Gastric Cancer in the Next-Generation Sequencing Era: Diagnostic and Therapeutic Strategies

Abstract

Gastric cancer (GC) is one of the most common malignancies and the fourth major cause of cancer-related deaths worldwide. There is growing interest in the role of genetic and epigenetic changes in the development of the disease. Next-generation sequencing (NGS) studies have identified candidate cancer-driving genes in the GC. Whole transcriptome sequencing and whole-genome sequencing analysis is also important methodology in discovering novel changes in GC. Importantly, cancer epigenetics has opened the way to reveal cancer-related genes in epigenetic machinery, including DNA methylation, nucleosome positioning, noncoding RNAs, and microRNAs, as well as histone modifications. The latest molecular research on GC may be a new diagnostic and therapeutic strategy in clinical practice. In this review, we will focus on recent advances in the description of the molecular pathogenesis of gastric cancer, underlying the use of these genetic and epigenetic alterations as diagnostic biomarkers and novel therapeutic targets.

1. Introduction

Gastric cancer is statistically the fifth most common cancer and the fourth most important cause of cancer-related death worldwide [1]. This malignancy is a highly heterogeneous disease caused by various factors, including genetic and environmental factors, diet, infection of Helicobacter pylori, or Epstein–Barr virus [2]. Based on Lauren’s classification, GC is divided into three histological types: intestinal, diffuse, and mixed [3]. The intestinal subtype is mostly characterized by intestinal metaplasia and Helicobacter pylori infection, whereas the diffuse-type is more aggressive, with resistance to treatment and a weak prognosis. The Lauren classification has been widely used in the past; however, the clinical importance is rather limited in terms of the molecular heterogeneity of the cancer. The application of next-generation sequencing (NGS) technologies to molecular patterns description of the cancer allowed us to describe the molecular heterogeneity of a disease further. The Cancer Genome Atlas and the Asian Cancer Research Group have identified four molecular subtypes of gastric cancer based on the genetic and epigenetic characteristics: microsatellite instability, chromosomal instability, genome stability, and EBV+ (Figure 1) [4]. These molecular signatures have been tested for clinical importance, but further identification of factors is needed to predict treatment effectiveness.

Figure 1.

Classification of gastric cancer molecular subtypes according to TCGA.

The development of new technologies allowed researchers to extend the knowledge about the genetic and genomic background of GC tumorigenesis. In particular next-generation sequencing become a powerful tool to describe the genetic alterations and anomalies across the whole-genome and panel of genes, related to specific signaling pathways, leading to disease development and progression. NGS is a more accurate and sensitive technology, in comparison to Sanger’s method, mostly by the fact that the percentage detection of allele frequency is 2–10% using high-throughput sequencing, whereas Sanger gives 15–25% [5]. Currently, there are different NGS-based approaches, like targeted sequencing of particular genes/pathways [6], whole-exome sequencing (WES) [7], whole-genome sequencing (WGS) [8], RNA-sequencing (RNA-seq) [9], epigenome sequencing [10], or even single-cell transcriptome profiling across the heterogeneous tumor tissue of GC [11]. Targeted sequencing represents a type of approach which allows for the analysis of exome, specific genes of interest, and so-called custom panels. This technique is fast and economical, allowing for the screening of targets within genes, or mitochondrial DNA. WGS covers the whole-genome analysis, displaying information about copy number changes, single nucleotide polymorphisms (SNPs), insertion/deletion (InDels) count, and large structural variants. WES is limited to the coding regions sequencing and exons. RNA-seq covers the detection of alternative gene-spliced transcripts, posttranscriptional modifications, SNPs, or alterations in the level of gene expression. This analysis can also be done at the resolution of single-cell.

In this paper, we investigated the screening of NGS-based techniques to describe the molecular pathogenesis of gastric cancer, determining genetic and epigenetic alterations, with levels of differentially expressed genes.

2. Targeted sequencing in the detection of mutations related to gastric cancer

Targeted next-generation sequencing is a method that is dedicated to screening regions of interest in the genome. It provides the aiming of specific genes, coding regions, or even chromosomal segments at deeper coverage than standard methods, like Sanger sequencing. To focus on the clinically significant targets of the genome or DNA sample, the method requires a presequencing DNA preparations step, where DNA sequences of interest are captured (hybrid capture-based) or frankly amplified (amplicon or multiplex PCR-based), to be further sequenced. Cristescu et al. [12] described four molecular subtypes of gastric cancer, and using targeted sequencing and genome-wide copy number microarrays, and they revealed important gene alterations among each of them. Tumors with microsatellite instability encompassed intestinal subtypes, hyper-mutated, and localized in the antrum, with the highest overall prognosis and decreased frequency of recurrence (22%). The mesenchymal-like type was classified as the type of gastric cancer with the worst prognosis, covering the diffuse subtype with the increased frequency of recurrence (63%). Types with tumor protein 53 (TP53)-active and TP53-inactive were assigned to the patients with an intermediate prognosis, where the TP53-active subtype displayed a better prognosis.

