Copy number alterations and epithelial‑mesenchymal transition genes in diffuse and intestinal gastric cancers in Mexican patients.

Gastric cancer (GC) is a common malignancy with the highest mortality rate among diseases of the digestive system, worldwide. The present study of GC alterations is crucial to the understanding of tumor biology and the establishment of important aspects of cancer prognosis and treatment response. In the present study, DNA from Mexican patients with diffuse GC (DGC), intestinal GC (IGC) or non‑atrophic gastritis (NAG; control) was purified and whole‑genome analysis was performed with high‑density arrays. Shared and unique copy number alterations (CNA) were identified between the different tissues involving key genes and signaling pathways associated with cancer. This led to the molecular distinction and identification of the most relevant molecular functions to be identified. A more detailed bioinformatics analysis of epithelial‑mesenchymal transition (EMT) genes revealed that the altered network associated with chromosomal alterations included 11 genes that were shared between DGC, IGC and NAG, as well as 19 DGC‑ and 7 IGC‑exclusive genes. Furthermore, the main molecular functions included adhesion, angiogenesis, migration, metastasis, morphogenesis, proliferation and survival. The present study provided the first whole‑genome high‑density array analysis in Mexican patients with GC and revealed shared and exclusive CNA‑associated genes in DGC and IGC. In addition, a bioinformatics‑predicted network was generated, focusing on CNA‑altered genes associated with EMT and the hallmarks of cancer, as well as precancerous alterations that may lead to GC. Molecular signatures of diffuse and intestinal GC, predicted bioinformatically, involve common and distinct CNA‑EMT genes related to the hallmarks of cancer that are potential candidates for screening biomarkers of GC, including early stages.

Introduction

According to the Global Cancer Observatory statistics, cancer is the leading cause of death in the world, with 9.9 million deaths in 2020; the incidence rate of cancer was 20% in the Caribbean and South America, with high mortality rates (14%). Worldwide, gastric cancer (GC) is estimated to be the fifth most common cancer type in both sexes, ranking sixth for new cases, with over one million cases per year and third in mortality (1).

In Mexico, according to statistics from the National Institute of Statistics and Geography, three out of 10 cancer-associated deaths among patients aged 30–59 years were due to cancer of the digestive system. From 2011 to 2016, four out of 10 and three out of 10 cancer-associated deaths respectively occurred in females and males aged >60 years and resulted from tumors in digestive organs (2).

GC refers to any malignancy originating in the region between the gastroesophageal junction and the pylorus. The World Health Organization and the Lauren classification system (3) have classified GC into two types: Intestinal GC (IGC) and diffuse GC (DGC). Intestinal or differentiated gastric cancer is characterized by localized and expansive growth, while DGC has an infiltrating growth pattern, is an undifferentiated adenocarcinoma and features dispersed cells with individual or group invasive capacity (4). The development of IGC is preceded by a precancerous process of several years and stages: Active chronic gastritis, multifocal atrophic gastritis, complete intestinal metaplasia, incomplete intestinal metaplasia, dysplasia and adenocarcinoma (5). GC has a multifactorial origin: Diet, lifestyle, genetics and socioeconomic factors, and it has been observed that 80% of cases of IGC are associated with previous Helicobacter pylori (H. pylori) infection (6,7). GC is characterized by a complex etiology with a set of factors, including genetic alterations and external factors. However, it has been reported that <3% of GC is due to heredity and includes hereditary DGC, proximal polyposis of the stomach and hereditary colorectal cancer not associated with polyposis (6). With respect to molecular pathogenesis, chromosomal instability (aneuploidy, chromosomal translocation, amplification, deletions and loss of heterozygosity), gene fusion and microsatellite instability (hypermethylation of gene repair promoters) are involved (7).

Copy number alterations (CNA) represent a class of genetic variation that involve cumulative somatic variations. CNA are defined as non-inherited genetic alterations that occur in somatic cells (8). These unbalanced structural variants usually contain gains or losses. Their interpretation and the CNA report continue to be a topic of interest in health and have an important role in GC (9,10).

The majority of gastric adenocarcinomas, similar to numerous other types of solid tumor, exhibit defects in the maintenance of genome stability, resulting in DNA CNA that may be analyzed using comparative genomic hybridization (CGH) (11). This is a widespread and common phenomenon among humans and several studies have focused on understanding these genomic alterations that are responsible for cancer and may be used for its diagnosis and prognosis (12).

