During human development, the microbiome adapted to dramatic variations that were reflective of the different body sites in which it was harbored. Within each site, the associated microbiome formed an enduring, stable, and distinctive microbial community. These unique biological entities respond to changes in their hosts, and the microbiome was involved in evolution (Rosenberg and Zilber-Rosenberg, 2018; Zmora et al., 2019; Suzuki and Ley, 2020). The intestinal compartment contained approximately 4 × 10¹³ bacteria that comprised the gut microbiota, which was roughly equal to the total number of human cells (Rosenberg and Zilber-Rosenberg, 2018). The number of microbial genes was approximately 100 times that of human genes (Gilbert et al., 2018; Rosenberg and Zilber-Rosenberg, 2018). The actual health state of humans can be estimated based on the structure and composition of the gut microbiota (Marchesi et al., 2016; Lavelle and Sokol, 2020). Due to the great diversity of microbial species and their contribution to host metabolism, the nervous system and other functions, the gut microbiota has become a novel biomarker for disease (DiGiulio et al., 2015; Gilbert et al., 2018).

The gut microbiota is influenced by a combination of genetic and environmental factors, and it changes dynamically during pregnancy (Koren et al., 2012; DiGiulio et al., 2015; Amir et al., 2020). The changes in the gut microbiota that occur over time are specific and vital for meeting the increasing energy and nutritional requirements of the fetus (Milani et al., 2017; Dicks et al., 2018). In addition, a series of complex metabolic alterations occur during the process of gestation, and are necessary for the successful maintenance of pregnancy (Lindsay et al., 2015; Liang et al., 2020). Although there is a consensus that the maternal gut microbiota and metabolism are crucial for the developing fetus and even neonates, our understanding of their correlations and involved mechanisms during the pregnancy remains incomplete.

Importantly, we have not yet clarified the variations in the gut microbiota, maternal metabolism and the microbial metabolome that occur during the third trimester of normal pregnancy or the potential influence on the onset of parturition. Moreover, the microbiota could function by producing metabolites that act as the intermediate compounds involved in the activities performed by both the host and microbiota (Marcobal et al., 2013). Therefore, an evaluation of the changes in the gut microbiota, the microbial metabolome and their correlations during normal pregnancy would be valuable. In this study, we collected the fecal samples and 4.0 ml peripheral venous blood samples from healthy pregnant women at 30–32 weeks of gestation and before parturition. The correlation networks among the gut microbiota, fecal metabolites and plasma metabolites were analyzed systematically. This study provides evidence for clinical management decisions and a more comprehensive understanding of the gut microbiota during pregnancy.

The actual variations in the gut microbiota that occur during normal pregnancy remain unknown. Accumulating evidence has suggested that abnormal variations in the gut microbiota are closely related to perinatal complications, such as preterm birth (DiGiulio et al., 2015), gestational diabetes mellitus (GDM; Ponzo et al., 2019; Hasain et al., 2020) and other complications (Lv et al., 2019). The transmission of microbiota from mother to child was a key determinant of the health of the child (DiGiulio et al., 2015; Younes et al., 2018). The gut microbiota plays an important role in maternal physiological activities. The interactions between the gut microbiota and host metabolism during pregnancy have not yet been elucidated. Thus, clarifying the changes that occur in the gut microbiota and identifying the key bacteria that can serve as biological markers during normal late pregnancy are valuable avenues of investigation. This study characterized the changes and correlations in the gut microbiota that occurred at various time-points. It also provided a corresponding basis for the gut microbiota as a target of clinical intervention or indicator for treatments targeting metabolism.

The alpha diversity indexes showed no difference between the two groups. As pregnancy progressed, the beta diversity of the gut microbiota increased significantly. This result was consistent with the findings of a previous study (Koren et al., 2012). There was a remarkable increase in the β-diversity of the gut microbiota with gestational age. The results also suggested that the composition and structure of the microbial community changed dynamically during pregnancy. However, some scholars have pointed out that the diversity of the gut microbiota is unrelated to gestational age during pregnancy and that the microbial structure remains stable (DiGiulio et al., 2015). The variations in β-diversity might be the result of hormonal changes or immune tolerance during pregnancy.

