The procedures of GDM model construction and probiotics supplement can be accessed from our previous study (Zheng et al., 2021). Briefly, after adaptive feeding, 48-week-old female Sprague–Dawley rats were randomly divided into normal pregnant group that fed with ordinary feed and GDM model group that received high-fat and high-sugar feed for 6 weeks. Then, two female rats were paired with one male rat overnight. Pregnant rats were validated by vaginal secretions and recorded as embryonic day 1. The pregnant rats were further injected with streptozocin (25 mg/kg) to induce the development of GDM. Fasting blood glucose values higher than 7.8 mmol/L at embryonic day 4 were considered indicative of GDM rats. The GDM rats were intragastrically administrated with or without probiotic supplements of 0.5 g (low-dose group) or 1 g (high-dose group) for 15 days. Each gram of probiotic supplement contained 5 × 10⁷ colony-forming units (CFU) of Lactobacillus rhamnosus LGG and 1 × 10⁸ CFU of Bifidobacterium animalis subspecies lactis Bb12 (Life-space^®, Australia). Finally, 10 rats in the normal pregnant group, eight rats in the GDM model group, eight GDM rats in the low-dose probiotics group, and nine GDM rats in the high-dose probiotics group were further studied. All animal procedures were approved by the Fujian University of Traditional Chinese Medicine (certificate number: SYXK 2019-0007; ethics approval number: FJTCM IACUC 2020020).

Abdominal aortic blood was collected on embryonic day 19. Before blood extraction, a dose of 5 ml/kg of 20% uratan solution was injected intraperitoneally for anesthesia. Abdominal aortic blood was placed at 25°C room temperature for 2 h, centrifuged at a low temperature to obtain the serum sample, and finally stored at −80°C. Concentrations of interleukin-6 (IL-6), tumor necrosis factor–α (TNF-α), nitric oxide (NO), insulin, and low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and lipid factors including total cholesterol (TC) and triglyceride (TG) were measured using enzyme-linked immunosorbent assay (ELISA) kits (Boster Biological Technology Co., Ltd., Wuhan, Hubei, China for IL-6; ProteinTech^® Group, Chicago, IL, United States, for insulin and TNF-α; and Nanjing Jiancheng Bioengineering Institute Co., Ltd., Nanjing, Jiangsu, China, for NO, LDL-C, HDL-C, TC, and TG). The value of homeostasis model assessment–estimated insulin resistance (HOMA-IR) was also calculated.

The pancreatic and colon tissues of rats were quickly collected and placed on ice. The tissues were fixed, embedded in paraffin, and stained with hematoxylin and eosin for histopathological analysis.

The procedures and analysis method for 16S rRNA sequencing have been described in a previous study (Zheng et al., 2021). Raw sequence data are available in the Sequence Read Archive under the BioProject accession number PRJNA770477. The 16S rRNA sequencing was performed using Illumina HiSeq platform (Illumina, California, United States). Raw data were cleared using FLASH, TrimMomatic, and UCHIME software. In addition, operational taxonomic units (OTUs) of 16S rRNA sequencing data were clustered using the UCLUST software based on 97% similarity. The obtained OTUs were then matched with the SILVA database for taxonomic assignment.

At embryonic day 19, fecal samples were collected from each rat and stored in a −80°C refrigerator until metabolomic analysis. For fecal sample pretreatment, 50 mg of sample was weighed into an Eppendorf tube, and 1,000 μl of extract solution (acetonitrile:methanol:water = 2:2:1, with an isotopically labeled internal standard mixture) was added. After subjecting to a 30-s vortex, the samples were homogenized at 35 Hz for 4 min and sonicated for 5 min in an ice-water bath. The homogenization and sonication cycles were repeated three times. The samples were then incubated for 1 h at −40°C and centrifuged at 12,000 rpm for 15 min at 4°C. The resulting supernatant was transferred to a fresh glass vial for further analysis. The quality control sample was prepared by mixing an equal aliquot of the supernatant from all samples.