At present, there are few published studies involving genotyping of GC samples using high-density microarrays (13–15); however, in those altered chromosomes, gains and losses have a phenotypic impact and different signaling pathways are involved. The presence of CNA changes the genetic dose and would modify several molecular mechanisms, such as epithelial-mesenchymal transition (EMT), which is the transformation of epithelial cells to mesenchymal cells and is a critical stage for the transition to metastasis (16). There are currently >1,184 genes in the EMT Gene Database (dbEMT 2.0), which are involved in other cancer-related processes, such as proliferative signaling, evasion of growth suppressors, avoidance of immune destruction, inactivation of replicative immortality, tumor-promoting inflammation, induction of angiogenesis, genomic instability, mutation, resting cell death, deregulation of cellular energetic activity, invasion and cell plasticity (17). EMT includes activation of transcription factors, expression of specific cell-surface proteins, reorganization and expression of cytoskeletal proteins and the production of extracellular matrix-degrading enzymes (18). EMT has been associated with the progression of cancer and increased stemness of tumors (18,19), and was observed to be involved in the formation, invasion and metastasis of GC (20,21). In addition, an association has been established between the presence of CNA and its effect on the expression level of EMT-associated genes in different cancer cell lines (22). Studies on CNA events involving Latin American populations are limited (23,24). In fact, at present, only a small number of studies have performed GC genotyping using whole-genome high-density microarrays in Mexican patients with GC (14,15). Therefore, the present study aimed to determine CNA in DGC and IGC to identify, through bioinformatics analyses, the main genes and signaling pathways involving EMT-associated genes.

Materials and methods

Samples

Institutional Review Board approval was obtained for the present study (approval no. 2008-785-001). The samples were collected at the Regional General Hospital No. 1 ‘Dr. Carlos MacGregor Sánchez Navarro’, Specialty Hospital ‘Dr. Bernardo Sepúlveda’ and Oncology Hospital from IMSS in Mexico City (Mexico). Clinical data and patient samples were processed following obtainment of written informed consent. All of the samples were collected over three years (April 2010 to May 2013) following standardized endoscopy preservation protocols (25). Histological assessment of the biopsies was performed by two trained pathologists independently. They assigned the phenotypic diagnosis of diffuse or intestinal tumors and non-atrophic gastritis (NAG) samples. Only samples with the same diagnosis (‘identical results’) by two independent expert pathologists were included in the analysis.

A total of 21 patients (5 females and 16 males) with tissue samples that met the criteria for DGC (n=7) and IGC (n=7) diagnoses, as well as subjects with NAG (n=7) as controls, were included. In the absence of an established measurement (gold standard), a value was arbitrarily determined to provide guidance to investigate relevant alterations. To identify the most relevant alterations for GC, the present analysis focused on alterations present in at least three patients (cut-off, ≥3 patients; ≥40% samples).

DNA extraction

DNA extraction was performed using a commercial kit (QIAamp® DNA Micro Kit; Qiagen GmbH) according to the manufacturer's protocol. The extraction was modified to include an initial incubation at 95°C for 15 min, followed by a 5-min incubation at room temperature, prior to digestion with proteinase K (Qiagen GmbH) for three days at 56°C in a water bath, and fresh enzyme was added at 24 h intervals, as described previously (26).

DNA quality assessment and preparation

The extracted DNA was quantified using spectrophotometry (Nanodrop 2000; Thermo Fisher Scientific, Inc.). Multiplex PCR was performed to assess the quality of DNA (Multiplex PCR kit; Qiagen GmbH) with a set of primers to amplify various regions of the GAPDH gene (27). Products were visualized using 1% agarose gel electrophoresis (RedGel® Nucleic acid gel stain; Biotium) and documented under an ultraviolet light transilluminator system (Syngene).

High-density whole-genome microarray analysis

The samples were analyzed using the Affymetrix® CytoScan™ microarray (Affymetrix; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol and with 250 ng DNA, except for the addition of five PCR cycles to increase the DNA sample. The PCR products (90 µg) were fragmented and labeled using additional PCR (https://assets.thermofisher.com/TFS-Assets/LSG/manuals/703038_cytoscan_assay_UG.pdf).

