This study was a clinical retrospective study. A total of 203 participants were recruited from the endoscopy center and gastrointestinal surgery of the first affiliated hospital of Guangzhou University of Chinese Medicine. All participants consisted of healthy people who underwent routine physical examination and patients diagnosed with chronic superficial gastritis, gastric ulcer, atrophic gastritis or gastric cancer by endoscopy and histopathological examination. The criteria for performing endoscopy referred to the ASGE guideline (Early et al., 2012). Subjects with a history of other serious systemic diseases and other malignancies were excluded. The healthy control group in this study included individuals with normal gastric mucosa or only mild chronic inflammation without other lesions. Gastric mucosa specimens of each subject were taken for subsequent Hp infection, histopathological changes of gastric mucosa, miRNA gene polymorphism and expression level detection. The study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Guangzhou University of Chinese Medicine [ethics approval No. (2015) 009]. All subjects signed the informed consent.

The detailed inclusion and exclusion criteria of patients were as follow: Inclusion criteria: 1) Those who meet the diagnostic criteria for those above gastric diseases; 2) Age between 18 and 80 years old; 3) Those who voluntarily participated in the study and signed the informed consent form. Exclusion criteria: 1) Patients combined with other serious systemic diseases; 2) Those who took traditional Chinese medicine, PPI or Hp eradication treatment in the past month; 3) Pregnant and breastfeeding women; 4) Those with a history of severe mental illness or intellectual disability who are unable to cooperate with the collection of medical history data and sign the informed consent form.

HiPure tissue DNA kit (Magen, Guangzhou, China) was used to extract genomic DNA from gastric mucosal tissues. Polymerase chain reaction was performed by PCR DSMIX kit (Dongsheng Biotech, Guangdong, China). The primer sequences of the five miRNAs were as follow: 5′-CGA​CTC​TCT​ATG​AGA​ATT​ATG​C-3′ (forward) and 5′-AAT​TGG​ATG​CCG​CAG​TGG​TC-3′ (reverse) for miR-499; 5′-TCCAGATAGATGCAAAGCTG-3′ (forward) and 5′-AGA​CTC​CTCTC​CCT​TAA​TCA​C-3′ (reverse) for miR-196a2; 5′-CAT​CTC​ATG​TCC​AGG​ACC​AC-3′ (forward) and 5′-AGTGAGCTGGTCCAAGACTCAG-3′ (reverse) for miR-149; 5′-GAC​CTG​GTA​CTA​GGA​AGC​AG-3′ (forward) and 5′-CTT​ATA​CCT​TCA​GAC​CTG​AG-3′ (reverse) for miR-146; 5′-TGT​GTT​TCA​GCT​CAG​TAG​GCA​C-3′ (forward) and 5′-CTG​TCA​CAA​ATC​ACA​TTGCC-3′ (reverse) for miR-27a. The PCR reaction procedure was as follows: for miR-499 and miR-149, pre-deformation at 94°C for 4 min, followed by denaturation at 94°C for 30 s, annealing at 60°C for 30 s, and 72°C for 30 s for a total of 40 cycles; finally, extension at 72°C for 10 min and storage at 10°C for 30 s. For miR-196a2 and miR-146a, pre-deformation at 94°C for 4 min, followed by denaturation at 94°C for 30 s, annealing at 60°C for 30 s, and 72°C for 30 s for a total of 32 cycles; finally, extension at 72°C for 10 min and storage at 10°C for 30 s. For miR-27a, pre-deformation at 94°C for 4 min, followed by denaturation at 94°C for 30 s, annealing at 60°C for 30 s, and 72°C for 30 s for a total of 35 cycles; finally, extension at 72°C for 10 min and storage at 10°C for 30 s. The PCR products were identified by 1% agarose gel electrophoresis. After purification, PCR products were cycle sequenced using BigDye Terminator v3.1 cycle Sequencing kit following ABI standard procedures.

