Precision medicine for gastrointestinal cancer: Recent progress and future perspective

Abstract

Gastrointestinal (GI) cancer has a high tumor incidence and mortality rate worldwide. Despite significant improvements in radiotherapy, chemotherapy, and targeted therapy for GI cancer over the last decade, GI cancer is characterized by high recurrence rates and a dismal prognosis. There is an urgent need for new diagnostic and therapeutic approaches. Recent technological advances and the accumulation of clinical data are moving toward the use of precision medicine in GI cancer. Here we review the application and status of precision medicine in GI cancer. Analyses of liquid biopsy specimens provide comprehensive real-time data of the tumor-associated changes in an individual GI cancer patient with malignancy. With the introduction of gene panels including next-generation sequencing, it has become possible to identify a variety of mutations and genetic biomarkers in GI cancer. Although the genomic aberration of GI cancer is apparently less actionable compared to other solid tumors, novel informative analyses derived from comprehensive gene profiling may lead to the discovery of precise molecular targeted drugs. These progressions will make it feasible to incorporate clinical, genome-based, and phenotype-based diagnostic and therapeutic approaches and apply them to individual GI cancer patients for precision medicine.

Core tip: Gastrointestinal (GI) cancer is one of the most common leading causes of cancer death worldwide. Hence, any effort in early diagnosis, choice of appropriate therapeutic strategies can have a pivotal role in reducing the disease related mortalities. Our review purpose to clarify the current advancement for precision medicine in GI cancer by elucidating the benefit of liquid biopsy, multiple gene panel, novel biomarkers and surgery in GI cancer.

INTRODUCTION

Precision medicine is a strategy designed to treat individual patients with the most suitable therapy at the most appropriate time based on the patient’s biologic and molecular features, using the analyses of genes of the patient’s cancer cells with next-generation sequencing (NGS). Such analyses can detect cancer-specific gene mutations, and molecular targeted drugs can be designed to be effective for one or more specific gene mutations. Precision medicine is thus a type of tailor-made and personalized therapy. The use of inappropriate medicine may not only do not benefit, but lead to cancer progression. As the accessibility to tumor genome sequencing technologies increases, genome-driven cancer treatment has emerged as a favorable approach[1]. The increasing number of patients who undergo multigene sequencing of their cancer can thus expect to be informed of their genomic alterations that could effectively be targeted with corresponding drugs[2].

Gastrointestinal (GI) cancer has a high tumor incidence and mortality rate worldwide[3]. Although colorectal cancer (CRC) could be largely managed, which results in long-term survival by a combination of drugs even in patients with widespread stage and GI lymphoma (e.g., MALT) may also be associated with good response and prolonged survival, the overall prognosis of patients with advanced GI cancer remains poor. Precision medicine approaches are currently being applied with molecular targeted and immune-based therapeutics across a variety of malignancies, such as advanced melanoma and non-small-cell lung cancer (NSCLC)[4,5]. Although GI cancer has been investigated with biomarkers (e.g., Ras and HER2 status), the development of biomarkers as well as targeted therapies for GI cancer has fallen behind compared to those developed for other malignancies. Analyses of liquid biopsies, multiple gene panels, and well-designed prospective trials are necessary to move the treatment of GI cancer forward. In this review, we summarize the progression of precision medicine in GI cancer in terms of specimens, assays, further biomarker information, surgery, and future perspectives.

LITERATURE SEARCH

We first conducted a search of the PubMed database for English articles using the medical subject heading terms in combination with “gastrointestinal cancer”, “esophageal cancer”, “gastric cancer”, “colorectal cancer”, “precision medicine”, “liquid biopsy”, “gene expression profiling”, “biomarker”, “molecular targeted therapy”, and “gene panel”. Relevant articles which were chosen from experimental studies and clinical trials since 1989 were involved as well as articles which were related to the disease processes. Articles which did not deal with the precision medicine of GI cancer were excluded from this review. Liver and pancreatic cancer and GI stromal tumor were not covered in this review due to the limited scope of the topic.

LIQUID BIOPSY

Conventionally, tissue biopsies have been used to access the molecular information of tumors, such as the histology and gene mutation[6]. However, the practical use of consecutive tissue biopsies to monitor for mutations is limited due to patient discomfort, pain, and risks associated with repeat tissue biopsies, and difficulty in capturing intra-tumor heterogeneity[6]. These shortcomings highlight the need for more innovative screening. One promising alternative to tissue biopsy is a new approach that may change the principles of cancer treatment. The term ‘liquid biopsy’ refers to the analysis of tumor-derived biomarkers identified from biological fluids of patients with malignancies. Even though peripheral blood is the major specimen for the liquid biopsy approach, tumor biomarkers can be isolated from various body fluids including urine, pleural effusions, ascites, and cerebrospinal fluid[7].

The liquid biopsy technique been studied to a great extent and is attracting further attention as it leads to efficient therapeutic interventions, reducing the therapeutic cost and significantly improving patient outcomes and overall survival[8]. Analyses of liquid biopsy specimens can provide comprehensive real-time data of the tumor-associated changes in an individual patient with a malignancy. These data can be used for cancer screening, the detection of minimal residual disease, drug selection (including sensitivity to anticancer agents), monitoring recurrence, and monitoring the patient’s response to targeted agents (including drug resistance)[9]. For example, an analysis of NSCLC patients’ plasma for epidermal growth factor receptor (EGFR) to determine the existence of a T790M mutation is widely used[10]. Liquid biopsies could become a new tool with a significant impact on cancer therapy.

Studies of liquid biopsy methodology have focused on the analysis of circulating tumor cells (CTCs), circulating tumor free (cf) DNA or RNA, and tumor-derived extracellular vesicles (exosomes)[11]. For the most effective discussion of the details of liquid biopsy methodology, it is essential to understand the different types of cancer-related biomarkers and their respective molecular aspects.

