Gut microbiota in early pregnancy among women with Hyperglycaemia vs. Normal blood glucose

In this study, 50 pregnant women in early pregnancy (< 20 weeks gestation) were recruited during their first prenatal visits at the Maternity and Child Health Center of Suzhou Municipal Hospital between 1 Nov 2015 and 31 Oct 2017. All participants were Han Chinese. Twenty-two pregnant women had different degrees of glucose metabolism disorders were signed into the HIP group. Among HIP group, 8 had type 2 diabetes mellitus before pregnancy who were only through diet/lifestyle to control glucose. Based on the diagnostic criteria recommended by the International Association of Diabetes and Pregnancy Study Groups (IADPSG) in 2010 [18], 5 overt diabetes and 9 GDM were diagnosed by hemoglobin A1c (HbA1c) or fasting glucose (FBG) and confirmed by an oral glucose tolerance test (OGTT). Twenty-eight pregnant women with fasting plasma glucose lower than 5.1 mmol/L were assigned into control group. Additionally, they had a normal glucose tolerance during 24–26 weeks’ gestation to comfirm they were healthy Patients were excluded if they met one of the following criteria: 1) history of antibiotic therapy within the last 3 months; 2) chronic diseases requiring medication; and 3) history of smoking. Written informed consent was obtained from all subjects before their participation. Ethics approval was obtained from the hospital.

The anthropometric data (pre-pregnancy weight (pre-weight), height, pre-body mass index (pre-BMI), waist circumference (waist) and hip circumference (hip)), lifestyle factors and history of medication and treatments were obtained from clinical medical records. Pre-BMI was calculated as the subject’s pre-weight (kg) divided by their height squared (m²). Waist and hip were measured in an erect position at the narrowest point between the iliac crest and the lower costal margin and at the level of the pubic symphysis, respectively. Systolic and diastolic blood pressure was measured twice (YuTu Model XJ11D, YuTu Company, Shanghai, China) with the participant in a sitting position, and the mean value was used for further analysis. The subjects were given instructions by a physician to how to be fasting prior to the collection of samples. After 8 h of fasting, venous blood samples were collected between 07:00 and 09:00 for the assessment of metabolic biomarkers (FBG, TC, TG, LDL-C, HDL-C, etc) using a fully automatic biochemical analyser (Hitachi 7000, Tokyo, Japan). HbA1c level was detected by using high-performance liquid chromatography (HLC-723G8, Tosoh Bioscience, Japan) Insulin content was measured using a Human Insulin ELISA Kit (Sigma-Aldrich, St. Louis, MO, USA).

Fresh stool samples were collected from pregnant women at home using sterile faecal collection tubes, and samples were immediately transferred to the lab within 6 h and then stored at − 80 °C under a uniform protocol before DNA extraction. Total microbial genomic DNA from 250 mg stool was extracted using the Qiagen QIAamp DNA Stool Mini Kit (Qiagen, Shanghai, China) following the manufacturer’s protocols. A NanoDrop 2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, USA) and 1% agarose gel were employed to evaluate the purification and quality of the DNA samples. The V3-V4 hypervariable regions of the bacterial 16S rRNA gene were amplified using 338F and 806R primers on a thermocycler PCR system (GeneAmp 9700, ABI, USA). PCR reactions (20 μL: 4 μL of 5 × FastPfu Buffer, 2 μL of 2.5 mM dNTPs, 1.6 μL of 5 μM primers, 0.4 μL of FastPfu Polymerase and 10 ng of template DNA) were performed according to the following cycling programme: denaturation at 95 °C for 3 min; followed by 27 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C 45 s; and 72 °C for 10 min. Each sample was analysed in triplicate. PCR products with clear electrophoresis strips and an expected size were extracted from 2% agarose gel, purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) and quantified using QuantiFluor™-ST (Promega, Madison, WI, USA).