LC-MS/MS analyses were performed using an ultrahigh-performance liquid chromatography system (Vanquish, Thermo Fisher Scientific) with a UPLC BEH amide column (2.1 × 100 mm, 1.7 μm) coupled to a Q Exactive HF-X mass spectrometer (Orbitrap MS, Thermo Fisher Scientific). The mobile phase consisted of 25 mmol/L of ammonium acetate and 25 mmol/L of ammonium hydroxide in water (pH = 9.75) and acetonitrile. The analysis was carried out with elution gradients as follows: 0–0.5 min, 95% B; 0.5–7.0 min, 95–65% B; 7.0–8.0 min, 65–40% B; 8.0–9.0 min, 40% B; 9.0–9.1 min, 40–95% B; and 9.1–12.0 min, 95% B. The column temperature was maintained at 30°C. The auto-sampler temperature was 4°C, and the injection volume was 3 μl.

The Q Exactive HF-X mass spectrometer was used for its ability to acquire MS/MS spectra in the information-dependent acquisition mode in the control of the acquisition software (Xcalibur, Thermo Fisher Scientific). In this mode, the acquisition software continuously evaluates the full-scan MS spectrum. Electrospray ionization source conditions were set as follows: sheath gas flow rate of 50 arbitrary units, aux gas flow rate of 10 arbitrary units, capillary temperature at 320^°C, a full MS resolution of 60,000, MS/MS resolution of 7,500, collision energy of 10/30/60 in the normalized collision energy mode, and spray voltage of 3.5 kV (positive) or −3.2 kV (negative). Meanwhile, raw data were converted to the mzXML format using the ProteoWizard application and processed with an in-house program, which was developed using R and based on the XCMS platform, for peak detection, extraction, alignment, and integration. Then, an in-house MS2 database (BiotreeDB) was used for metabolite annotation. The cutoff for annotation was set at 0.3.

The hierarchical clustering analysis of the gut metabolites was performed using the pheatmap function in R software.

The Orthogonal Partial Least-Squares Discrimination Analysis (OPLS-DA) was used to analyze the metabolic patterns. The OPLS-DA permutation test was used to test the effect of the model fit. The R²Y (cum) index of the OPLS-DA permutation test was used to determine the accuracy of the model fit. The Q² (cum) index refers to the predictive power of the model, and it is an evaluation parameter that is model-validated to prevent random fitting or overfitting. When the R²Y (cum) and Q² (cum) indexes are closer to 1, the model fits well. In addition, if they are greater than 0.5, then the model is acceptable. On the basis of the threshold criteria of the projection of the first principal component of OPLS-DA > 1, an absolute fold change of < 0.5 or > 2 and a P-value of < 0.05 indicated a significant difference.

Metabolic signaling pathways of gut metabolites were obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database.¹ The signaling pathways included more than four metabolites.

The correlations of metabolites and the relative abundance of gut microbiota were analyzed using the Spearman’s correlation test through the psych package in the R software. The co-occurrence network of metabolites and gut microbiota was constructed using the Cytoscape software (version 3.1.1).

Statistical analysis and plotting of figures were conducted using GraphPad Prism (version 7.0), SPSS (version 25.0), and R programming language software (version 4.4.1). A normality test was performed to determine whether the data obeyed a normal distribution. If data obeyed a normal distribution, then a t-test or post-hoc test was performed to analyze the data. Otherwise, a non-parametric test was conducted. Statistical significance was set at P < 0.05, and data were presented as mean ± standard deviation (mean ± SD).

