Different Metabolites of the Gastric Mucosa between Patients with Current Helicobacter pylori Infection, Past Infection, and No Infection History

Helicobacter pylori (H. pylori) alters metabolism during the gastric carcinogenesis process. This study aimed to determine the metabolites in the gastric mucosa according to the status of the H. pylori infection. Patients who visited the outpatient clinic for a gastroscopy and H. pylori tests were included. Gas chromatography–time-of-flight mass spectrometry (GC-TOF-MS) analysis was performed using gastric biopsied specimens from the corpus. Twenty-eight discriminative metabolites were found in the gastric mucosa of 10 patients with current H. pylori infection, in 15 with past infection, and in five with no infection history. The relative abundances (RAs) of amino acids and sugars/sugar alcohols were higher in patients with no infection history than in patients with current or past infection. The current infection group showed higher RAs of organic acids and lower RAs of fatty acids and lipids compared with the other groups. The RA of inosine was highest in the past infection group. Based on GC-TOF-MS analysis findings, metabolites differed not only between the infected and non-infected patients, but also between those with and without infection history. Amino acid and sugars/sugar alcohol metabolites decreased in patients with current or past infection, whereas fatty acid and lipid metabolites decreased only during current infection.


Introduction
Helicobacter pylori (H. pylori) infection induces the carcinogenesis process in the human stomach. Some precancerous lesions including gastric atrophy and intestinal metaplasia (IM) may persist even after the regression of H. pylori; therefore, the risk of gastric cancer is higher in stomachs with infection history than in H. pylori-naive stomachs [1,2]. Furthermore, the risk increases with the degree of atrophy and IM, which extends from the antrum to the corpus [3,4].
H. pylori alters metabolism after infection and changes metabolic dynamics when eradicated [5]. H. pylori infection leads to the loss of carboxylic acids and amino acids in the corpus and antrum [6]. In the latter study, the corpus exhibited significant metabolite changes related to stress, tissue damage, and nutrient depletion in contrast to the antrum. Furthermore, 43 plasma metabolites involved in amino acid, lipid, and fatty acid metabolism differed between H. pylori-infected, non-infected, and gastric cancer patients [7]. Although studies on metabolomic profiling of gastric cancer are increasing, main findings are inconsistent [8]. Moreover, no study has evaluated the differences in metabolism between patients with current infection, past infection, and no infection history.
In this study, we attempted to determine the metabolites found in the gastric mucosa according to the H. pylori infection status. We hypothesized that metabolites observed in the corpus would differ not only between the infected and non-infected patients, but

Gastroscopic Examination and H. pylori Tests
The gastroscopy was performed by a gastroenterologist (Dr. Lee S.-Y.) using GIF-H290 (Olympus, Tokyo, Japan). A rapid urease test (PyroPlus, ARJ Medical, Inc., Oldsmar, FL, USA) was performed using gastric biopsy specimens taken from the greater curvature side of the upper body. Pathology findings were reported according to the updated Sydney classification as previously described [9]. Briefly, no infiltration was scored as 0, mild degree as 1, moderate as 2, and marked as 3. H. pylori serology and serum pepsinogen (PG) assays were performed as described in our previous study [10].
Endoscopic findings were scored from 0 to 8 based on the Kyoto classification scoring system for gastritis [11]. Gastric atrophy was scored 0 for no atrophy or closed-type I (atrophy confined to the antrum), 1 for closed-type II or III (atrophy exhibiting atrophic border in the lesser curvature side of the body), or 2 for open-type atrophy [12]. IM was scored 0 for no IM, 1 for limited IM confined to the antrum, or 2 for extensive IM observed in both the corpus and antrum. Hypertrophic gastric folds were scored 0 (absent) or 1 (present). Nodular gastritis was scored 0 (absent) or 1 (present). Diffuse redness in the corpus was scored 0 (none), 1 (mild), or 2 (severe).

Confirmation of H. pylori Infection Status
Current infection was diagnosed if the rapid urease test or Giemsa staining findings were positive. Past infection was defined as a successful history of eradication in patients with negative H. pylori test findings. The remaining patients without a history of eradication were classified as patients with no infection history. If there was a discrepancy between the H. pylori serology and invasive test findings (rapid urease test, histology, and Giemsa staining), the 13 C-urea breath test (POCone ® , Otsuka Electronics Co., Ltd., Osaka, Japan) was performed for confirmation.