Copy number processing

The raw intensity files (.CEL), retrieved from the commercial platform, were analyzed using their proprietary software, Chromosome Analysis Suite (ChAS) v3.2 and NetAff 33 Libraries, based on the construction of the hg19 genome (February 2009) as a reference model.

Data processing was based on the segmentation algorithm, where the Log2 ratio for each marker was calculated relative to the reference signal profile. To calculate the copy number variation (CNV), the data were normalized to baseline reference intensities using the reference model (provided by ChAS), including 270 HapMap samples and 96 healthy individuals. The Hidden Markov Model, available in ChAS, was used to determine the CN state and their breakpoints. The customized high-resolution condition was used as a filter for the determination of CNV: CN gains with a 50-marker count and 400 Kb, and CN losses with a 50-marker count and 100 Kb. The median absolute pairwise difference (MAPD) and the single nucleotide polymorphism quality control (SNP QC) score were used as the quality control parameters. Only samples with values of MAPD >0.25 and SNP QC <15 were included in the further analysis.

Bioinformatics analysis

A Perl script was developed to load the CNV segment data files generated by ChAS for each sample to compare the files to generate a list of genes that contained event types (gains or losses), frequencies of altered regions, including chromosomes and cytogenetic bands and Online Mendelian Inheritance in Man information, and to incorporate additional information from different databases (haploinsufficiency information from the DECIPHER database of genomic variation, genes reported at dbEMT 2.0 and genes affected in gastric adenocarcinoma from Harmonized Cancer Datasets; Table SI).

The genes altered in at least three patients (cut-off, ≥3) with DGC, IGC or NAG were included for analysis and visualizations were performed using R v4.0.2 and Bioconductor v3.12 packages (Table SII). The karyotype was created with KaryoploteR and the Bioconductor software annotation package (BSgenome.Hsapiens.UCSC.hg19 v1.4.0). The comparison among samples was performed by generating Venn diagrams with the jvenn server and a heatmap with gplots. Gene Ontology (GO) analyses were performed with the ClusterProfiler v3.16.1 packages (org.Hs.eg.db v3.11.4, enrich plot v1.8.1 and GOplot v1.0.2), with the support of functional enrichment analysis using the database for annotation, visualization and integrated discovery (DAVID) v6.8 resource (Table SII). The profile of altered molecular function (MF) terms in GC was summarized according to the proportion of CNA-associated genes and the MF GO terms from the DAVID database, adjusted by the false discovery rate. Dot plots, heatmaps and chord plots were utilized to visualize the GC CNA profiles for DGC, IGC and NAG.

To identify the main genes and signaling pathways involving CNA EMT-associated genes, GC CNA-associated genes (cut-off, ≥3) were analyzed and compared according to those previously reported in the dbEMT 2.0, accessed on 12th October, 2020.

Finally, to establish the profile-associated hallmarks of cancer involving DGC, IGC and NAG EMT-associated genes, an interaction network was generated using CNA type (gains and losses) based on genetic and physical interactions and biological pathways. Furthermore, associations were determined using the GeneMANIA prediction server and Cytoscape v.3.8.2, including the manual annotation of their corresponding cancer hallmarks [adhesion, angiogenesis, inflammation, migration, metastasis, morphogenesis, proliferation and survival (28)], with punctual scrutiny and assistance from databases, such as The Human Protein Atlas. Table SII provides information on the databases, protocols, software and specific packages used.

Results

Sample characteristics

Samples from 21 patients with GC from Mexico (third-generation Mexicans) between 35 and 91 years of age (mean ± SD, 59.61 ± 15.94 years), without any previous cancer treatment (naïve) were included in the present study. The samples included seven cases who had DGC, seven who had IGC and seven who had NAG (control samples). The raw data were deposited in the NCBI Gene Expression Omnibus database (ID no. GSE117093). There are seven adjacent tissue files (.CEL); however, these files were not included in the data analysis, as certain adjacent tissues were contaminated with cancer cells or these were not of the quality required for subsequent analyses.

Table I presents the ID and the percentage of neoplastic cells for tumor tissues ranging between 50 and 70%. Blood agar culture indicated that one patient with IGC and three patients with NAG were positive for H. pylori (data obtained from our biobank database). The patient data are also presented in Table I (29–31).

Table I.

Characteristics of GC and NAG cases analyzed in the present study (n=21).