After purification and equivalent mixing of the PCR products, paired-end (2 × 300 bp) Illumina MiSeq (Illumina, San Diego, USA) sequencing was carried out at Nanjing Decode Genomics BioTech Co., Ltd. (Nanjing, China) according to the standard protocols. We trimmed and filtered the reads (FastQ files) by using Trimmomatic [19] and merged the paired reads with FLASH [20]. Operational taxonomic units (OTUs) were clustered using USEARCH [20, 21](version 7.1 http://​drive5.​com/​uparse/​) at the 97% similarity level. UCHIME was performed to filter out the chimeric sequences [22]. Taxonomic classification of OTUs was performed using the RDP Classifier algorithm [23] (http://​rdp.​cme.​msu.​edu/​) by aligning representative sequences to the Silva (SSU123) 16S rRNA database [24].

Continuous data are expressed as the median (interquartile range). Comparisons between groups were performed using Mann-Whitney U tests or Fisher’s exact test with adjusted age. The alpha diversity (including the Chao1 index and Shannon diversity index) of bacteria within individuals was used to evaluate the number of different taxa, species richness and evenness of bacterial taxa. Microbial community composition differences (beta diversity) between groups were analysed using permutational multivariate analysis of variance (PMANOVA) of unweighted or weighted UniFrac distances (Bray-Curtis distance method, 999 permutations), and the results are illustrated by principal coordinates analysis (PCoA). Canonical correspondence analysis (CCA) was used to assess the impact of each of the factors listed on microbial abundance. Differences in microbial abundance (at the phylum, family, genus and species levels) were analysed using the Mann-Whitney U test. P-values were adjusted for false discovery rate (FDR) using the Benjamini and Hochberg method. P < 0.05 was considered statistically significant. Bacterial community diversity and composition were illustrated using Boxplot graphs and cluster and heatmap diagrams in R (ggplot2 package). Statistical analyses were performed using Graph Pad Prism 6 software (GraphPad, San Diego, USA) or SPSS 19 software (IBM, New York, USA).

The clinical characteristics of the 50 pregnant women at < 20 weeks gestation are presented in Table 1. Twenty-eight controls and 22 women with HIP were matched for maternal age. Fasting serum levels of seven biochemical variables (fasting blood glucose (GLU), triglycerides (TGs), high-density lipoprotein (HDL), low-density lipoprotein (LDL), HbA1c and C-reactive protein (CRP)) were measured in all participants. There were significant differences in GLU, TGs, HDL, LDL, HbA1c and CRP between the two groups, while TC did not show significant differences (Table 1), with subjects in the HIP group having a more disturbed metabolic profile.

A total of 912 OTUs (with 97% similarity) were clustered in this study. The PMANOVA results showed significant differences between the HIP and control groups (P = 0.001). However, we also found that the samples from 10 pregnant women (cluster 2) showed a bacterial composition distinct from those of the rest of the study cohort (cluster 1) (pMANOVA P = 0.001, Fig. 1). There were no significant differences in anthropometric or biochemical variables between cluster 1 and cluster 2 (Table S1). Based on microbial analysis, it was found that the abundance of the Bacteroidaceae family was higher in Cluster 1 (32.1 [18.2–45.8] vs. 7.6 [0.9–14.2] %, P < 0.0001), whereas the proportion of the Prevotellaceae family was more abundant in Cluster 2 (2.6 [0–7.9] vs. 38.5 [17.0–59.9] %, P < 0.0001). Consistent with previous enterotype structure reports [25], Cluster 1 and Cluster 2 were classified as Bacteroides and Prevotella enterotypes. However, no significant relationship was found between enterotype and disease status (P = 0.103, Fisher’s exact test). Furthermore, we examined the principal components with a contribution greater than 4% and found that the first and second principal components were not only significantly correlated with enterotype (P < 0.0001, Mann–Whitney U test) but were also significantly correlated with disease status (P < 0.05, Mann–Whitney U test). These results suggested that in addition to enterotype, disease status was also an important factor affecting the gut microbial composition in our study (Table S2).

