1. Introduction
Gastric cancer is widespread because it is difficult to diagnose and treat. In 2018, it was the fifth most common neoplasm in the world, causing 782,685 deaths [
1], and is the tenth most common cause of death in Mexico [
2]. Even when diagnosed early, the only treatment for gastric cancer is gastric resection. Currently, two types of gastric cancer have been identified: the diffuse or poorly differentiated type; and intestinal cancer [
3,
4]. The principal risk factor for the progression of gastric cancer is chronic infection by
Helicobacter pylori [
5], which impacts over 50% of the population worldwide. It is established as a lifelong infection in the human stomach after its acquisition during childhood and produces chronic inflammation of the gastric mucosa [
6]. Between 2–5% of those infected acquire chronic atrophic gastritis characterized by the loss of acid secretion and increased levels of pepsinogen I and gastrin [
6,
7].
Intestinal cancer progresses in a series of sequential histological events described by Correa in 1992. The normal mucosa changes in chronic gastritis, leading to atrophic gastritis, intestinal metaplasia, dysplasia, and ultimately gastric cancer [
5,
6,
7]. Gene expression in the early stages of gastric cancer has been studied in response to infection with
H. pylori, although most of the research has been conducted with in vitro models of epithelial cell culture, animal models, or both. In a study of human gastric tissue, in which the presence or absence of
H. pylori infection is not considered, Hippo et al. evaluated the expression of 6800 genes from 30 samples of gastric cancer and adjacent tissue using high-density oligonucleotide microarrays. These authors found genetic alterations in the expression of genes related to the DNA damage and repair mechanisms, regulation of the cell cycle, activation of oncogenes, and inactivation of the tumor suppressor genes implicated in gastric carcinogenesis [
8]. Additionally, Kim et al. investigated the early stages of gastric cancer by examining the expression of 25,100 genes using DNA microarrays in 24 tissue samples from chronic atrophic gastritis, intestinal metaplasia, and normal mucosa, without considering the presence or absence of
H. pylori infection. Comparing the characteristic expression profiles of chronic atrophic gastritis with those of intestinal metaplasia disorders, the authors conducted a bioinformatics analysis to compare the profiles with the expression of genes of gastric carcinogenesis [
9]. The gene expression patterns found in these studies provided new comparative information in chronic atrophic gastritis and intestinal metaplasia, which may play an important role in the development of gastric cancer. To gain a molecular understanding of
H. pylori infection in gastric carcinogenesis, we conducted the analysis of gene expression profiles in chronic atrophic gastritis and gastric cancer of subjects infected with this bacterium. The gene expression patterns observed in this study allowed the identification of CLDN1 and MMP9 proteins as promising biomarkers of early stages of gastric cancer development.
2. Materials and Methods
To accomplish the study objectives, two groups of samples were used, including gastric biopsies for the microarray analysis and qPCR. To validate the results obtained, we used tissues embedded in paraffin for immunohistochemistry analysis.
2.1. Ethical Approval and Consent to Participate
This study was approved by the Investigation and Ethics Committee of the School of Medicine of the UNAM, National Institute of Medical Sciences and Nutrition Salvador Zubirán (INCMNSZ), (Registry numbers: 019-2009 and 209, respectively). All participants gave their written informed consent, in accordance with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards, prior to sample collection.
2.2. Gastric Biopsy Samples
The gastric biopsies were collected from patients with follicular gastritis, chronic atrophic gastritis, and gastric cancer. After obtaining consent, subjects with gastric complaints who were programmed for an exploratory endoscopy to determine the source of their symptoms were recruited. After the endoscopic procedure, body and antrum samples taken for diagnosis were submitted to the Pathology Department, and an expert pathologist performed the histological examination according to the Sydney classification and confirm the presence of
H. pylori. Gastric cancer was classified according to its histological type by the Lauren and Macroscopic classification as well as frequency by growth form (tumor type) [
10,
11]. For the study, patients with an
H. pylori infection and a diagnosis of follicular gastritis, chronic atrophic gastritis, and gastric cancer were selected. All the biopsies from the gastric lesions were collected and stored until use in RNAlater (Ambion, Austin, TX, USA) at –70 °C for nucleic acid preservation.