Sample Preparation for Metabolite Analysis
The extraction of gastric mucosa metabolites was performed using the methods described by Park et al., with a few modifications [13]. Gastric mucosa samples were homogenized and extracted in 100% methanol (400 µL) containing an internal standard solution (10 µL; 2-chloro-phenylalanine, 1 mg/mL in water) using an MM400 mixer mill (Retsch ® , Haan, Germany) at a frequency of 30 Hz for 10 min with sonication for another 10 min. Subsequently, the extracted samples were incubated for 1 h at 4 • C and centrifuged at 13,000 rpm for 10 min at 4 • C. Then, the supernatants were filtered using 0.2 µm polytetrafluorethylene (PTFE) filters (Chromdisc, Daegu, Korea). The filtered samples were completely evaporated in a speed vacuum concentrator (Biotron, Seoul, Korea). The completely dried extracts were reconstituted with 50 µL of 100% methanol.

Gas Chromatography-Time-of-Flight Mass Spectrometry Analysis
For gas chromatography-time-of-flight mass spectrometry (GC-TOF-MS) analysis, the reconstituted samples (50 µL) were dried using a speed vacuum concentrator. The dried samples were derivatized using the following protocol. First, 25 µL of methoxyamine hydrochloride (20 mg/mL in pyridine) was added to the dried samples and placed on a thermomixer for 90 min at 30 • C to protect the ketone and aldehyde groups. Then, 25 µL of N-methyl-N-trimethylsilyl trifluoroacetamide (MSTFA), being used as a silylating agent, was incubated for 30 min at 37 • C. One microliter of each derivatized samples was injected into an Agilent 7890A GC system (Santa Clara, CA, USA) equipped with an L-PAL3 autosampler and Pegasus ® HT TOF-MS system (LECO Corp., St. Joseph, MI, USA). Metabolites were separated using an RTX-5MS column (30 m length × 0.25 mm inner diameter × 0.25 µm particle size; Restek Corp., St. Joseph, MI, USA) with a constant flow of helium (1.5 mL/min) as the carrier gas. The analytical parameters and program were adopted from our previous study [13]. All samples were run in a randomized manner to reduce systematic errors and bias.

Data Processing and Statistical Analysis
MS data processing and multivariate statistical analyses were performed as described in our previous study [13]. The raw GC-TOF-MS data were converted to netCDF (*.cdf). Then, the .cdf format files were processed using MetAlign software (RIKILT-Institute of Food Safety, Wageningen, The Netherlands) to determine baseline correction, peak selection, peak area normalization, peak mass (m/z), and retention time (min). Multivariate statistical analysis was performed using SIMCA-P+ software (version 12.0, Umetrics, Umea, Sweden). Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were performed to compare significantly different metabolites among the three groups. The significance of the OPLS-DA models was determined through the analysis of variance testing of cross-validated predictive residuals (CV-ANOVA) derived from the SIMCA-P+ program. Discriminative variables were selected based on the variable importance in the projection (VIP) scores derived from the OPLS-DA model. The discriminative metabolites obtained from GC-TOF-MS were tentatively identified by comparing their retention time, mass spectrum, and mass fragment pattern with available databases such as the National Institute of Standards (NIST) and Technology database (version 2.0, 2001, FairCom, Gaithersburg, MD, USA), Wiley 9, the Human Metabolome Database (HMDB, http://www.hmdb.ca/ accessed on 19 January 2022), and in-house libraries and standard compounds analyzed under identical GC-TOF-MS analysis conditions. Significant differences were evaluated through ANOVA or the Student's t-test using PASW Statistics 18 software (SPSS Inc., Chicago, IL, USA).
For continuous variables, differences between the three groups were analyzed using ANOVA with Bonferroni correction, and findings were presented as mean ± standard deviation. For continuous variables with asymmetrical distribution, the Kruskal-Wallis test was used, and findings were presented as medians and ranges. Categorical variables were analyzed using the Chi-squared test with Bonferroni correction, and findings were presented as percentage. Furthermore, correlation analysis was performed to verify the link between the RA of each metabolite and the degree of gastric atrophy and IM. The findings were presented as Pearson's correlation coefficient (r) values.

Infection Status of the Included Patients
Among the 30 included patients, 10 showed positive H. pylori test findings and were classified into the current infection group. Fifteen patients with negative H. pylori test findings had an eradication history; hence, they were classified into the past infection group. The remaining five patients with negative H. pylori test findings were classified into the no infection history group.

