The Clinical Landscape of Circulating Tumor DNA in Gastrointestinal Malignancies

Technologies for genomic analyses have revealed more details in cancer biology and have changed standard treatments for cancer, including the introduction of targeted gene-specific therapy. Currently, liquid biopsies are increasingly being utilized in clinical trials and research settings to analyze circulating tumor DNA (ctDNA) from peripheral blood. Several studies have shown the potential of ctDNA in the screening, prognostication, molecular profiling, and monitoring of gastrointestinal malignancies. Although limitations continue to exist in the use of ctDNA, such as method standardization, the sensitivity, concordance with tumor tissue, and regulatory issues, this field offers promising benefits for cancer treatment. A deeper understanding of tumor biology via ctDNA analyses and ctDNA-guided clinical trials will lead to the increasing use of ctDNA in clinical practice in the near future; this development will result in the improvement of outcomes among patients with gastrointestinal malignancies.

iNTRODUCTiON Over several decades, the gold standard in the diagnosis and screening of tumors has been tissue biopsy (1). However, conventional tissue biopsies are invasive, painful, and carry a risk of complica tions such as bleeding, local infection, and damage to neighboring tissues. Moreover, a tissue biopsy cannot always reflect tumor dynamics or response to treatment. The recent era of precision medicine, which represents a paradigm shift in cancer, has challenged the gold standard in diagnosis by intro ducing a transition from tissue biopsy to liquid biopsy. Compared with tissue biopsies, liquid biopsies carry minimal potential risk and can be repeatedly performed in routine practice during cancer treatment by using peripheral blood. Furthermore, liquid biopsies have the potential to provide more complete information regarding the biology of whole tumors despite tumor heterogeneity. Liquid biopsies include the testing of soluble factors, such as circulating tumor DNA (ctDNA) and circula ting cellfree DNA (cfDNA), as well as proteins and tumor markers (2). cfDNA is highly fragmented DNA that is released from necrotic or apoptotic cells into the bloodstream (3)(4)(5). cfDNA consists of DNA from healthy cells and tumor cells, whereas ctDNA is defined as DNA that is derived only from primary or metastatic tumor cells.
Since Mandel and Metais (6) reported fragmented DNA in blood for the first time in 1948, technolo gies for cfDNA quantification have changed over 70 years from quantitative polymerase chain reaction to complex BEAMing and deep nextgeneration sequencing (NGS), thus achieving improvements in the sensitivity and specificity of cfDNA detection. With the development of sensitive techniques that can detect rare mutations, the heterogeneous landscape of tumors can be determined using blood samples. In fact, National Comprehensive Cancer Network guideline for nonsmall cell lung cancer (version 4. 2018) states that plasma biopsy should be considered if repeat biopsy is not feasible (7).
Here, we review ctDNA in gastrointestinal malignancies by focusing on clinical utility and future perspectives.

ctDNA AND ReLATeD TeCHNOLOGieS
The presence of cfDNA in the blood is a wellestablished fact, and DNA fragments are released from dying cells because of cel lular turnover or other types of cell death (2). In cancer patients, a fraction of cfDNA is tumor derived and is termed ctDNA. ctDNA originates from primary tumors, metastatic tumor cells, or circulating tumor cells. ctDNA molecules are shorter than nonmutant cfDNA molecules in plasma, as demonstrated by PCR (8,9) and sequencing (10,11).
Representative approaches for analyzing ctDNA are sum marized in Table 1 (12). Mutationspecific realtime or endpoint PCR has been used for the detection of point mutations in ctDNA (13)(14)(15)(16)(17). More recently, digital PCR methods such as BEAMing and droplet digital PCR have been developed to improve the identification of genomic alterations in ctDNA (18)(19)(20). The recent implementation of NGS has allowed the direct sequence based detection of chromosomal alterations in plasma DNA (21)(22)(23); however, it is necessary to distinguish the relatively few somatic alterations in ctDNA from the larger numbers of structural variants present in the germline cells of all individuals. Bioinformaticsbased filters that enrich highconfidence somatic structural alterations while eliminating germline and artifactual changes have been developed (12). In addition, importantly, amplification in ctDNA can be depend on both the amount of ctDNA in the plasma due to high tumor burden and high copy number of specific gene. Commercially available kits for the NGS assays of ctDNA are summarized in Table 2.