CTCs

CTCs are tumor cells that are mainly detached from primary or metastatic lesions. They circulate through the body fluid to metastatic sites, either as a single cell or in clusters, which lead to the establishment of one or more secondary tumor foci[12]. The United States Food and Drug Administration (FDA)-cleared CellSearch system has enabled the enumeration of CTCs in cancer patients, and this has made it possible to determine disease activity and patients’ treatment responses, which rely on the expressions of epithelial cell adhesion molecule and cytokeratin on cancer cells in blood[13]. The authors of a previous study described the establishment of colon CTC cultures and permanent cell lines which provided in vivo experimental models. These experiments may provide genetic and epigenetic information on tumor biology, and they may help assess the cells' sensitivity to anticancer drugs[13]. However, the number of CTCs is generally low in patients with GI cancer[14], and this limits the clinical applications of CTC analyses in site of the progression of various methods[14].

Circulating tumor DNA (ctDNA)

ctDNA has emerged as another component of liquid biopsies as a quantitative marker of tumor DNA, reflecting genomic alterations in the blood[15,16]. Compared to the detection of CTCs, the ctDNA-based approach provides more information about a patient-specific disease and treatment. Further benefits of the use of ctDNA as a marker is that ctDNA measurements can provide the real-time pathology of the patient’s disease and higher sensitivity for the early detection of cancers[17]. A previous study showed a significantly broad range for ctDNA among patients with CRC (22–3922 ng/mL of blood) compared to healthy subjects (5-16 ng/mL of blood)[18]. Liquid biopsy analyses may take the place of tissue testing for assessing the mutational status of RAS in patients with CRC. The OncoBEAM RAS CRC Assay identifies the cfDNA of the most frequent KRAS and NRAS mutations by using BEAMing technology[19].

MicroRNAs (miRNAs)

In addition to the quantification of cfDNA, circulating transcriptome is also detectable in the serum of individuals with malignancies. The circulating transcriptome consists of both coding and noncoding RNAs, such as miRNAs or long noncoding RNAs (lncRNAs)[20]. Although RNA is generally unstable in blood, microRNA (miRNA) comprises stable, short, noncoding molecules made of 18-25 nucleotides. This endogenous, single-stranded RNA mediates the expression of nearly 30% of protein-encoding genes in humans[21]. MiRNAs can be analyzed by targeted or RNA sequencing methods, with miRNA signatures observed to be significantly deregulated in cancer patients compared to healthy parsons, and these analyses may become useful in cancer diagnosis and prognosis.

Exosomes

Exosomes are nanosized vesicles (40-150 nm)[22]. These small, membrane-bound vesicles can transport a number of biomolecules which lead to the modification of the activity of recipient cells[22]. Compared to CTCs and ctDNA, exosomes have advantages in several aspects, including their homogeneous size distribution. In addition, due to the particular form of exosomes, they can be distinguishable by electron microscopy. Previous studies have obtained evidence that the exosome-mediated recruitment and manipulation of the tumor microenvironment is a critical step in the formation process of metastasis[23].

Liquid biopsy in GI cancer: Toward clinical applications

The clinical utility of a liquid biopsy has been studied in different clinical phases of GI cancer, from the screening for this disease to the identification of outcome factors in early GI cancer, the detection of minimal residual tumor, drug selection, and monitoring for recurrence and the patients’ response to targeted agents. Current advances of liquid biopsy as diagnostic, monitoring and predictive markers in GI cancer are summarized in Table 1 and Table 2.

Table 1.

Current progress of circulating tumor cells, circulating tumor DNA and stool DNA as diagnostic, monitoring and predictive markers in gastrointestinal cancer

Liquid biopsy	Patients/ controls	Organs	Source of fluid	Abnormalities	Technology	Target	Clinical setting	Ref.
CTCs	140/0	EC	B		FIHC	CK19, CD45	Prognosis	Li et al[82], 2016
CTCs	NA	EC	B		ISET	NA	Prognosis	Han et al[83], 2019
CTCs	116/31	GC	B		FAST-disc	EpCAM, CK, CD45-	Diagnostic	Kang et al[84], 2017
CTCs	81/31	GC	B		ISET	CK8/18/19, Vimentin, CD45	Prognostic	Zheng et al, 2017
CTCs	101/31	GC	B		CellSearch and IF-FISH	EpCAM, CK8, CK18, CK19, CD45-, HER2	Predictive	Mishima et al[86], 2017
CTCs	121/0	CRC	B		Cyttel method/imFISH	CD45	Prognostic	Wang et al[87], 2019
ctDNA	11/0	EC	P, T, NT	Mutation	WES and NGS panel		Diagnostic /Therapeutic	Luo et al[88], 2016
ctDNA	13/0	EC	P, T	Mutation	NGS panel		Predictive	Ueda et al[89], 2016
ctDNA	63/0	EC	P	Copy number status	qPCR	CCND1	Predictive	Komatsu et al[90], 2014
cfDNA	32/0	GC	P	Copy number status	cfDNA NGS testing	ERBBB2	Therapeutic	Kim et al[91], 2018
ctDNA	277/0	EC/GC	P, T	Mutation	MassARRAY	TP53, PIK3CA, ERBB2, KRAS	Diagnostic /Prognostic	Kato et al, 2018
ctDNA	70/0	GC	P, T	Mutation	NGS panel	HER2	Therapeutic	Gao et al[93], 2017
cfDNA	60/30	GC	P	Mutation	Droplet digital PCR	HER2	Therapeutic	Shoda et al[94], 2017
ctDNA	1/0	GC	P, T	Mutation	NGS panel	MET	Therapeutic	Du et al30], 2017
ctDNA	230/0	CRC	B	Mutation	Safe-SeqS assay	NA	Prognostic	Tie et al[95], 2016
cfDNA	22/0	CRC	S	Mutation	NGS/dPCR	TP53, KRAS, APC, PIK3CA, BRAF, FBXW7, NRAS	Diagnostic /Prognostic	Furuki et al[96], 2018
cfDNA	3/0	CRC	P	Mutation	BEAMing	RAS, BRAF, PIK3CA	Predictive	Klein-Scory et al, 2018
cfDNA	20/0	CRC	P	Mutation	Droplet digital PCR	APC, TP53, KRAS, PI3CA	Predictive	Vandeputte et al[97], 2018
Stool DNA	71/22	CRC	Stool	Methylation	QIAamp DNA Stool Mini Kit	SDC2	Diagnostic	Oh et al[98], 2017