2.3. RNA Isolation and Expression Microarray Analysis
The biopsies selected from the exploratory set were removed from RNAlater and put into a lysis solution (RNAqueous Kit; Ambion, Austin, TX, USA). Each gastric biopsy was homogenized with a tissue homogenizer (Cole-Parmer, Vernon Hills, IL, USA) until its complete lysis. Total RNA was isolated from the lysate using the RNAqueous commercial kit (Ambion, Life Technologies, Carlsbad, CA, USA) according to the manufacturer’s instructions. The RNA quality was measured with a microvolume spectrophotometer (ND-1000; NanoDrop Technologies, Wilmington, DE, USA). RNA integrity was assessed with 28S:18S ribosomal RNA (rRNA) ratios to calculate the RNA Integrity Number (RIN) using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). The microarray assays were carried out using the Human Gene 1.0 ST Array (Affymetrix, Santa Clara, CA, USA) according to the manufacturer’s instructions.
2.4. Microarray Assays
For microarray experiments, target complementary DNA (cDNA) from each biopsy was prepared according to the WT Expression Sense Target Kit (Ambion, Austin, TX, USA). Briefly, one µg of RNA was converted into first-strand cDNA. Next, second-strand cDNA synthesis was performed, followed by in vitro transcription to generate cRNA. The cRNA products were used as templates for a second-cycle cDNA synthesis where dUTP was incorporated into the new strand. Purified sense-strand cDNA (with incorporated dUTP) was fragmented and labeled using the Affymetrix GeneChip WT Terminal Labeling Kit (Affymetrix, Santa Clara, CA, USA). The cDNA was fragmented using Uracil-DNA Glycosylase (UDG) and Apurinic/Apyrimidic Endonuclease 1 (APE1). The fragments (40–70 mers) were then marked by biotin-labeled deoxynucleotide terminal addition reaction using a terminal deoxynucleotidyl transferase (TdT). Finally, using the GeneChip Hybridization, Wash, and Stain Kit, each fragmented and labeled cDNA sample was hybridized onto Affymetrix Human Gene 1.0 ST arrays (Affymetrix, Santa Clara, CA, USA), respectively. After hybridization, the 13 arrays were washed, stained for biotinylated cDNA, and scanned per the manufacturer’s recommendations to obtain CEL files for each microarray.
2.5. Microarray Analysis and Selection of Differentially Expressed Genes
Samples were classified into three main groups: (1) samples from patients with Follicular Gastritis [Control group (Ctrl)], (2) samples from patients with Chronic Atrophic Gastritis (CAG), and (3) samples from patients with Gastric Cancer (GC). All possible pairwise comparisons among the three groups generated three comparisons of interest: CAG vs. Ctrl; GC vs. Ctrl, and GC vs. CAG. We performed a low-level data analysis in which the raw microarray data were background-corrected using the robust multi-array average or the RMA method [
12] and normalized using the quantile normalization approach [
13,
14], both executed in R v. 3.1.3 (
http://www.cran.r-project.org, accessed 18 March 2018). Differential expression was determined by fitting a linear model on each gene using the Limma package with an empirical Bayesian approach [
15,
16]. Correction for multiple hypotheses was applied by controlling the false discovery rate (FDR). Genes were selected as differentially expressed based on |logFCh| ≥ 0.85 and statistical significance considering a B-statistic ≥ 1 with associated FDR adjusted
p-values ≤ 0.01. The expression matrix created with this procedure was employed for the selection of differentially expressed genes, the enrichment analyses, and the evaluation of the functional properties of the genes.
