ABSTRACT

   For embryo implantation and fetal development, the maternal immune
   system undergoes dramatic changes. The mechanisms involved in inducing
   alterations of maternal immunity have not been fully clarified. Gut
   microbiome and metabolites were thought to influence the host immune
   response. During normal pregnancy, notable changes occur in the gut
   microbiota and metabolites. However, the relationship of these
   alterations to immune function during pregnancy remains unclear. In
   this study, we examined gut microbiota, fecal metabolites, plasma
   metabolites, and cytokines in pregnant women and non-pregnant women.
   Our findings revealed that, in comparison to non-pregnant women,
   pregnant women exhibit a significant increase in the relative abundance
   of Actinobacteriota and notable differences in metabolic pathways
   related to bile acid secretion. Furthermore, there was a marked
   reduction in pro-inflammatory cytokines levels in pregnant women.
   Correlation analyses indicated that these alterations in cytokines may
   be linked to specific gut bacteria and metabolites. Bacteria within the
   same microbial modules exhibited consistent effects on cytokines,
   suggesting that gut bacteria may function as functional groups.
   Mediation analysis further identified that certain bacteria might
   influence cytokines through metabolites, such as bile acids and
   arachidonic acid. Our findings propose potential biological connections
   between bacteria, metabolites, and immunity, which require further
   validation in future studies.

IMPORTANCE

   A great number of studies have focused on diseases induced by
   intestinal microecological disorders and immune imbalances. However,
   the understanding of how intestinal microbiota interacts with immunity
   during normal pregnancy, which is fundamental to studying pathological
   pregnancies related to intestinal microbiota disturbances, has not been
   well elucidated. Our study employed multi-omics analysis to discover
   that changes in gut microbiota and metabolites during pregnancy can
   impact immune function. In addition, we identified several metabolites
   that may mediate the effect of gut microbes on plasma cytokines. Our
   study offered new insights into our understanding of the connections
   between the gut microbiome, metabolome, and the immune system during
   pregnancy.

   KEYWORDS: pregnancy, gut microbiota, metabolism, cytokines, 16S rRNA

INTRODUCTION

   Pregnancy represents a distinct physiological state characterized by
   fluctuations in hormone levels, alterations in body structure, and
   modifications in immune status. These changes play a pivotal role in
   facilitating embryo implantation and fetal growth ([36]1). During the
   first and early second trimesters of pregnancy, the maternal immune
   system tends to exhibit a pro-inflammatory profile, which transitions
   to an anti-inflammatory state in the subsequent phase ([37]1). As labor
   approaches, there is a shift back to a pro-inflammatory state in
   maternal immunity, which serves to promote labor ([38]1). However, the
   precise mechanisms governing these dynamic immunological changes during
   pregnancy remain incompletely elucidated. Factors such as the presence
   of the fetal allograft, fluctuations in hormone levels, and the
   influence of placental cytokines have been implicated in modulating the
   immune response ([39]2, [40]3). Recent findings from an animal
   experiment have shed light on the role of the microbiome in
   pregnancy-related immune regulation and outcomes ([41]4). This study
   compared the immune adaptations and pregnancy outcomes of germ-free
   mice and conventional mice following pregnancy. Notably, despite having
   a similar total number of fetuses, germ-free mice exhibited a markedly
   lower count of living fetuses compared to their conventional mice
   ([42]4). These results highlight the potential significance of the
   microbiome in shaping both immune responses during pregnancy and
   pregnancy outcomes.

   The gut microbiome is the largest microbial community in the human body
   ([43]5, [44]6). It plays a key role in the regulation of host
   functions, such as fermentation of dietary fibers ([45]7), pathogen
   offense ([46]8), metabolic function ([47]9), and immune maturation
   ([48]10). The gut microbial community in early pregnancy was similar to
   that of non-pregnant women. In late pregnancy, the abundance of
   Firmicutes decreased, while the abundance of Proteobacteria and
   Actinobacteria increased ([49]11). Significant evidence has
   demonstrated that disturbed gut microbiota promotes preeclampsia by
   affecting autophagy and M2 polarization of macrophages ([50]12). This
   study underscores the significance of the gut microbiome in modulating
   the immune response during pregnancy. However, the precise mechanisms
   underlying the interaction between the altered gut microbiota in normal
   pregnant women and host immunity remain elusive, necessitating further
   research. A deeper understanding of the associations between the gut
   microbiome and pregnancy immunity may provide new ideas for the
   treatment of pathological pregnancy.

   The microbiome can account for only approximately 10% of the
   variability in cytokine levels ([51]13). The mutual immune regulation
   between the microbiome and the host is primarily mediated through
   metabolites, rather than direct interactions between specific microbes
   and immune cells ([52]13). A recent study reported that numerous plasma
   metabolites undergo pronounced shifts during pregnancy ([53]14). The
   question of whether and how these metabolites, which undergo
   significant changes during pregnancy, interact with immune cells
   remains a subject that requires further investigation.

   In this study, we analyzed the gut microbiome, fecal metabolites,
   plasma metabolites, and cytokines of pregnant women, and compared them
   with those of healthy non-pregnant women. We used the integrated method
   to reveal the associations among maternal microbiome, metabolites, and
   cytokines. This study expands our current knowledge regarding how
   altered gut microbiota and metabolites interact with the host during
   pregnancy.

