Alterations in the Gut Microbiota in Pregnant Women with Pregestational Type 2 Diabetes Mellitus

ABSTRACT Human gut dysbiosis is associated with type 2 diabetes mellitus (T2DM); however, the gut microbiome in pregnant women with pregestational type 2 diabetes mellitus (PGDM) remains unexplored. We investigated the alterations in the gut microbiota composition in pregnant women with or without PGDM. The gut microbiota was examined using 16S rRNA sequencing data of 234 maternal fecal samples that were collected during the first (T1), second (T2), and third (T3) trimesters. The PGDM group presented a reduction in the number of gut bacteria taxonomies as the pregnancies progressed. Linear discriminant analyses revealed that Megamonas, Bacteroides, and Roseburia intestinalis were enriched in the PGDM group, whereas Bacteroides vulgatus, Faecalibacterium prausnitzii, Eubacterium rectale, Bacteroides uniformis, Eubacterium eligens, Subdoligranulum, Bacteroides fragilis, Dialister, Lachnospiraceae, Christensenellaceae R-7, Roseburia inulinivorans, Streptococcus oralis, Prevotella melaninogenica, Neisseria perflava, Bacteroides ovatus, Bacteroides caccae, Veillonella dispar, and Haemophilus parainfluenzae were overrepresented in the control group. Correlation analyses showed that the PGDM-enriched taxa were correlated with higher blood glucose levels during pregnancy, whereas the taxonomic biomarkers of normoglycemic pregnancies exhibited negative correlations with glycemic traits. The microbial networks in the PGDM group comprised weaker microbial interactions than those in the control group. Our study reveals the distinct characteristics of the gut microbiota composition based on gestational ages between normoglycemic and PGDM pregnancies. Further longitudinal research involving women with T2DM at preconception stages and investigations using shotgun metagenomic sequencing should be performed to elucidate the relationships between specific bacterial functions and PGDM metabolic statuses during pregnancy and to identify potential therapeutic targets. IMPORTANCE The incidence of pregestational type 2 diabetes mellitus (PGDM) is increasing, with high rates of serious adverse maternal and neonatal outcomes that are strongly correlated with hyperglycemia. Recent studies have shown that type 2 diabetes mellitus is associated with gut microbial dysbiosis; however, the gut microbiome composition and its associations with the metabolic features of patients with PGDM remain largely unknown. In this study, we investigated the changes in the gut microbiota composition in pregnant women with and without PGDM. We identified differential taxa that may be correlated with maternal metabolic statuses during pregnancy. Additionally, we observed that the number of taxonomic and microbial networks of gut bacteria were distinctly reduced in women with hyperglycemia as their pregnancies progressed. These results extend our understanding of the associations between the gut microbial composition, PGDM-related metabolic changes, and pregnancy outcomes.

with PGDM. Consistent with those reported in clinical practice, the times of delivery were earlier in the PGDM group than in the healthy controls, while the neonatal birth weights and numbers of admissions to neonatal intensive care units (NICUs) were similar between the two groups. The PGDM cases had higher levels of fasting blood glucose (FBG) and glycated albumin (GA) throughout the pregnancies than the control group, but the levels remained within the range that is considered acceptable in the PGDM guidelines of the American College of Obstetricians and Gynecologists. Moreover, the liver and renal functions and C-reactive protein levels did not differ between the groups. These findings indicate that the hyperglycemic pregnancies included in our study may be defined as cases with managed PGDM.
Altered gut microbiota composition in women with PGDM. A total of 13,252,913 high-quality reads were detected in 234 fecal samples collected from all of the participants from T1 to T3, with an average yield of 56,636 reads per sample. The samples were checked for quality and clustered into 9,454 operational taxonomic units (OTUs) (see Table S1 in the supplemental material). Alpha diversity analysis showed that the women with PGDM had lower richness (lower Chao1 index, P , 0.001 in all trimesters) and diversity (lower Shannon index, P , 0.001 in all trimesters; lower Simpson index, P = 0.001 in T1, P = 0.004 in T2, and P , 0.001 in T3) during pregnancy than the healthy controls ( Fig. 2A). A permutational multivariate analysis of variance (PERMANOVA) detected marked dissimilarities in the gut microbial composition between the two groups in different trimesters (P , 0.001 in all trimesters) (Fig. 2B). Principal-coordinate analysis (PCoA) plots further showed group-specific clustering in the samples from women with or without PGDM (P , 0.001 in all trimesters) (Fig. 2C).