EC: Esophageal cancer; GC: Gastric cancer; CRC: Colorectal cancer; NT: Normal tissue; B: Blood; P: Plasma; S: Serum; T: Tumor tissue; PLF: Peritoneal lavage fluid; NGS: Next-generation sequencing; WES: Whole exome sequencing; FIHC: Fluorescent immunohistochemistry; NA: Not avaibable; EpCAM: Epithelial cell adhesion molecule.

Table 2.

Current progress of microRNAs and exosome as diagnostic, monitoring and predictive markers in gastrointestinal cancer

Liquid biopsy	Patients/ controls	Organs	Source of fluid	Abnormalities	Technology	Target	Clinical setting	Ref.
MiRNAs	231/0	EC	Peripheral blood lymphocytes	Polymorphism	SNPShot	KIAA0423 rs1053667, GEMIN3 rs197412	Prognostic	Faluyi et al[99], 2017
MiRNAs	3156/0	EC	S/P	Upregulation/Downregulation	NA	miR-15a, miR-22, miR-31, miR-451, miR-506, miR-613, miR-1297	Diagnostic /Prognostic	Yao et al[100], 2018
MiRNAs	125/0	EC	S/P	Upregulation/Downregulation	RT-PCR	miR-21, miR-223, miR-100, miR-25, miR-375	Diagnostic /Prognostic	Zhang et al[101], 2018
MiRNAs	250/538	GC	Gastric juice	Upregulation	miScript RT kit	miR-421, miR-21, miR-106a, miR-129	Diagnostic	Virgilio et al[102], 2018
MiRNAs	20/20	GC	S	Upregulation	TaqMan OpenArray assays	miR-331 and miR-21	Diagnostic	Sierzega et al[103], 2017
MiRNAs	The miRNA expression profile (GSE29298)	CRC	NA -	Upregulation	NA	miR-198, miR-765, miR-630, miR-371-5p, miR-575, miR-202, miR-513a-5p	Predictive	Zhu et al[104], 2017
MiRNAs	232/0	CRC	S	Upregulation	NA	miR-21, miR-29b, miR-92.	Diagnostic	Carter et al[105], 2017
MiRNAs	61/0	CRC	P	Upregulation	miRVANA PARIS kit	miR-20b, miR-29b, miR-155	Prognosis /Predivtive	Ulivi et al[106], 2018
Exosome	66/20	EC	P	Upregulation	AChE activity	Exosomes	Prognostic	Matsumoto et al[107], 2016
Exosome	30/0	GC	PLF	Upregulation	MiRNA microarray	miR-21, miR-1225-5p	Diagnostic /Therapeutic	Tokuhisa et al[108], 2015
Exosome	232/20	GC	P	Downregula-tion	Taqman microRNA assays	miR-23b	Prediction /Prognostic	Kumata et al[109], 2018
Exosome	227/28	CRC	S	Upregulation/Downregula-tion	qRT-PCR microarray	miR-17, miR-18a, miR-19a, miR-19b, miR20a, miR-92a, hsa-miR-25-106b, hsa-miR-17-92a	Predictive /Prognosis	Matsumura et al[110], 2015
Exosome	108/0	CRC	S	Downregula-tion	The total exosome isolation kit	miR-548c-5p	Prognosis	Peng et al[111], 2018

EC: Esophageal cancer; GC: Gastric cancer; CRC: Colorectal cancer; NT: Normal tissue; B: Blood; P: Plasma; S: Serum; T: Tumor tissue; PLF: Peritoneal lavage fluid; NGS: Next-generation sequencing; WES: Whole exome sequencing; FIHC: Fluorescent immunohistochemistry; NA: Not avaibable.

Cancer screening: The noninvasive nature of a liquid biopsy makes this approach ideal for the early detection of cancer. The evaluation of molecular biomarkers in early-stage cancer patients is necessary for the development of more personalized monitoring and treatment schedules. However, the possibility of detecting a malignancy at an early stage with a liquid biopsy is somewhat limited by the low concentration of circulating biomarkers associated with the low tumor burden. With respect to CRC, screening has been impacted using colonoscopy as the gold standard, mainly because of its high sensitivity and specificity for detecting cancerous and precancerous lesions. Despite its strengths, colonoscopy has certain disadvantages and limitations (e.g., bowel preparation, sedation, aspiration, perforation, and splenic injury).Therefore, continued progress in novel assays, such as fecal immunochemical test, fecal DNA and other molecular markers, can be expected to further displace screening colonoscopy[24]. The Epi proColon^® 2.0 assay (also referred to as the mSEPT9 assay), which was FDA-approved for CRC screening in April 2016, is a qualitative in vitro diagnostic polymerase chain reaction (PCR) test for the detection of mutated methylated septin9 DNA in EDTA plasma derived from patient whole-blood specimens[25].

Detection of minimal residual disease: One of the major fields of the application of liquid biopsy would be the detection of minimal residual disease in patients with surgically treatable tumors. The tumor burden of GI cancer at diagnosis is acknowledged as a pivotal factor of disease assessment before the beginning of treatment. A recent study indicated that somatic KRAS- and BRAF-mutated DNA in the peripheral blood of CRC patients may be a good estimate of CTCs and of surgical clearance of the disease[26].