Previously, we successfully constructed a GDM rat model and demonstrated that Lactobacillus and Bifidobacterium probiotic supplements can reduce the fasting blood glucose level of GDM rats by restoring the diversity of gut microbiota (Zheng et al., 2021). Furthermore, using the same GDM model, we tested the effects of Lactobacillus and Bifidobacterium probiotic supplements on the concentrations of IL-6, TNF-α, NO, insulin, LDL-C, HDL-C, TC, and TG and estimated the value of HOMA-IR in the serum of GDM rats. In total, 10 normal pregnant rats, eight GDM rats, eight GDM rats with low-dosage probiotic supplements, and nine GDM rats with high-dosage probiotic supplements were tested. Compared to the normal pregnant rats, the concentrations of TNF-α, NO, insulin, LDL-C, TC, and TG (P = 0.002, P = 0.003, P = 0.041, P = 0.000, P = 0.000, and P = 0.000, respectively) and the value of HOMA-IR were increased in GDM rats (P = 0.000). Moreover, high-dose Lactobacillus and Bifidobacterium probiotic supplements significantly alleviated the concentrations of TNF-α, insulin, and LDL-C (P = 0.009, P = 0.011, and P = 0.007, respectively) and the value of HOMA-IR in GDM rats (P = 0.000). However, high-dose Lactobacillus and Bifidobacterium probiotic supplements did not alleviate the concentrations of NO, TC, and TG in GDM rats (P = 0.653, P = 0.046, and P = 0.002, respectively) (Figure 1A). In contrast, the concentrations of HDL-C were decreased in GDM rats (P = 0.000) and could not be further increased by the Lactobacillus and Bifidobacterium probiotic supplements (P = 0.794 and P = 0.358, respectively) (Figure 1A). Furthermore, the concentration of IL-6 was not significantly different among normal pregnant rats, GDM rats, and GDM rats with probiotic supplements (P < 0.05) (Figure 1A).

Lactobacillus and Bifidobacterium probiotic supplements could also influence the affected morphological and structural characteristics of pancreatic tissues in GDM rats. Compared to normal pregnant rats, the pancreatic islands of GDM rats were irregular and cyto-reduced and exhibited vacuolar degeneration. In addition, the capillary vessels of the pancreatic islands had undergone congestive expansion. After probiotic supplementation, the pancreatic islands of the rats were clearer, had less vacuolar degeneration, and showed cellular increase compared to these parameters in GDM rats (Figure 1B).

The morphological and structural characteristics of the colon tissues of GDM rats were also analyzed (Figure 1B). The colon mucous membrane and cell–cell junction of the epithelial cells of GDM rats were significantly damaged. The cryptostructures of colon tissues partially disappeared, and the beaker cells were reduced. After probiotic supplementation, the colon mucous membrane of rats was more complete and displayed queuing discipline compared to that in the GDM model group, and the numbers of cryptostructures and beaker cells were further increased.

One possible mechanism of Lactobacillus and Bifidobacterium probiotic supplements restoring the normal serum parameters and pancreatic and colon histology influenced by GDM is the regulation of gut metabolites. Fecal metabolomic studies facilitate the interpretations of the metabolic interactions between the host, diet, and gut microbiota and provide functional data on the microbiome. Therefore, in this study, we further explored the influence of probiotic supplements on gut metabolites in the fecal samples of GDM rats using LC-MS/MS.

In total, 999 gut metabolites were detected. Normal rats were clearly distinguished from the GDM rats based on the levels of 999 gut metabolites by using the unsupervised clustering heatmap (Figure 2A). In addition, in the OPLS-DA score plots, normal pregnant rats and GDM rats were clearly separated from each other (Figure 2B). Similarly, the OPLS-DA permutation test indicated that the classifications of normal and GDM rats were robust and exhibited good fitness and prediction [(R²Y (cum) = (0, 0.71), Q2 (cum) = (0, −1.01)] (Figure 2C). Moreover, compared to normal rats, the levels of most gut metabolites were increased in GDM rats (Figure 2A).

However, in the unsupervised clustering heatmap, the GDM rats were indistinguishable from the GDM rats treated with low- or high-dose Lactobacillus and Bifidobacterium probiotic supplements (Figure 2A). In addition, in the OPLS-DA score plots and in the OPLS-DA permutation test, GDM rats and GDM rats with low probiotic supplements could not be distinctly separated from each other (Figures 2B,C). In contrast, GDM rats with high probiotic supplements represented a separate group in the OPLS-DA score plots and in the OPLS-DA permutation test (Figures 2B,C).

Although most of the altered metabolites in GDM rats could not be restored by Lactobacillus and Bifidobacterium probiotic supplements, partial metabolites were significantly influenced by probiotic supplements. We detected 44 metabolites that were increased in GDM rats, and their levels were alleviated by high-dose Lactobacillus and Bifidobacterium probiotic supplements (Figure 3A).