Comparison of the compositions of the top 10 bacterial phyla in the two groups revealed that the gut microbiota mainly comprised Bacteroidetes, Firmicutes, Proteobacteria, and Actinobacteria at the phylum level ( Fig. S1A; Table S2). In the PGDM group, the abundance of Bacteroidetes increased between trimesters (Fig. S1B). The Firmicutes/Bacteroidetes ratio significantly decreased in T2 (P = 0.0037) and T3 (P = 0.0065) compared with that in the healthy controls (Fig. S1C). We further identified the top 10 OTUs in the pregnant women with or without PGDM. OTU1 (Bacteroides vulgatus) and OTU2 (Prevotella copri) were the most prevalent OTUs, followed by OTU3 (Faecalibacterium prausnitzii), OTU4 (Eubacterium rectale), OTU5 (Faecalibacterium), OTU6 (Bacteroides stercoris), OTU7 (Bacteroides uniformis), OTU8 (Bacteroides plebeius), OTU9 (Bacteroides), and OTU10 (Bifidobacterium adolescentis) (Fig. 2D). Most of these OTUs were comparable between the two groups. Conversely, the relative abundances of OTU1 (Bacteroides vulgatus), OTU3 (Faecalibacterium prausnitzii), OTU4 (Eubacterium rectale), and OTU7 (Bacteroides uniformis) were lower in the patients with hyperglycemic pregnancies than in the healthy controls ( Fig. 2E; Table S3).  Linear discriminant analysis revealed multiple taxonomic biomarkers in the two groups in different trimesters ( Fig. 2F; Table S4). In T1, OTU12 (Subdoligranulum), OTU16 (Megamonas), OTU23 (Bacteroides), OTU27 (Bacteroides), OTU53 (Blautia sp.), OTU73 (Eubacterium hallii), OTU80 (Bacteroides), OTU88 (Roseburia intestinalis), OTU91 (Faecalibacterium), OTU104 (Eubacterium siraeum), and OTU136 (Subdoligranulum) were more abundant in the PGDM group than in the healthy controls. Among these OTUs, OTU16 (Megamonas), OTU23 (Bacteroides), OTU27 (Bacteroides), OTU80 (Bacteroides), and OTU88 (Roseburia intestinalis) were consistently enriched in patients with hyperglycemia from T1 to T3. Conversely, 25 biomarkers had a high relative abundance in the healthy controls in T1.  within and between the PGDM (red) and control (blue) groups from T1 to T3. (B) Bacterial community dissimilarities within and between the PGDM (red) and control (blue) groups from T1 to T3. (C) PCoA of the Bray-Curtis dissimilarities of the two groups from T1 to T3; ellipses represent 95% confidence intervals (CIs). (D) Heat map of the top 10 OTUs in both groups in different trimesters. (E) Comparison of the relative abundances of the top 10 OTUs within and between the PGDM (red) and control (blue) groups from T1 to T3. (F) Linear discriminant analysis score of the taxonomic biomarkers for the women with PGDM and healthy controls. The colors of the bars indicate the log 2 fold changes in the relative abundances for the taxonomic biomarkers between the women with and without PGDM. The color of the y axis label indicates the phylum of each OTU. (G) Comparison of the relative abundances of the taxonomic biomarkers within and between the PGDM (red) and control (blue) groups from T1 to T3. The dotted line of the graph with each box plot represents the average. F, Friedman test. Significances based on P and Q values are separated by a slash. ns, not significant (P or Q . 0.1); *, P or Q , 0.05; **, P or Q , 0.01; ***, P or Q , 0.001.