Clinical trials of advanced gastric cancer (AGC) are based mostly on the background of the protein expression level or amplification of relevant genes. Kuboki et al. performed the NGS study, including a panel of 409 cancer-related genes, on a cohort of formalin-fixed, paraffin-embedded tumor samples from 121 stage III/IV GC patients [13]. Among the analyzed group, at least one mutation was found in 93.4% of patients. The most repeated mutated gene was TP53 (36.4%). Additionally, alterations in oncogenes, such as PIK3CA (7.4%), ROS1 (5.0%), ERBB2 (4.1%), EGFR (1.7%), MET (1.7%), FGFR2 (1.7%), BRAF (1.7%), and ALK (1.7%) were revealed. ERBB2 mutations were V842I and V777L, as previously published; PIK3CA alterations were detected in exon 9 or 20. Two mutations of BRAF were described: non-V600E mutations (N581Y, R682Q). Other mutations in SYNE1 (10.7%), CSMD3 (9.1%), CDH1 (9.1%), ARID1A (8.3%), MLH1 (1.7%), and MSH2 (0.8%) were detected. Hirotsu et al. investigated the study on a cohort of 20 GC patients, both males and females, aged 60–87 years [14]. They created two custom panels of selected genes: A selective hotspot panel and a comprehensive panel, for the detection of mutations related to GC. The authors were able to identify 21 somatic mutations by the selective hotspot panel and 70 mutations by Comprehensive Panel, including all detected by the selective panel. Somatic mutations in TP53 (43%), APC (29%), MUC6 (33%), and SYNE1 (24%) were the most common among the analyzed cohort, detected in more than 20% of cases. Other alterations were identified in CTNNB1 (5%) and KRAS (5%) genes, with less frequency.

Cai et al. performed NGS sequencing, using a panel of 612 cancer-associated genes, on a cohort of 153 gastric cancer patients [15]. Identification of 35 importantly mutated genes was conducted, and among them, the top five genes were altered: TP53 (59.09%), DRD2 (14.29%), CDH1 (13.64%), AKAP9 (14.93%) and ATM (11.69%). TP53 was the most common mutated gene in the studied population. The list of significantly mutated genes found by the authors, exploiting the custom gene panel, was compared to the TCGA cohort of GC. Thirteen importantly mutated genes in GC, reported by TCGA were also displayed by the authors of this study, but only six of them were analyzed as significantly mutated, including P53 (59.09%), PTEN (13.64%), FBXW7 (2.60%), CDH1 (13.64%), SMAD4 (7.79%), and APC (5.84%). In comparison to the TCGA database, authors found 29 novel significantly mutated genes. KEGG enrichment analysis underlined the affected signaling pathways assigned to p53 signaling, MAPK signaling, Ras signaling, PI3K–Akt signaling, VEGF signaling, ErbB signaling, JAK-STAT pathway, and cell movement-related pathways. Yu et al. studied a cohort of 529 gastric cancer patients, with an average age at diagnosis of 60 years [16]. For the detection of somatic mutation, they applied the panel of 450 cancer-associated genes within the exons and certain introns of 39 genes. They revealed the importance of mutations in 449 genes. The most common mutated genes were TP53 (59.7%), ARID1A (21.9%), LRP1B (14.7%), PIK3CA (13.8%), ERBB2 (13.4%), CDH1 (13.0%), KRAS (11.7%), FAT4 (11.5%), CCNE1 (10.6%), KMT2D (10.4%), and RNF43 (10.4%). The most frequent alteration was C > T, accounting for 62.6% of the total SNVs. On the other hand, the most commonly amplified regions and genes were assigned to: CCNE1 (n = 55, 10.4%), ERBB2 (n = 44, 8.3%), 11q13 (including CCND1, FGF19, FGF4, and FGF3; n = 30, 5.7%), GATA4 (n = 26, 4.9%), and FGFR2 (n = 25, 4.7%). Increased frequency of mutated signaling pathways was observed within PI3K/Akt, Wnt, and cell-cycle signaling pathways.

Toal et al. evaluated the 115 tumor biopsies across 32 study cases, performing targeted next-generation sequencing [17]. They found a group of mutated genes: ATAD2, ATR, BRCA2, CSDE1, CSMD3, DLC1, EGFR, ELF3, ERBB4, FGFR2, KLF5, TRPA1, TSHZ2, GNAS, MYC, and MMP9 for MSS tumors and ATM, CDC27, ESR1, KMT2E, and NEB for MSI tumors, which have been formerly described as driver genes by TCGA/non-TCGA investigations for different types of cancer, excluding GC. Among these findings, authors detected six genes in MSS tumors (EYS, FAT4, FSIP2, PCDHA1, RAD50, and RECQL4) and two in MSI tumors (EXO1 and FSIP2), that were clonally mutated across the cohort of studied patients, not detected in the past as driver genes for GC development. Multiple of these genes occurs in main processes affected in GC tumorigenesis, like homologous recombinant repair (RAD50 and RECQL4), extracellular matrix (EYS), or cell adhesion (FAT4 and PCDHA1).

3. Whole-exome and whole-genome studies in GC development

Next-generation sequencing facilitates the sequence of large amounts of DNA, including all the pieces of a patient’s DNA, that supply the information for protein making. Most of the known mutations that are responsible for gastric cancer development and progression occur in exons, thus, this method of sequencing is perceived as an efficient approach to characterizing disease-causing alterations. Additionally, it is known that some DNA variations outside the exons might alter the gene activity and protein production; therefore, whole-genome sequencing is also applied in GC studies to uncover alterations in any part of the genome. In this section, we described several studies, including whole-exome and whole-genome sequencing of different subtypes of gastric cancer, highlighting the type of studied population, frequently mutated genes, and affected pathways, which are presented in Table 1.