Total RNA extraction was performed using Hipure Universal miRNA Kit (Magen). The isolated RNA was reverse transcribed to cDNA using the ToYoBo qPCR RT Kit (ToYoBo). KAPA Probe Fast qPCR Master Mix (KAPA BIOSYSTEMS) was used for fluorescent qPCR reactions. The expression level of U6 gene was used as the internal reference for target miRNA genes. The primer sequences of target genes and U6 were shown in Supplementary Table S1. The PCR reaction conditions were: 95°C for 3 min, followed by 40 cycles (95°C for 3 s, 55°C for 30 s) and 30°C for 30 s. Three replicate measurements were performed for each sample. 2-ΔΔCq method was used to calculate the expressions of miRNAs (Livak and Schmittgen, 2001).

The Kyoto consensus (Sugano et al., 2015) as well as the New Sydney system (Dixon et al., 1996) were consulted for the diagnostic criteria of Hp infection. Rapid urokinase test and methylene blue staining were applied to detect Hp infection. As long as one of these tests was positive, the subject was considered to be Hp-infected. According to the distribution of Hp, the degree of Hp infection was classified into four levels: negative, mild, moderate and severe.

The histopathological changes of gastric mucosa were observed by hematoxylin-eosin staining. All slides were reviewed blindly by two trained pathologists. According to the new Sydney system (Dixon et al., 1996), the severity of inflammation, inflammatory activity, atrophy, intestinal metaplasia and dysplasia in subjects’ gastric mucosa was divided into four grades: none, mild, moderate, and severe. Based on the above results and research needs, we divided all subjects into five different pathological groups as follows: 1) Normal control group (NOR): those with basically normal gastric mucosa or only with mild inflammation; 2) Chronic inflammation group (CI): Those with moderate to severe inflammation in gastric mucosa; 3) Gastric atrophy group (GA): Those whose gastric mucosa with glandular atrophy as the main manifestation, not accompanied by intestinal metaplasia as well as dysplasia; 4) Gastric precancerous lesion group (GPL): Gastric mucosa atrophy with intestinal metaplasia or/and dysplasia; 5) Gastric cancer group (GC): Patients diagnosed with gastric cancer. Considering that severe dysplasia is more likely to become cancerous in precancerous lesions of gastric cancer, the gastric precancerous lesion group (GPL) was subdivided into two subgroups: early precancerous lesion group (EPL) and severe dysplasia group (SE). Early precancerous lesion group (EPL): Those with gastric atrophy accompanied by intestinal metaplasia and/or mild to moderate dysplasia; Severe dysplasia group (SE): those with severe dysplasia in gastric mucosa.

Based on normality of data, continuous variables were expressed as mean ± standard deviation or median (first quartile, third quartile). Categorical variables were analyzed by χ² test or Fisher exact test. Kruskal-Wallis H test was used to estimate the differences in miRNA expression levels between various groups. Hardy-Weinberg equilibrium test of SNPs was performed using χ² test or Fisher exact test. Binary logistic regression was used to evaluate the odds ratio (OR) and 95% confidence interval (CI) for the associations between miRNA gene polymorphism and GPL as well as GC risk. Spearman correlation analysis was used to examine the correlations between miRNA expressions and severity of Hp infection as well as histopathological changes of gastric mucosa. Gene-gene and gene-environment interactions were explored by multifactor dimensionality reduction (MDR) method. 1000 permutation test was used to assess the statistical significance of MDR interaction models. The above data analysis was carried out using SPSS 21, GraphPad Prism 8.0, open source multifactor dimensionality reduction (MDR) software (V3.0.2), and MDR permutation test software BETA version 0.4.6. p < 0.05 was considered statistically significant.

The baseline characteristics of subjects were shown in Table 1. The differences in the distribution of Hp infection severity between different pathological groups were statistically significant. The rate of Hp infection was significantly higher in GPL and GC group than in NOR group (all p < 0.01). Specifically, there were more moderate to severe Hp infected subjects in GPL group than in CI and NOR group (p = 0.001, p < 0.001). There was no significant difference in age and sex distribution among pathological groups. Figure 1 demonstrated different Hp infection situations and histopathological features in gastric mucosa of subjects.

Different genotypes of all miRNA SNPs were detected by PCR direct sequencing, and the results were shown in Figure 2.