2.6. Enriched Gene Ontology Terms
Three tools were utilized to evaluate the functional properties and pathway analysis of the differentially expressed genes. We used gene annotation enrichment analysis within the set of significant genes, employing the Database for Annotation, Visualization and Integrated Discovery bioinformatics tool (DAVID) v6.7 (
http://david.abcc.ncefcrf.gov/, accessed 7 October 2018), of the National Institute of Allergy and Infectious Diseases, NIH (NIAID) [
17,
18]. Enrichment analysis was also conducted using the Ingenuity Pathway Analysis (IPA) platform Ingenuity
® v.26127183 (Redwood City, CA, USA) as well as the Gene Set Enrichment Analysis software tool (GSEA) of the Broad Institute of Harvard and M.I.T. [
19,
20].
2.7. Amplification of Gene Expression by qPCR
Total RNA from each sample of the exploratory set was reverse transcribed using a SuperScript III Reverse Transcriptase Kit (Invitrogen, Carlsbad, CA, USA). The resulting complementary DNA (cDNA) was used in a real-time polymerase chain reaction (qPCR) to validate the genes of interest. We employed TaqMan probes and primers for claudin-1 (CLDN1), claudin-7 (CLDN7), matrix metalloproteinase-9 (MMP9), MYC (C-MYC), and olfactomedin-4 (OLFM4) from Applied Biosystems (Carlsbad, CA, USA) (assay ID: Hs00221623_m1; Hs00600772_m1; Hs00234579_m1; Hs00905030_m1, and Hs00197437_m1, respectively). The expression assays for each gene were conducted in triplicate on 96-well optical plates employing the ABI Prism 7000 sequence detection system and ABI Prism® 7000 SDS 1.2.3 with RQ (Applied Biosystems). The typical protocol utilized was the following: initial denaturing for 2 min at 95 °C, 40 cycles at 95 °C for 10 min, 95 °C for 15 s, and 60 °C for 1 min, and then a final extension at 72 °C for 7 min. Actin beta gene (ACTB, assay ID: Hs99999903_m1) was used as an internal control for the expression of independent sample-to-sample variability. Each gene-of-interest and the housekeeping gene were tested in triplicate. Relative gene expressions were determined from the obtained cycle threshold (Ct) values and the 2−∆∆Ct method.
2.8. Validation of Gene Assays
To validate the results obtained in the gene expression microarray and qPCR assays, the archives of the Pathology Department from INCMNSZ were reviewed to obtain pathological records and paraffin-embedded gastric biopsies (FFPE) samples from the past 10 years. We studied a total of 82 gastric tissues: 18 gastric biopsies without histological alterations, 20 of chronic follicular gastritis, 20 of chronic atrophic gastritis and 24 of gastric cancer. In addition, liver explant and right colon tissues were included as positive controls. A gastric biopsy from each study group were used as a negative control, in which the primary antibody was omitted. Glass slides stained with hematoxylin and eosin were checked for histological types according to Lauren classifications [
3]. Tumor samples were also obtained from patients undergoing gastric surgical resection. All samples were verified by a second pathologist. From the FFPE blocks corresponding to gastric tissues, serial 3-µm slices were taken from each block to perform an immunohistochemical analysis of CLDN1 and MMP9.
2.9. Immunohistochemistry of CLDN1 and MMP9
Descriptive analysis for the claudins and metalloproteinases in each study group was conducted. A positive signal indicating the expression of CLDN1 and MMP9 outside and inside the cells (positive signal either in the plasma membrane in the case of CLDN1 and in the extracellular matrix or in the nucleus for MMP9) was analyzed, and the distribution of the signal along the gastric tissue was evaluated. In brief, the FFPE samples from the validation set were dewaxed and rehydrated. Anti-CLDN1 antibody (DBS, Pleasanton, CA, USA) diluted at 1:100 was used for CLDN detection. Anti-MMP9 [MMP9/2025R] (Abcam, Cambridge, UK) diluted at 1:500 was used for MMP detection. The Mouse/Rabbit InmunoDetector HRP/DAB Detection System (Bio SB, Santa Barbara, CA, USA) was employed to conduct the immunohistochemistry (IHC). Tissue sections were briefly counterstained with hematoxylin and observed under the Olympus BX61VS microscope; images were acquired with Olympus OlyVIA 2.9 (Build 13771) software (Tokyo, Japan, Model BX-UCB). The negative control for the IHC was established by omitting the primary antibody against CLDN1 and MMP9.
