The metabolites mainly composed of lipids in tongue coating are non-invasive potential biomarkers for chronic gastritis

The changes in tongue coating metabolites in patients with chronic gastritis (CG) under different gastroscopy indicators were analyzed, and these metabolites were screened for potential non-invasive biomarkers to assist in the diagnosis of chronic gastritis. The technology of gas chromatography and liquid chromatography combined with mass spectrometry has been used to more comprehensively detect tongue coating metabolites of 350 CG patients. Spearman correlation analysis and random forest algorithm were used to screen metabolites that can serve as potential biomarkers. Compared with healthy individuals, CG group showed significant changes in the content of 101 metabolites, with an increase in the content of 54 metabolites and a decrease in the content of 47 metabolites. These differential metabolites are mainly composed of 47 lipids and lipid like substances. 1 metabolite was associated with bile reflux, 1 metabolite was associated with gastric mucosal erosion, 10 metabolites were associated with atrophy, 10 metabolites were associated with intestinal metaplasia, and 3 metabolites were associated with Helicobacter pylori infection. The ROC model composed of 5 metabolites can distinguish between CG group and healthy individuals, with an accuracy of 95.4%. The ROC model composed of 5,6-Dihydroxyindole can distinguish between chronic superficial gastritis group and chronic atrophic gastritis group, with an accuracy of 75.3%. The lipids and lipid like metabolites were the main abnormal metabolites in patients with chronic gastritis. It was worth noting that the content of Sphinganine 1-phase, 4-Ipomenol, and Nervonic acid in tongue coating increased, and the content of 1-Methyladenosine and 3-Hydroxycapric acid decreased, which helped to identify CG patients. The decrease in the content of 5,6-dihydroxyindole reminded patients that the development trend of CG was shifting from superficial to atrophic or even intestinal metaplasia. The detection of these metabolic markers of tongue coating was expected to be developed as a non-invasive and convenient technology in the future to assist us in monitoring and diagnosing the occurrence and development of CG.


Ethics approval and statement
The ethics approval for the study was granted by the Ethics Committee of Shanghai University of TCM, with the approval obtained in December 2018.Prior to sample collection, all the participants provided their informed consent by signing consent agreements, indicating their voluntary participation in the study.The experimental procedures were carried out in strict accordance with the Declaration of Helsinki and local laws.

Criteria
With reference to the first and second editions of the Kyoto Classification of Gastritis, patients were subjected to endoscopic examination and biopsies were performed from suspected lesions of gastric antrum, gastric angle, gastric body, and cardia 7 .The gastroscopic report includes six gastroscopic and pathological indicators: bile reflux, gastric mucosal erosion, inflammatory activity, atrophy (mild, moderate, or severe), intestinal metaplasia (mild, moderate, or severe), and Helicobacter pylori (Hp) infection (mild, moderate, or severe).
The inclusion and exclusion criteria followed our previous study which were shown in Supplementary file and Fig. S1 8 .

Tongue coating collection
The tongue coating was collected using our previous collection method 8 .The specific acquisition method was shown in Supplementary file.

Metabolomics testing
In the GC-TOF-MS testing, the reagents used were listed in Table S3a.The experimental steps were presented in Supplementary file.The original data were analyzed by using Chroma TOF (V 4.3x, LECO) software 9 .The mass spectrum and retention index identification of metabolites were performed by using the LECO-Fiehn Rtx5 database.The peaks detected in QC samples were less than half of the peaks were removed in QC samples with RSD > 30% 10 .
In the UHPLC-QE-MS testing, the reagents used were listed in Supplementary Table S3b.The experimental steps were presented in Supplementary file.The cutoff value for annotation was set to 0.3 11 .