The levels of N-methyl-α-aminoisobutyric acid, L-proline, L-threonine, isovalerylalanine, 1-aminocyclopropanecarboxylic acid, alanyl-gamma-glutamate, and histidinyl-valine were shown in the box plots. The concentrations of N-methyl-α-aminoisobutyric acid, valyl-proline, L-threonine, isovalerylalanine, 1-aminocyclopropanecarboxylic acid, alanyl-gamma-glutamate, and histidinyl-valine concentrations were increased in GDM rats (P = 0.008, P = 0.002, P = 0.000, P = 0.004, P = 0.015, P = 0.000, and P = 0.008, respectively). Moreover, high-dose Lactobacillus and Bifidobacterium probiotic supplements significantly reduced the concentrations of these metabolites in GDM rats (P < 0.05) (Figure 3B), suggesting that probiotic supplements restored normal serum parameters and pancreatic and colon histology influenced by GDM through the regulation of these metabolites.

The changes in metabolites in the GDM rats were further studied. There were 499 metabolites that significantly changed in GDM rats compared to normal pregnant rats. Furthermore, the top 20 most significantly changed metabolites in GDM rats vs. normal pregnant rats were shown in Table 1. Those significantly changed metabolites in GDM rats was dihydroisoalantolactone, followed by (E)-2-Butyl-2-octenal and 4-Hydroxy-4-(3-pyridyl)-butanoic acid.

Next, the KEGG signaling pathways associated with the significantly altered metabolites in GDM rats vs. normal pregnant rats were studied. Although these metabolites were diverse, they were mainly concentrated in signaling pathways of amino acids, organic acids, and bile acids (Figure 4A). Metabolites including amino acids, organic acids, and bile acids and glucuronide metabolites were further depicted in the heatmap (Figure 4B). The concentrations of most of the metabolites were increased in GDM rats. For example, pipecolic acid, l-lysine, 5-aminopentanoic acid, L-phenylalanine, L-glutamic acid, and cholic acid from porphyrin and chlorophyll metabolism, biosynthesis of amino acids, and bile secretion signaling pathways were all significantly increased in GDM rats compared to their levels in normal pregnant rats (P = 0.000, P = 0.006, P = 0.006, P = 0.000, P = 0.003, and P = 0.015, respectively) (Figure 4C).

The changes in metabolites in GDM rats treated with Lactobacillus and Bifidobacterium probiotics were further studied. Compared to the metabolites in the GDM rats, low-dosage probiotic supplements induced the changes of 56 metabolites and high-dosage probiotic supplements induced the changes of 102 metabolites. Moreover, the top 20 most significantly changed metabolites in the GDM rats vs. GDM rats with low-dosage probiotic supplementation and GDM rats vs. GDM rats with high-dose probiotic supplementation are shown in Tables 2, 3, respectively. LysoPE and 2-aminoacrylic acid were significantly changed after low-dose probiotic supplementation, whereas [4]-gingerdiol 3,5-diacetate and propionylcarnitine were significantly changed after high-dose probiotic supplementation.

Next, the KEGG signaling pathways associated with the significantly changed metabolites in GDM rats vs. GDM rats with low-dose probiotic supplementation and GDM rats vs. GDM rats with high-dose probiotic supplementation were analyzed. Compared to GDM rats, the metabolites induced by low- or high-dose probiotic supplementation were concentrated on the metabolism pathways of porphyrin and chlorophyll (Figures 5A,B). Interestingly, the metabolism pathways of porphyrin and chlorophyll were also associated with the altered metabolites in GDM rats vs. normal pregnant rats (Figure 4A). We further showed that the levels of the metabolites L-threonine, I-urobilin, D-urobilin, and L-urobilin from the metabolism pathways of porphyrin and chlorophyll were decreased by high-dose probiotic supplementation in GDM rats (P = 0.001, P = 0.002, P = 0.002, and P = 0.005, respectively) (Figure 5C). In addition, D-urobilin and L-urobilin levels were also decreased by low-dose probiotic supplementation in GDM rats (P = 0.005 and P = 0.006, respectively) (Figure 5C), whereas L-glutamic acid level was decreased by low-dosage probiotic administration in GDM rats (P = 0.026) (Figure 5C).