We compared the gut microbial composition between 22 HIP and 28 healthy women in early pregnancy. We found that the HIP group showed significantly lower microbial richness and α-diversity than the healthy controls (P < 0.05 for both indexes, Mann–Whitney U test; Fig. 2, Fig. S1). PMANOVA analysis showed that there was a significant distance between the HIP and control groups (P < 0.001). Consistent with the PMANOVA results, CCA also showed a marked separation of women with HIP from the controls (Fig. S2).

At the phylum level, Firmicutes and Bacteroidetes were the dominant phyla, accounting for ~ 91.5% of all sequences in individuals, whereas 0.04% was accounted for by unknown bacteria (Table S3). The ratio of Firmicutes to Bacteroidetes was similar in patients and controls (1.33 vs. 1.27, P = 0.44). Other less-prominent phyla including Saccharibacteria and Fusobacteria were enriched in the HIP group, while Lentisphaerae and Tenericutes were enriched in the healthy controls (P < 0.05, Fig. 3). The Fusobacterial class and Fusobacteriales order were also enriched in the HIP group.

At the family level, women with HIP had a significantly higher proportion of Nocardiaceae, Saccharibacteria_norank, and Fusobacteriaceae, while the proportions of Christensenellaceae, Clostridiales_vadinBB60_group, Mollicutes_RF9_norank, Oxalobacteraceae, Victivallaceae, Rhodospirillaceae, and Coriobacteriaceae were higher in healthy controls (P < 0.05; Fig. 4). In addition, many genera showed significantly lower abundances in women with HIP than in controls, and these genera included Ruminococcaceae_UCG-014, Christensenellaceae_R-7_group, Ruminococcaceae_UCG-003, and Ruminococcaceae_UCG-002. Differences at the species level were also found between women with HIP and healthy controls and were consistent with the findings at the genus level (Table S3). These findings suggested that women with HIP had gut microbial dysbiosis.

To assess associations between clinical characteristics and the gut microbiota, Spearman’s correlation coefficient was calculated. At the family level, pre-BMI, pre-weight and waist were significantly negatively correlated with the Clostridiales_vadinBB60_group (Fig. 5a). pre-BMI was also negatively associated with Ruminococcaceae_UCG-005 and Ruminococcaceae_UCG-014. In addition, waist was negatively correlated with the Gastranaerophilales_norank and positively correlated with Erysipelotrichaceae.

The relationship between gut microbial composition and blood glucose metabolism was also assessed. HbA1c was positively correlated with the relative abundance of Proteobacteria, Fusobacteria, and Saccharibacteria and negatively correlated with Lentisphaerae (P < 0.05). Furthermore, HbA1c was positively associated with the abundance of the Bacteroidaceae family from the Bacteroidetes phylum (Fig. 5a). A positive correlation between HbA1c, GLU and the Enterobacteriaceae family was observed (P < 0.05) (Fig. 6a). A higher HbA1c level was correlated with a lower abundance of multiple families, such as Ruminococcaceae, Christensenellaceae, Victivallaceae, Rhodospirillaceae, and Micrococcaceae, while GLU exhibited a similar trend (Fig. 5 a, Fig. 6 b, C). At the genus level, Bacteroides, Bilophila and Lachnospiraceae_uncultured were positively correlated with HbA1c, while several genera of the Ruminococcaceae family were negatively correlated with HbA1c and GLU (Fig. 5 b).

Low-grade chronic inflammation plays a key role in the pathophysiology of glucose disorders. CRP is mainly used as a marker of inflammation and has been proven to be associated with diabetes. In our study, CRP was positively correlated with the relative abundance of Proteobacteria, Fusobacteria, and Saccharibacteria and negatively correlated with Lentisphaerae (P < 0.05). Fusobacteriaceae was the main family of the Fusobacteria phyla. Further analysis within Fusobacteria indicated that the Fusobacteriaceae family and the Fusobacterium genus were also positively associated with CRP (P < 0.05) (Fig. 5a, Fig. 6d). We also found that the abundance of the Bacteroidaceae family from the Bacteroidetes phylum was positively associated with CRP.