Statistical analysis
The Simca-p+13.0 software (Umea, Sweden) was used for OPLS-DA.To further validate the model, a drive substitution experiment was carried out.The significance of metabolite differences between two groups was determined using various statistical criteria, including the false discovery rate (FDR) of the rank-sum test, P-values (P < 0.05), variable importance in the projection (VIP) of the first principal component of the OPLS-DA model (VIP > 1) 11 , similarity value (SV) identified through GC-TOF-MS (SV > 700) 12 , and MS2 score obtained from UHPLC-QE-MS detection (MS2 score > 0.6) 13 .Metabolite peak areas' mean values were compared between the two groups, and the log fold change (FC) value was subsequently calculated to assess the magnitude of differences.
The significantly different metabolites of the two groups were randomly combined, and the receiver operating characteristic (ROC) curve was used to evaluate model accuracy.Finally, the combination with the highest AUC value was selected as the diagnostic model.
The Spearman algorithm was used to analyze correlation between gastroscopic indicators and differential metabolites.The Corr matrix, correlation P-values were calculated for subsequent analysis and graphically plotted.
Among the metabolites that have significant correlation with gastroscopic indicators, Analysis of variance (ANOVA) was used to compare the standardized peak areas of these metabolites in patients with chronic gastritis when there were different levels of pathological indicators.

Ethics approval and consent to participate
The Ethics Committee of Shanghai University of TCM approved our study.The study obtained informed consent from all participants.

Basic information analysis results
It can be seen from Table S1 that there was no significant difference in the sex ratio between the CG group and healthy controls, but there was significant difference in the age.Therefore, the orthogonal projections to late structures-discriminant analysis (OPLS-DA) were used to observe the interference of age factors between the two groups.From Fig. S2, it can be shown that the metabolites were not significantly distinguished, which suggested that age does not affect the reliability of our results.
The horizontal axist[1]P represents the predicted component scores of the first principal component, while the vertical axis t[1]O represents the scores of the orthogonal component.The scatter plot uses different shapes and colors to represent distinct groups.According to the results from the OPLS-DA score plot, it is evident that the differentiation between sample groups is significant, and the samples generally fall within the 95% confidence interval (show in Fig. 1).
The Q2 values of the random model were consistently lower than those of the original model.The intercept of the regression line between Q2 and the vertical axis was less than zero.Simultaneously, as the permutation retention rate gradually decreased, the proportion of permuted Y variables increased, leading to a gradual decline in the Q2 of the random model.This indicates that the original model exhibits good robustness and does not show signs of overfitting (show in Fig. 1).

Different metabolic MS peaks in tongue coating samples
As depicted in Fig. 2a-c, discernible differences in MS peaks were observed between CG patients and healthy controls.
Applying the criteria of P-value < 0.05 and VIP > 1, the analysis using GC-TOF-MS identified 32 peaks in the tongue coating of CG patients (20 increased and 12 decreased).Similarly, UHPLC-QE-MS positive analysis revealed 98 peaks (62 increased and 36 decreased), while UHPLC-QE-MS negative analysis detected 14 peaks (10 increased and 4 decreased).
When employing the criteria of similarity (SV > 700) or MS2 score (> 0.6), GC-TOF-MS analysis unveiled 19 peaks (11 increased and 8 decreased) in the tongue coating of CG patients.For UHPLC-QE-MS positive analysis, 72 peaks were identified (36 increased and 36 decreased), and UHPLC-QE-MS negative analysis revealed www.nature.com/scientificreports/ 10 peaks (7 increased and 3 decreased).These findings highlight the distinct metabolic profiles present in the tongue coating of CG patients compared to healthy controls.

Diagnostic model of CG
As shown in Fig. 3, the diagnostic model was constructed by identifying distinctive metabolites of 350 CG patients and 50 healthy individuals.The optimal diagnostic model comprised 5 tongue coating metabolites.
Comparative analysis with healthy controls revealed significant upregulation of 4-Ipomeanol, Nervonic acid, and Sphinganine 1-phosphate in CG patients, whereas 1-Methyladenosine and 3-Hydroxycapric acid exhibited significant downregulation in CG group.These findings underscore the potential of these identified metabolites as valuable markers for distinguishing CG patients from healthy individuals in the diagnostic context.
In Hp infection patients, the contents of 2,4-dichloro-1-(2-chloroethenyl)-benzene, 4-methylcatechol, and Methylmaleic acid in the three stages of Hp infection were higher than those in the healthy controls.The contents Table 2.The standardized peak areas of metabolites in the tongue coating of chronic gastritis patients at different pathological stages.*There were significant differences between mild and moderate, P < 0.05.  of Methylmaleic acid and 4-methylcatechol of Hp infection patients showed an upward trend (Fig. 8a).The changing trend of 2,4-dichloro-1-(2-chloroethenyl)-benzene content of Hp infection patients was first decreased and then increased (Fig. 8b).