Methodology	Studied cohort	Frequently mutated genes	Top perturbed signaling pathways	Conclusions	Reference
Whole-exome sequencing	Fifteen gastric adenocarcinomas	TP53 (11/15 tumors), PIK3CA (3/15) and ARID1A (3/15) Frequent alterations in chromatin remodeling genes (ARID1A, MLL3, and MLL)	Cell adhesion pathway	Somatic inactivation of FAT4 and ARID1A could be the main tumorigenic events in GC	[18]
Whole-exome sequencing	Four samples from patients with early gastric cancer (EGC), and compared to advanced gastric cancer (AGC)	DYRK3, GPR116, MCM10, PCDH17, PCDHB1, RDH5 and UNC5C genes are recurrently mutated in EGCs		EGC and AGC share common somatic mutations AGC is related to an additional cumulative genetic mutations in cell adhesion and chromatin remodeling genes	[19]
Whole-genome sequencing	One hundred tumor-normal pairs	Identification of previously known (TP53, ARID1A and CDH1) and novel (MUC6, CTNNA2, GLI3, RNF43 and others) importantly altered driver genes Indication of RHOA mutations in 14.3% of diffuse-type GC tumors but not in intestinal subtype	Adherens junction and focal adhesion signaling pathways, in which RHOA and other altered genes play a key role	Underlining the molecular heterogeneity and complexity of GC leading to improve genome-guided personalized therapy	[8]
Whole-genome and whole-exome sequencing	Eighty-four clinical biopsy tumor samples (including matched pre- and posttreatment) from 35 cases with gastric cancer, with described responses to neoadjuvant chemotherapy	The top mutated genes: TP53, PI3KCA, RNF43, ARIDA1, and KRAS, as previously detected in other GC studies Mutations associated with the response of the tumor to chemotherapy: C10orf71 and IRS1 mutations, and MYC and MDM2 amplifications MSI status as a clinical biomarker to direct therapy in patients with gastric cancer	MYC signaling activated in the response group Upregulation of DNA repair pathway in the response group One top subnetwork consisted of IRS1, IRS2, PIK3CA, JAK1, and IL6ST IRS1 is involved in transmitting signals from the insulin and insulin-like growth factor-1 receptors to the phosphatidylinositol 3-kinase/AKT pathway	Identification of molecular markers assigned to the tumor response to neoadjuvant chemotherapy	[20]
Whole-exome sequencing	Thirty-eight gastric cancer patients, 26 metastatic and 12 nonmetastatic	Somatic mutation of ATAD3B assigned to the metastatic stage Rare germline mutations associated with GC survival or metastasis: FANCM, PDGFRA, and POLE CCNE1 and ERBB2 were displayed to be amplified CNVs of several genes including MMP9, PTPN1, and SS18L1 were significantly related to metastasis	TP53 signaling pathway Base excision repair (BER) pathway Pathways of cell growth control and response to interferon stimulation	Detection of potential new predictive molecular markers of survival and metastasis	[7]
Whole-exome sequencing	One hundred five cases of alpha-fetoprotein-producing gastric carcinomas (AFPGC)	Thirty-four significantly mutated genes identified Among them, the most frequently altered genes were: TP53 (69%), PCLO (21%), CSMD3 (19%), and KMT2C (19%)	RTK/RAS/PI(3)K, p53/cell cycle, and JAK/STAT signaling pathways	A large genomic landscape of AFPGC, a step forward to understanding the disease mechanisms	[21]

Table 1.

Whole-exome and whole-genome studies among different subtypes of gastric cancer.

4. The chromatin landscape of gastric cancer tumor

In gastric cancer development and progression, the impact of aberrant DNA methylation of a promoter CpG island (CGI) is significant because Helicobacter pylori infection leads to aberrant methylation [22]. Park et al. compared the methylation status in normal and gastric cancer tissues using different methods including high-throughput sequencing [23]. They found that CpG island hypermethylation in promoters influences changes in gene expression profile. Hypermethylation of the 5′-end of coding exons appeared in cancer and affected the progression of the disease. Methylation status changes in low-range epigenetic silencing (LRES) regions, and younger repetitive elements suggested a direction for early disease detection and specific targeting strategies. The authors also revealed that within LRES regions, there were hypermethylated genes, like MDM2, DYRK2, and LYZ. Dysregulation of the MDM2-mediated pathway at the epigenetic level was displayed in some cancer samples. Moreover, the methylation status of HOX genes was assigned to tissue-specific manners in GC and might be an important target during tumorigenesis. Zouridis et al. investigated the analysis of DNA methylation signatures in gastric carcinoma among 240 tumor samples and 94 matched healthy tissues by genome-wide CG dinucleotide (CpG) methylation profiles [24]. Epigenetic variabilities were present in 44% of CpGs, including both hyper- and hypomethylation. Cancerous gene expression levels were correlated with 25% of changes in methylation status. Authors revealed a subgroup of CpG island methylator phenotype (CIMP) with specific signatures, like young age, broad hypermethylation status, unfavorable patient outcome in a disease phase-autonomous way, and long-range regions of epigenetic silencing (LRESs). Moreover, they found regions of long-range tumor hypomethylation, assigned to the higher level of chromosomal instability. To conclude these studies, the authors indicate that for gastric cancer patients that are included in the CIMP group, silenced DNA provides a target for accessible drugs.

