung cancer[71], lymphoma[72], ovarian cancer[73], and pancreatic cancer[74,75].

For CRC, studies have accumulated over the past five years that focus on profiling circulating blood plasma or serum miRNA and validating the findings with RT-qPCR. Ng et al[76] was the first group to profile 95 miRNA using a real-time PCR-based array on 5 CRC patients and 5 controls (presumably from the Chinese population in Hong Kong) and to validate the results with 90 CRC patients and 50 healthy controls. They identified miR-17-3p and miR-92 to be elevated significantly in CRC patients with 89% sensitivity and 70% specificity. Wang et al[77] profiled 742 miRNA using a miRNA microarray on 10 CRC patients and 10 normal controls from the Chinese population and validated the results with 90 CRC patients, 43 AA patients, and 58 healthy donors. They found miR-601 and miR-760 to be decreased in both CRC and AA patients when compared to healthy controls with 83.3% sensitivity and 69.1% specificity. Giráldez et al[78] performed a genome-wide profiling of 743 miRNA using a miRNA microarray on 21 CRC patients, 20 AA patients, and 20 healthy controls from the Spanish population, and they validated the findings using RT-qPCR with 42 CRC patients, 40 AA patients, and 53 controls. They identified a six-miRNA panel (miR-15b, miR-18a, miR-19a, miR-19b, miR-29a, and miR-335) as being able to differentiate CRC patients from healthy individuals with 78.57% sensitivity and 79.25% specificity, and miR-18a could also differentiate AA patients from healthy individuals with both 80% sensitivity and specificity. Luo et al[79] used a TaqMan MiRNA array to profile 667 miRNAs on 50 CRC patients and 50 controls from the German population and validated the results with new cohorts of 80 CRC patients compared to 144 controls and 50 AA patients compared to 50 controls. They identified nine miRNA (miR-18a, miR-20a, miR-21, miR-29a, miR-92a, miR-106b, miR-133a, miR-143, and miR145) to be differentially expressed in CRC patients and controls with the area under the accompanying receiver operating characteristic curve reported to be 0.745. The panel of miRNA did not, however, differentiate AA patients from the controls. Kanaan et al[80] screened for 380 miRNA using microfluidic TaqMan array technology on 20 CRC patients, 9 AA patients (referred to as colorectal adenomas), and 12 healthy donors of mixed racial background in the United States. They then validated the findings with a new cohort of 45 CRC patients, 16 AA patients, and 26 healthy controls; they derived an eight-miRNA panel (miR-15b, miR-17, miR-142-3p, miR-195, miR-331, miR-532-5p and 532-3p, and miR-652) that can distinguish AA patients from controls with 88% sensitivity and 64% specificity, and a three-miRNA panel (miR-431, miR-15b, and miR-139-3p) to differentiate stage IV CRC patients from controls with 93% sensitivity and 74% specificity. Ahmed et al[81] performed a profiling using miRNA microarray chips covering miRNA based on the published miRBase v17 list (presumed to be 1733 human miRNA) and validated their results using TaqMan RT-qPCR to analyze a panel of miRNA expression both in CRC patient plasma and tissues. They found nine miRNA (miR-7, miR-17-3p, miR-20a, miR-21, miR-92a, miR-96, miR-183, miR-196a and miR-214) to have increased expression and six miRNA (miR-124, miR-127-3p, miR-138, miR-143, miR-146a, and miR-222) to have reduced expression in both CRC patient plasma and tissues with 90% sensitivity and 95% specificity.

A few studies selected their miRNA markers based on published literature and re-confirmed the results with RT-qPCR assays. Huang et al[82] measured the levels of twelve miRNAs (miR-17-3p, -25, -29a, -92a, -134, -146a, -181d, -191, -221, -222, -223, and -320a) studied in the literature in 120 CRC patients, 37 AA patients, and 59 healthy controls from the Chinese population, and they confirmed miR-29a and miR-92a as potential indicators for CRC with 83% sensitivity and 84.7% specificity. Similarly, Liu et al[83] measured the levels of five miRNAs (miR-18a, -21, -31, -92a, and -106a) in serum samples from 200 CRC patients, 50 AA patients, and 80 healthy controls from the Chinese population and identified miR-92a along with miR-21 to be both significantly higher in CRC patients with 68% sensitivity and 91.2% specificity. Pu et al[84] measured miRNA expression levels of three target miRNAs (miR-21, -221, and -222) in 103 CRC patients and 37 controls from the Chinese population and found elevated expression of miR-221 in CRC patients with 86% sensitivity and 41% specificity. Wang et al[85] screened three miRNAs (miR-29a, -92a, and -17-3p) in 38 metastatic CRC and 36 primary CRC patients, assumed to be from the Chinese population, but did not utilize healthy controls. They found miR-29a to be higher in CRC patients with liver metastases than in primary CRC patients with sensitivity and specificity of 75%, and hence miR-29a may be useful in discriminating metastatic from non-metastatic CRC patients. Cheng et al[86] screened three miRNAs (miR-21, -92, and -141) using a cohort of 102 CRC patients and an age-matched cohort of healthy donors of mixed racial background from the United States population, validated their findings using 156 CRC patients and matched controls from the Chinese population, and found miR-141 to be higher in cases of advanced CRC (stage IV) with 90.9% sensitivity and 77.1% specificity.