Discussion
We detected the tongue coating metabolites of CG patients by metabolomics, compared with the controls, the number of lipids and lipid-like molecules of CG patients was the largest (there were 47 substances in total, of which 22 up regulated and 25 down regulated), accounting for 55.29% of the total metabolites.Researchers have found that Lipid oxidative stress can lead to a variety of chronic inflammatory diseases 14 .Some researchers also have found that lipids and lipid-like molecules as the main different serum metabolites in the CG patients may promote the development of CG to gastric cancer.This study 15 focuses on the diagnostic significance of lipid metabolites in serum.Linoleamide, which is mentioned in the paper as a metabolite upregulated in both chronic gastritis (CG) and gastric cancer (GC) patients, was also observed to change in the tongue coating of CG patients.Sphinganine 1-phosphate, identified as a candidate biomarker in the paper, was also found in our study to have increased levels in CG patients and is correlated with pathological indicators.However, our study involves a richer variety of lipid compounds with greater abundance variations.We speculate that the microbial metabolism in the oral cavity and digestive tract may affect the types and abundance of lipid compounds, leading to significant differences from those found in serum.Further research is needed to confirm this hypothesis.
Among the different metabolites of lipids and lipid-like molecules there was a substance called Sphinganine 1-phosphate, which was significantly elevated compared to healthy individuals.Sphinganine 1-phosphate was one of the markers of the diagnostic model of chronic gastritis.It has also been found to have significant correlation with atrophy and intestinal metaplasia which was positive.Studies have also shown that Sphinganine 1-phosphate may inhibition of apoptosis 16 , It can also significantly inhibit liver necrosis and cell apoptosis caused by liver ischemia-reperfusion 17 .In addition, we found that the level of Sphinganine 1-phosphate decreased gradually with the severity increases of gastric mucosal atrophic and intestinal metaplasia in CG patients.We speculated that the decrease of this substance may be related to the aggravation of the disease.Among the different metabolites screened by the diagnostic model of CG, except Sphinganine 1-phosphate, there was another substance belonging to the different metabolites of lipids and lipid-like molecules, which was named Nervonic acid.Some researchers found Nervonic acid in the mucosa of gastric cancer patients 18 .Nervonic acid exists not only in the upper digestive tract but also in the lower digestive tract.Moreover, as one kind of lipid, Nervonic acid can increase fatty acid oxidation in the liver and reduce the level of circulating triglycerides leading to obesity 19 .
As shown in Fig. 5c the content of Smilanippin A of severe atrophic patients was higher than that of mild and moderate atrophic patients.The content of Lactosylceramide (d18:1/26:0) of severe atrophic patients was lower than that of mild and moderate atrophic patients.In addition, according to Fig. 7d, the content of Lactosylceramide (d18:1/26:0) gradually decreased with the aggravation of intestinal metaplasia.Lactosylceramide can promote the growth of colon cancer cells 20 , and such substance also exists in the serum of children with inflammatory bowel disease 21 .This substance will aggravate digestive system diseases.Moreover, oxidative stress www.nature.com/scientificreports/environment can promote inflammation 22 .Glycosphingolipids readily lead to the occurrence of free radical induced oxidation, this oxidation may be related to the increase of lactosylceramide content 23 .In Figs.5b and 7b, the content of Chondrillasterol 3-[glucosyl-(1->4)-glucoside] of severe atrophic and intestinal metaplasia patients was higher than that of patients with mild and moderate atrophic and intestinal