Drug selection: Chemotherapy is often administered for patients with metastatic disease (e.g., metastasis of regional lymph nodes) in a resected tumor specimen. Although there are a number of different chemotherapeutic agents that can be combined in a variety of chemotherapeutic regimens, the effect of chemotherapy on a specific patient cannot be predicted. Specific ctDNA identification has also been used as guidance for specific systemic chemotherapy and targeted agents. For instance, emerging RAS mutations during therapy with anti-EGFR antibody revealed resistance in patients with metastatic CRC (mCRC)[27]. Some studies found that undetectable low-frequency KRAS-mutant clones may be selected for anti-EGFR treatment by assessing ctDNA in the blood of mCRC patients during anti-EGFR therapy[28,29]. In similar, resistance to crizotinib has been emerged by using serial ctDNA measurements in gastric cancer (GC)[30].

Monitoring recurrence: One of the most challenging tasks in GI oncology is the identification of patients who will benefit from postoperative adjuvant chemotherapy after curative surgery. The histopathologic and molecular tumor features correlated with greater relapse risk (e.g., the TNM classification) only imply a tendency for metastasis; they do not reveal whether metastatic cells were seeded during surgery. The identification of postoperative ctDNA is a definite sign that occult tumor cells remain in the patient.

The authors of a recent study proposed that in patients with CRC, the postoperative detection of ctDNA can be used to monitor the patients for residual disease and predict their future relapse risk with high probability[31]. Moreover, serial ctDNA serves as a tool for the early detection of recurrence during patient follow-up and for the patient’s response to relapse intervention[31]. In CRC, the novel BCAT1/IKZF1 blood test was found to be more sensitive for recurrence compared to carcinoembryonic antigen (CEA) as a marker, and the likelihood of recurrence given a positive BCAT1/IKZF1 result was twice that compared to a positive CEA result[32].

Monitoring patients’ responses to cytotoxic and targeted agents: The most potentially beneficial application of the liquid biopsy approach is the possibility of using this approach to monitor patients' therapeutic responses. In general, ctDNA has seemed to be an early biomarker that can be used to deduce the tumor burden of patients with CRC during chemotherapy and to predict the early therapeutic reaction. Molecular alterations that are related to drug resistance can be identified at an early stage by evaluating ctDNA, and this evaluation can be performed easily for the same patient at different time intervals.

A single-arm phase II trial (Erbitux Study of CPT11, Oxaliplatin, UFToral Targeted-therapy) was carried out in patients with previously untreated KRAS wild-type advanced CRC, using a regimen of irinotecan, oxaliplatin, and tegafur-uracil with leucovorin and cetuximab. The stratification of patients by the CTC count can identify the patients who might benefit the most from an intensive four-drug regimen, avoiding the use of high-toxicity regimens in low-CTC groups[33].

GENE PANEL SEQUENCING IN GI CANCER

Sequencing is often performed to identify cancer-associated gene mutations in patients with advanced cancer. Sequencing panels allow the targeting of multiple genes simultaneously, quickly and accurately through comprehensive bioinformatics in order to exploit the useful information from a single study. The NGS of tumor sample DNA can lead to the optimal clinical treatment by offering diagnostic and/or prognostic data and by contributing to the selection of potential treatment regimens (e.g., molecular-targeted and immune checkpoint blockade therapies). Recent advances in NGS has enabled the performance of whole-genome sequencing, whole-exome sequencing, whole-transcriptome sequencing and RNA sequencing, as well as the detection of enormous genetic aberrations[34].

Due to the progress in sequencing technologies, tissue comprehensive genome profiling has become more widely available in clinical practice. For example, the current National Comprehensive Cancer Network guidelines recommend comprehensive genome profiling in patients with advanced non-small-cell lung adenocarcinoma[4]. Currently, NGS provides faster, cheaper, and more accurate whole-genome sequencing. The Cancer Genome Atlas has revealed the genome profiles of many cancers, including GI cancer[35,36]. Current progress of multiplex gene panels in GI cancer is summarized in Table 3.

Table 3.

Current progress of multiplex gene panels in gastrointestinal cancer

Organs	Panel tested	Number of genes tested	Number of patients	The type of sample	Companion diagnostic indications	Ref.
EC	HiSeq2000	N/A	144	Tumor tissue DNA	CCND1, CDKN2A, FBXW7, MLL2, EP300, CREBBP, TET2, NOTCH1, NOTCH3, FAT1, YAP1, AJUBA, PIK3CA, EGFR, ERBB2	Sawada et al[44], 2016
EC	Exiqon miRNA qPCR panel	168miRNA	140	Serum miRNA	miR-20b-5p, miR-28-3p, miR-192-5p, miR-223-3p, and miR-296-5p	Huang et al[50], 2017
EC	Ion AmpliSeq Custom DNA Panel	12	27	Tumor tissue/Serum DNA	BRAF, DDR2, ERBB2, HRAS, KEAP1, KRAS, NFE2L2, NRAS, PIK3CA, PTEN, RHOA	Pasternack et al[112], 2018
GC	Illumina HiSeq 2000	38	138	Tumor tissue DNA	RHOA, CDH1, PIK3CA, CTNNB1, APC, ARID1A, KMT2C, KRAS	Kakiuchi et al[42], 2014
GC	Illumina HiSeq 2000	N/A	100	Tumor tissue DNA	ARID1A, CDH1, MUC6, CTNNA2, GLI3, RNF43, RHOA	Wang et al[43], 2014
GC	CANCERPLEX	435	207	Tumor tissue DNA	ARID1A, CDH1, ERBB2, CCNE1, KRAS	Ichikawa et al[41], 2017
GC	Ion-Proton sequencer	50	29	Tumor tissue DNA	APC, CTNNB, KRAS, NPM1, FBXW7 ERBB2, FGFR2, KIT	Yoshida et al[113], 2019
CRC	CANCERPLEX	415	201	Tumor tissue DNA	ERBB2, APC, CDKN2A, NRAS, ATM, BLM, BRCA2, NBN, NRE11A	Nagahashi et al[39], 2016
CRC	IT-PGM seqencing	22	77	Tumor tissue DNA	RAS, PIK3CA, FBXW7, BRAF, SMAD4, MET, FGFR1	Capalbo et al [114], 2019
CRC	OncoAim™ DNA panel	39	648	Tumor tissue DNA	KRAS, APC, PIK3CA, SMAD4, BRAF, FBXW7, NRAS	Wang et al[115], 2018
CRC	MiSeq	207	22	Tumor tissue DNA	KRAS, PIK3CA, FBXW7, PTEN, SMAD4, BRAF, CTNNB1, NRAS	Gao et al[116], 2019
CRC	cfDNA panel	14	101	Plasma cfDNA	AKT1, BRAF, CTNNB1, EGFR, ERBB2, FBXW7, GNAS, KRAS, MAP2K1, NRAS, PIK3CA, SMAD4, APC,	Osumi et al, 2018
CRC	TruSight Cancer Sequencing Panel	42	N/A	Blood ctDNA	MLH1, MSH6, PMS2 APC, SMAD4, TP53, BRIP1, CHEK2, MUTYH, HNF1A, XPC	Seifert et al[117], 2019