Loh et al. did a high-throughput methylation analysis on FFPE primary tumor and tumor-adjacent gastric tissue samples among a cohort of 60 patients with gastric cancer [25]. The set of 219 CpG islands within 147 genes was displayed to be differentially methylated. Almost all of these genes, apart from (CHFR, DAB2IP, DLC1, SFRP1, TCF4, and TFPI2) are new methylation biomarkers in gastric cancer that might be useful for early screening of disease. Six methylation subgroups were isolated within two different clusters, including 72% high methylation (H) and 28% low-methylation (L) of GC tumors. Differential analysis indicated HOXA5 as the most differentially methylated gene between two subgroups of gastric tumor and tumor-adjacent gastric tissue. The H subgroup displayed methylated genes with increased incidence of polycomb occupancy and H3K4+/H3K27+ bivalent marks, showing the correlation between chromatin dysregulation and hypermethylated phenotype. Chong et al. performed a wide screening of methylation patterns on fresh tumor and nontumor samples from EGC patients [26]. Further to validate methylation data in 3 GC histological subtypes (intestinal, diffuse, and mixed), authors conducted pyrosequencing with 12 genes, selected from the methylation screening. The methylation assay displayed 169 differentially methylated regions between histological subtypes of GC. Pyrosequencing with 12 genes of interest indicated the abnormal methylation patterns of DVL2 and ETS1, which were corresponding to both diffuse and mixed subtypes, while C19orf35 and CNRIP1 were assigned to the diffuse-type, and GAL3ST2 and ITGA3 were related to the mixed-type. Status of several other methylated genes: CCDC57, CLIP4, MAML3, SDC2, and XKR6, was related to factors like tumor location, age, or H. pylori infection.

Yoda et al. investigated to analyze the gastric cancer-related pathways affected by epigenetic variations, using wide screening with a bead array with 485,512 probes for CpG and non-CpG sites [27]. Among a group of 50 gastric cancer patients, they found that tumor-suppressor genes, such as CDH1, CDKN2A, and MLH1, were inactivated mostly by epigenetic changes. Moreover, repression of negative regulators (DKK3, NKD1, and SFRP1) of the WNT signaling pathway by epigenetic alterations occurred in all study cases. The cell-cycle regulation pathway was perturbed by the abnormal methylation of CDKN2A and CHFR in 13 cases. In two gastric carcinomas, they found that mismatch repair is affected by the aberrant methylation of the MLH1 gene. Abnormal methylation patterns of downstream genes in 38 cases, caused inactivation of their p53 pathway. Sepulveda et al. screened nonmetaplastic mucosa, intestinal metaplasia, and gastric cancer with methylation arrays and bisulfite next-generation sequencing [28]. In gastric cancer cases, 13 genes had higher CpG methylation status, in comparison to nonmetaplastic mucosa, including: BRINP1, CDH11, CHFR, EPHA5, EPHA7, FGF2, FLI1, GALR1, HS3ST2, PDGFRA, SEZ6L, SGCE, and SNRPN. Additionally, the hypermethylation status in most of these genes correlated with lower expression levels, indicating that they could mediate neoplastic transformation from nonmalignant intestinal metaplasia to cancer. Authors observed hypermethylation and decreased expression of the FLT1 gene in GC, suggesting a tumor-suppressor role, which was previously shown in breast cancer studies [29]. Hypermethylated genes BRINP1 and SGCE were matched with better survival in GC.

Alterations in histone H3 lysine 27 (H3K27me3) in gastric cancer are not fully uncovered. Zhang et al. performed chromatin immunoprecipitation linked to the microarray (ChIP-chip) approach, to describe changes in H3K27me3 in CpG island regions, among eight gastric cancer patients and matched healthy tissues [30]. Important H3K27me3 distinctions were displayed between normal and tumorous tissues, among 128 genes (9 lowered and 119 increased H3K27me3). Additionally, abnormal DNA methylation was also discovered on arbitrarily picked genes: AFF3, MMP15, RB1, SHH, and UNC5B. This study highlighted the importance of H3K27me3, as a biomarker and target for epigenetic-based therapies. Muratani et al. investigated Nano-ChIP-seq in primary gastric carcinomas to display histone modifications and their associated regulatory elements [31]. Authors were able to uncover for the first time the landscape of promoters and putative enhancer elements, placed in noncoding regions of the genome, that are somatically changed in primary GCs. They found a huge proportion of promoters assigned with cancer that were cryptic, which indicated activation of noncanonical promoters, ending with changed transcriptional usage of 5′ exons. They revealed germline variants, placed within somatically changed regulatory elements, showing allelic bias, which might predispose to GC development.