Dynamics of the gut microbiota composition from T1 to T3. Our analysis of the intertrimester dynamics of the gut microbiota from T1 to T3 in both groups revealed no significant differences in the alpha diversity in the PGDM group in different trimesters. Conversely, the Chao1 index significantly decreased (P = 0.016, Friedman test) and the Shannon and Simpson indices increased (P = 0.005; P = 0.003, Friedman test) ( Fig. 2A) from T1 to T3 in the healthy controls. Furthermore, the dissimilarities in the gut microbiome during pregnancy showed an increase with gestational age in women with PGDM (PERMANOVA, P = 0.004). Conversely, it distinctly decreased in the healthy controls from T1 to T3 (PERMANOVA, P , 0.001) (Fig. 2B).
At the phylum level, the top 10 phyla in the women with PGDM in different trimesters were not significantly different. Verrucomicrobiota, however, was less abundant in T3 than in T1 (P = 0.003). In the control group, Acidobacteria (P = 0.001), Gemmatimonadetes (P = 0.001), and Verrucomicrobia (P = 0.008) were less abundant in T3 than in T1 ( Fig. S1B; Table S2). In the PGDM and control groups, the shift in the Firmicutes/Bacteroidetes ratio from T1 to T3 was comparable (P = 0.223 in the PGDM group and P = 0.296 in the control group) (Fig. S1C).
A longitudinal investigation into the changes in the OTUs in both groups as the pregnancies progressed revealed that 25 OTUs in the PGDM group presented different relative abundances with gestational ages and at a nominal significance threshold. None of these OTUs withstood the correction for multiple testing. In the control group, 218 stage-specific OTUs were found, and 157 OTUs remained after multiple testing corrections (Fig. S1D), 13 of which were also identified as taxonomic biomarkers between the two groups ( Fig. 2G; Table S3). In the women with normoglycemia, the relative abundances of nine PGDMenriched taxa, namely, OTU23 (Bacteroides), OTU27 (Bacteroides), OTU53 (Blautia sp.), OTU73 (Eubacterium hallii), OTU80 (Bacteroides), OTU91 (Faecalibacterium), OTU104 (Eubacterium siraeum), OTU136 (Subdoligranulum), and OTU189 (Bacteroides), increased from T1 to T3, whereas the relative abundance of one taxon in the healthy group, namely, OTU25 (Roseburia inulinivorans), decreased from T1 to T3. Among these taxa, three OTUs, namely, OTU23 (Bacteroides), OTU27 (Bacteroides), and OTU91 (Faecalibacterium), withstood the correction of the LASSO regression analysis (Fig. S2B; Tables S2 and S5). These taxonomic changes resembled those in the gut microbiota composition in the PGDM group.
Patterns in bacterial interactions in PGDM and normoglycemic pregnancies. We further constructed concurrence networks to investigate the patterns in gut microbiota interactions in the women with PGDM and healthy controls. The bacterial interactions exhibited relatively stable dynamics within each group from T1 to T3, and most were positively correlated with each other (Fig. 4). However, we found distinct differences in the microbial correlation networks between the two groups. The hyperglycemic pregnancies exhibited fewer nodes and edges and lower average degrees (the average number of connections per node) but higher modularity (indicator of how well a network decomposes into modular communities) than the control group, indicating that the overall bacterial interactions in the women with PGDM were weaker than those in the women with normoglycemia. Of note, in the 10 most connected OTUs in the PGDM group, the proportion of taxa belonging to Bacteroidetes increased from T1 to T3, whereas that belonging to Firmicutes decreased in comparison with the control group (Table S6).