TP53 was commonly implicated in all references except 28035762, 30297788 and 30523343 (PMID)[50,112,117]. EC: Esophageal cancer; GC: Gastric cancer; CRC: Colorectal cancer; NGS: Next-generation sequencing; FFPE: Formalin-fixed paraffin-embedded; N/A: Not available; EBV: Epstein-Barr virus; MSI: Microsatellite instability; NGS: Next-generation sequencing.

Gene panaels contains the most commonly mutated genes or candidate actionable genes in many cancers. In CRC, KRAS, BRAF, PIK3CA, TP53, CTNNB1, APC, SMAD4, and PTEN are among the most commonly altered genes[37,38]. Patients with CRC in Japan were recently studied using an NGS - based comprehensive genomic panel test[39]. Significant differences in ERBB2, APC, TP53, CDKN2A, and NRAS mutations were identified in the Japanese patients compared to United States patients. Genomic alterations in DNA repair genes (e.g., ATM, BLM, BRCA2, NBN, NRE11A), which are observed in a significant proportion of CRC patients, were also detected. A novel, positive correlation between APC and TP53 mutations with tumors that presented on the left side was reported. A study through deep sequencing in patients with mCRC presented that mutations in TP53, KRAS, APC, KRAS, GNAS, and SMAD4 genes were detected in 69.3%, 39.6%, 23.7%, 16.8% and 13.8% patients, respectively. The mutations in KRAS, GNAS, and SMAD4 were significantly associated with lung metastasis[40].

In GC, comprehensive genomic sequencing using a 435-gene panel in Japanese gastric cancers (GCs) showed that the most frequently mutated gene was TP53 (53.1%), followed by ARID1A (15.9%) and CDH1 (14.0%); ERBB2 amplification (12.1%) was the most frequently observed somatic copy number alteration, followed by CCNE1 (7.2%) and KRAS (5.8%) amplification[41]. Specific subcategories of GCs harbor characteristic genetic aberrations, such as somatic mutations in RHOA and a chimeric gene fusion of CLDN18-ARHGAP26 in diffuse-type GCs[42,43]. The landscape of esophageal cancer (EC)-related gene mutations that regulate the cell cycle (TP53, CCND1, CDKN2A, FBXW7), epigenetic processes (MLL2, EP300, CREBBP, TET2), and the signaling pathways involving NOTCH (NOTCH1, NOTCH3), WNT (FAT1, YAP1, AJUBA) and receptor-tyrosine kinase-phosphoinositide 3-kinase (PIK3CA, EGFR, ERBB2) has been described[44].

Current advances in cancer genome analyses using NGS have revealed an increased mutation burden (a high rate of somatic mutation) in some solid tumors. In GI cancers, one of the leading causes of hypermutation - which is closely related to the generation of neo-antigens - is a defect in DNA mismatch repair (MMR), leading to microsatellite instability (MSI). Several research groups have stated that the tumor mutated burden correlates with the clinical response to immunotherapy[45,46]. GI cancer patients with MMR deficiency and a subsequent hypermutated phenotype achieved outstanding outcomes after anti-PD-1 therapy[47]. This highlights the clinical significance of identifying hypermutated tumors for immunotherapy treatment.

In CRC, mutations in transforming growth factor-beta (TGF-β) signaling genes and BRAF were markedly increased in hypermutated tumors[35]. Mutations in DNA polymerase D1 (POLD1) and DNA polymerase E (POLE) genes have also been described as a cause of hypermutated CRC[48]. The mutation rate of MSI-High GCs was significantly higher than that of MSS tumors[41]. TGFBR2, ACVR2A, SMAD4, and ELF3 as well as the TGF-β pathway are frequently mutated, suggesting a pivotal role in GC pathogenesis, including MSI[43,49].

Given the advances in NGS, it may well become possible in the near future to identify the predominant cancer genes and pathways and tumor-specific genes and pathways. Several multigene assays are available to estimate the risk of relapse after definitive surgery, including the MSK-IMPACT, NCC Oncopanel, Todai OncoPanel, Oncomine Dx Target test, Foundation OneCDx, and CANCERPLEX.

A recent study using the Exiqon panel identified miR-20b-5p, miR-28-3p, miR-192-5p, miR-223-3p, and miR-296-5p as significantly upregulated in the serum of patients with EC, suggesting that these 5-miRNA signatures may serve as potential diagnostic biomarkers for ECs[50]. Similarly, the expressions of seven miRNAs (miR-103a-3p, miR-127-3p, miR-151a-5p, miR-17-5p, miR-181a-5p, miR-18a-5p, and miR-18b-5p) were significantly higher in CRC compared to normal controls[51].