ID	Age, years	Sex	Cancer type	% CC	H. pylori	TNM	Treatment
3CG-008	72	M	Intestinal	70	Positive	IB T1 N1 M0	Naïve
3CG-126	80	M	Intestinal	60	Negative	IIA T4 N0 M0	Naïve
3CG-128	91	M	Intestinal	70	Negative	IIA T3 N2 M0	Naïve
3CG-046	52	F	Intestinal	60	Negative	IV T4 N2 M0	Naïve
3CG-099	59	M	Intestinal	50	Negative	II T3 N0 M0	Naïve
3CG-146	71	M	Intestinal	60	Negative	IIB T3 N2 M0	Naïve
3CG-104	69	M	Intestinal	60	Negative	III A T4 N0 M0	Naïve
3CG-047	58	M	Diffuse	70	Negative	IV T4 N3 M0	Naïve
3CG-173	76	M	Diffuse	70	Negative	III A T2 N3 M0	Naïve
8CG-004	76	M	Diffuse	70	Negative	II T1 N0 M0	Naïve
1CG-001	45	M	Diffuse	60	Negative	IV T4N2M1	Naïve
3CG-035	55	M	Diffuse	60	Negative	IV T4N2M0	Naïve
3CG-042	64	M	Diffuse	50	Negative	IV T4, N2 M0	Naïve
3CG-064	38	M	Diffuse	50	Negative	IV T4 N2 M0	Naïve
4GB-001	64	M	NAG	0	Negative	NA	NA
4GB-031	62	M	NAG	0	Negative	NA	NA
4GB-015	35	F	NAG	0	Negative	NA	NA
4GB-025	39	M	NAG	0	Positive	NA	NA
4GB-033	76	F	NAG	0	Positive	NA	NA
4GB-036	38	F	NAG	0	Positive	NA	NA
4GB-042	77	F	NAG	0	Negative	NA	NA

ID, identification code; CC, cancer cells; M, male; F, female; NA, not applicable; TNM-T, extent of the primary tumor; TNM-N, absence or presence and extent of regional lymph node involvement; TNM-M, absence or presence of distant metastasis; NAG, non-atrophic gastritis.

Genomic detection of CNA

The total number of CNAs was obtained and they were classified as either gains or losses for each chromosome in the GC and NAG samples. From the total CNAs, DGC had more CNAs compared to IGC (3,505 and 2,781, respectively), while there were 828 events in the NAG samples. With respect to the tissue, more gains than losses were observed in both cancer types, DGC (2,310 and 1,195, respectively) and IGC (1,550 and 1,231, respectively), but the opposite was observed in NAG (375 and 453, respectively) (Table SIII).

To identify the most relevant CNA in GC and NAG, alterations occurring in at least three patients (cut-off, ≥3) were analyzed. This comparison indicated a similar pattern for total CNA, with more events in DGC (n=710) than in IGC (n=590) or in NAG (n=332). In addition, more gains than losses were observed in DGC (516 and 194, respectively), IGC (314 and 276, respectively) and even in NAG (196 and 136, respectively), which was different when all of the patients were included. Furthermore, DGC had the highest number of gains and IGC had the highest number of losses (Table SIII). Table II lists chromosomes and sizes with gain and loss numbers, representative and summarized.

Table II.

Principal affected chromosomes by CNA cumulative length in DGC, IGC and NAG.

Type/chromosome	Gains	Losses	Length, Mb-cl
DGC
1	327	-	117.9
4	-	155	40.8
5	-	148	74.23
IGC
1	-	148	33.78
8	365	-	139.8
X	-	66	167.1
NAG
6	-	28	0.40
7	21	-	0.20
14	10	-	3.02
17	-	15	1.86
X	-	87	0.47
X	207	-	1.20

The locations of copy number alterations are presented in Table SIV. DGC, diffuse gastric cancer; IGC, intestinal gastric cancer; NAG, non-atrophic gastritis; Mb-cl, Megabases cumulative length.

To visualize the distribution of DGC and IGC chromosome gains and losses, the identified CNA present in a karyogram (cut-off, ≥3) was plotted, which displayed alterations according to the coordinates of the Human genome hg19 (Fig. 1). The top five altered cytobands are provided in Tables III and SIV.

Figure 1.

Karyogram with CNA distribution in DGC and IGC. CNA events (gains or losses) were present from chromosomes 1 to 22, as well as X and Y. Gains (blue and dark blue) and losses (orange and red) are plotted for ≥3 samples from patients with DGC or IGC. Cytobands (gray, black or white bars) and centromeres (green bars) are presented. CNA, copy number alteration; DGC, diffuse gastric cancer; IGC, intestinal gastric cancer; Chr, chromosome.

