This nested cohort study was conducted at Peking University Shenzhen Hospital from April 2023 to November 2023. Participants were eligible if they were aged 18–45 years, had a gestational age less than 14 weeks at enrollment, and provided informed consent. Exclusion criteria comprised pre-pregnancy diabetes, hypertension, intrahepatic cholestasis of pregnancy, chronic diarrhea, acute or chronic infectious diseases, and use of antibiotics/probiotics within the last 3 months. Participants diagnosed with GDM were assigned to the GDM group, while those without GDM were placed in the non-diabetic (ND) group (Figure 1). Based on the BMI in the first trimester, participants were further categorized into normal weight and overweight/obese subgroups within GDM and ND groups. According to a previous Chinese study, 14 cases of GDM were sufficient to demonstrate significant GM differences in the second trimester between GDM and control groups (Li et al., 2023). Given the 15% GDM incidence in China (Juan and Yang, 2020), a sample size of 93 participants were calculated. Accounting for a 15% dropout rate, a total of 110 women were recruited. This study received approval from the Ethics Committee of Peking University Shenzhen Hospital (No. 2022-007).

A 5-mL venous blood sample was collected from all participants after a 10-h fast between 11–13 weeks and 6 days of gestation. This sample was used to measure fasting plasma glucose (FPG), total bile acids, and glycosylated hemoglobin levels. During 24–28 weeks of gestation, a 75 g oral glucose tolerance test (OGTT) was performed to determine the presence of GDM. Venous blood samples were taken at fasting (0-h), 1-h, and 2-h after consuming a 75 g glucose solution. FPG, 1-h blood glucose, and 2-h blood glucose levels during the OGTT were measured using the hexokinase method with a Glucose assay kit on a Beckman AU5800 analyzer (Agencourt, Beckman Coulter, Brea, CA, United States). GDM was diagnosed according to the criteria of the International Association of Diabetes and Pregnancy Study Group (IADPSG) (He et al., 2022), where OGTT 0-h blood glucose ≥5.1 mmol/L, 1-h blood glucose ≥10.0 mmol/L, or 2-h blood glucose ≥8.5 mmol/L indicated GDM. HbA1c was measured using the high-performance liquid chromatography (HPLC) method on a Glycosylated hemoglobin analyzer D-100 (Bio-Rad Laboratories, Berkeley, CA, United States) with an HbA1c Kit. Total bile acids (TBA) were examined using the Total Bile Acid Assay Kit (Leadman bio, Beijing, China) with the enzymatic cycling method on a Beckman AU5800 analyzer.

Body weight was recorded in the first trimester (11–14 weeks) and second trimester (21–24 weeks), with BMI calculated as weight (kg) divided by height (m²). BMI in the first trimester between 18.5 and 24 kg/m² was considered normal weight, while BMI ≥ 24 kg/m² indicated overweight or obesity. Data on maternal age, ethnicity, parity, occupation, and education were also collected and compared between the GDM and ND groups.

Fecal samples were collected using a disposable sampling kit (Miraclean Tech Co., Ltd., Shenzhen, China) during the first trimester (11 weeks to 13 weeks and 6 days) and the second trimester (21 week to 23 weeks and 6 days). Over 20 grams of feces were collected under aseptic conditions and immediately placed into cryotubes containing preservation solution. Samples were stored at −80°C within 24 h and transported to the laboratory (Wekemo Tech Group Co., Ltd., Shenzhen, China) for subsequent testing. During transport, the samples were kept on dry ice.

Total fecal genomic DNA was extracted from approximately 250 mg-500 mg of feces using Magnetic Stool DNA Kit (Cat# DP712, TIANGEN Biotech Co. Ltd., Beijing, China), following the manufacturer’s instructions. After concentrating and purifying the DNA, primers 515F (5’-GTGCCAGCMGCCGCGGTAA-3′) and 907R (5’-CCGTCAATTCCTTTGAGTTT-3′) were used to amplify the V4 to V5 regions of 16S rRNA genes via PCR. PCR products were purified using a Universal DNA Kit (Cat# DP214, TianGen, China). Sequencing libraries were generated using the NEBNext® Ultra DNA Library Prep Kit (Cat# E7370L, Illumina, San Diego, CA, United States), following the manufacturer’s guidelines. Library quality was assessed using the Agilent 5400 system (Agilent Technologies Co. Ltd., United States). Sequencing was performed on an Illumina platform, producing 250 bp paired-end reads.