BIOMARKERS FOR GI CANCER

Convincing biomarkers are a crucial aspect of precision medicine, used to match appropriate patients with the right treatment at the right time. Clinically relevant biomarkers are genetic, epigenetic, proteinic, or cellular alterations that are intrinsic to cancer cells. These biomarkers can be used to predict patients' responses to chemotherapy, targeted therapy, or immune checkpoint inhibitors. To date, the most reliable molecular marker in clinical practice is the KRAS gene for patients receiving EGFR - targeted therapy for CRC metastatic disease and HER2 overexpression for patients with HER2-positive GC[52,53]. Detection of BRAF mutation status was also recommended due to the ineffectiveness of anti-EGFR therapy for CRC patients with BRAF mutations[54]. Although there is a crucial need for novel diagnostic and prognostic biomarkers to improve GI cancer prognosis, these tools are still being investigated. In this section, we summarize the current advances of biomarkers in GI cancer, with a focus on the development of new biomarkers that are of predictive and/or prognostic values.

Another biomarker for therapeutic target in GI cancer may be MET. A multicenter phase II study demonstrated antitumor activity of small-molecule MET inhibitor was shown in MRT-amplifier gastric/gastroesophageal/esophageal adenocarcinoma[55]. A recent study using whole-exome sequencing characterized KDR/VEGFR2 somatic mutations as potential genetic biomarkers of patients’ responses to antiangiogenic cancer therapies[56]. Interestingly, a recent cohort study presented that ALK, ROS1, and NTRK rearrangements classified a new subtype of mCRC with particularly poor outcome[57]. Rearrangements of ALK, ROS1, and NTRK were more frequently observed in elderly patients with right-sided tumors and node-spreading, RAS wild-type, and MSI-high cancers. As noted above, ctDNA and RNA-based biomarkers provide high specificity and are ideal as predictive markers for monitoring patients' responses to chemotherapy as well as tumor progression[52]. MMR-deficiency deficiency has emerged as another meaningful biomarker. MMR deficiency has been shown to be positively prognostic for outcome in patients with GC and CRC[58,59]. Notably, MMR deficiency is a variety of cancer predictor for response to anti-PD-1/PD-L1 blockade therapies[60]. Tumor-infiltrating lymphocytes (TILs) are the major type of infiltrating immune cells[36]. The density of TILs is considered to be an indication of the host immune response against tumor cells. To date, the density of TILs have been investigated as a useful prognostic factor in GI cancer[61]. Collectively, research has moved towards the identification of mutations in key genes involved in the progression of GI cancer. In the meanwhile, large-scale prospective clinical studies for evaluating the sensitivity and specificity of these biomarkers are required before their application in clinical practice, due to their low mutational burden and insufficient specificity. The approved biomarkers and candidate biomarkers of GI cancer are summarized in Table 4.

Table 4.

Current progress of biomarkers associated with diagnosis, prognosis, prediction of therapeutic response in gastrointestinal cancer (excluding liquid biopsy)

Market	Tumor type	Alteration	Clinical setting	Ref.
HER2	GC, CRC	Amplification, Overexpression	Predictive	Bang et al[118], 2010; Sartore-Bianchi et al[119], 2016
KRAS	CRC	Activating mutation within catalytic RAS domain	Predictive	Wormald et al, 2013; Febbo et al, 2011; Schmoll et al[122], 2012; Locker et al[123], 2006
NRAS,	CRC	Overexpression	Prognostic/Predictive	Hu et al[124], 2018
BRAF	CRC	Mutation	Prognostic/Therapeutic	Tie et al[54], 2011
KDR	CRC	Mutation	Predictive	Loaiza-Bonilla et al[125], 2016
VEGF-D	CRC	Overexpression	Predictive	Tabernero et al[126], 2018
AKT	GC	Activation	Predictive	Ito et al[127], 2017
PTEN	GC	Downregulation	Predictive	Kim et al, 2017
NTRK fusion	CRC	Overexpression	Predictive	Drilon et al[129], 2018
ALK	CRC	Rearrangement	Prognostic	Pietrantonio et al[57], 2017
POLE	CRC	Mutation	Predictive	Domingo et al[130], 2016
MMR	GC, CRC		Predictive	Llosa et al[131], 2015
PD-L1	CRC	Mutatoin	Prognostic	Eriksen et al[132], 2019
Tumor infiltrating lymphocyte	GC, CRC	Overexpression	Prognostic	Iseki et al[133], 2018
CagA	GC	Upregulated	Diagnostic	Saju et al[134], 2016
Gastrokine 1	GC	Downregulated	Diagnostic	Altieri et al[135], 2017
MEK	CRC	Activation	Predictive	Martinelli et al[136], 2017
PIK3CA	CRC	Mutation	Prognostic/ Therapeutic	Jehan et al[137], 2019; Schmoll et al[122], 2012
TP53	EC, GC, CRC	Mutation	Prognostic	Schmoll et al[122], Guo et al[138], 2017
CTNNB1	CRC EC, GC	Mutation Overexpression	Prognostic Prognostic	Gao et al[116], 2019; Szász et al[139], 2016; Ishiguro et al[140], 2016
APC	CRC	Mutation	Prognostic	Liang et al[141], 2017; Chen et al[142], 2013
IGFR-!R	CRC	Upregulation	Prognostic	Codony-Servat et al[143], 2017
SFRP2	CRC	Hypermethylation	Diagnostic/Prognostic	Tang et al, 2011
UGT1A1	CRC	Hypermethylation	Predictive	Crea et al[145], 2011
SMAD4,	EC, GC, CRC	Downregulation	Prognostic/Predictive	Salem et al[146], 2018; Wasserman et al[147], 2019
MET	EC, GC	Amplificatoin	Predictive	Van Cutsem et al[55], 2018
CDKN2A	EC,	Methylation	Diagnostic	Zhou et al, 2017
ATM	GC, CRC	Mutaion/Downregulation	Prognostic	Randon et al[149], 2019; Han et al[83], 2017
BLM,	CRC	Mutaion/Polymorphisms	Diagnostic	de Voer et al[150], 2015; Frank et al[151], 2010
BRCA1/2,	CRC	Mutaion	Diagnostic	Oh et al[152], 2018
ARID1A	GC	Mutation	Predictive	Wei et al[153], 2014
	CRC	Overexpresion	Prognostic	Ronchetti et al[154], 2017
CDH1	GC	Mutation	Diagnostic	Hansford et al[155], 2015
	CRC	Polymorphism	Diagnostic	Grünhage et al[156], 2008
CCNE1	GC	Amplification	Therapeutic	Ooi et al[157], 2017
RHOA	GC, CRC	Overexpression	Prognostic	Chang et al[158], 2016
CCND1	EC	Amplification/Overexpression	Diagnostic	Hu et al[159], 2016
	CRC	Polymorphism	Diagnostic	Grünhage et al[156], 2008
FBXW7	CRC	Mutation	Prognostic	Korphaisarn et al[160], 2017
NOTCH1	EC	Mutation	Prognostic	Song et al[161], 2016
	CRC	Gene copy number	Prognostic	Arcaroli et al[162], 2016
NOTCH3	CRC	Overexpression	Predictive	Ozawa et al[163], 2014
YAP1	EC, GC, CRC	Overexpression	Prognostic	Zhang et al[164], 2018