eARLY DeTeCTiON OF CANCeR
The early detection of cancer is one of the most important issues in reducing cancerrelated deaths. In many cases, gastrointestinal cancer is detected via endoscopy or CT scans conducted for symp toms such as anorexia, abdominal pain, or constipation. ctDNA may have a potential role in the noninvasive early diagnosis and screening of gastrointestinal cancer. Even localized cancers shed DNA into circulation; therefore, ctDNA can be detected in patients with localized cancers, in addition to patients with advanced or metastatic cancers.
In a study across several early and latestage cancers, ctDNA was detected in 73, 57, and 48% of patients with colorectal cancer (CRC), gastroesophageal cancer, and pancreatic cancer, respec tively (25). The use of several biomarkers in ctDNA including the levels of overall ctDNA, ALU247 fragment concentration (26), KRAS mutations (27,28), TP53 mutations (29,30), BRAF mutations (28), and septin 9 (SEPT9) methylation (31)(32)(33)(34) have been demonstrated for the diagnosis of CRC. Also, detection of methylated SEPT9 DNA in plasma is US FDA approved as a blood test for CRC screening. Compared with biomarkers for CRC, bio markers for the diagnosis of gastric cancer (GC) and esophageal cancer have been assessed in a relatively small number of cohorts (25,35). For the early detection or screening of cancers including CRC, GC, and esophageal cancer, the sensitivity of ctDNA analy sis needs to be improved. Analysis that can be performed using a few milliliters of blood would be suitable for cancer screening; however, increasing the analytical sensitivity beyond 0.1% may not provide clinical benefits because it also leads to difficulties in distinguishing oncological mutations and sampling noise. In fact, cancerassociated genomic alterations have been found in plasma from healthy individuals (36). In addition, because many cancers share common gene mutations such as TP53 mutations and KRAS mutations, ctDNA presents challenges in the detec tion of the specific organ sites of malignancies. To overcome these issues in ctDNA, the methylation profiling of cfDNA has been investigated in cancer diagnosis. Methylation haplotyping in plasma is a promising strategy for the early detection of a tumor and its primary growth site (37). Studies have reported the utility of methylation scores from over 9,000 CpG sites in cfDNA for cancer detection, with 76.3% accuracy for the prediction of cancer type (38).
Despite the above hurdles to the use of ctDNA in cancer screening, it is expected that the clinical use of ctDNA is less than a decade away because of its utility and convenience in cancer screening (Figure 1).

PROGNOSiS AND DeTeCTiON OF ReSiDUAL DiSeASe
Following curative therapy for gastrointestinal malignancies, ctDNA may be a potential biomarker for minimal residual disease. The detection of ctDNA even in the absence of any other clinical  evidence of disease may mean that the patient has higher risk of relapse. In a cohort of 230 patients with stage II CRC, the assess ment of ctDNA using the SafeSeqS NGS method at the first visit after surgery indicated that recurrencefree survival at 3 years was 0% in a ctDNApositive group and 90% in a ctDNAnegative group (39). Other studies have also demonstrated that the per sistent detection of ctDNA after local therapy (surgery or radical radiotherapy) predicts a high risk of relapse in patients with colon cancer (40,41). In addition, methylated BCAT1/IKZF1 have been evaluated as biomarkers for CRC (42,43 (44). Another study demonstrated that the level of ctDNA was associated with tumor recurrence in patients who underwent curative surgery for GC (45). Similarly, several studies have reported the tumorassociated mutations in ctDNA and the prognosis of patients with esophageal cancer (46,47); however, these studies included a limited number of patients, and further investigations are warranted. Almost all of these studies followed a retrospective design and provided limited validation for clinical use in gastrointestinal malignancies. One of the ideal applications of ctDNA is in the early detection of residual disease or recurrence compared with CT imaging and tumor markers. A more attrac tive idea is patientspecific ctDNA panels in patients who have undergone curative surgery (41,48). Individual surgical tumor samples may provide a great opportunity to obtain tumor DNA from each patient to guide the design of patientspecific ctDNA panels from peripheral blood samples. Despite hurdles such as tumor heterogeneity, validation, regulatory issues, and quality control of individual panels, patientspecific ctDNA panels are potential biomarkers for postoperative monitoring.

biOMARKeRS OF CHeMOTHeRAPY ReSPONSe AND ReSiSTANCe iN MeTASTATiC DiSeASe
Another clinical potential of ctDNA is in the determination of systemic chemotherapy regimens, the prediction of response to chemotherapy, and the identification of resistance mechanisms. The short halflife of ctDNA enables the realtime monitoring of cancer burden, in contrast to radiological imaging or tumor markers. Indeed, a prospective study of 53 patients with metastatic CRC demonstrated that early changes in ctDNA during firstline chemotherapy predicted later radiologic response. Significant reductions in ctDNA levels were observed before cycle 2 and were correlated with CT response at 8-10 weeks (odds ratio 5.25 with a 10fold ctDNA reduction; p = 0.016) (49). Bloodbased monitoring is an ideal strategy during cancer treatment because of its minimal invasiveness and avoidance of radiation exposure.