The results of Hardy Weinberg equilibrium test showed that the distribution of different genotypes in GPL, GC and NOR groups were in Hardy Weinberg equilibrium (all p > 0.05) (Table 2). The frequencies of different genotypes and alleles of miR-499 rs3746444, miR-149 rs2292832, miR-196a2 rs11614913, miR-146a rs291016 and miR-27a rs895819 in GPL, GC group and NOR group were presented in Table 2. The frequency of CT genotype at miR-196a2 rs11614913 locus was significantly higher in GPL and GC group than in NOR group [CT vs. CC: adjusted OR = 6.16, 95%CI (1.46–26.03), p = 0.01; adjusted OR = 11.83, 95%CI (1.65–84.72), p = 0.01, respectively]. In addition, the frequency distribution of CT genotype and C allele of miR-27a rs895819 in GC group was significantly higher than in NOR group [CT vs. TT: adjusted OR = 10.14, 95%CI (2.25–45.77), p < 0.01; C vs. T: adjusted OR = 3.71, 95%CI (1.46–9.44), p = 0.01]. And the differences in genotypes and alleles distribution of the other three SNPs between GPL, GC group and NOR group were not statistically significant.

We analyzed the frequency distribution of allele combinations at the five miRNA (miR-499/miR-149/miR-196a2/miR-146a/miR-27a) SNPs in GPL, GC group and NOR group to assess the potential synergistic effects among miRNAs. Compared with NOR group, the frequency of A-T-C-C-T allele combination was significantly lower and the frequencies of A-T-T-G-C and A-C-T-C-T allele combination were significantly higher in GC group (all p < 0.05) (Table 3). While in GPL group, the frequencies of A-T-C-G-T, A-T-T-G-T, G-T-T-C-T and G-T-T-G-T allele combinations were significantly lower and the frequencies of A-C-C-C-T, A-C-C-C-C, A-C-T-C-T and A-C-T-C-C allele combinations were significantly higher (all p < 0.05) (Table 3).

The MDR method was used to explore whether there were gene-gene and gene-environment interactions between miRNA SNPs and Hp. As illustrated in Table 4, the second-order interaction model of rs3746444-rs11614913 was the best model with highest testing balanced accuracy of 56.4% among gene-gene interaction models for GPL susceptibility (Table 4). And within the gene-gene interaction models for GC susceptibility, the third-order interaction model of rs2292832-rs11614913-rs2910164 had the largest testing balanced accuracy of 61.1% and cross-validation consistency of 100% (Table 4), which was the best model. However, neither of the above two best models was statistically significant based on 1000 permutation tests. According to the results of gene-environment interaction analysis, only the second-order interaction model of rs11614913-Hp was statistically significant in two-order and higher-order models for GPL susceptibility (Table 5).

To investigate the effect of Hp infection on miRNA expression levels in each pathological group, we divided all subjects into two parts for discussion according to whether Hp was infected or not. As displayed in Figure 3, expression levels of miR-499, miR-196a2 and miR-27 were statistically significant in different pathological groups among Hp positive subjects (p = 0.031, p = 0.002, p < 0.001, respectively). In particular, expression level of miR-499 was significantly higher in GC group than in EPL group (p = 0.029). And expression level of miR-27a was significantly up-regulated in GC group compared to CI, GA and EPL groups (p = 0.017, p = 0.021, p = 0.011). What’s more, expression level of miR-196a2 was significantly higher in GC group than in NOR, CI and EPL groups (p < 0.001, p = 0.015, p = 0.002). However, there was no significant difference in miRNA expression levels between different pathological groups among Hp-negative subjects (Supplementary Figure S1).

The results were revealed in Supplementary Table S2. The CC genotype of miR-146a rs2910164 was associated with lower miRNA expression level than CG/GG genotype in NOR, CI and GC groups (p = 0.025, p = 0.013, p = 0.024).

Spearman correlation analysis showed that miR-196a2 expression level was positively correlated with the severity of inflammation, inflammatory activity, gastric atrophy, intestinal metaplasia and dysplasia (all p < 0.05) (Figure 4). Moreover, the expression level of miR-146a was positively correlated with degree of Hp infection and inflammation in gastric mucosa (all p < 0.05) (Figure 4). Conversely, miR-149 expression level were negatively correlated with severity of Hp infection, inflammation, inflammatory activity, atrophy and dysplasia (all p < 0.05) (Figure 4).