2.10. Statistical Analysis
Clinical characteristics of patients were presented as mean ± standard error (SE) using statistical software (GraphPad Prism version 5.00 for Windows, GraphPad Software, San Diego, CA, USA). Statistical significance was determined by p-value (p ≤ 0.05). To confirm whether the genes were present in all gastric pathologies that proceed to gastric cancer, we used the Kruskal–Wallis test.
4. Discussion
In this study, we identified 770 genes differentially expressed in CAG vs. GC associated with
H. pylori infection by analyzing microarray assays. GSEA, DAVID, and IPA identified several bio-functions of genes (
Supplementary Figure S4) and the interaction networks for genes expressed differentially in each study group (
Figure 3 and
Figure 4, and
Supplementary Figure S5). Our data also suggests that multiple signaling pathways (
NF-kB,
RAS and
C-MYC) and genes (
CLDN1,
CLDN7,
MMP9,
OLFM4 and
C-MYC) could participate in the development of gastric cancer.
Among 541 genes found only in gastric cancer (GC vs. Ctrl) and 229 genes found in chronic atrophic gastritis (GCA vs. Ctrl), the genes (
CDH17,
OLFM4,
MUC17,
SI,
TM4SF20,
MUC12,
CLDN7,
ANPEP,
CST1 and
CLDN1) were found to be overexpressed in GC. This finding agrees with other studies which suggest that the overexpression of
C-MYC [
23,
24],
CLDN1, and
CLDN7 [
25,
26,
27,
28] is associated with the progression of GC. In the case of chronic atrophic gastritis (GCA vs. Ctrl), we found that the genes (
OLFM4,
CLCA1,
SI,
TMPRSS15,
CDH17,
TM4SF20,
DMBT1,
REG4,
ANPEP and
FABP2) were overexpressed, suggesting that the alteration of
OLFM4 and
CLDN7 may be associated with the development of CAG. When we performed the third contrast (GC vs. CAG), we found eight overexpressed genes (
CST1,
MXRA5,
PCDHB9,
SNX10,
MMP9,
C5AR1,
SOD2 and
TNFSF4). This result indicates that the significant increase in
MMP9 expression is likely related to the progression of GC. Interestingly, by confirming all these data by qPCR of the genes induced in CAG (
CLDN7 and
OLFM4) and GC (
CLDN1,
MMP9 and
C-MYC), we found that the genes mainly expressed in CAG and GC were
C-MYC and
MMP9 (
Figure 4).
It is well known that the claudin family plays a crucial role in the structure of tight-junction function in normal epithelial cells. The expression profile found in GC vs. Ctrl exhibits an increase in the claudin family (
Figure 1B). Although there are several reports on
CLDN1 expression in gastric cancer, there is still no agreement on the relationship between this expression and clinicopathological parameters [
25].
CLDN1 produces epithelial-mesenchymal transition (EMT) through activation of the c-Abl-ERK signaling pathway. In contrast, its overexpression in intestinal-type gastric cancer cogenerates an increase in lymph node metastases and tumor-node-metastasis (TNM) stages in patients with intestinal-type gastric cancer [
26]. Other reports propose that the overexpression of
CLDN1 is related to the carcinogenesis of invasive and metastatic gastric cancer [
25,
27]. In our study,
CLDN1 was one of the upregulated genes that we found in
H. pylori-exposed gastric epithelial cells, as previously observed [
29], indicating a probable relationship between chronic exposure to
H. pylori and
CLDN1 upregulation in gastric mucosa. We also found that
CLDN7 was upregulated in the tumor samples and that this gene could also be involved in gastric carcinogenesis [
28,
30,
31,
32], suggesting that this bacterium could regulate, in some way, the expression of
CLDN1 and
CLDN7.