The analysis followed the QIIME2 tutorial.¹ Raw FASTQ files were imported into a QIIME2-compatible format using the QIIME tools import program. Sequences from each sample were demultiplexed and filtered for high quality, trimmed, and de-noised. Chimeric sequences were identified and removed using the QIIME2 dada2 plugin to obtain the amplicon sequence variant (ASV) feature table. ASV sequences were aligned to the pre-trained Greengenes2 database² (McDonald et al., 2024) using the QIIME2 feature-classifier plugin to generate a taxonomy table. Contaminating mitochondrial and chloroplast sequences were filtered out. Relative abundance (RA) was used to reflect the proportion of a certain genus or phylum in the total microbiota, with bacterial flora having RA > 0.1% defined as dominant flora. Kruskal-Wallis, Linear discriminant analysis Effect Size (LEfSe), DESeq2, and Circos analysis were applied to identify bacteria with different RA among groups (Figure 1). Diversity metrics were calculated using the core-diversity plugin within QIIME2. Alpha diversity was assessed using the Chao1 and Shannon indices, while beta diversity was investigated using nonmetric multidimensional scaling (NMDS) and principal coordinate analysis (PCoA). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (Langille et al., 2013) predicted potential functional Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways associated with changes in gut microbiota composition in the GDM group (Figure 1). Samples were divided into first-trimester GDM (GDM_FT), first-trimester ND (ND_FT), second-trimester GDM (GDM_ST), and second-trimester ND (ND_ST) groups for microbiota abundance comparison.

In this study, variance of the relative abundance (VRA) and ratio of the relative abundance (RRA) were used to analyze microbiota differences between the first and second trimesters in the GDM and ND groups. VRA was defined as the difference in RA between the second and first trimesters for a given bacterial flora, with the average VRA compared between the GDM and ND groups. RRA was calculated as the ratio of the mean RA in the second trimester to the mean RA in the first trimester in the GDM group, divided by the corresponding ratio in the ND group, using the formula: RRA = (RA^GDM_ST / RA^GDM_FT) / (RA^ND_ST / RA^ND_FT). RRA reflects the change in VR of a given flora between trimesters, adjusted for background changes. An RRA > 1.5 or < 0.8 was considered statistically significant.

Statistical analysis was performed using SPSS 26.0 software (SPSS Inc., Chicago, IL, United States). Categorical variables were expressed as frequencies and compared using the chi-square test. Quantitative data following a normal distribution were presented as mean ± standard deviation and compared using the t-test. For non-normally distributed data, the median (Q1, Q3) was provided, and comparisons were made using the Mann–Whitney U test or Kruskal-Wallis test. For relative abundance comparisons, p-values or Q-values (false discovery rate adjusted) < 0.05 were considered significant. In the LEfSe analysis, microbiotas with linear discriminant analysis (LDA) scores >2 were identified as potential biomarkers. Statistical significance was defined as p < 0.05.

Of the 157 pregnant women who met the inclusion criteria, 93 successfully provided fecal samples in both the first and second trimesters and were included in the study (Figure 1). The participants’ ages ranged from 19 to 42 years, with an average age of 31.7 ± 3.4 years. The gestational age at the time of the first sample collection ranged from 11 weeks and 3 days to 13 weeks and 6 days, while for the second sample, it ranged from 21 weeks to 23 weeks and 5 days. In the second trimester, 11 women (11.83%) were diagnosed with GDM. The GDM group exhibited significantly higher BMIs and OGTT blood glucose levels compared to ND group (p < 0.05). However, no significant differences were observed between the groups in terms of maternal age, ethnicity, parity, occupation, educational status, gestational weeks at either sampling time point, fasting plasma glucose, total bile acid, or glycosylated hemoglobin levels (Table 1).

Sequencing the 16S rRNA genes from 186 fecal samples produced a total of 21,280,018 high-quality reads, averaging 114,408.7 clean reads per sample. Using a 100% similarity threshold, clean reads were clustered via sklearn species classification, resulting in 80,829 amplicon sequence variants (ASVs). These ASVs were subsequently matched to the Greengenes2 database for taxonomic annotations. At a confidence coefficient of 0.7, all the ASVs were identified, with 60,908 classified at the genus level.