In gastric cancer, regulatory enhancer elements are still a target for new studies. Ooi et al. performed chromatin profiling using ChIP-seq to uncover the enhancer landscape in primary GC [32]. Samples including healthy tissues, primary tumors of GC, and cell lines underwent epigenome characteristics, which revealed 36,973 predicted enhancers and 3759 predicted super-enhancers. Super-enhancers were categorized by their somatic mutation profile, into somatic loss, gain, and nonmutated subgroups, which were enriched in various transcriptional, epigenetic, and pathway signatures. Somatic gain-predicted super-enhancers had an impact on proximal and distal gene expression and were perceived as significant regulators of abnormal gene expression in GC. The authors also showed that somatic gain-predicted super-enhancers are correlated with the occurrence of CDX2 and HNF4α. This study contributed to the discovery of complexity among enhancers and super-enhancers reprogramming during tumorigenesis of GC. Aberrant methylation at promoter parts was described for many genes; however, it is still not well recognized the function of DNA methylation signatures at distal regulatory parts in GC. Baek et al. did ChIP-seq study to extend the knowledge about epigenetic alterations in the proximal and distal regulatory regions [33]. They found multiple enhancers with abnormal DNA methylation status. The analysis displayed genes that were over-expressed and hypo-methylated at their promoters, including CDX1, HNF4A, MUC4, MUC6, and MAGE family genes (MAGEA1, MAGEA6, MAGEC1, and MAGEC2). The authors also displayed that hypo-methylated-enhancer regions were enriched with the TEAD4 motif. Additionally, the analysis of methylation changes at the promoter regions of lncRNAs allowed for the identification of presumed lncRNAs, like EBV1-AS, HOXD-AS1, HOXD-AS2, and MALAT1. Four subtypes of GC were also analyzed (CIN+, EBV+, GS+, and MSI+) for methylation patterns, and the EBV+ subtype was significantly hypermethylated among other subtypes.

The epigenomic promoter landscape of GC was analyzed by Qamra et al. [34], as promoter elements are significant factors in cell-type-specific expression profiles [34]. In this study 110 chromatin profiles (H3K4me1, H3K4me3, and H3K27ac) were investigated, among gastric cancer cell lines, primary tumors, and healthy samples. Different alternative promoters were found in previously described as well as novel genes in GC, like alternative promoters at the EPCAM gene locus, which was active in GC samples. Another alternative promoter that was revealed for the first time was localized in the RASA3 gene, and N-terminal Var RASA3 increased migration and invasion in studied gastric cell lines. The connection between tumor immunity and somatic promoters was also revealed, as isoforms of alternative promoters with increased expression in GC displayed depletion of N-terminal peptides, with immunogenic properties.

5. Analysis of gastric cancer transcriptome

The next-generation sequencing applications, like RNA-sequencing or gene expression microarrays, facilitate the study of wide-scale functional genomics to cancer investigations and its application in clinical settings. These techniques provide individual gene expression profiles, which might be used as molecular markers in the treatment of patients with GC [35, 36, 37, 38]. Several studies implicated transcriptome profiling to uncover differentially expressed genes between healthy and tumorous tissues in GC patients [39, 40]. Additionally, various studies have described stage and histological-specific gene expression profiles [41]. Moreover, databases with transcriptome studies data are a source of published results, which allow analysis in the context of disease development. In Table 2 are listed several studies on the transcriptome of GC, with an indication of the studied group, DEGs, and modulated signaling pathways.