On the other hand,
CLDN1 is considered as a marker of epithelial-to-mesenchymal transition (EMT) [
33] which intervenes in cellular processes such as migration, invasion, and matrix metalloproteinase (MMP) activation [
34]. Previous reports indicated that
CLDN1 inducted and activated
MMP2, which improve cell invasion and metastasis [
35,
36].
The matrix metalloproteinase family is important because it is involved in degrading components of the extracellular matrix (ECM). It also participates in many functions, such as physiological and pathological events, and stomach diseases such as gastritis and gastric cancer [
37,
38,
39] (
Figure 3). The MMP family also regulates the immune response, angiogenesis, invasion, cell growth, survival, and the EMT [
39]. Various studies have reported high levels of MMP9 in humans with gastrointestinal cancer through immunoassays and the observation of enzymes by electrophoretic techniques. In a TGF-β-signaling-deficient colon cancer model, tumor cells are capable of secreting
CCL9 that induces the enrollment of CCR1+ myeloid cells, producing MMP2 and MMP9 and facilitating the invasive growth of tumors [
40]. IL-1 induces the expression of genes (VEGF, IL-6, IL-8, TGF-β and MMP) involved in metastasis and inflammation [
41]. It has also been reported that IL-8 is able to increase the expression of MMPs by developing metastases [
42]. Another study showed that some human primary tumors are capable of recruiting neutrophils that secrete MMP9 and favor angiogenesis and intravasation of cancer cells [
43]. It is interesting to note that in post-surgical infections, these can activate neutrophil-producing traps (NET) containing high levels of active MMP9 [
44]. Tumor-associated macrophages are the main producers of proteases such as cathepsin and MMP, which can degrade the ECM, generating a tumor microenvironment and promoting the development of metastases [
45]. For example, under hypoxic conditions, prostate-cancer cells are capable of releasing exosomes charged with proinflammatory cytokines such as TNF, IL-6, proteinases MMP2 and MMP9, which enhance the invasiveness and metastasis of cancer cells [
46]. According to the results of these reports, our study shows that the expression of
MMP9 messenger RNA (mRNA) is significantly increased in the GC vs. Ctrl; the same occurs with C-MYC but to a lesser extent (
Figure 4). Through an in-depth analysis of
MMP9 expression in follicular gastritis, chronic atrophic gastritis, and gastric cancer, we found MMP9 expressed in polymorphonuclear cells that participate in the inflammation of lesions in gastritis. This signature could potentially be associated with
H. pylori infection. Interestingly, both
MMP9 and
CLDN1 began to express themselves from chronic atrophic gastritis and increased in gastric cancer (
Figure 5,
Table 6, and
Supplementary Table S5). In this way, CAG could be associated with a greater progression to malignant lesions.
Our results suggest that
H. pylori can use several mechanisms to interact with the gastric mucosa, which could eventually lead to the development of atrophy or gastric cancer. For example, CagA virulence factors and the peptidoglycan of
H. pylori induce signaling cascades of NF-κB, RAS, MEK, ERK, and C-MYC, resulting in the increased transcription of
CLDN1,
CLDN7,
OLFM4,
C-MYC, and
MMP9 genes, which promote cell proliferation, differentiation, survival, and eventually gastric carcinogenesis. In
Figure 6A, we present interactions between genes and the signaling cascade that is activated to favor the overexpression of
CLDN7 and
OLFM4.
H. pylori by CagA virulence factors can interact with
NF-κB and generate greater
CLDN7 expression; chronic inflammation may also contribute to the overexpression of the claudin family. The peptidoglycan of
H. pylori is recognized by NOD1, activates NF-κB, and increases the expression of
OLFM4 in chronic atrophic gastritis. In the gastric cancer scaffold (
Figure 6B), we observed that CagA could interact with NF-κB, RAS, and C-MYC to favor the expression of genes such as
CLDN1,
CLDN7,
C-MYC, and
MMP9. The augmented expression of
CLDN1 and
C-MYC could increase cell proliferation, whereas the increased expression of
MMP9 causes deterioration of the ECM that promotes cell migration and invasion in intestinal gastric cancer.