At the phylum level, the dominant microbiotas (with abundance over 1%) were Bacteroidota (43.59%), Firmicutes_A (36.28%), Proteobacteria (7.19%), Firmicutes_C (5.37%), Actinobacteriota (2.97%), and Firmicutes_D (2.83%), collectively accounting for approximately 98.24% of the total microbiota. In the first trimester, Bacteroidota (47.75%) and Firmicutes_A (34.53%) were the most prevalent phyla. In the second trimester, Bacteroidota (39.43%) and Firmicutes_A (38.03%) remained dominant (Supplementary Table S1).

At the genus level, the dominant genera (with relative abundance over 1%) were Phocaeicola_A_858004, Prevotella, Faecalibacterium, Bacteroides_H, Gemmiger_A_73129, Escherichia_710834, Lachnospira, Bifidobacterium_388775, Blautia_A_141781, Acinetobacter, Megamonas, Collinsella, Agathobacter_164117, Phascolarctobacterium_A, Parabacteroides_B_862066, Dialister, Anaerostipes, Fusicatenibacter, Streptococcus, Enterocloster, Acetatifactor, Roseburia, and Klebsiella_724518, which together constituted 74% of the total abundance. In the first trimester, the most abundant genera (>5%) were Phocaeicola_A_858004, Prevotella, Faecalibacterium, and Bacteroides_H. In the second trimester, Gemmiger_A_73129 joined this group (>5%) (Supplementary Table S2).

In the first trimester, the GDM group showed slight, non-significant variations in the dominant phyla compared to the ND group (Figure 2A). While Prevotella increased in the GDM group, this increase was not statistically significant (Figure 2B). However, Bacteroides_H (p = 0.020) (Figure 2C), Acetatifactor (p = 0.049) (Figure 2D), and Megasphaera_A_38685 (p = 0.018) (Figure 2E) significantly decreased in the GDM group. Alpha diversity (Figures 2F,G) and beta diversity (Figures 2H,I) showed no significant differences between the groups. The volcano plot indicated a significant increase in CAG_417 and a decrease in Scybalousia, Megasphaera_A_38685, UBA1394, Burkholderia, Choladousia, and JC017 in the GDM group (Figure 2J). Anaeroglobus was identified as a GDM marker, while Bacteroides_H, Acetatifactor, Megasphaera_A_38685, Enterococcaceae, and Enterococcus_E were markers for the ND group (Figure 2K). The Circos plot suggested a weaker association of Acetatifactor, Scybalousia, Megasphaera_A_38685, and Burkholderia with GDM (Figure 2L).

In normal-weight women, the GDM group showed a significant increase in Escherichia_710834, Blautia_A_141780, Massilia, Thermus_A, Campylobacter_B, Oxalobacter_566322, and Slackia_A, while Megasphaera_A_38685, UBA3402, and Acetatifactor decreased (p < 0.05) (Supplementary Table S3). The volcano plot highlighted a significant increase in Massilia and a decrease in Megasphaera_A_38685 (Figure 2M). Linear discriminant analysis (LDA) identified 26 marker bacterial communities in the GDM group and 3 in the ND group (Figure 2N).

In overweight or obese GDM women, the abundance of Vescimonas, UBA1417, CAG_510, Gordonibacter, Brevundimonas, Eubacterium_O_258270, and Scatomonas significantly increased, while Haemophilus_D_735815 decreased (p < 0.05) (Supplementary Table S4). CAG_417 and UBA7173 were notably higher, and Klebsiella_724518 was lower in GDM patients, as shown by volcano plot (Figure 2O). LDA identified 8 marker bacterial communities in the GDM group and 3 in the ND group (Figure 2P). No significant difference in microbiota diversity was observed between the GDM and ND groups, regardless of BMI, in the first trimester (Supplementary Figures S1–S4).