EC: Esophageal cancer; GC: Gastric cancer; CRC: Colorectal cancer; MMR: mismatch repair; PD-L1: programmed death ligand 1; PD-1: programmed death-1; POLE: DNA polymerase 1; HER2: human epidermal growth factor receptor type2; EGFR: epidermal growth factor receptor.

Future research may identify biomarkers that enable cost-effective and noninvasiveness treatments for GI cancer. It is also necessary to determine the best prognostic panel of biomarkers and to find predictive biomarkers to help in the selection of the most suitable therapy.

PRECISION SURGERY IN GI CANCER

Precision medicine is a general concept and is thus not limited to genetic detection. Although surgery is the most effective treatment for localized GI cancer and is often curative, an insufficient removal of a tumor results in secondary tumor foci for which the existing chemotherapeutics and/or radiation would be ineffective. In this finally section, we would like to discuss the progress of the precision treatment of GI cancers through surgery.

Fluorescence-guided surgery for GI cancer

Surgery has been said to provide the most benefit for patients with GI cancer. When R0 resection was carried out in a series of GI cancer patients, the local 5-year relapse rate was significantly improved[62]. The reported rates of local recurrence and distant metastasis were high at 2.6% and 30% of patients who underwent an R0 resection[63,64]. Real-time imaging to find positive surgical margins during a surgical procedure may be useful to diminish the rates of recurrence. Intraoperative fluorescence imaging, or fluorescence-guided surgery (FGS), can offer highly reliable tumor visualization for localization and margin identification[65]. The targeted fluorescent labeling of cancer cells may therefore alter the ways we detect and treat cancer.

Indocyanine green (ICG) is applied clinically to define liver tumor margins and biliary anatomy. The authors of a recent meta-analysis stated that intraoperative ICG fluorescence angiography has been demonstrated to reduce anastomotic leakage rates after colorectal resection[66]. In CRC, ICG fluorescence lymphangiography can be used to detect the primary tumor, its lymphatic drainage, and potentially malignant nodes, which may change the operative plan[67]. FGS can thus serve as a surgical guide with the potential to provide benefits for patients with GI cancer.

Sentinel node navigation surgery

Many investigators have described the potential usage of sentinel node (SN) navigation surgery in patients with early-stage EC and GC who have no lymph node metastasis preoperatively[68,69]. In early stage upper GI cancer, SN mapping provides significant information about an individual patient’s metastatic situation and enables the modification of the patient’s surgery. Several single-institution investigations have noted pivotal benefits of SN mapping for early EC, especially when using the radio-guided method[70]. Clinically T1 esophageal cancers were suitable targets for SN mapping, because in T3 or T4 tumors as well as those with lymph node metastasis, the original lymphatic routes can be obstructed, which leads to a high rate of false-negative outcomes. SNs were detected in 95% of patients, and the accuracy was as high as 94%[71]. Moreover, SNs were identified widely from the cervical area to the abdominal area, which allows the partial resection of the distal esophagus via the laparoscopic trans-hiatal approach without extensive mediastinal lymph node dissection when the SNs are identified only in the abdominal region and are pathologically negative in cT1N0 cases of the distal esophagus[71]. The precise indications for laparoscopic surgeries (e.g., partial resection and segmental gastrectomy for cT1N0 GC) based on the SN status could be individually determined. SN navigation surgery could be a strategy to ensure a better prognosis than conventional operative strategies.

FUTURE PERSPECTIVES

Precision medicine is the application of the latest biological technology that takes into account the patient’s living environment along with the patient’s clinical data (as well as molecular imaging techniques and bioinformatics technology) to achieve accurate diagnoses and treatments. It is difficult to determine the precise clinical and biological significance for each individual patient because of the inconsistency in biological features on the human genome[71]. Moreover, the complexity of the NGS data-analysis process makes it impractical for oncologists to understand the meanings and uncertainties of the results easily. A systematic and easily interpreted system with an accessible database is immediately necessary for detecting specific genomic alterations and genotype-matched therapeutic options with clinical practice. Although it would be impossible to completely prepare a treatment plan for each individual case, more suitable treatment based on the unique genomic changes of each patient's tumor could be adapted.