metaplasia.Moreover, the content of this substance in the three stages of atrophic and intestinal metaplasia was lower than that in the tongue coating of healthy controls.The content of Azelaic acid gradually decreases with the aggravation of atrophic, and its content in the tongue coating of subjects with different degrees of atrophic was lower than healthy people.Moreover, the content of Azelaic acid in severe intestinal metaplasia subjects was lower than that in mild and moderate patients, and the content in the three stages of intestinal metaplasia was lower than that in the tongue coating of healthy controls.Azelaic acid is a marker of lipid peroxidation 24 .Researchers found that the content of Azelaic acid in the plasma of gastric cancer patients was significantly increased 25 .However, we found that the content of Azelaic acid in the tongue coating of gastric precancerous lesion patients was lower than healthy people.This phenomenon deserves our in-depth study in the future.
Organic acids and derivatives were the second largest metabolites.There were 13 metabolites (7 up regulated and 6 down regulated), among which pantothenic acid was up regulated.Some researchers have found that this substance in gastric cancer was a potential biomarker 26 .The different metabolite 3-Hydroxycapric acid screened by the diagnostic model of CG also belongs to Organic acids and derivatives.
In other metabolic species, l-histidinol expression was up regulated.Some studies have shown that histidinol can inhibit gastric acid secretion and alleviate stomach pain.In addition, it can also inhibit the growth and metastasis of lung cancer cells 27,28 .Dictagymnin expression was up regulated, and this substance has anti-inflammatory and tumor growth inhibitory effects 29 .Patients with chronic gastritis were probably undergoing self-repair by the body during their illness.From the correlation between chronic gastritis and pathological indexes, we found that the differential metabolites belonging to Organoheterocyclic compounds were 5,6-Dihydroxyindole and alpha-Carboxy-delta-decalactone.In Figs.5d and 7a, the content of 5,6-Dihydroxyindole of severe atrophic and intestinal metaplasia patients was lower than that of patients with mild and moderate atrophic.The content of alpha-Carboxy-delta-decalactone of moderate atrophic and intestinal metaplasia patients was higher than that of patients with mild and severe atrophic.
Among the different metabolites screened by the diagnostic model of chronic gastritis, 4-Ipomeanol belongs to Organic oxygen compounds, and 1-Methyladenosine belongs to Nucleosides, nucleotides, and analogues.1-Methylladenosine was also found in the results of correlation of pathological indexes.In the trend figures of chronic gastritis with intestinal transformation, the content of 1-Methylladenosine increased with the increase of intestinal degree.
In addition, as shown in Fig. 8a.The contents of methylmaleic acid and 4-methylcatechol gradually increased with the aggravation of Hp infection.The content of 2,4-dichloro-1-(2-chloroethenyl)-benzene of moderate intestinal metaplasia patients was lower than mild and severe patients.Experiments on rats showed that 4-methylcatechol and Hp infection may cause gastric cancer 30,31 .We suggest that 4-methylcatechol might be relate to the transformation from CG with Hp infection to gastric cancer.In Table 2, we also found that the content of Conduritol-beta-expoxide decreases significantly with the severity increases of atrophic and intestinal metaplasia.However, we have not found the content change of this substance in digestive system diseases.
With further investigation, these metabolites present in the tongue coatings of CG patients hold promise as potential non-invasive diagnostic markers, augmenting gastroscope-based diagnosis and monitoring of patients with chronic gastritis.Continued research in this area may offer valuable insights into improving diagnostic methods and enhancing patient care for CG.