Table III.

Top five altered cytobands in DGC, IGC and NAG.

Type/cytoband	Gains	Length, Mb-cl	Number of patients
DGC
Xq28	37	11.5	6
8q24.22	25	15.8	5
1q32.1	26	12.16	5
8q24.3	25	17.46	6
1q23.3	22	15.26	6
IGC
8q24.3	24	12.662	4
13q34	18	4.9018	4
8q24.21	18	6.2619	4
8q24.22	16	6.4856	4
8q12.1	16	5.8523	4
NAG
Xq26.2	24	34.026	6
Xq21.2	22	33.586	7
Xp22.33	19	167.504	7
Xq23	16	63.79	6
Xq26.3	13	217.774	7

DGC, diffuse gastric cancer; IGC, intestinal gastric cancer; NAG, non-atrophic gastritis; Mb-cl, Megabases cumulative length.

Of note, in DGC and IGC, the most frequent CNA lengths were between 100 and 200 Kb, while lengths of 1–50 Kb were more common in NAG, with respect to gains and losses (Table SV).

GC genes associated with CNA

Overall, 2,441 CNA-associated genes were identified in DGC, IGC and NAG. GC had 2,420 affected genes (99%), while only 108 genes (4%) were affected in NAG; of these alterations, certain candidates were shared between GC and NAG. There were 1,317 unique CNA-associated genes in DGC, 596 in IGC and 21 in NAG. Furthermore, both cancer types shared 420 genes, while 60 genes were shared between GC and NAG. In addition, 19 genes in NAG were shared with DGC and eight genes with IGC (Fig. 2A; Table SVI).

Figure 2.

Profile of CNA-associated genes from ≥3 patients with GC. (A) Venn diagram presenting the frequencies of specific and shared genes with CNA among the DGC, IGC and NAG subjects. (B) Heatmap displaying hierarchical clustering of gains (blue), losses (red) and no alterations (light blue and light red). DGC, diffuse gastric cancer; IGC, intestinal gastric cancer; NAG, non-atrophic gastritis; CNA, copy number alteration.

To identify the possible emerging patterns among the samples, hierarchical clustering heatmaps were generated (Fig. 2B). The results provided the molecular signature and hierarchical clustering of samples according to the 2,441 genes. The emerging pattern of altered genes affected by CNA distinguishes DGC and IGC from NAG.

GO analysis of GC

The functional profile for GC was generated through enrichment and GO analysis of the 2,420 GC-altered genes; 1,317 genes were only altered in DGC and 596 only altered in IGC. To identify the principal MFs altered in GC, these CNA-associated genes were categorized, independently of the GC type, into two groups: Gains and losses (Fig. 3A and B). The top 10 MFs associated with CNA gains or losses revealed that transcription activator, tyrosine kinase activity, growth factors and hormone binding, as well as intracellular signal transduction genes were enriched in GC. Gene losses mainly involved transcription coactivator and serine/threonine kinase activity, as well as several receptors binding to hormone, steroid hormone, nuclear receptor, β-catenin, intermediate filament and mitogen-activated protein kinase binding genes. In addition, the principal CNA-associated genes affecting the MF by GC type (DGC and IGC) were identified (Figs. 3C and D and S1).

Figure 3.

MF profiles of GC CNA-associated genes. Enrichment analysis of genes affected by CNA-associated genes in ≥3 patients with GC. Data are summarized by CNA type, with dot plots for (A) gains and (B) losses, according to the proportion of CNA-associated genes and MF GO terms, adjusted by the false discovery rate. The CNA-associated genes from the MF networks reveal the possible association between CNA (C) gains and (D) losses within GC. The gene ratio (M/N) is the proportion between CNA genes in ≥3 patients with GC (M) and the collection of genes from the GO term database function (N). Counts (circle sizes) represent the number of CNA genes associated with MF. GC, gastric cancer; MF, molecular function; GO, Gene Ontology.