Methodology	Studied group	Differentially expressed genes	Modulated pathways	Conclusions	Reference
Gene expression microarray	Twenty-six samples of primary gastric carcinomas and matched healthy tissues	2371 differentially expressed mRNAs, 1142 downregulated, and 1229 upregulated between gastric cancer and control tissues The genes such as GKN2, PGC, MUC6, CHIA, PSCA, and FBP2 were in the group of top 20 downregulated, while KLK8, SFRP4, INHBA, CLDN1, CST1, FAP, SPP1, OLFM4, and KRT17 were among the top 20 upregulated	Upregulated genes were involved in angiogenesis, tumorigenesis, migration, and microenvironment formation, while downregulated genes were assigned to metabolism	Better understanding of gastric cancer carcinogenesis Key genes as a target in antitumor therapy	[42]
RNA-seq data (GSE36968) downloaded from the Gene Expression Omnibus Database	Six healthy tissues GC stage I: 5 samples, stage II: 5 samples, stage III: 8 samples, and stage IV: 6 samples	3576 genes with stage-specific expression patterns Kinesin family member C1, KIFC1; and septin 2, SEPT2) were highly expressed in stage I and II Neuropilin-2, NRP2; collagen triple helix repeat containing-1, CTHRC1; secreted protein, acidic, cysteine-rich, osteonectin, SPARC; matrix metalloproteinase 17, MMP17; and collagen, type VI, alpha 3, COL6A3) were specific for stage IV GC	Two regulatory pathways in stage IV GC: HOXA4-GLI3-RUNX2-FGF2 and HMGA2-PRKCA	Understanding the pathogenesis of GC by stage-specific gene profile	[43]
RNA-seq	Fifteen cases with advanced or metastatic GC Patients were underwent the therapy with ramucirumab	Three genes were differentially expressed in the tumors for responders (to ramucirumab) versus nonresponders: CHRM3, LRFN1, and TEX15 CHRM3 was upregulated in the responders Downregulation of CDC42 activators, such as RAP1A, RAP1B, and SRC, could be related to tumor response to ramucirumab	Nectin adhesion pathway This pathway was more active in the nonresponders, correlated with a higher expression of the RAP1A, RAP1B, and SRC genes	RNA-sequencing might be used to individualize the recommendation of ramucirumab for GC patients	[44]
RNA-seq	Twenty-four patients with gastric cancer Two main subgroups according to histopathology classification: intestinal and diffuse	Detection of 2064 differentially expressed genes between healthy and cancer samples 772 of them assigned to the intestinal subtype, 407 specific for the diffuse subtype In the intestinal subtype enrichment of CXCR2, CXCR1, FPR2, CARD14, EFNA2, AQ9, TRIP13, KLK11, and GHRL was displayed In the diffuse-type, low levels of CXCR2 and increased levels of CARD14 mRNA were negative predictors of 4 years of survival	Increased modulation of MAPK, RAS, PI3K–AKT–mTOR, JAK/STAT, NF-kB, VEGF Modulation of cell-cycle regulators, chemokine, and cytokine signaling Modulation of immune and proliferation pathways	Therapeutic strategy for gastric cancer patients (diffuse and intestinal) by targeting AQP9 and CXCR2	[45]
RNA-seq data from the TCGA database	TCGA gastric cancer contained a total of 407 samples, 375 tumor samples, and 32 normal samples	Classification into prognostic metabolic subgroups: cholesterolemia, glycolytic, mixed, and quiescent, based on the expressed genes, related to cholesterol synthesis and glycolysis A total of 1966 DEGs between cholesterol and glycolysis subtypes mRNA levels of mitochondrial pyruvate carriers 1 and 2 (MPC1/2) were subtype-specific Glycolytic subtype showed increased PDCD1 expression	Glycolysis subtypes were associated with: immune-related T cell receptor signaling, B cell receptor signaling, natural killer cell-mediated cytotoxicity, and primary immunodeficiency pathways	Genes correlated with glucose and lipid metabolism have an impact on gastric cancer development	[46]
RNA-seq	Two male patients with early-stage gastric adenocarcinoma	1677 DEGs and 111 differentially expressed novel transcripts and noncoding transcripts were found to be expressed in the GC tumoral tissues in comparison to healthy ones Dysregulation of 22 genes was confirmed by the TCGA dataset: ATP4A, ATP4B, GKN1, GKN2, and gastric type LIPF, which were expressed only in the stomach, while ghrelin, GHRL, and SLC5A5 were expressed in the stomach but also in many other tissues Novel downregulated noncoding RNAs including GATA6 antisense RNA 1, antisense to LYZ, antisense P4HB, overlapping ACER2	Gastric acid secretion was the most significantly enriched pathway Drug metabolism and transporters, molecular toxicology, O-linked glycosylation of mucins, immunotoxicity, glycosylation	Key genes and regulatory pathways involved in GC tumorigenesis	[9]
RNA-seq	Paired tumor-normal samples from four GC patients	148 highly significant DEGs CLDN7, SELL, CLDN4, HLA-DOA, and CLDN1 genes enriched in the CAM pathway ATP4A, ATP4B, KCNE2, KCNJ16, and SLC26A7 genes enriched in the gastric acid secretion pathway Two upregulated genes, APOC1 and SALL4 with prognostic importance	CAM pathway, gastric acid secretion, and mineral absorption pathways	SALL4 as a potential molecular marker candidate in GC	[47]
RNA-seq data from TCGA and GEO databases	Gastric cancer samples	Identification of 25 DEGs Six secretory genes (APOC1, OLFM4, CST1, CEMIP, COL4A1, and CD55) with diagnostic importance Higher COL4A1 expression might be correlated with a poor prognosis	Connective tissue development, collagen fibrous tissue-related processes, extracellular structure, extracellular matrix (ECM) tissue, focal adhesion, and PI3K-Akt signaling pathway	COL4A1 could be the molecular marker in GC diagnosis	[48]
RNA-seq	Six gastric cancer samples (stage I, II III) and matched healthy tissues	2207 differentially expressed genes, including 972 upregulated genes and 1235 downregulated Top genes in stage I: KRT5, CDH3, ARNT2, EFNA3, PRKAR2B, VIT, ACACB Top genes in stage II: S100A2, SLC28A2, CDC25A, FBLN2, ARHGEF19, LRRC-66, ACTL8, and SSTR1 Key genes in stage III: SH3BP5, MSRB3, SGPL1, PRKACB, DRAM1, SLCO2A1, MAPK11 and NCEH1	Stage I: chemical carcinogenesis, drug and xenobiotics metabolism by cytochrome P450, fat digestion Stage II: complement and coagulation cascades and nitrogen metabolism Stage III: chemokine and TNF signaling pathway	Deeper understanding of molecular pathogenesis of GC	[49]

Table 2.

Transcriptome studies in the development of gastric cancer stages.