DISCUSSION
To our knowledge, this study is the first to demonstrate that pregnant women with PGDM have altered gut microbial compositions compared with women who have healthy pregnancies. Differences in taxon composition and microbial networks between the women with PGDM and healthy controls were identified from T1 to T3. Furthermore, the taxonomic shifts in the gut bacteria distinctly decreased in the women with hyperglycemia as their pregnancies progressed. These findings contribute to the understanding of the association between the gut microbiota and the clinical outcomes of T2DM during pregnancy. Our data revealed profound remodeling of the gut microbiota in the participants with normoglycemia during pregnancy. In the healthy controls, the alpha and beta diversities decreased, the Firmicutes/Bacteroidetes ratio increased, and the relative abundances of specific OTUs associated with insulin resistance and T2DM, such as Bacteroides and Blautia sp., increased as the pregnancies progressed. Moreover, one OTU, namely, Roseburia inulinivorans, was enriched in the healthy individuals, and its abundance significantly decreased from T1 to T3 (11,16,21). These taxonomic changes were consistent with those reported in previous studies on the changes in the gut microbiota during pregnancy (15)(16)(17)(18)(19).
The pregnant women with hyperglycemia had a lower richness in the different trimesters and fewer changes in alpha diversity with gestational age than the women with healthy pregnancies. The hyperglycemic women also showed an opposite trend in beta diversity compared to the healthy controls. In the PGDM group, the intertrimester variability of the gut microbiota decreased markedly. The lack of broad dynamic changes in microbial diversity in the women with PGDM suggested that such changes may be independent of the effects of pregnancy. Moreover, Megamonas, a genus of Bacteroidetes, was consistently enriched in the PGDM group during pregnancy. The abundance of Megamonas has been found to be positively correlated with FBG and GA levels in patients with T2DM (22,23). It is also associated with obesity and inflammation (24)(25)(26), suggesting that it may play a role in T2DM pathophysiology. However, we detected a negative correlation between Megamonas and OGTT 2-h glucose levels, which was contradictory to the findings in previous studies. Therefore, further studies are needed to determine the role of Megamonas in the metabolic status of T2DM during pregnancy. In addition, bacteria such as Faecalibacterium prausnitzii, Roseburia inulinivorans, Subdoligranulum, Veillonella dispar, Christensenellaceae, Lachnospiraceae, and Prevotella melaninogenica were significantly depleted in the women with PGDM. This observation was consistent with that in patients with T2DM compared with healthy individuals (11,21,(27)(28)(29). Most of these bacteria can produce short-chain fatty acids, particularly butyrate, which can improve insulin sensitivity by maintaining normal intestinal functions, regulating gut permeability, Alterations in the Gut Microbiota in Women with PGDM mSystems increasing gut hormone secretion, and suppressing proinflammatory cytokine production (11,21,30,31). We also noticed a distinct difference in the gut microbial networks between the women with and without hyperglycemia. In general, most of the gut bacteria were positively correlated with each other in both groups. However, the PGDM group exhibited weaker microbial interactions in comparison with the healthy controls. Moreover, the women with PGDM had a greater abundance of Bacteroidetes, while the proportions of Firmicutes, particularly those of butyrate-producing bacteria, decreased. These opposite interaction patterns between the two groups suggest that normoglycemic individuals may have a positive feedback mechanism that can maintain normal gut function by generating a sufficient amount of butyrate, while pregnancies affected by PGDM lack these beneficial microbial networks but have more intense positive interactions involving Bacteroidetes, which may be related to the reduced production of butyrate and consequent increased risk of low-grade inflammation and gut permeability. Further studies are needed to determine whether gut microbial networks are related to PGDM metabolic status and pregnancy outcome.
Our study enhances the understanding of the correlations between specific OTUs and maternal clinical indices. As expected, the OTUs that were overrepresented in the PGDM group were more likely to be correlated with abnormal glycemic traits. In addition, some of the genera showed divergent correlations with the same indices between the two groups, suggesting that different strains within a genus may be involved. Shotgun-based sequencing studies on the gut microbiome in PGDM may be performed in the future to better elucidate these findings.