In the current study, the relative abundance of Proteobacteria was approximately three-fold higher than that in the full-term stage at the phylum level. At the genus level, the relative abundance of Lactobacillus, Pantoea, and Klebsiella showed a downward trend, whereas Sellimonas and Megamonas of Firmicutes increased significantly by the time of full-term pregnancy. Lactobacillus can colonize the vagina and gastrointestinal tract and provides substantial benefits from its appropriate colonization (Mu et al., 2018; Barrientos-Durán et al., 2020). Thus, it is often supplemented in the diet as an important component of probiotics (Liévin-Le Moal and Servin, 2014; Dobrut et al., 2018). Lactobacillus is a gram-positive facultative anaerobe (Mu et al., 2018). The time-dependent decrease in Lactobacillus abundance during pregnancy might lead to the increase in the levels of some fastidious anaerobes. Lactobacillus can take part in food fermentation and produce related metabolites, including lactic acid, bacteriocins, non-bacteriocins, and nonprotein molecules with direct bactericidal effects (Liévin-Le Moal and Servin, 2014). The relative abundance of rectal Lactobacillus in healthy pregnant women in Poland has been reported to decrease significantly during late pregnancy (Dobrut et al., 2018), which was basically consistent with our results. Interestingly, the abundances of Lactobacillus species obtained from gut samples that have a certain degree of homology decreased with gestational age, while the vaginal abundance of Lactobacillus was found to be increased in a metagenomic analysis (Goltsman et al., 2018). The decrease in Lactobacillus and its transformation during late pregnancy should be investigated and confirmed in more clinical experiments. Pantoea is a gram-negative and facultative anaerobe. It belongs to the Enterobacteriaceae family and Proteobacteria, which are widely distributed and can be isolated from water, soil, humans, animals and plants (Wang et al., 2017; Tambong, 2019). Pantoea has been studied as an opportunistic pathogen and is often associated with occupational exposure and community-acquired or hospital-acquired infections (Mani and Nair, 2021). In addition, it is closely related to premature rupture of membranes and preterm birth (Tiwari and Beriha, 2015). Similar to Pantoea, Klebsiella is ubiquitous in the environment. It can colonize the skin, or the gastrointestinal and respiratory tracts of healthy people (Patro and Rathinavelan, 2019; Herridge et al., 2020). However, it can infect and spread among neonates, elderly individuals and immunocompromised patients (Bengoechea and Sa Pessoa, 2019). More importantly, Klebsiella is associated with the occurrence of urinary tract infections during pregnancy (Lee et al., 2019) and gestational diabetes mellitus (Kuang et al., 2017). It is an important pathogen causing recurrent abortion (Omwandho et al., 2006), premature rupture of membranes and neonatal infection (Bonasoni et al., 2021). In our study, all the participants were healthy pregnant women in the third trimester of pregnancy. The differences in the abundances of Pantoea, Klebsiella or even Proteobacteria from 30 to 32 weeks of gestation to full term might be an adaptive change that occurs during pregnancy. With the progress of pregnancy, the relative abunance of Sellimonas and Megamonas tended to increase. Sellimonas is an anaerobic and gram-positive microbe in the family Lachnospiraceae of Firmicutes (Seo et al., 2016; Muñoz et al., 2020). The relative abundance of Sellimonas increases gradually with the improvement of intestinal barrier function, such as during the recovery of the intestinal barrier after chemotherapy for colorectal cancer or therapeutic splenectomy for liver cirrhosis. Sellimonas might be a candidate biomarker for intestinal homeostasis (Muñoz et al., 2020). Acetic acid could be produced by Sellimonas during glucose fermentation (Seo et al., 2016). A higher abundance of Megamonas was observed at full term in our study. This enriched genus might play a vital role in maternal metabolism. Its actual impact on pregnancy remains controversial. Megamonas could act as a beneficial microbe. However, the overgrowth of Megamonas could be involved in the occurrence of gestational anemia (Wei et al., 2021). Increased Megamonas abundance has been demonstrated to be associated with major depressive disorder (Wang et al., 2021). The relative abundance of Megamonas showed a positive correlation with BMI during pregnancy (Zhan et al., 2021; Abdullah et al., 2022). There was a downward trend of Megamonas in intrahepatic cholestasis of pregnancy (ICP) patients (Zhan et al., 2021), while the trend in GDM patients was exactly the opposite (Abdullah et al., 2022). The change in Megamonas might be associated with metabolite alterations that occur during late pregnancy.

Additionally, it is worth mentioning that the microbial network observed during the antepartum period became less complex and the correlations were reduced. The correlations among the microbes changed in the third trimester of pregnancy. The variation in the microbial structure should be considered holistically rather than based on specific bacterial species. This result indicated that the correlations among the microbes associated with the microbial community may be profoundly affected by metabolism in the third trimester of pregnancy. The regulatory role of the microbial network during pregnancy has not yet been explicitly elucidated in the literature from a biological perspective.