As summarized in Table 1, there are a total of 38 miRNA that have been studied and proposed as potential biomarkers for CRC in the publications discussed above. In general, most of these studies focused on early stage CRC patients while some also included borderline AA patients. When pooling from all the studies mentioned here, sensitivities in the range of 68%-91% were reported, but the majority (in 9 out of 12 cases) observed sensitivities in the 83%-91% range. Reported specificities were in the range of 41%-95%, but the majority (also in 9 out of 12 cases) were in the 70%-95% range. Some miRNA, including miR-15b, miR-17-3p, miR-18a, miR-20a, miR-21, miR-29a, and miR-92a, have been proposed by more than one group of investigators. One unique miRNA, miR-21, might actually be a useful pan-cancer marker as it is similarly up regulated in other cancers[87]. However, most of these studies have not yet been evaluated beyond the proof-of-principle and pilot stage, and not all miRNA markers were subsequently studied and confirmed by other groups. For example, Faltejskova et al[88] was not able to confirm the potency of miR-17-3p, miR-29a, miR-92a, and miR-135b as biomarkers for CRC. Luo et al[79] and Ahmed et al[81] found differential miR-143 expression in their respective studies. Other potential markers such as miR-17-3p, miR-18a, miR-21, miR-92, and miR-221 were not confirmed in follow-up studies by other groups[82-84,86].

Clearly, it is comprehensible that different experimental designs, procedures and methods, endogenous controls, patient populations, instrumentation and lab personnel could contribute to the seemingly contradicting results that have been published thus far. Nevertheless, the 38 candidate miRNA markers together can be further investigated using currently available technology, such as the TaqMan RT-qPCR profile platform already utilized by some of the research groups. It is possible, therefore, to coordinate a multicenter clinical trial involving different research groups and incorporating patient populations from a wide variety of backgrounds. It would be critical to synchronize specimen collection, processing procedures, and storage conditions for the collected specimens. The experimental design should also be based on a coordinated and synchronized set of experimental procedures and instrumentation that utilize the same endogenous control(s). The validity of each of the 38 miRNA markers as a tool for diagnosing CRC can then be evaluated for their potential future application.

Given the increased availability of RNA-seq technology, it is now possible to study the lncRNA, which was dismissed as “junk” in the past but has now been found to regulate gene expression and cellular functions[105]. LncRNA, like its miRNA counterpart, plays major roles in tumor suppression and oncogenic functions and has been found to be dysregulated in human cancers[106]. Therefore, its potential role as biomarkers for cancer and other diseases has been investigated extensively[107,108]. As an example, Prostate cancer antigen (PCA3, also known as DD3) is a non-coding RNA that is highly sensitive and now used as a biomarker for the urine diagnostic test of prostate cancer[109-111].

In terms of CRC, research is currently focused on the role of lncRNA as tissue biomarkers. Ge et al[112] found that Prostate cancer-associated ncRNA transcripts 1 was upregulated in CRC tissue but not in adjacent normal tissue. Zhai et al[113] found that long intergenic noncoding RNA-p21 was upregulated in CRC tissue, and the expression level seemingly correlated with tumor progression (higher expression in later stages). Ling et al[114] showed a novel lncRNA-CCAT2 was highly overexpressed in CRC, and it was shown to be promoting tumor growth, metastasis and chromosomal instability. Kogo et al[115] demonstrated that expression of lncRNA-HOTAIR, which is known to reprogram chromatin organization and promote breast cancer metastasis[116], is also higher in stage IV CRC patients with liver metastases. Xu et al[117] found the lncRNA-human metastasis associated lung adenocarcinoma transcript 1 (MALAT-1) to be dysregulated in cancer, and the mutation on the 3’ end of MALAT-1 is apparently tumorigenic. It is conceivable that RNA-seq technology can help facilitate further investigation into lncRNA functions and exploration of blood-circulating lncRNA as potential biomarkers for CRC and other cancers in the future.