6. Single-cell atlas of expression programs in gastric cancer

Single-cell transcriptomics (single-cell RNA-sequencing) has a broad potential to reveal the novel basis of cancer development and progression. The diversity of cells at the individual level might be better studied by single-cell transcriptomics. This method allows for sequencing millions of cells with proper accuracy and viability in a short period. Transcriptional heterogeneity in primary gastric carcinomas was investigated by Zhang et al. [50]. The authors performed scRNA-seq analysis on 27,677 cells derived from 9 GC tumors and 3 healthy samples. Results showed five main populations of cells, with unique expression profiles. The variability in cell composition between tumor samples and within them was displayed. They revealed two characteristic groups with distinct transcriptome features. One subgroup was characterized as GA-FG-CCP, expressing chief-cell markers (LIPF and PGC), and RNF43 with WNt/B-catenin signaling pathway. The other subgroup expressed immune-specific genes (eg, LYK6 and major histocompatibility complex class II), including the Epstein-Barr virus. Moreover, analysis of nonmalignant epithelium uncovered a prospective transition from chief cells to MUC6 + TFF2 + spasmolytic polypeptide expressing metaplasia.

Kim et al. studied the diversity of the cell population of precancerous lesions and gastric cancer using scRNA-seq technology [51]. They described 10 cell populations in GC and displayed the histology-based composition of GC subtypes, including intestinal and diffuse types. Interestingly, intestinal and diffuse-type cancer cells had various metaplastic cell lineages: diffuse-type cancer cells resemble de novo pathways, while intestinal subtype cancer cells differentiated along the intestinal metaplasia lineage. In the diffuse-type lineage, single-cell patterns were linked with intratumoral CAFs and might develop into various cell sets to survive. Additionally, the authors distinguish a population of cancer cells, the EmyoT type, displaying a characteristic gene expression profile. The EmyoT tumor cell type was correlated with a poor prognosis, diffuse marker gene expression, and weak EMT signature. Moreover, CCND1_mut and iCAFs might have a significant impact on alterations in GC. In the study performed by Yang et al. [52], the analysis of scRNA-seq data of GC, displayed 3385 various cell characteristics expressed by 4110 EGC cells. The authors detected gastric cells, memory T cells, and plasmacytoid dendritic cells [52]. The top 8 expressed genes were: CCL5, CHGA, FABP1, KRT7, OLFM4, SRGN, TFF3, and TTR. The important relationship between the DEGs and TNF signaling pathway, oxidative phosphorylation, and endoplasmic reticulum protein processing was found. Moreover, they discovered a prognostic marker for GC, FABP1, which regulated the fat digestion, PPAR signaling pathway, hormone-sensitive lipase (HSL)-mediated triacylglycerol hydrolysis, and absorption in GC progression. The expression of FABP1 was correlated with the age of patients’ diagnosis and an increased level of FABP1 was assigned with a lower survival rate in GC.

A study across clinical stages and histology classification of GCs allowed for a wide-range analysis of 48 samples from 31 patients and displayed a single-cell atlas of over 200,000 cells [53]. Identification of 34 various cell-lineage states was displayed, encompassing new rare cell populations. In the diffuse subtype of GC, there was a higher level of plasma cell proportions, associated with epithelial-resident KLF2 and cancer-associated fibroblast populations, with increased INHBA and FAP coexpression. Therefore, cancer-associated fibroblast subtype with INHBA-FAP-high cell populations might be an indicating factor of poor clinical prognosis. Importantly, high expression of Epilnt1 was specific for the subpopulation of intestinal-type epithelial cells, suggesting that this set of cells could be significant in the transition into malignancy from metaplasia. Li et al. [54] profiled the transcriptome of nine patients with GC, revealing the composition of 47,304 cells [54]. Among identified populations, Treg cells were enriched in GC tissues, with higher expression of genes related to immune suppression. Tumor endothelial cells specifically expressed the ACKR1 gene, which was related to the poor prognosis, and could be a novel target in GC treatment. Limitation of immunotherapy in GC patients might be supported by the obtained results, which indicated the deficit of separate exhausted CD8+ T cell cluster, and the decreased expression level of exhaustion markers CTLA4, HAVCR2, LAG-3, PDCD1, and TIGIT in those specific cells. Gastric cancer metastasis (liver, ovary, peritoneum, and lymph node) with primary tumors and nontumoral adjacent samples were analyzed by Jiang et al. using sc-RNA-seq [55]. They discovered several phenotypes, including malignant epithelial clusters assigned with invasion, dormancy-like features, intraperitoneal metastasis propensity, and epithelial–mesenchymal transition-induced tumor stem cell phenotypes. Immune and stromal cells showed cellular heterogeneity and constructed an immunosuppressive microenvironment. Authors revealed not only malignant tumor cells, but also endothelial subcluster, plasmacytoid dendritic cells, T cell-like B cells, mucosal-associated invariant T cells, macrophages, monocytes, and neutrophils may contribute to HLA-E-KLRC1/KLRC2 interaction with cytotoxic/exhausted CD8+ T cells and/or natural killer (NK) cells, suggesting new therapeutic opportunities in GC.

7. Liquid biopsy in gastric cancer diagnosis

Cell-free DNA (cfDNA) is released from cells into the circulatory system throughout the body. It might be detected in plasma and body fluids, such as saliva, urine, pleural fluid, cerebral spinal fluid, and others. In some pathological states, like cancer, organ transplantation, or pregnancy, the affected tissues could release additional DNA into the peripheral circulation. Detection of cfDNA/ctDNA in peripheral blood might be useful for the identification of abnormalities in individuals in a noninvasive manner [56]. CfDNA/ctDNA might be used for gastric cancer diagnosis, prognosis, prediction of efficiency, and resistance to therapy.