CNA-EMT genes in DGC and IGC

To identify the main genes and signaling pathways involving the CNA-EMT genes in GC and NAG, GC CNA-associated genes were compared against a comprehensive and annotated database of EMT genes (dbEMT 2.0). A total of 551 CNA-EMT genes were found in DGC, 619 in IGC and 28 in NAG. Using the cut-off ≥3, 112 genes in DGC, 66 in IGC and 5 in NAG were obtained. The complete data of EMT-associated genes for DGC, IGC and NAG, with chromosome and cytoband locations, CNA type (gain or loss), and the P-values are provided in Table SI.

GO analysis of the EMT-associated genes

GO enrichment analysis was performed to determine the MF of the main CNA-EMT genes affected in DGC, IGC and NAG (Fig. 4). The results indicated that gains in the CNA-EMT genes in DGC were associated with transmembrane receptor tyrosine kinase, DNA and RNA binding and receptor binding for insulin, growth factors, Toll-like receptors, hormone, as well as SMAD, p53, chromatin, calcium ion binding and microtubule binding (Fig. 4A). Losses in CNA-EMT genes included associations with DNA and chromatin binding, nuclear hormone receptor binding, β-catenin, steroid hormone, mitogen-activated protein binding, intermediate filament binding, p53 binding and RNA polymerase II-specific DNA binding (Fig. 4B). Furthermore, gains in CNA-EMT genes in IGC were associated with insulin receptor substrate and phosphatase binding, kinase regulation, neurotrophin receptor binding, 1-phosphatidylinositol-3-kinase activity, transmembrane receptor protein tyrosine kinase adaptor activity and VEGF receptor binding, while losses in CNA-EMT genes in IGC were only associated with coenzyme binding and transcription coactivator activity (Fig. 4C). On the other hand, the main MF for gains in the CNA-EMT genes in NAG included transcription regulatory region DNA, transcriptional activator activity RNA, armadillo repeat and C2H2 zinc finger domain binding, γ- and β-catenin binding, as well as cysteine-type endopeptidase inhibitor activity involved in apoptotic process and estrogen receptor, as well as steroid hormone receptor activity, while losses in CNA-EMT genes in NAG were associated with damaged DNA, WW domain, p53 binding and DNA-binding transcription activator activity (Fig. 4D).

Figure 4.

Principal MF associated with CNA-EMT genes in GC and NAG. The main MFs associated with CNA-EMT are displayed as chord plots and dot graphics for (A) DGC gains, (B) DGC losses, (C) IGC gains or losses and (D) NAG gains or losses. Chord plots (left panels) present associations between genes and MFs, indicating their CNA type by color (blue for gains and red for and losses). Dot plots (right panels) provide an enrichment analysis of MF and loss or gain genes counts in the samples. CNA, copy number alteration; EMT, epithelial mesenchymal transition; DGC, diffuse gastric cancer; IGC, intestinal gastric cancer; NAG, non-atrophic gastritis; MF, molecular function.

CNA-EMT genes associated with the hallmarks of cancer

Based on the main molecular profile of altered CNA-EMT genes in GC and NAG, the functional network between 39 previously selected unique CNA-EMT genes (19 genes for DGC, 7 for IGC, 11 common to GC and two for NAG; cut-off, ≥3 patients) was generated. Gained genes, with the highest degree and at least four interactions per gene, were EGFR, MICAL2, MYC, NDRG and PIK3R1, while lost genes included GLI2, EP300 and PTPN11. The principal functions associated with these CNA-EMT genes have been previously associated with several hallmarks of cancer: Adhesion, angiogenesis, inflammation, migration, metastasis, morphogenesis, proliferation and survival (Fig. 5).

Figure 5.

GC and NAG CNA/EMT-associated gene network and associated hallmarks of cancer. The functional interactions among CNA/EMT genes identified in DGC (rectangles), IGC (ovals) and shared genes (hexagons) were identified by their CNA type (blue represents gains and red represents losses) and associated hallmarks of cancer (colored dots). DGC, diffuse gastric cancer; IGC, intestinal gastric cancer; NAG, non-atrophic gastritis; EMT, epithelial-mesenchymal transition; CNA, copy number alterations.