Different Test Findings According to the Status of H. pylori Infection
The current infection group showed higher serology titers, higher serum PG II levels, and lower PG I/II ratios than those in the past infection and no infection history groups. The total Kyoto classification score was highest in the current infection group, whereas it was lowest in the no infection history group. The findings of the serum assays and gastroscopy are summarized in Table 1.  Duodenal ulcer scar 2 (6.7%) 0 2 (13.3%) 0 * Significant difference compared with the current infection group (p < 0.05). ** Significant difference compared with the past infection group (p < 0.05). For continuous variables, ANOVA with Bonferroni correction was used, and findings are presented as mean ± standard deviation (SD). For variables with asymmetrical distribution, the Kruskal-Wallis test was used, and findings are presented as medians and ranges. For categorical variables, the Chi-squared test with Bonferroni correction was used.

Gastric Mucosa Metabolites Observed in Patients with Current H. pylori Infection, Past Infection, and No Infection History
To determine significantly distinguished metabolites among the three groups, nontargeted metabolite profiling of the gastric mucosa was conducted using GC-TOF-MS followed by multivariate statistical analysis, including unsupervised PCA and supervised OPLS-DA. The OPLS-DA score plot obtained from the GC-TOF-MS dataset showed a clearly distinguished pattern among the CI, PI, and NI groups according to OPLS1 (4.96%) ( Figure 1A). The statistical model value of OPLS-DA was determined by R 2 X (cum) = 0.316, R 2 Y (cum) = 0.560, Q 2 (cum) = 0.240, and p < 0.05, which indicated the model validation, prediction accuracy, fitness, and cross-validation analysis, respectively. Nonetheless, the unsupervised PCA score plots were not clearly separated from the three groups ( Figure S1). Distinguished metabolites among the three groups were identified and selected based on the VIP value (>1.0) from the OPLS-DA model. One-way ANOVA was applied to determine statistical significance (p < 0.05).
Biomedicines 2022, 10, x FOR PEER REVIEW 5 of 12 R 2 Y (cum) = 0.560, Q 2 (cum) = 0.240, and p < 0.05, which indicated the model validation, prediction accuracy, fitness, and cross-validation analysis, respectively. Nonetheless, the unsupervised PCA score plots were not clearly separated from the three groups ( Figure  S1). Distinguished metabolites among the three groups were identified and selected based on the VIP value (>1.0) from the OPLS-DA model. One-way ANOVA was applied to determine statistical significance (p < 0.05). A total of 28 metabolites including four organic acids, nine amino acids, four sugars and sugar alcohols, six fatty acids and lipids, four others, and one unknown compound were tentatively identified ( Table 2). The relative contents of 28 significantly different metabolites were converted into fold changes and displayed on a heat map ( Figure 1B). Based on this, the relative abundance (RA) of organic acids, except for pyruvic acid, GABA, salicylic acid, and uric acid, was relatively higher in the current infection group than the RAs in the past infection and no infection history groups. Intriguingly, the RAs of fatty acids and lipids in the current infection group were relatively lower than those in the other two A total of 28 metabolites including four organic acids, nine amino acids, four sugars and sugar alcohols, six fatty acids and lipids, four others, and one unknown compound were tentatively identified ( Table 2). The relative contents of 28 significantly different metabolites were converted into fold changes and displayed on a heat map ( Figure 1B). Based on this, the relative abundance (RA) of organic acids, except for pyruvic acid, GABA, salicylic acid, and uric acid, was relatively higher in the current infection group than the RAs in the past infection and no infection history groups. Intriguingly, the RAs of fatty acids and lipids in the current infection group were relatively lower than those in the other two groups. The RA of inosine was significantly higher in the past infection group. The RAs of amino acids (valine, leucine, isoleucine, asparagine, and cystine) and four sugars and sugar alcohols were higher in the no infection history group. Collectively, the current infection group showed higher RAs of organic acids and lower RAs of fatty acids/lipids Biomedicines 2022, 10, 556 6 of 11 compared to those in the past infection and no infection history groups. Moreover, the RAs of amino acids and four sugars/sugar alcohols were higher in the no infection history group than those in the other two groups.

Gastric Mucosa Metabolites Observed in Patients with a History of Gastric Neoplasm or Peptic Ulcer Disease
The included patients did not have active or healing-stage ulcers. Duodenal ulcer scars were observed in two patients in the past infection group. Gastric ulcer scars were observed in four patients owing to previous endoscopic resection for gastric neoplasm. Three patients in the past infection group underwent gastric adenoma resection, whereas one patient in the current infection group underwent early gastric cancer resection. The latter patient developed reinfection after successful H. pylori eradication.
Among the 28 metabolites, the RAs of amino acids and sugars/sugar alcohols (except sucrose) were relatively lower in the patients with a history of gastric neoplasm than those in their counterparts (Figure 2A). Conversely, the RAs of amino acids (except GABA and cystine) and inosine were relatively higher in patients with duodenal ulcer scars than those in their counterparts ( Figure 2B).