The intestinal microbial ecosystem takes part in the synthesis of vitamins and metabolism of the host for maintaining the intestinal and immune homeostasis (Kasubuchi et al., 2015). Metabolites are essential in the modification of the microbial signatures. They can serve as intermediaries for microbial functional activities and the interactions with the host (Marcobal et al., 2013; Xu et al., 2020). In this study, some metabolites with significant changes in their levels in the fecal and plasma samples obtained from pregnant women at full term were identified. The differential metabolites consisted mainly of lipids and lipid-like molecules, or organic acids and derivatives. The levels of microbiota-generated metabolite, hydroxyisocaproic acid tended to decrease, while 21-hydroxypregnenolone excreted by the host in the fecal samples tended to increase with gestational age. Hydroxyisocaproic acid is an end product of leucine metabolism (Liebich and Först, 1984; Sakko et al., 2012). Hydroxyisocaproic acid produced from the fermentation of Lactobacillus has shown potential as a topical antibiotic, and it also exerts anti-inflammatory properties (Sakko et al., 2021). Our results suggested that the reduction in the level of hydroxyisocaproic acid during full-term pregnancy was associated with a decrease in the abundance of Lactobacillus. 21-hydroxypregnenolone could originate from the fetuses and its high concentration played an essential role in deoxycorticosterone synthesis during pregnancy (Guerami et al., 1988). It was reported that 21-hydroxypregnenolone could be a biomarker of hypothalamic–pituitary–adrenal (HPA) axis abnormalities and immune dysfunction. This molecule has been found to help improve memory, prevent fatigue, and relieve stress (Shao et al., 2017). In our study, the increase in 21-hydroxypregnenolone was the obvious manifestation of pregnancy progression. In addition, an increase in the levels of long-chain fatty acids (palmitic acid and 20-hydroxyarachidonic acid) was observed in our study. Palmitic acid is a straight chain saturated fatty acid that can be absorbed from food or generated from the endogenous synthesis of some other fatty acids, carbohydrates and amino acids (Innis, 2016). It was reported that palmitic acid may have adverse impacts on conception and preimplantation embryo development (Innis, 2016; Rampersaud et al., 2020; Schindler et al., 2020). Interestingly, fetuses were found to accumulate palmitic acid during the last trimester of gestation. The accumulation of palmitic acid during late pregnancy is essential for the growth of membrane lipids, protein palmitoylation and the production of palmitoylated signaling molecules (Innis, 2016). Previous studies (Aparicio et al., 2021; Li et al., 2021) also reported that palmitic acid increased significantly in the third trimester of normal pregnancy, which was consistent with the findings of our study. Nevertheless, the abnormal accumulation of palmitic acid might give rise to inflammation at the maternal–placental interface, which resulted in pregnancy complications (such as preeclampsia, GDM and unexplained recurrent pregnancy loss; Nandakumaran et al., 2000; Rampersaud et al., 2020; Sano et al., 2020).

20-Hydroxyarachidonic acid (20-HETE) is a key factor in the development of endocrine metabolic diseases and it was one of the arachidonic acid metabolites produced by the cytochrome P450 (CYP) enzyme. The elevated concentration of 20-HETE was found to be related to the dysfunctions of endothelial and smooth muscle cells, resulting in hypertension or inflammation (Kroetz and Xu, 2005; Lu et al., 2014; Szczuko et al., 2020). The excretion of 20-HETE was found to be negatively correlated with the insulin level. A high level of insulin might result in the dysfunctional synthesis of 20-HETE. This could be caused by the inhibition of phospholipase A2 (Szczuko et al., 2020). The level of 20-HETE fluctuated drastically during late pregnancy. Although an increased level of 20-HETE in the microsomes isolated from placental vessels was observed, the circulating level of 20-HETE in preeclamptic patients showed no significant difference compared with healthy pregnant women (Plenty et al., 2018). The potential effect of 20-HETE on sodium excretion could lead to a reduction in blood pressure in the second and third trimesters of pregnancy (Kikut et al., 2020). Moreover, 20-HETE was shown to have a similar effect as oxytocin in inducing uterine muscle contractions. Our results suggested that an uneventful late pregnancy was characterized by an increase in the levels of long-chain fatty acids in fecal samples. This finding demonstrated that the changes of some crucial fatty acids might have the immunomodulation effects, resulting in the transformation to a proinflammatory immune state during the third trimester, which is important for labor.