Overall, we investigated the dynamics of the gut microbiota during pregnancy in women with PGDM and healthy controls and identified the taxa with differential abundances in different trimesters, which may be correlated with maternal metabolic status during pregnancy. However, this study was limited by the sample size of the case group and the lack of fecal samples before pregnancy. Given that this was an observational study, we could not exclude the existence of certain confounding factors. For instance, 20% of the women with PGDM in our study used oral metformin combined with insulin injections as their antidiabetic therapies. Metformin markedly affects the intestinal microbiota composition (32); however, because of the small size of the PGDM group, we could not correct the potential confounding factor of metformin during the assessment of the gut microbiome of the patients with PGDM.
In summary, this study reveals the distinct characteristics of the gut microbiota composition with gestational ages between normoglycemic and PGDM-affected pregnancies, including the altered gut microbiota composition, reduced intertrimester variability, certain associations between specific taxa abundances, and changes in gut microbial networks.
Our findings indicate that interventions modulating the gut microbiota may be promising strategies for the clinical management of pregnant women with hyperglycemia. Further longitudinal studies involving women with T2DM at preconception stages and investigations using shotgun metagenomic sequencing should be performed to elucidate the relationships between specific bacterial species and PGDM metabolic statuses during pregnancy and thereby identify potential therapeutic targets for these conditions.

MATERIALS AND METHODS
Patient recruitment and sample collection. This study was approved by the institutional review board of the Peking Union Medical College Hospital, Chinese Academy of Medical Science (date of approval, 27 March 2018; reference number JS-1535). Written informed consent was obtained from all of the eligible participants in accordance with the principles of the Declaration of Helsinki. This study was organized and reported in accordance with the principles of the STORMS reporting guidelines.
Twenty patients with PGDM were consecutively enrolled in Peking Union Medical College Hospital from April 2018 to December 2020. Pregnant women $18 years old were recruited during their first trimester (7 to 14 weeks) if diabetes had been diagnosed prior to pregnancy with the standard diagnostic criteria of a hemoglobin A1C (HbA1C) value of $6.5%, fasting plasma glucose of $7.0 mmol/L, or 2-h glucose of $11.0 mmol/L in a 75-g OGTT. Exclusion criteria were as follows: antibiotic treatment, proton pump inhibitors in the last 3 months, prebiotic or probiotic intake in the last 3 months, and presence of intestinal bowel disease. Control participants with normal blood glucose levels throughout their pregnancies Alterations in the Gut Microbiota in Women with PGDM mSystems were recruited and matched to the confirmed PGDM cases at a ratio of 1:3 with respect to age (63 years), gestational age (61 week), and sample collection date (61 month). The 75-g OGTT results for all of the controls were in compliance with the standards proposed by the American College of Obstetricians and Gynecologists for gestational diabetes mellitus, which require fasting, 1-h, and 2-h blood glucose values to be lower than 5.1, 10.0, and 8.5 mmol/L, respectively (33,34). Clinical data collection. The following data were recorded for each participant: age, family history, diagnostic duration of T2DM, complications, gravidity, parity, method of conception, expected date of confinement, antibiotic treatments, height, weight at the first visit, preconception weight, weight gain during pregnancy, blood pressure, delivery method, and neonatal outcome (gender, birth weight, and NICU admission).
Laboratory data were extracted from electronic medical records, including 75-g OGTT results, fasting insulin levels, GA, HbA1C, fasting venous blood glucose, postprandial venous blood glucose, C-reactive protein, alanine aminotransferase, serum creatinine, and serum lipid profiles.
Sample collection. Stool samples from the participants with PGDM and healthy controls were collected during T1, T2, and T3. The samples were delivered to the laboratory within 3 h after collection and stored at 280°C until further processing.