Discussion

To the best of our knowledge, the present study was the first whole-genome high-density array study on GC in Mexican patients with DGC and IGC, as well as NAG as non-cancerous controls. Using this experimental strategy, it was possible to generate a karyogram and obtain molecular signatures for DGC and IGC, and their association with CNA-EMT genes, independent of age, sex, percentage of cancer cells, presence/absence of H. pylori infection, TNM and treatment (naïve samples in the present study). In addition, the genomic analysis was focused on the molecular profile of GC, particularly involving alterations of EMT-associated genes, given their role in cancer progression, as epithelial cell transformation to mesenchymal cells is fundamental to metastasis (32) and chemoresistance (33,34). The results of the present study are consistent with those previously reported in the literature (detailed above), which provides validity and robustness to the results and enables the reporting of novel data or data not yet investigated to identify potential diagnostic, prognostic and treatment response markers.

Globally, the alteration profile in GC was dominated by gains. This phenomenon, where gains are more abundant than losses, has been previously reported in different tumor cell lines, including gastric cancer cell lines (35). Chromosomal gains in cancer may result in increased gene functions, providing cancer cells with a competitive advantage for the development of metastasis (36), while chromosomal losses may involve the downregulation of tumor suppressor genes (37), disrupting homeostasis and accelerating cancer progression. The most affected CNA chromosomes for DGC were 1, 4 and 5; for IGC 1, 8 and X, and for NAG 6, 7, 14, 17 and X. The altered cytobands associated with GC observed in the present study are in agreement with previous studies. For instance, 8q24 has been associated with the development of different types of tumor (38). The highest frequencies of gains in advanced GC were found at 8q24.21 (65%) and 8q24.3 (60%), and the pattern of CNA in advanced GC was different from that in early GC. This increase in CNA numbers is associated with disease progression from early to advanced GC (39). The 8q24 cytoband has also been reported in Latin American countries, such as Brazil (40) and Venezuela (41), as well as in Asian countries, including Korea (42).

Of note, the most frequent CNA length in GC was 100–200 Kb, in both DGC and IGC compared with 1–50 Kb in NAG. The biological implications of this difference in length in GC compared with non-cancerous tissues, such as NAG, is yet to be determined. Furthermore, it is important to highlight that a resolution of 100–200 Kb versus Mb is an advantage of molecular resolution approaches over classical cytogenetics (CGH and fluorescence in situ hybridization) to discover ‘small’ potentially important alterations in cancer samples.

The cumulative length averages (Megabases, Mb) of these alterations were 183.44 Mb for DGC, 113.56 Mb for IGC and 1.19 Mb for NAG. These lengths, whether gained or lost, describe the magnitude of global alterations per tissue; however, their relevance lies in the MFs, biological processes and interaction networks in which they participate.

In the present study, a molecular profile that distinguishes GC from NAG was identified based on 2,441 genes affected by CNA. They are associated with GC, as well as the differences and similarities among histological subtypes (undifferentiated DGC and well-differentiated IGC) compared with that in non-cancerous tissue, such as NAG (43). Of note, 60 affected genes shared between GC and NAG were identified; 19 genes were shared exclusively with DGC, while only eight were shared with IGC. This emerging pattern of shared altered genes between cancerous and non-cancerous tissues should be further studied to identify possible CNA-dependent oncogenic pathways and progression trajectories from NAG to either GC subtype, particularly in conjunction with environmental factors, such as H. pylori infection, diet and lifestyle, that may be associated with the spread patterns affecting patient survival (43).

In the heatmap, a separation between NAG and GC was observed, exhibiting clusters based on the molecular profiles of CNA-associated genes. There is a greater heterogeneity among the IGC samples in clusters but there were more genes affected in DGC. The front-line tool for IGC distinction has been based on different criteria, such as the Lauren histopathological classification system. However, due to challenges, including disagreements in the correct assignment, diagnosis and treatment, new criteria have been proposed, such as molecular characterization according to The Cancer Genome Atlas (TCGA) Research Network, which divides GC into four subtypes (44). The results of the present study agree with the requirement for new proposals for the classification of GC, which includes defined subgroups, with the integration of several genomic and genetic parameters where CNA are present.

In the present study, the MF profile of GC CNA-associated genes was analyzed and determined. With respect to gains, there were increased alterations involving transcription, signaling, tyrosine kinases, growth factors, hormones and insulin, while with respect to losses, molecules involved in transcription, serine/threonine and MAP kinases, steroid hormones, β-catenin binding and filament binding were decreased. These gene sets are important in GC biology. There were 13 CNA genes in IGC (including CDH1, LAST1, ROCK1 and WWOX) and 49 CNA genes in DGC (including CRIM1, EGFR, MIR9-1, MUC1, MYC, NDRG1, SCRIB, SNAI2, VEGF and ZEB2); therefore, these genes were further analyzed with the intention of comparing the results of the present study with those of others and organizing the data in a biologically coherent context. For instance, CDH1 codes for E-cadherin and, from a simplified viewpoint, E-cadherin maintains the epithelial phenotype; if CDH1 is lost, this promotes the mesenchymal phenotype, i.e., it favors loss of adhesion and metastasis (32).