Aberrant DNA methylation has been associated with tumorigenesis as a consequence of the alteration it causes in gene expression[122,123]. For example, hypermethylation of tumor suppressor promoter genes would cause inappropriate gene silencing and therefore lead to cancer[124]. In general, DNA methylation is thought to be associated with an early event in tumorigenesis and has therefore been proposed as a potential early cancer detection marker[123,125]. The research strategy typically focuses on using methylation specific PCR (MSP) to study hypermethylation of methylation sites, in CpG dinucleotides or in CpG islands, in the promoters of tumor suppressor genes[124,126]. In the context of CRC, Nakayama et al[127] and Lecomte et al[128] both monitored the hypermethylation of the promoter of tumor suppressor gene p16 and found the plasma in 21 of 31 (68%) patients and 31 of 45 (69%) patients, respectively, to be positive. Grady et al[129] found aberrant hypermethylation of the human MutL homolog 1 (hMLH1) promoter in the sera of 9 out of 19 (47%) cases of CRC. Leung et al[130] monitored promoter hypermethylation in three genes, adenomatous polyposis coli (APC), hMLH1, and helicase-like transcription factor, and found at least one of the three genes with methylated promoter DNA in the sera of 28 out of 49 (57%) CRC patients. Additional genes monitored for tumor-related promoter hypermethylation, including the putative metastasis suppressor gene death-associated protein kinase, the detoxification gene glutathione S-transferase P1, the DNA repair gene O⁶-methylguanine-DNA-methyltransferase, and p14-ARF in other cancers exhibit a detection rate that is generally in the range of 42% to 73%[131-133]. It is conceivable that NGS technology can be coupled with MSP to identify a pool of tumor suppressing genes silenced in association with early stage CRC and AA, test their corresponding promoter methylation, and generate a set of candidate markers based on epigenetic changes as a screening panel for CRC in the future.

The NGS technology has been employed for somatic mutation analysis in CRC[134], particularly on several high mutation frequency genes, such as K-RAS[128,135,136], TP53[137], and APC[138]. However, the percentage of circulating tumor DNA is relatively low when compared to wild-type DNA[139]. For example, Diehl et al[138] has shown that in advanced CRC, the mutated APC DNA fragment is found to be in the range of 1.9% to 27% of cfDNA but only 0.01% to 0.12% in early stage CRC. Even with direct sequencing technology, it does not allow reliable detection of less than 25% mutant signal in a background of wild-type DNA[140]. Furthermore, the tumor-associated mutations are often unique with each patient[141,142], and therefore, based on the current available technology, it is less likely to develop a low cost and highly sensitive comprehensive test to cover all somatic mutations for early cancer detection.

Microsatellite alterations, which include microsatellite instability (MSI) and loss of heterozygosity (LOH), are known to be associated with tumorigenesis and cancer progression and therefore were proposed as potential tumor markers detectable in cfDNA[143]. MSI analysis focuses on measuring the specific polymorphic tetranucleotide repeat and/or dinucleotide markers that are located in regions frequently shifted or altered in cancer, and LOH analysis focuses on the loss of specific chromosomal regions bearing tumor suppressors. Hibi et al[141] examined microsatellite alterations and found LOH or microsatellite shift of at least one locus (18a, 17p, and 8p) in 35 of 44 (80%) primary CRC tumors, but none of the LOH or microsatellite shifts were detected in the corresponding serum DNA. Several other groups focused on different cancers with most success in metastatic cancers[143,144]. In general, microsatellite alteration analysis exhibits relatively low sensitivity and specificity in detecting early stage cancer.

There are generally a few hundreds of copies of mitochondrial DNA in each cell[145]. Due to its multi-copy nature, mtDNA is frequently found to be heteroplasmic, with a heterogeneous mixture of polymorphic variants. In cancer cells, mtDNA harbor further heteroplasmic alterations associated specifically with cancer, especially in the highly variable D-loop (displacement loop) region. With the NGS, the approaches generally focused on either differential copy number of mtDNA versus gDNA, or mtDNA alteration and tumor-associated mtDNA mutations[146]. For CRC, Hibi et al[147] has studied mtDNA alternation in early CRC patients and found that 7 out of 77 (9%) CRC tissues contained true somatic mutations in the D-loop region, but only one out of these 7 positive patients (14%) were noted to have mtDNA alterations in their serum DNA. Due to of the relatively low detection rate of early stage cancer, most studies therefore focused on its potential application in metastatic cancers[148-153].