The recent progress in the use of precision medicine in GI cancer was summarized in this review. Regarding treatment, we expect that the narrowing down of the number of eligible patients in accord with dose setting, schedule setting, and the selection of concomitant drugs based on the mechanism of molecular targeted agents will lead to effective therapy customized to each individual. For GI cancers, there is an urgent need for preclinical models to identify and select suitable target for therapy. Recent developments in stem cell biology have enabled the in vitro generation of complex three-dimensional (3D) multicellular stem cell-derived constructs that mimic their corresponding organ in vivo[72]. These organ-like structures denoted as organoids. Patient-derived organoids (PDOs) may be an attractive candidate for an appropriate cancer model that is able to identify the most effective therapy for individual patients with currently available drugs in a timely manner, but also the future of regenerative medicine. therapies, 3D organoids have been advanced for several cancer types and been shown to effectively recapitulate tumor specific characteristics, which may lead to facilitate the development of precision medicine[73]. A recent study demonstrated that the feasibility of GC PODs from endoscopic biopsies and also suggest that endoscopic-derived PDOs may serve as an precise surrogates of the primary lesion of tumor, which may lead to possess the superiority to drug sensitivity screening and precision therapies[74]. Other study using patient-derived CRC organoids presented that of all RASGTPases activating proteins, only neurofibromin (NF1) deficiency facilitate cell survival and prompted EGF-independent tumor cell growth in human CRC samples, suggesting that NF1 protein levels should be measured in CRCs prior to initiate of targeted therapy against the MAPK pathway[75].

Our understanding of the fundamental biology of GI cancer is continually advancing. GI cancer is a heterogeneous disease with significant differences between patients in prognosis and therapeutic response. Part of these differences can be explained by the molecular diversity detected in GI cancer. So as to provide a more overall insight into this complexity, biologically distinct molecular subtypes of GI cancer based on gene expression analyses were defined and validated. EC is classified into three distinct molecular subgroups based on gene analysis findings[76]. The first subgroup (ESCC1) includes tumors that respond poorly to chemoradiotherapy, leading to poor prognoses. The principal gene alteration identified is NRF2 pathway disruption. The second subgroup is ESCC2, characterized by the mutation of NOTCH1, ZNF750, KDM6A, KDM2D, PTEN, PIK3R1, and CDK6 amplification. This subgroup is also associated with white blood cell infiltration. The last molecular subgroup (ESCC3) is characterized by PI3K pathway disruption. Similarly, GC is sub-classified into four major subtypes based on the molecular pattern; the EBV group, MSI group, chromosomal instability group, and genomically stable group[36]. In CRC, four consensus molecular subtypes (CMS) were shown. CMS1 is enriched for MSI tumors that reveal marked immune activation. CMS2 reflects the classical subtype encompassing higher CIN and strong WNT/MYC-driven tumors with epithelial characteristics, whereas CMS3 is enriched for KRAS-mutated tumors with activation of metabolic pathways. CMS4 has mesenchymal features, shows a high stromal content and activation of TGF-β and VEGFR pathways[77]. Apparent clinical distinctions are distinct with poor prognosis for CMS4 and a relatively good prognosis for CMS1. A study classifying CRC by both tumor side and location using NGS panel presented that RAS mutations are seen in 70% of cecal tumors but only 57% of ascending colon and 43% of hepatic flexure tumors. BRAFV600 mutations occur in 10% of cecal, 16% of ascending colon, and 22% of hepatic flexure tumors. PIK3CA mutations are seen in 26% of descending colon but only 14% of sigmoid and 9% of rectosigmoid tumors. CTNNB1 mutations are almost absent in the sigmoid (1%), rectosigmoid junction (0%), and rectum (1%), but are still present in the descending colon (6%). This study also revealed increasing rates of CMS2 moving from right to left, accompanied by a fall in CMS1, while CMS3 and CMS4 were relatively stable when we compared CMS by tumor side[78]. In summary, the region from the sigmoid colon to the rectum appears unique and the transverse colon appears distinct from other right sided locations.

Another study define the colorectal cancer intrinsic subtypes (CRIS) distinguished by specific molecular, functional and pathogenic features; (1) CRIS-A: Mucinous subtype, glycolytic metabolism, with marked MSI, mutated BRAF or KRAS; (2) CRIS-B: Active TGF-β signaling, epithelial–mesenchymal transition, bad prognosis; (3) CRIS-C: High EGFR signaling, and to EGFR inhibitors (i.e., cetuximab); (4) CRIS-D: High WNT signaling, IGF2 gene amplification/ overexpression; and (5) CRIS-E: Paneth-like phenotype and TP53-mutated genotype[79]. Recent work revealed that subtype-specific analysis can be used to predict therapy response, which provides a great opportunity to improve patients’ management regarding precision medicine[80,81].

Although subclassification systems proposed for each GI cancer type have also possessed major challenges and caused important questions that need to be further investigated still it is applied for patient care timely, there is the possibility that these subgroup analyses revolutionize our approach towards precision medicine. Advances in tumor genomics and the immunologic landscape based on “big data” will allow the identification of expanding indications for molecular target drugs and chemotherapy in GI cancer and its predictive biomarkers. Clinical trials for targeted therapies, coupled with genomic profiling for optimum patient selection, are required to demonstrate clinical utility, including treatment outcomes and cost-effectiveness. Investigations of the safety and efficacy of clinical cancer therapies may reveal novel research directions for treating GI cancer. Increasing our knowledge of the signaling that mediates the driver mutations in GI cancer will improved our understanding of GI cancer and serve to guide future precision medicine applications for this disease. At present, we are in the very early phases of this transition towards precision and personalized medicine. We hope that this review can be a guideline for clinical and bench investigators to further develop precision medicine.