In the second trimester, the composition of dominant phyla (Figure 3A) and genera (Figure 3B) differed between the GDM and ND groups. At the phylum level, Fusobacteriota (Figure 3C) and Firmicutes_D (Figure 3D) were significantly higher in the GDM group (p < 0.05). At the genus level, Fusobacterium_A (Figure 3E), Scatomonas (Figure 3F), and Fournierella (Figure 3G) significantly increased, while Anaerotruncus (Figure 3H), Coprobacter (Figure 3I), and Angelakisella (Figure 3J) significantly decreased in the GDM group (p < 0.05). However, alpha diversity (Figures 3K,L) and beta diversity (Figures 3M,N) did not significantly differ between the GDM and ND groups. The Volcano plot (Figure 3O) showed that in the second trimester, Butyribacter and Allisonella were significantly increased, while OM05_12, QALR01, Fenollaria, Duncaniella, Scatocola, Ezakiella, Limisoma, Sutterella, Coprobacter, Acidaminococcus, Cryptobacteroides, and Onthocola_B were significantly decreased in the GDM group. LEfSe analysis identified 13 microbiotas as biomarkers for the GDM group and 15 for the ND group (Figure 3P). The Circos plot suggested a close association between Butyribacter, Duncaniella, and Allisonella with the GDM group (Figure 3Q).

In normal-weight women, Christensenella, Coprenecus, and Intestinibacter were significantly more abundant in the GDM group during the second trimester (p < 0.05) (Supplementary Table S5). However, no significant difference was found between the GDM and ND groups in the volcano plot (Figure 3R). LDA identified 2 marker bacterial communities in the GDM group and 6 in the ND group (Figure 3S).

For overweight or obese participants, Enterocloster, Clostridium_Q_135822, and Adlercreutzia_404257 increased, whereas Dielma, Phocaeicola_A_858004, and Bacteroides_H decreased in the GDM group (Supplementary Table S6). The volcano plot indicated significant increases in Flavobacterium, Adlercreutzia_404257, CAG_269, Butyribacter, Asticcacaulis, Enterocloster, Gemmiger_A_73129, Eubacterium_R, Allisonella, Brevundimonas, and Achromobacter, while Sutterella, Cryptobacteroides, Megamonas, Bacteroides_H and Dielma significantly decreased (Figure 3T). LDA identified 11 marker communities in the GDM group and 5 in the ND group (Figure 3U). No significant difference in GM alpha or beta diversity was observed between the GDM and ND groups in the second trimester, regardless of initial BMI (Supplementary Figures S5–S8).

Matched comparative analyses were used to investigate longitudinal changes in the microbiota from the first trimester (FT) to the second trimester (ST). In the GDM group, the relative abundance of 14 genera during the second trimester was markedly higher than in the first trimester; however, only 8 of these genera showed a significant increase in the ND group. Specifically, the abundance of Streptococcus rose from 0.288 to 1.308% in the GDM group, compared to an increase from 0.521 to 1.166% in the ND group (Supplementary Table S7).

At the genus level, compared with the ND group, the variances of relative abundance (VRAs) of Eubacterium_R (Figure 4A), Allisonella (Figure 4B), Fournierella (Figure 4C), Intestinibacter (Figure 4D), and CAG_41 (Figure 4E) significantly increased in the GDM group (p < 0.05). Conversely, Angelakisella (Figure 4F) exhibited a significant decrease in VRA (p < 0.05) (Supplementary Table S8). In overweight or obese participants, the VRA of Megamonas in the GDM group was significantly lower than in the ND group (Figure 4G), whereas this difference was not observed in women of normal weight. Among normal-weight women, the VRA of Faecalibacterium was higher (Figure 4H), while the VRA of Escherichia_710834 was lower (Figure 4I), compared to the ND group.

The relative abundance ratios (RRAs) of Fusobacteriota, Verrucomicrobiota, and Desulfobacterota_I in the GDM group all exceeded 1.5 (36.22, 4.50, and 1.88, respectively) (Figure 4I), indicating a higher magnitude of increase from the first to the second trimester in the GDM group compared to the ND group. In contrast, the RRA of Bacteroidota was only 0.75, indicating a lesser change in relative abundance in the GDM group (Figure 4J; Supplementary Table S9). At the genus level, the RRAs of Staphylococcus, CAG_269, Streptococcus, Enterocloster, Mediterraneibacter_A_155507, Megamonas, Collinsella, and Faecalibacillus were all greater than 1.5 (23.72, 4.48, 2.03, 2.01, 1.86, 1.69, 1.69, and 1.53, respectively), while the RRAs of Phocaeicola_A_858004 and Bifidobacterium_388775 were both less than 0.8 (0.74 and 0.70, respectively) (Figure 4K; Supplementary Table S10).