Correlation between the Metabolites and Degree of Atrophy and IM
Among the 28 significantly different metabolites, the RAs of valine (r = −0.481, p = 0.011), leucine (r = −0.414, p = 0.032), and isoleucine (r = −0.404, p = 0.038) were inversely correlated with the degree of gastric atrophy. With regard to the degree of IM, only the RA of valine exhibited inverse correlation (r = −0.481, p = 0.011). The RAs of salicylic acid (r = 0.406, p = 0.044) and sucrose (r = 0.523, p = 0.003) showed positive correlations with the degree of IM.

Gastric Mucosa Metabolites Observed in Patients with a History of Gastric Neoplasm or Peptic Ulcer Disease
The included patients did not have active or healing-stage ulcers. Duodenal ulcer scars were observed in two patients in the past infection group. Gastric ulcer scars were observed in four patients owing to previous endoscopic resection for gastric neoplasm. Three patients in the past infection group underwent gastric adenoma resection, whereas one patient in the current infection group underwent early gastric cancer resection. The latter patient developed reinfection after successful H. pylori eradication.
Among the 28 metabolites, the RAs of amino acids and sugars/sugar alcohols (except sucrose) were relatively lower in the patients with a history of gastric neoplasm than those in their counterparts (Figure 2A). Conversely, the RAs of amino acids (except GABA and cystine) and inosine were relatively higher in patients with duodenal ulcer scars than those in their counterparts ( Figure 2B).

Gastric Mucosa Metabolites Observed in Patients with H. pylori-Negative Gastritis
Inflammatory cell infiltration was observed in all 10 patients with H. pylori infection, whereas only seven (46.7%) patients showed mononuclear cell or neutrophil infiltration among the 15 patients with past infection. Among the five H. pylori-naive patients, two (40%) showed a mild degree of inflammatory cell infiltration. Therefore, nine patients were considered to have H. pylori-negative gastritis. The RAs of fatty acids/lipids, amino acids (except GABA and cystine), and sugars/sugar alcohols (except sorbitol) were relatively higher in the H. pylori-negative gastritis group than those in the H. pylori gastritis group ( Figure S2).