The KEGG analysis revealed that the pentose phosphate pathway was enriched in the fecal samples. The PPP is a branch of glycolysis and plays a pivotal role in synthesizing ribonucleotides and nicotinamide adenine dinucleotide phosphate (NADPH). The PPP is closely correlated with high cell proliferation and is essential for the biosynthesis of fatty acids and sterols (Giacomini et al., 2020). A longitudinal cohort study of 232 healthy pregnant women also indicated that the levels of the intermediate product of glycolysis or gluconeogenesis increased, while intermediates involved in the tricarboxylic acid cycle decreased significantly in the second and third trimesters of pregnancy (Wang et al., 2018). In our study, the levels of D-gluconate, which is an important metabolite of the PPP, showed an upward trend during late pregnancy. The increased activation of the pentose phosphate pathway suggested that the absorption and utilization of carbohydrates and the synthesis of other substances by maternal cells increased to meet the demands of normal fetal growth and development. Combined with the findings regarding the biological correlation of the gut microbiota, these findings of our study might indicate the involvement of the underlying mechanism of insulin resistance during late pregnancy, in addition to with the reduction in the aerobic breakdown of glucose.

Furthermore, a decrease in the levels of some essential amino acids and derivatives was detected in both the fecal and plasma samples. KEGG analysis revealed that the pathways of phenylalanine metabolism were enriched during late pregnancy. This result suggested that a large amount of amino acids in the plasma of pregnant women were transported to the fetus and enriched in the placental tissue for further utilization. In addition, the placenta made great use of leucine, valine and isoleucine, some of which produced α-ketoic acid and ammonia under the action of transaminase (Manta-Vogli et al., 2020). Caprylic acid is an eight-carbon straight-chain fatty acid and is also known as octanoic acid. It can be found naturally in numerous foods and breast milk (Lemarié et al., 2018). Caprylic acid was found to have strong antibacterial activity and a positive effect due to changes in intestinal epithelium structure or in the microbiota of the gastrointestinal tract (Świątkiewicz et al., 2020). In our study, the level of caprylic acid presented a downward trend in the plasma at the full term. The decrease of caprylic acid might have been the consequence of the diffusion across membranes to placenta and metabolism by the fetus. It was reported that the placenta expressed the enzymatic machinery required for mitochondrial fatty acid (FA) oxidation. The placental FA oxidation pathway has several physiological effects in pregnancy. Octanoic acid oxidation was found to be 10-fold higher than palmitic acid oxidation in term human placentas undergoing labor (Mendez-Figueroa et al., 2013). The decreased mitochondrial β-oxidation of medium-chain fatty acids might be attributed to the development of fatty liver during late-term pregnancy (Grimbert et al., 1993). The levels of other substances in the plasma or fecal samples gradually increased with gestational age. This finding might have been the result of a complex adaptation for pregnancy. Generally, there were some specific metabolic features observed in the fecal and plasma samples obtained from pregnant women at different stages.

Few studies exist on the differential metabolites and gut microbiota features that are involved in physiological activities during pregnancy. We assumed that the correlations between the gut microbiota and metabolism were altered during late pregnancy and contributed to the onset of delivery. The alterations in fecal metabolites were caused by a combined effect of the microbial community instead of a specific microbe (Figure 4). There was no obvious correlation between the levels of plasma metabolites and differential microbes in our study. The differential microbes of Proteobacteria were closely associated with the metabolites involved in the pentose phosphate pathway and fatty acids metabolism. As the microbes associated with the gut microbiota and their metabolites are interconnected, pregnancy might be regulated at the level of the gut microbiota or gut microbiota−metabolite and metabolite−metabolite levels. It is possible that specific microbial or microbiota-metabolite relationships play a vital role in biological pathways during the late pregnancy. Additionally, fecal metabolites might further disturb the level of some specific plasma metabolites, leading to variations in host metabolism. For instance, the increased 20-hydroxyarachidonic acid level in the fecal samples was proportional to the levels of phenylalanine, norleucine and valine in the plasma samples. These essential amino-acids play a crucial role in the metabolic network and help maintain metabolic balance for the growth of the fetus. Notably, fatty acid metabolism was crucial during late pregnancy. As the gestational week progressed, the concentration of metabolites (hydroxyisocaproic acid and caprylic acid) which have anti-inflammatory effects, showed a declining trend. In contrast, the increase in 20-hydroxyarachidonic acid and palmitic acid levels might give rise to inflammation during full-term pregnancy. Our earlier study showed that the levels of IL-1β, IL-2, IL-12, and IFN-γ increased from baseline to labor and that microbes had immunomodulatory effects, which helped in the switch to a pro-inflammatory state in the third trimester (Chen et al., 2019). This current study further revealed that the specific metabolites also presented some vital changes during late pregnancy. To the best of our knowledge, the associations between metabolites and the gut microbiota in the third trimester of normal pregnancy have not been studied. This requires further investigation with larger sample sizes.