RAS Mutations in Metastatic CRC
The assessment of RAS status has been mandatory in patients with metastatic CRC to predict the response of cetuximab and panitumumab to antiEGFR antibodies (50)(51)(52)(53)(54)(55). A double blinded prospective study of 106 patients with mCRC has been performed to compare the KRAS mutation status assessed using tumor tissue via routine goldstandard methods to that assessed using plasma DNA via qPCRbased methods; the resultant speci ficity and sensitivity for the detection of KRAS point mutations were 98 and 92%, respectively, resulting in 96% concordance (28). In addition, the retrospective exploratory analysis in a biomarker subgroup of the CORRECT trial, which was a phase III trial investigating the efficacy and safety of regorafenib in patients with mCRC, confirmed the utility of detecting KRAS, BRAF, and PIK3CA mutations in ctDNA. Plasma DNA detected with BEAMing in 503 patients demonstrated that mutation status in ctDNA changed dynamically during chemotherapy and differed from that in pretreatment archival tissue (56). Currently, the OncoBEAM RAS CRC assay is the only European committee in vitro diagnostic test for RAS mutations in ctDNA. This assay is a qualitative PCRbased test and allows for the detection of 34 mutations within exons 2, 3, and 4 of KRAS and NRAS genes from a single blood sample. Four large cohort studies have been reported to achieve high concordance of approximately >90% (range, 89.7-93.3%) between OncoBEAM using plasma ctDNA and tumor tissue analysis in patients with CRC (57)(58)(59)(60).
Acquired resistance to antiEGFR antibody therapy has also been found by using ctDNA analyses. ctDNA from 28 patients receiving panitumumab monotherapy was assessed using qPCR, and 9 out of 24 (38%) patients whose tumors were initially KRAS wildtype showed KRAS mutations in ctDNA after panitumumab treatment (61). This study suggested that the emergence of KRAS mutations is a mechanism of resistance to antiEGFR therapy and that these mutations may be detected in ctDNA as a more sensi tive monitoring tool than radiological imaging. More recently, other studies have also demonstrated mutations associated with the resistance and decline of mutant KRAS clones after the with drawal of antiEGFR therapy (62,63).

bRAF Mutations in Metastatic CRC
BRAF mutations have been confirmed to be associated with poor prognosis in patients with metastatic CRC; moreover, the limited efficacy of antiEGFR therapy in patients with BRAFmutant metastatic CRC has been shown in several studies (64)(65)(66). The analysis of BRAF mutations in ctDNA by using qPCRbased methods has shown specificity and sensitivity of 100% (28). Based on preclinical studies (67,68), the clinical trials of dual EGFR and MAPK signaling pathway inhibition in patients with BRAFmutant metastatic CRC is ongoing. In a phase Ib study of a combination therapy of dabrafenib, trametinib, and panitumumab, BRAF V600E mutant burden in ctDNA was more markedly reduced in responders than in nonresponders, and the emergence of RAS mutations was seen with disease progression in 9 of 22 patients (41%) (69). This exploratory analysis sug gested that the monitoring of BRAF V600E mutant fraction in ctDNA could effectively predict response to combination therapy including a BRAF inhibitor and that overcoming the emergence of RASmutant subclones is important in combating resistance to this combination therapy.

Other Alterations in Metastatic CRC
HER2 or MET amplification is also known as a mechanism of resistance to antiEGFR therapy in patients with metastatic CRC. The patientderived xenograft models of HER2amplified CRC showed resistance to antiEGFR therapy (70,71). In addition, the frequency of HER2 amplifications increased from approximately 3% in treatmentnaïve patients to over 10% in patients who were administered antiEGFR therapy (72).
Although there are few studies regarding the concordance of HER2 status between ctDNA and tissue samples, 4 of 18 (22%) patients exhibited HER2 amplification in ctDNA by digital PCR after cetuximab therapy despite being negative for HER2 amplification prior to antiEGFR therapy (73). The promising results of trastuzumab and TDM1 combination therapy in the HERACLES trial (71) have encouraged clinical trials in patients with HER2positive metastatic CRC such as the MyPathway trial (74) and the TRIUMPH trial (75); notably, the TRIUMPH trial includes patients with HER2 amplification detected using not only tissue samples but also ctDNA analysis using an NGSbased method.
Another important alteration that causes resistance to anti EGFR therapy is MET amplification. A preclinical model of MET amplified CRC also showed resistance to antiEGFR therapy (76). In fact, MET amplification in ctDNA was detected using NGS in 12 of 53 (22.6%) patients who showed disease progression with antiEGFR therapy; no such amplification was detected in patients before cetuximab therapy. Furthermore, MET amplifica tion in ctDNA was not detected in patients with RAS mutations after cetuximab therapy, thus suggesting that MET amplification is one of the mechanisms (other than RAS mutations) that cause resistance to antiEGFR therapy (77). In a phase Ib trial of cabo zantinib and panitumumab combination therapy, the preliminary evidence of efficacy in patients with METamplified metastatic CRC was reported (78).