Discussion
In this study, we found that metabolites of the gastric corpus differed not only between H. pylori-infected and non-infected patients, but also between H. pylori-exposed patients (current and past infection groups) and non-exposed patients (no infection history group). From a metabolic perspective, amino acids and sugars/sugar alcohols were decreased in patients with an exposure history. Fatty acid and lipids were decreased only during current infection and seemed to be restored after the regression of H. pylori. To the best of our knowledge, this is the first study to show differences in gastric corpus metabolites between the patients with current infection, past infection, and no infection history.
Fatty acid and lipid metabolites decreased in patients with current infection. During H. pylori infection, the gastric mucosal barrier becomes fragile due to accelerated fatty acid metabolism in gastric mucosal cells [14]. H. pylori infection altered fatty acid metabolism by increasing the concentrations of arachidonic acid and prostaglandin E2 in their study. These changes diminished with the regression of H. pylori; hence, decreased fatty acid and lipid metabolites were no longer observed in the absence of infection. Our findings are supported by a previous study showing that lipid peroxide increased during H. pylori infection due to gastric mucosal oxidative inflammation [15]. From the perspective of oxidative stress, microvascular leucocyte activation and chemokine and myeloperoxidase activities increased during H. pylori infection, with significant changes in lipid metabolites.
In a study using plasma metabolites, the loss of linoleic acid and palmitic acid was correlated with the progression of IM after adjusting for age, sex, H. pylori infection, and histological findings [16]. Another study using plasma metabolites of gastric cancer patients found no difference according to the presence of infection, although significant metabolites were found in men and smokers [17]. These two studies suggest that metabolic deregulation of fatty acids and lipids is involved in gastric carcinogenesis; however, they did not find a correlation with current infection. Different from their studies, two Korean studies reported under-expressed lipid metabolites in gastric cancer tissues [18,19]. Because the proportion of current infection, past infection, and no infection is 75.2%, 22.5%, and 2.3% in Korean gastric cancer patients [20], these Korean studies support our results showing that fatty acid and lipid metabolites decrease during H. pylori infection. Recent Chinese studies using gastric cancer tissues also reported a significant decrease in lipid metabolites [21,22]. Therefore, when analyzing metabolic profiles, it is important to obtain the gastric tissue and to discriminate patients with and without an infection history.
Notably, metabolites on sugars and sugar alcohols were decreased in patients with current or past infection. In an in vivo study, 11 metabolites of glycolysis, tricarboxylic acid cycle, and amino acid metabolism were altered by H. pylori infection [23]. In human studies, several types of glycans showed decreased levels due to altered glycolysis in patients with gastric cancer [24,25]. Similarly, amino acid metabolites were decreased in patients with current or past infection. In an animal study, the glycolytic pathway, tricarboxylic acid cycle, and choline pathway were upregulated, whereas the urea cycle was downregulated after H. pylori infection [26].
Gastric cancers are more common in the severe atrophy group than in the moderate, mild, and none atrophy groups [27]. In their study, advanced atrophy, aging, uric acid levels, and ulcers were risk factors of gastric cancer. In our study, the RA of uric acid was increased in the current infection group; however, there was no correlation with the degree of atrophy or IM. Instead of uric acid, valine, leucine, and isoleucine metabolites decreased with the progression of gastric atrophy or IM. Our findings can be supported by a recent study showing that valine, leucine, and isoleucine synthesis and glycolysis are disturbed during gastric carcinogenesis due to the increased energy requirement [28]. Another study on tongue coating metabolites also reported that degradation of valine, leucine, and isoleucine is observed in patients with gastric precancerous lesions [29]. Together with our findings, the downregulation of valine, leucine, and isoleucine seems to be involved with the progression of atrophy or IM.
Interestingly, patients with H. pylori-negative gastritis did not show loss of fatty acids, lipids, amino acids, sugars, and sugar alcohols in our study. This finding can be supported by our previous study, which showed that H. pylori-negative gastritis was not associated with atrophy and IM [9]. Only H. pylori-related gastritis can induce endoscopic gastritis, which requires gastric cancer surveillance. Similarly, patients with duodenal ulcer scars showed increased amino acid metabolite levels, whereas patients with a history of gastric adenoma or cancer resection showed decreased amino acid and sugar/sugar alcohol metabolite levels. A recent study showed that the level of metabolites related to glutathione cycle is lower in epithelial cells infected with gastric cancer-derived H. pylori strains than those infected with duodenal ulcer-derived strains [30]. They demonstrated that the oxidative stress induced by gastric cancer-derived H. pylori strains was stronger than that induced by duodenal ulcer-derived strains. Altogether, patients with H. pylorinegative gastritis or duodenal ulcer history rarely show gastric carcinogenesis, in which valine, leucine, isoleucine synthesis and glycolysis are disturbed.
This study had several limitations. First, in an endemic area of H. pylori infection, there is a risk of unintended eradication in patients without a history of infection. Older patients may have been infected in the past. Second, our findings may not be observed outside the East Asia, because 94.4-96.2% of Koreans with H. pylori infection show positive East Asian type CagA findings [31,32]. Third, biopsy samples were obtained only from the greater curvature side of the upper body, because false-negative findings are not rare when samples are obtained from other sites [33,34]. Despite these limitations, we found significant differences in fatty acid/lipid, amino acid, and sugar/sugar alcohol metabolites between patients with current infection, past infection, and no infection history.

Conclusions
Non-targeted metabolite profiling of gastric mucosa tissues reveals that the loss of amino acids and sugar/sugar alcohols is observed in patients with current or past infection. Loss of fatty acids and lipids seems to recover after regression of H. pylori, because it is only observed during current infection. Collectively, gastric mucosa tissue derived from H. pylori infection or exposure history affected metabolite variation. Our multi-omics data would be useful for not only determining the status of H. pylori infection but also discriminating significant gastritis related to gastric cancer. New insights into delaying atrophy and IM might be provided by understanding the downregulation of valine, leucine, and isoleucine metabolites. Furthermore, the efficacy of novel targeting agents for gastric cancer prevention should be evaluated by measuring the downregulation of the H. pylorirelated metabolome. Comprehensive multi-omics approaches are needed to understand the meaning of gastric mucosa-related metabolic changes observed during H. pylori-induced gastric carcinogenesis.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/biomedicines10030556/s1. Figure S1. Principal component analysis score plot derived from GC-TOF-MS data of gastric mucosa samples from the three different groups (CI, PI, and NI group). Symbol: current infection (CI, •), past infection (PI, •), and no history of infection (NI, •). To determine the statistical significance, p <0.05 obtained from one-way ANOVA was applied. Figure S2. Heat map analysis for the relative abundance of different metabolites (VIP >1.0) derived from the GC-TOF-MS analysis. The colored squares (blue to red) indicate the fold changes normalized by the average of each metabolite. To determine the statistical significance, p <0.05 obtained from one-way ANOVA was applied. Hp, H. pylori.