There were some limitations in our study. First, the cohorts analyzed in our study were ultimately not large enough to obtain remarkable results. Second, the traditional analysis of microbiome datasets presented many challenges and biases. The generated microbiome datasets were compositional (Gloor et al., 2017). The biases in sequencing and microbial analysis could be caused by various factors, including the collection and preservation of samples, the methods of DNA extraction (Yang et al., 2020), and the choice of PCR primers for amplification (Sipos et al., 2007; Plauzolles et al., 2022). Amplicon sequencing of the 16S rRNA gene performed by utilizing primers binding to highly conserved regions is the most common sequencing approach used to evaluate the diversity of bacterial communities in clinical samples. The amplicon sequencing of the 16S rRNA gene could provide microbial information to the phylum level but failed to identify all the bacteria to the species level (Heravi et al., 2020). The observed gut microbiota might not accurately reflect actual taxon densities. Third, the assessment of relative abundance or correlation analysis might have involved bias (Gloor et al., 2017). The relative abundances of microbes in the microbial community were assessed by 16S rRNA high-throughput sequencing. We could not determine the absolute abundance and actual contributing species in the microbial community. Multiple tools or combined methods are necessary for a comprehensive understanding of the microbiome. Further shotgun metagenomic sequencing analysis that covers the whole genomic content of microorganisms could provide greater specificity and sensitivity at the species level and other biological information. The changes in the microbial community should be studied in the context of the actual clinical phenotypes. Furthermore, metabolism during pregnancy is affected by many factors, such as hormonal changes and diet. Metabolites should be monitored on a more continuous basis to investigate the real dynamic changes in metabolism. The structural identification of metabolites was achieved by using a self-built database. The differential metabolites, such as M795T29 and M415T107_2, should be identified definitively with targeted metabolomics. Moreover, it was exceptionally challenging for us to accurately distinguish host-derived metabolites from microbially generated metabolites. Thus, we focused on the differential microbes at the genus level and their correlations with metabolism in the third trimester of pregnancy.

Despite these limitations, our study indicated that there were significant changes in the composition of the gut microbiota and metabolic phenotypes during the third trimester of pregnancy. Evaluation of the interplay between the gut microbiota and metabolism during the third trimester of pregnancy revealed the possibility of executing some early interventions for pregnancy complications and to provide better prenatal care.

Overall, the correlation networks of the gut microbiota changed dramatically as pregnancy progressed. The biomarkers of the gut microbiota during late pregnancy were far from satisfying the requirement of improving the precision and specificity of diagnoses. The metabolic network in late pregnancy is individualized and complex. The metabolism of the host and microbiota was influenced by various factors. In the third trimester of pregnancy, the excretion of free amino acids, glycolysis products and some specific fatty acids was affected by microbial activities. There was a distinctive metabolic system in late pregnancy. The identified microbiota and metabolites have promising value in further interventions for pregnancy complications and better prenatal care. Further in-depth study on the relationships of the gut microbiota and related metabolites in pregnancy complications will be necessary.

The data presented in the study are deposited in the China Nucleotide Sequence Archive (CNSA: https://db.cngb.org/cnsa) repository, accession code CNP0002979.

The studies involving human participants were reviewed and approved by the Ethical Committee of the First Affiliated Hospital of Jinan University (approval number: 2019-011). The patients/participants provided their written informed consent to participate in this study.

XX and XL conceived and designed the study. XL, HL, and XX collected the samples and clinical data. XX and LL supervised the experiments. XL and RW described the methodology, performed the analysis, interpreted the data, and prepared the manuscript. YL and XC polished the manuscript. XX, LL, and XL approved the final manuscript. All authors agreed to the submission and had full access to the final version of the manuscript.

This work was supported by grants from the National Natural Science Foundation of China (Nos. 81771664 and 81871217) and the Science and Technology Program of Guangzhou, China (No. 201904010390).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.