In the present study, an enrichment analysis of unique CNA-associated genes for all tissues was performed and several shared MFs, such as protein binding, were obtained. Several gained-genes that encode for RNA-binding proteins (45) have diverse targets and participate in tumor progression by regulating homeostasis and changing expression patterns. Chromatin binding is another altered function in GC that participates in regulating eukaryotic gene expression, methylation profile modulation, and genome stability maintenance (46). EMT is a process that involves changes in histone modification, DNA methylation and chromatin accessibility. These changes may be promoted through transcription, allowing the cell to have an identity or to have a mesenchymal-epithelial transition-EMT conversion (47). The kinase function in DGC and IGC gains have recently been considered key regulators in the development of cancer (48). Numerous kinases were associated with the initiation and progression of carcinogenesis and are one of the main therapeutic targets for the development of inhibitors in the clinical field. Kinases are able to promote EMT and enhance invasion, migration and evasion of apoptosis (49). In the present study, PIK3R1 and PIK3CA were associated with IGC. The PI3K pathway is a key regulatory hub for cell growth, survival and metabolism (50). Activation of PI3K is a frequent hallmark of cancer, highlighted by the prevalence of somatic mutations in genes encoding key components of this pathway (51). These enzymes are responsible for transferring a phosphate group; however, the reverse process is performed by phosphatases, which are also affected in IGC. PIK3R1 is a gene frequently affected by mutations or copy numbers in various types of cancer, according to the TCGA project. These genes converge with the PI3K/AKT/mTOR pathway, which is involved in the regulation of several processes (51).

To date, the differences between DGC and IGC have been insufficiently investigated and understood; differences in etiology, location, incidence and genetic profiles have been observed (52). In the present study, a CNA-EMT network for GC was generated with relevant genes according to different criteria: Frequency among patients, genetic connections, reported pathways and experimental associations with several databases [dbEMT 2.0, The Human Protein Atlas, COSMIC (53)] and Cancer Hallmark Genes (54). Shared and exclusively altered genes were observed for each tissue type. The common CNA-EMT genes between DGC and IGC include GLI2, which has been associated with proliferation (55); EP300, with multiple functions as an inhibitor of antitumor immune response via metabolic modulation (56); PTPN11, associated with GC progression; and NDRG1, associated with metastasis and poor prognosis in GC (57). Another relevant gene in DGC is EGFR, which, due to its association with CNV GC, is now the target for the development of anti-GC therapies (58). IGC-associated EMT genes include MICAL2, MYC and PIK3R1. MICAL2, a destabilizing F-actin in cytoskeletal dynamics, has been associated with poor prognosis in GC (59). MYC gains have also been reported in several GC studies, as expected for a common oncogenic gene (60) associated with proliferation, differentiation and apoptosis (61). PIK3R1 participates in the PI3K/AKT signaling pathway, with roles in apoptosis and cell survival, as well as chemotherapy resistance in GC (62).

A large amount of data remains to be analyzed, including loss of heterozygosity, mosaicism and other gene sets that participate in different hallmarks of cancer. Another limitation of the present study was the absence of a transcriptomic analysis to validate the GC EMT signature, particularly for DGC and IGC. Yet, the concordance of CNA with expression alterations in EMT-associated genes is plausible, as previously observed for multiple types of cancer from TCGA (35). In addition, further inclusion of precancerous stages would allow further analysis of the ‘profile’ of IGC progression. The results of the present genomic approach coincide with those already reported in the literature, which provides validity and solidity to the results. After all, this strategy allowed us to report novel or thus far scarce data, or those not previously investigated, to identify differential GC CNA, identify associations with relevant MFs associated with the hallmarks of cancer and predict the EMT signature for DGC and IGC. It may be hypothesized that these networks will potentially provide treatment targets, as well as diagnostic and prognostic markers. In addition, the use of NAG as a non-malignant control allowed for investigation of the molecular and cellular events of GC and the identification of potential biomarkers for the ‘early’ stages of GC.