HeR2 Amplification in GC
The amplification of the HER2 gene or overexpression of the HER2 protein, which contributes to cancer progression, has been reported in approximately 20% of patients with advanced GC (79,80). According to the results of the ToGA trial, HER2 is a key biomarker of HER2targeted therapy using trastuzumab for advanced GC (79). The goldstandard diagnostic method for detecting HER2 positivity and suitability for trastuzumab therapy is an immunohistochemistry score of 3+ or 2+ with a positive ctDNA in Gastrointestinal Malignancies Frontiers in Oncology | www.frontiersin.org July 2018 | Volume 8 | Article 263 result in fluorescence in situ hybridization. A retrospective study of 52 patients with advanced GC and 40 healthy volunteers demonstrated that the plasma HER2-RPPH1 ratio (with RPPH1 as a reference gene) was significantly higher in patients with HER2positive tumors than those with HER2negative tumors (81). More recently, the droplet digital PCR of HER2 copy number in ctDNA has been reported. In a study of 60 patients with GC, including 17 patients who developed recurrence and 30 healthy volunteers, preoperative plasma HER2 ratio correlated with tumor HER2 status; postoperative plasma HER2 ratios were high during the recurrence of tumors, which were diagnosed as HER2negative tumors in surgery samples (82). Considering that HER2 status may be altered after recurrence, the HER2 copy number analysis in ctDNA enables the realtime evaluation of HER2 status and leads to more effective treatment choices with HER2targeted agents.

FUTURe PeRSPeCTiveS
Overall, the data generated in all studies discussed above support the potential role of ctDNA in the diagnosis and treatment of patients with gastrointestinal malignancies. Despite a few limi tations, including the standardization of detection and ctDNA quantification, the sensitivity, and concordance between ctDNA and tissue biopsies that currently hinder the routine use of ctDNA in clinical trials and clinical practice, its use would allow a deeper understanding of cancer biology and enable better cancer treatment, thus leading to improvements in patient survival.
In the context of clinical trials for metastatic disease treatment, several studies are ongoing or have been conducted using eligibility criteria based on gene alterations in ctDNA. A prospective study on the comprehensive ctDNAguided treatment of advanced GC and lung cancers is ongoing in Korea (83). Another trial called the Targeted Agent and Profiling Utilization Registry, which is a large basket/umbrella trial sponsored by the American Society of Clinical Oncology, is accepting patient selection on the basis of ctDNA analysis (NCT02693535). In addition, an umbrella trial in patients with mCRC based on the molecular profiling of ctDNA, including the status of HER2, BRAF V600E, BRAF nonV600E, MET, or high tumor mutation burden, is ongoing in Japan. If promising results are obtained in these clinical trials, ctDNA will be used in routine clinical practice and in clinical trials in the near future.
Economic and regulatory issues still hinder the practical use of ctDNA. Although most guidelines recommend that comprehen sive molecular profiling should be conducted, the substantial costs of NGS assays lead many community oncologists to rely on PCR tissue tests and do not understand the added benefit of a com prehensive genomic test. In addition, emerging ctDNAguided clinical trials are essential to obtain approval for the use of ctDNA in clinical practice. These barriers need to be challenged, perhaps initially in patients with CRC, which is one of the most prevalent gastrointestinal malignancies worldwide. Simultaneously, further studies are needed on other gastrointestinal malignancies such as esophageal cancer and GC to identify the best gene biomarkers that are detectable in ctDNA for the diagnosis, prognosis, and prediction of therapy response.

CONCLUDiNG ReMARKS
The potential role of ctDNA in gastrointestinal malignancies has been shown in basic studies, retrospective studies, and limited prospective studies. A paradigm shift in cancer diagnosis and treatment in ctDNAbased clinical trials and clinical practice will occur in the near future, thus leading to the availability of more DNA sequence information compared with that in the past decade. Although some limitations continue to exist on the use of ctDNA in clinical practice and clinical trials, ctDNAbased personalized therapy promises to improve patient outcomes and quality of life.

AUTHOR CONTRibUTiONS
This review was drafted by KS and DK and was revised by HB.