Although the significantly altered metabolites in GDM rats vs. GDM rats with Lactobacillus and Bifidobacterium probiotic supplementation were not enriched with metabolism signaling pathways of amino acids and bile acids, some metabolites from these pathways were still influenced by probiotic supplements. We showed that the concentration of 2-aminoacrylic acid was increased in GDM rats following supplementation with low or high dosages of Lactobacillus and Bifidobacterium probiotics, whereas that of 1-aminocyclopropanecarboxylic acid was decreased in GDM rats following supplementation with low or high dosages of probiotics (Figures 6A,B). Moreover, the concentrations of valyllysine, L-glutamic acid, serylvaline, thyrotropin-releasing hormone, clupanodonyl carnitine, and falcarindiol were significantly increased after the administration of low-dose Lactobacillus and Bifidobacterium probiotic supplements (Figure 6A). In contrast, the concentrations of 1-aminocyclopropanecarboxylic acid, benzyl-gentiobioside, L-threonine, isovalerylalanine, aminoadipic acid, alanyl-gamma-glutamate, phenylacetylglycine, L-leucine, and norvaline were significantly reduced by high-dose probiotic supplements (Figure 6B).

Fecal metabolomics may reflect the functions of the gut microbiome. The altered metabolites in GDM rats may be perturbed by the gut microbiome during GDM pathogenesis. Previously, we studied gut microbiota in GDM rats and GDM rats with Lactobacillus and Bifidobacterium probiotic supplements (Zheng et al., 2021). Next, we explored the functional correlations between gut microbiota perturbations and metabolite changes. We observed strong correlations between the gut bacterial families and metabolites assigned to the metabolism of amino acids and bile acids in GDM rats, as determined by the Spearman’s correlation coefficient. The gut bacterial OTUs belonging to the genus Allobaculum displayed strong positive correlations (P < 0.05), whereas the genera Bryobacter and Gemmatimonas displayed strong negative correlations with the metabolism of amino acids and bile acids, such as L-glutamic acid, pipecolic acid, alanyl-gamma glutamate, isovalerylalanine, L-phenylalanine, 5-aminopentanoic acid, L-threonine, glycocholic acid, and lysine (P < 0.05) (Figure 7). Moreover, the metabolites pipecolic acid, 5-aminopentanoic acid, and cholic acid were associated with most of the gut bacterial genera in GDM rats (P < 0.05) (Supplementary Figure 1A).

To further explore the reciprocal synergistic relationships between gut microbiota and metabolites, we constructed a co-occurrence network based on the correlations of the gut bacterial genus and its metabolites in GDM rats (Figure 7). We found that metabolites 5-aminopentanoic acid, alanyl-gamma glutamate, benzyl-gentiobioside L-phenylalanine, and the gut bacterial genera Phascolarctobacterium, Barnesiella, and Blautia were at the hub of the co-occurrence network, whereas other gut bacterial genera and their metabolites were connected with those hub factors (Supplementary Figure 1A).

Lactobacillus rhamnosus LGG and Bifidobacterium lactis Bb12 were used as probiotic supplements in this study. Therefore, we tested the correlations of the Lactobacillus and Bifidobacterium genera with gut metabolites. The Spearman’s correlation coefficient of the OTU of Lactobacillus and Bifidobacterium genera and concentrations of 999 detected gut metabolites were determined. On the basis of r-values of > 0.6 and P-values of < 0.01, the Lactobacillus genus was positively correlated with 90 gut metabolites and the Bifidobacterium genus was positively correlated with 71 gut metabolites. None of the gut metabolite showed a negative correlation with the OTUs of Lactobacillus or Bifidobacterium genera. The correlations of 26 gut metabolites with Lactobacillus and Bifidobacterium genera are shown in a heatmap (Figure 8); these include the amino acid metabolites arginyl-phenylalanine, L-phenylalanine, D-proline, lysyl-valine, valyl-lysine, lysyl-leucine, L-tyrosine, N-α-acetyl-L-arginine, and tryptophyl-arginine. The co-occurrence network also revealed the connections of Lactobacillus and Bifidobacterium probiotic supplements with amino acid metabolism (Supplementary Figure 1B). In addition, other metabolites, such as histamine, oleamide, fenpropimorph, dhurrin, and tyramine, were also correlated with the OTUs of the Lactobacillus or Bifidobacterium genera (Supplementary Figure 1B).