Connections Between The Human Gut Microbiome And Gestational Diabetes Mellitus

ObjectiveThis study used the second generation of metagenomics technology to analyze composition and functional characteristics of gut microbiome between Gestational diabetes patients and healthy pregnant women.Then,based on the combination of gut microbiome composition,pathways of gestational diabetes mellitus(GDM)and machine learning algorithm,we established a variety of machine learning models to predict GDM in order to improve the early diagnosis of GDM.Methods1.We compared the gut microbial composition of 50 GDM patients and 55 healthy pregnant women via whole-metagenome shotgun sequencing of their fecal samples which were collected from the First Affiliated Hospital of Guangdong Pharmaceutical University.we also collected their corresponding clinical data to compare the basic clinical phenotypes of gestational diabetes patients and healthly pregnant women.In addition,a metagenomic database of gut microbiome about GDM was established based on standard DNA extraction protocols and metagenomic data building procedures.2.Through the taxanomy annotation of gene sequences and comparison with public data,we can see the distribution of gut microbiota in the whole study subjects,and find the differences in the whole gut microbiota between GDM group and the healthly controls.3.Wilcoxon rank sum analysis and Random Forest model were used to compare the composition of gut microbiota between GDM patients and NGT mothers in order to find the characteristic species of gut microbiome of GDM.4.Based on the taxa of intestinal flora and the clinical phenotype we conducted sparse partial least squares discriminant analysis(sPLS-DA)to investigate the correlation between taxanomy and clinical phenotype.5.Based on the Kyoto Gene and Genomic Encyclopedia(KEGG)database and the MetaCyc database,we compared the pathways of gut microbiome between GDM women and NGT women in order to explore the functional characteristics of gut microbiome about GDM.Combined with the changes of GDM characteristic microbiome,the pathological mechanism of GDM and related gut microbiota may be explored.6.Based on the composition and pathways of gut microbiome about gestational diabetes as well as the latest machine learning algorithm,we attempted to establish a variety of linear models and nonlinear models to predict gestational diabetes.And then we used the receiver operating characteristic curve(ROC)analysis to assess the performance of the model.ResuLts1.Compared with healthy pregnant women,pregnant women with GDM had significantly higher blood glucose values in OGTT test.In addition,insuLin,HOMA,and Hb1Ac values associated with glucose metabolism were significantly increased.Triglycerides,low density lipoproteins,and free fatty acids associated with lipid metabolism were all significantly increased in GDM cohort.At the same time,the age and weight of GDM patients were significantly different from those of healthy pregnant women.2.Abundance at the phylum,order family and class levels was similar between GDM patients and healthy controls;however,At the genus level,GDM patients had a significantly higher abundance of Megamonas,while healthy controls were significantly enriched for Bacteroidales noname,Prevotella,Streptococcus,Faecalibacterium,and SubdoligranuLum.We also found a number of bacterial species that differed significantly between GDM patients and healthy controls consistent with the genus-level observations.3.Megamonas sp.and Megamonas rupellensis with 1hPBG,2hPBG,insuLin and HOMA-IR are positively correlated.They can fermenate in the gut and then produce acetic acid and propionic acid,but the mechanism of those meterials how to reguLate of blood glucose is not clear.4.In the functional annotation and analysis of gut microbiome,we can see that the microbiome of GDM was enriched in the pathway class of energy metabolism,amino acid metabolism,vitamin metabolism and energy metabolism,while the gut microbiome of healthy controls was enriched in the pathway class of genetic information(p<0.05).It is worth mentioning that the gut microbiome of GDM patients is rich in the inflammation-related metabolic pathways of lipopolysaccharide biosynthesis(LPS)and the metabolic pathways related to amino acids such as β-alanine,histidine,glutathione and phenylalanine,while the gut microbiome of healthly controls was enriched in the tryptophan metabolic pathway(p<0.05).however,there was no significant difference in P values in all KEGG metabolic pathways an pathway categories after FDR and Bonferroni correction.Combined with characteristic changes of gut microbiome in GDM patients,it is specuLated those differences may induce gestational diabetes through muLtiple pathways of inflammation(LPS),short-chain fatty acids(SCFAs)and aromatic amino acids(AAA).5.The AUC values of the gradient boosted regression trees model(GBRT)and Adabosting model based on gut microbiome composition were 0.91 and 0.88,respectively,which mean those models can effectively predict gestational diabetes,and the Wilcoxon rank sum test was used to find the important bacterial species with significant differences between two groups(Prevotella stercorea,Faecalibacterium prausinistizii,and Megamonas sp.).Conclusions1.These findings suggest significant dysbiosis of the gut microbiota among GDM patients,which may be seen as future potential clinical diagnostic markers of GDM and may induce GDM by participating in multiple molecular mechanisms.2.The GBRT model and Adabosting model based on intestinal microbes have a good predictive effect on GDM.