In normal-weight women, the RRAs of Staphylococcus, Enterocloster and Collinsella were all less than 0.8. The RRAs of Streptococcus, Mediterraneibacter_A_155507 and Faecalibacillus were between 0.8 and 1.5, while only CAG_269 and Megamonas maintained RRAs greater than 1.5 (Figure 4L). However, in overweight or obese women, the RRAs of CAG_269, Streptococcus, Enterocloster, Mediterraneibacter_A_155507, and Collinsella were all larger than 1.5 (Figure 4M).

Spearman analysis revealed positive correlations between the variances of relative abundance (VRAs) of certain microbiotas and blood glucose levels during OGTT. Specifically, Fusobacterium_A (r = 0.258, p = 0.015), Gemella (r = 0.285, p = 0.007), Schaedlerella (r = 0.225, p = 0.035) and Porphyromonas_A_859426 (r = 0.251, p = 0.018) showed positive relationships with OGTT-0 h blood glucose levels. Similarly, Fusobacterium_A (r = 0.223, p = 0.037) and Fusobacterium_C (r = 0.211, p = 0.048) were positively correlated with OGTT-1 h blood glucose levels, while Erysipelatoclostridium (r = 0.224, p = 0.036), Fusobacterium_C (r = 0.216, p = 0.043), and Peptococcus (r = 0.240, p = 0.025) were positively correlated with OGTT-2 h blood glucose levels (Figure 5).

Additionally, the VRAs of Erysipelatoclostridium (r = 0.251, p = 0.020), Mediterraneibacter_A_155590 (r = 0.271, p = 0.012), Coprobacillus (r = 0.261, p = 0.016), and Turicibacter (r = 0.227, p = 0.037) positively correlated with HbA1c levels in the first trimester (p < 0.05). Conversely, the VRAs of Sellimonas (r = −0.235, p = 0.028), Oliverpabstia (r = −0.265, p = 0.013), CAG_269 (r = −0.243, p = 0.023), Alistipes_A_871404 (r = −0.225, p = 0.035), Coprenecus (r = −0.259, p = 0.015), and Vescimonas (r = −0.232, p = 0.029) were negatively related to OGTT-0 h blood glucose (Figure 5).

PICRUSt2 software analysis predicted differences in KEGG pathways between the groups. In the first trimester, one pathway [Methane metabolism (p = 0.030)] was significantly up-regulated, while five KEGG pathways [ECM-receptor interaction (p = 0.022), Furfural degradation (p = 0.002), Lipoarabinomannan (LAM) biosynthesis (p = 0.003), Lipoic acid metabolism (p = 0.038), Neomycin, kanamycin and gentamicin biosynthesis (p = 0.026)] were significantly down-regulated in the GDM group compared to the ND group (Figure 6A).

In the second trimester, however, nine pathways, including African trypanosomiasis (p = 0.018), Bacterial invasion of epithelial cells (p = 0.016), Chloroalkane and chloroalkene degradation (p = 0.034), Endocytosis (p = 0.038), Glycerophospholipid metabolism (p = 0.006), Naphthalene degradation (p = 0.008), Prion diseases (p = 0.013), Staphylococcus aureus infection (p = 0.042), and Thyroid hormone signaling pathway (p = 0.016), were up-regulated in the GDM group. Concurrently, one pathway—Steroid hormone biosynthesis (p = 0.022)—was down-regulated in the GDM group. Notably, four of the nine up-regulated pathways were related to inflammation: African trypanosomiasis, bacterial invasion of epithelial cells, prion diseases, and Staphylococcus aureus infection (Figure 6B).

PICRUSt2 did not predict any pathways with significant between-group differences among individuals with normal weight during the first trimester (Supplementary Table S11). However, among overweight or obese individuals in the first trimester, PICRUSt2 predicted 25 significantly up-regulated KEGG pathways and 33 significantly down-regulated pathways in the GDM group (Figure 6C). Among the significantly up-regulated KEGG pathways, two were inflammation-related: African trypanosomiasis (p = 0.012) and Staphylococcus aureus infection (p = 0.046) (Figure 6C).