Figure 1 .
Figure 1.OPLS-DA analysis of the chronic gastritis group and healthy controls.(a) GC-TOF-MS scatterplot of OPLS-DA model scores.(b) UHPLC-QE-MS positive ion modes scatterplot of OPLS-DA model scores.(c) UHPLC-QE-MS negative ion modes scatterplot of OPLS-DA model scores.(d) GC-TOF-MS plot of permutation test results for the OPLS-DA model.(e) UHPLC-QE-MS positive ion modes plot of permutation test results for the OPLS-DA model.(f) UHPLC-QE-MS negative ion modes plot of permutation test results for the OPLS-DA model.

Figure 2 .
Figure 2. Mass spectrum peaks of chronic gastritis group and healthy control group.(a) GC-TOF-MS, (b) UHPLC-QE-MS negative ion modes, (c) UHPLC-QE-MS positive ion modes.As shown in (a-c), there were some different mass spectrum peaks between the chronic gastritis patients and healthy control people.

Figure 3 .
Figure 3. Diagnostic model of tongue coating metabolites in patients with chronic gastritis.The accuracy of the diagnostic model is 95.4%, the specificity is 87.4%, and the sensitivity is 88.0%.

Figure 4 .
Figure 4. (a) Correlation figure between metabolic markers of tongue coating and gastroscopic indexes and pathological indexes in patients with chronic gastritis analyzed by GC-TOF-MS.Draw a thermodynamic diagram to show the correlation analysis results, which were represented by red (corr = 1), blue (corr = − 1) and white (corr = 0).Data with correlation P values less than 0.05 were marked with "*" in the figures.Data with correlation P values less than 0.01 were marked with "+" in the figures.The abscissa was the differential metabolic marker and the ordinate was the gastroscopic characteristic indicator.(b) Correlation figure between metabolic markers of tongue coating and gastroscopic indexes and pathological indexes in patients with chronic gastritis analyzed by UHPLC-QE-MS.Draw a thermodynamic diagram to show the correlation analysis results, which were represented by red (corr = 1), blue (corr = − 1) and white (corr = 0).Data with correlation P values less than 0.05 were marked with "*" in the figures.The abscissa was the differential metabolic marker, and the ordinate was the gastroscopic characteristic indicator.

Figure 5 .
Figure 5. (a) Compounds with increasing and decreasing then content in atrophic gastritis of different degrees.The changing trend of Lactosylceramide and alpha-Carboxy-delta-decalactone (d18:1/26:0) content of atrophic patients was first increased and then decreased.(b) Compounds with decreasing and then increasing content in atrophic gastritis of different degrees.The changing trend of Octadecanol, Conduritol-beta-expoxide, Chondrillasterol 3-[glucosyl-(1->4)-glucoside], and Tetracosanoic acid contents of atrophic patients was first decreased and then increased.(c) Compounds with an increasing trend in content in atrophic gastritis of different degrees.The content of Smilanippin A in the tongue coating of atrophic patients showed an upward trend.(d) Compounds with a decreasing trend in content in atrophic gastritis of different degrees.The contents of 5,6-Dihydroxyindole, Sphinganine 1-phosphate, and Azelaic acid of atrophic patients showed a downward trend.

Figure 6 .
Figure6.Diagnostic model of tongue coating metabolites in patients with with chronic superficial gastritis and chronic atrophic gastritis.The accuracy of the diagnostic model is 73.50%, the specificity is 60.00%, and the sensitivity is 82.93%.

Figure 7 .
Figure 7. (a) Compounds with increasing and decreasing then content in intestinal metaplasia (IM) of different degrees.The changing trend of Azelaic acid, alpha-Carboxy-delta-decalactone, and 5,6-Dihydroxyindole contents of intestinal metaplasia patients was first increased and then decreased.(b) Compounds with decreasing and increasing then content in intestinal metaplasia (IM) of different degrees.The changing trend of Tetracosanoic acid, Chondrillasterol 3-[glucosyl-(1->4)-glucoside], and Conduritol-beta-expoxide contents of intestinal metaplasia patients was first decreased and then increased.(c) Compounds with an increasing trend in content in intestinal metaplasia (IM) of different degrees.The content of 1-Methyladenosine of intestinal metaplasia patients showed an upward trend.(d) Compounds with a decreasing trend in content in intestinal metaplasia (IM) of different degrees.The contents of Trimethylaminoacetone, Sphinganine 1-phosphate, and Lactosylceramide (d18:1/26:0) of intestinal metaplasia patients showed a downward trend.

Figure 8 .
Figure 8.(a) Compounds with an increasing trend in content in HP of different degrees.The contents of Methylmaleic acid and4-methylcatechol of Hp infection patients showed an upward trend.(b) Compounds with decreasing and increasing then content in HP of different degrees.The changing trend of 2,4-dichloro-1-(2chloroethenyl)-benzene content of Hp infection patients was first decreased and then increased.

Table 1 .
Identification of significant different metabolites in tongue coating by comparison of chronic gastritis patients and healthy controls.GC, GC-TOF-MS; LC +, UHPLC-QE-MS positive ion; LC −, UHPLC-QE-MS negative ion.