Zhong et al. studied the clinical value of plasma cell-free DNA and its role as a biomarker for advanced gastric cancer [57]. CfDNA concentration in advanced gastric cancer patients was increased in comparison to healthy cases. Importantly the concentration of cfDNA among cases with disease progression (PD), displayed an increasing level over time. There was no valuable correlation between cfDNA concentration and traditional biomarkers like age, gender, pathological type, CA125, CA199, CA724, AFP, and CEA. Plasma cfDNA concentration was higher in patients with gastric cancer, and its diagnostic efficacy was better than common tumor biomarkers. Kandimalla et al. investigated a genome-wide DNA methylation analysis of 1781 gastrointestinal stromal (GI) tumors and adjacent normal tissues and identified the differentially methylated regions (DMR) between analyzed groups [58]. A panel of 67,832 tissue DMRs was prioritized and validated in 300 cfDNA specimens. The authors supported the first evidence for a cfDNA methylation assay, that provides robust diagnostic accuracy for GI cancers. Yang et al. investigated the evaluation of molecular residual disease (MRD) by ctDNA in 46 resected gastric cancers with stage I-III [59]. Sixty tumor samples and 296 plasma samples were enrolled for targeted deep sequencing and ctDNA profiling. Cases with detected ctDNA in the immediate postoperative period, experienced recurrence. CtDNA occurrence during longitudinal postoperative follow-up was associated with worse postoperative disease-free (DFS) and overall survival (OS).

Exosomes are small (30–140 nm) membrane-bounded extracellular vesicles, that are secreted by large multivesicular bodies and might be detected in blood, urine, cerebrospinal fluid, and other body fluids. They are released by multiple cell types, such as epithelial cells, neuronal cells, hematopoietic cells, fibroblasts, adipocytes, and tumor cells. Noncoding RNAs, like microRNAs (miRNAs), circular RNAs (circRNAs), and long noncoding RNAs (lncRNAs), which in normal conditions are degraded, might be packed into exosomes, which provide them stability [60].

Exosomal miRNAs (EmiRs) could be used for the prediction of GC development. Qian et al. investigated miRNA sequencing to identify key members of EmiRs in GC [61]. The exosome samples derived from blood and urine were taken from 7 GC cases and 3 healthy donors. For GC cases, authors found three upregulated differentially expressed miRNAs (DEmiRNAs): hsa-miR-130b-3p, hsa-miR-151a-3p, and hsa-miR-15b-3p in the serum exosomes, and one upregulated DEmiRNA (hsa-miR-1246) in the urinary exosomes. Further analysis showed the commonly enriched ontology terms, including GO BP terms like cell surface receptor signaling pathway involved in cell-cell signaling, positive regulation of the catabolic process, and morphogenesis of an epithelium. Four key exosomal miRNAs and their targets (TAOK1, CMTM6, SCN3A, WASF3, IGF1, CNOT7, GABRG1, and PRKD1) could be a reference of the molecular mechanisms in gastric cancer development. Tang et al. used NGS sequencing to identify exosomal miRNAs in serum, considered early diagnostic markers for GC [62]. A total of 66 up and 13 downregulated exosomal miRNAs were detected in the studied cohort. Increased levels of serum exosomal miR-92b-3p, let-7 g-5p, miR-146b-5p, and miR-9-5p were found to be importantly correlated with early-stage GC. Moreover, serum content of exosomal miR-92b-3p was significantly correlated with poor cohesiveness, let-7 g-5p and miR-146b-5p with nerve infiltration, and miR146b-5p with tumor invasion depth in early-stage GC. In conclusion, serum exosomal miR-92b-3p, -146b-5p, -9-5p, and let-7 g-5p could be candidates for noninvasive biomarkers for early diagnosis of GC.

8. Conclusion

Remarkable advances in next-generation sequencing technologies allowed for the characterization of the genetic, epigenetic, and transcriptomic landscape of gastric cancer. In this review, we summarized the studies, indicating multiple molecular signatures and distinct molecular subtypes, which might serve as a future roadmap for patient treatment and trials of targeted therapies. The fast advances in NGS approaches will shortly continue to display driver mutations of GC, for further understanding the GC carcinogenesis and will improve the individual tumor therapy. Predictive biomarkers are important in precision oncology. In the past few years, multiple studies have investigated GC treatment. While targeting HER2 remains the key therapy strategy for a limited number of patients with advanced GC, new targets have been explored, mostly those for immune checkpoint molecules. Currently, CLDN18.2 is being investigated among various other targets, and new results are expected to be revealed shortly. The main barrier to accurate medicine for gastric cancer is intratumoral heterogeneity, which has an impact on tissue-based diagnostics, and this could cause primary and secondary drug resistance. Recent studies displayed blood-based biomarkers that could be used as diagnostic indicators and for monitoring postsurgical minimal residual disease. Among these biomarkers, we have circulating DNA, RNA, extracellular vesicles, and proteins. Detection of accurate diagnostic markers for GC that have high sensitivity and specificity will, in the future, improve survival rates and contribute to precision medicine.