DNA extraction and 16S rRNA gene amplicon sequencing. Total genomic DNA was extracted from the fecal samples (200 mg) using the QIAamp Fast DNA stool minikit (Qiagen, Hilden, Germany) per the manufacturer's instructions. The hypervariable V4 region of the bacterial 16S rRNA gene was amplified using specific 515F and 806R primers with their barcodes (35). PCR was performed using the 2Â Kapa library amplification ReadyMix (Roche, Boston, MA, USA). The PCR products were purified using the QIAquick gel extraction kit (Qiagen, Hilden, Germany) following the manufacturer's instructions and quantified using Qubit 4.0 (Invitrogen, Carlsbad, CA, USA). Ultrapure water was used as a negative control during the extraction and PCR procedures to detect potential contamination. A sequencing library was constructed using the TruSeq DNA PCR-Free sample preparation kit (Illumina, Heyward, CA, USA) per the manufacturer's instructions and sequenced on the Illumina HiSeq 2500 platform to generate 250-bp paired-end reads.
Processing of the 16S rRNA sequencing data. Raw paired-end reads were obtained using FLASH (version 1.2.11) and filtered with mothur (version 1.31.2). High-quality sequences with $97% similarity were assigned to the same OTUs using USEARCH (version 7.0.1090). Chimeras were filtered against the GOLD database (v20110518) using UCHIME (v4.2.40). The seed sequences for each OTU were annotated using taxonomic information from the SILVA database (V138; https://www.arb-silva.de) with the RDP classifier (version 2.2).
Statistical analysis. The data were analyzed using R version 4.0.2 and SPSS version 20.0 (SPSS Inc., Chicago, IL, USA). For the clinical data analysis, quantitative data were reported as means and standard deviations or medians and interquartile ranges, where appropriate. Normally distributed quantitative data were examined using Student's t tests; otherwise, the Wilcoxon rank sum test (also called the Mann-Whitney U test) was performed. Qualitative data are presented as numbers (percentages) and were evaluated using the x 2 test or Fisher's exact probability test. Data with P values of ,0.05 were considered statistically significant.
Alpha diversity indices, including Chao1, Shannon-Wiener, and Simpson's indices, were calculated based on the OTU levels (R vegan 2.5-7). The Friedman test was performed to compare the data from different trimesters, and the Wilcoxon rank sum test was conducted to compare the data from the PGDM and case groups, including OTU/phylum relative abundances and alpha diversity indices. The Bray-Curtis dissimilarities of the relative abundances of the OTUs between the groups or among the trimesters were calculated using R vegan. PERMANOVA was used to assess statistical significance based on the Bray-Curtis dissimilarity. The Firmicutes/Bacteroidetes ratios between two groups were evaluated using Welch's t test.
Significant OTU features between the PGDM and case groups were identified using a linear discriminant analysis (LDA) effect size with an LDA score above 2. Significant OTU features between the PGDM and case groups adjusted for confounding variables were identified using LASSO regression: (i) we chose features that occurred in at least 10% of the samples and had a mean abundance of at least 0.01% and then centered and scaled the OTU data; (ii) we performed 10-fold cross-validation and identified lambda.1se within 1 standard error of the minimum; (iii) we fitted the LASSO regression model and considered unbalanced variables, such as preconception BMI and PCOS, using the glmnet (v4.1-3) function of the R package. Significant features between the PGDM and case groups and significant OTU features among the different trimesters were identified using the Friedman test based on features that occurred in at least 10% of the samples and had a mean abundance of at least 0.01%. P values were corrected for multiple testing using the Benjamini-Hochberg method, and data with false discovery rates (FDRs) of ,0.05 were considered statistically significant.
Spearman's correlation coefficients were calculated to investigate associations between the OTUs and clinical indices in the different trimesters. The package pheatmap in R was used to visualize the results.
Associations between the OTUs were assessed using Spearman's correlation coefficients with Benjamini-Hochberg FDR corrections for multiple testing (significance threshold [Q] , 0.05). All of the included OTUs were detected based on a relative abundance of at least 0.001 in .50% of the samples in each group. Significant correlations with high absolute correlation coefficients .0.3 were visualized using the R package igraph (v1.2.11). The characteristics of the microbial correlation networks were also calculated using igraph (v1.2.11).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.