RESULTS

Clinical characteristics of participants

   This study recruited 30 healthy pregnant women and 15 healthy
   non-pregnant women as controls. The average age of the pregnant
   participants was 26.67 ± 2.15 years, compared to 26.40 ± 1.84 years for
   the control group. Furthermore, the mean pre-pregnancy body mass index
   (BMI) for the pregnancy group stood at 19.90 ± 1.46 kg/m^2, while it
   was 19.57 ± 0.67 kg/m^2 for the controls. There were no statistically
   significant differences in age (P = 0.684) and BMI (P = 0.312) between
   the two groups (Table S1).

Variation of plasma cytokines in pregnant women

   The immune profiles of pregnant and non-pregnant women exhibited
   notable differences. Pregnant women displayed reduced levels of
   pro-inflammatory cytokines, including interleukins IL-1β, IL-2, IL-6,
   IL-12, interferon gamma (IFN-γ), and tumor necrosis factor alpha
   (TNF-α), while exhibiting elevated levels of the anti-inflammatory
   cytokine IL-4 ([54]Fig. 1). Conversely, IL-10, another
   anti-inflammatory cytokine, increased in the control group ([55]Fig.
   1). This finding suggested that the immune state during pregnancy is
   characterized by a relative immunosuppression compared to the
   immunological status observed in the non-pregnant condition.

Fig 1.

   [56]Fig 1
   [57]Open in a new tab

   Comparison of plasma cytokines between the pregnancy group and control
   group. Differences between groups were calculated using the Wilcoxon
   rank-sum test. ***P < 0.001.

The relationship between gut microbiota and plasma cytokines

   Previous studies have reported distinct gut microbiota between pregnant
   women and non-pregnant women ([58]11, [59]15, [60]16). Therefore, we
   compared the gut microbiota between the pregnancy group and the control
   group. We observed a total of 2,660 unique operational taxonomic units
   (OTUs) in the pregnancy group. In contrast, the control group exhibited
   a significantly lower count with only 630 OTUs identified ([61]Fig.
   2A). Furthermore, Firmicutes dominated as the most prevalent phylum in
   the gut microbiota of both the pregnancy group and the control group
   (65.28% vs 67.82%, P = 0.80) ([62]Fig. 2B). The relative abundance of
   Bacteroidota was lower in the pregnancy group compared to the control
   group (18.61% vs 24.09%, P = 0.23) ([63]Fig. 2B). Conversely, the
   pregnancy group exhibited a higher relative abundance of
   Actinobacteriota and Proteobacteria compared to the control group
   (Actinobacteriota: 9.15% vs 2.98%, P = 0.002; Proteobacteria: 4.04% vs
   3.48%, P = 0.76) ([64]Fig. 2B), which is broadly consistent with the
   findings of Koren et al. ([65]11).

Fig 2.

   [66]Fig 2
   [67]Open in a new tab

   Characteristics of gut microbial community in pregnancy group and
   control group. (A) Venn diagram indicates the number of unique and
   shared OTUs between the pregnancy and control groups. (B) Comparison of
   the relative abundance of the top four phyla in the gut microbiota of
   the pregnancy and control groups. (C) A heatmap illustrating that the
   relative abundance of microbial biomarkers varied in pregnant women and
   controls. (D–G) Comparison of the relative abundance of modules 1–4
   between the pregnancy and control groups. (H) Heatmap of Spearman
   correlation analysis between gut microbiota and plasma cytokines. The
   colors on the heatmap represent the magnitude of the correlation
   coefficients, with orange/red indicating a positive correlation and
   green/blue indicating a negative correlation. *P < 0.05, **P < 0.01,
   and ***P < 0.001.

   Given OTU was the basic unit of taxonomy, we established a random
   forest model based on OTU level to identify the microbial biomarkers
   that can distinguish between these two peoples (Fig. S1; Table S2).
   Microbes may survive and adapt as guilds to respond to environmental
   perturbations. In microbial ecology research, it is common to
   preliminarily identify functional microbial communities by partitioning
   microbial modules or co-abundance groups ([68]17[69]–[70]19).
   Therefore, we selected the top 100 OTUs with the highest contribution
   to distinguish the two groups in random forest model and clustered them
   into four modules by using WGCNA methods ([71]Fig. 2C; Table S3). In
   the pregnant group, the abundance of modules 1, 2, and 4 was
   significantly higher than that of the control group, while module 3
   exhibited a significantly lower abundance compared to the control group
   ([72]Fig. 2D through G). Furthermore, we conducted Spearman correlation
   analysis to investigate their relationships with cytokines. As
   illustrated in [73]Fig. 2H, most of the OTUs enriched in modules 1, 2,
   and 4 in the pregnant group exhibited a negative correlation with
   pro-inflammatory cytokines such as IL-2 and TNF-α, while most of the
   OTUs in module 3, depleted in the pregnant women, showed a positive
   correlation with pro-inflammatory cytokines. This suggests that
   microbes within the same module may have a broadly consistent impact on
   the immune system. Specifically, OTU_4 and OTU_6, belonging to
   Bifidobacterium, displayed significantly higher abundance in the gut of
   pregnant women compared to the control group (Fig. S2), and they
   exhibited a negative correlation with TNF-α, IFN-γ, and IL-1β.
   Similarly, OTU_16 and OTU_82, belonging to the Ruminococcus genus,
   showed significantly higher abundance in pregnant women (Fig. S2).
   OTU_16 was negatively correlated with TNF-α, and OTU_82 exhibited
   negative correlations with cytokines such as TNF-α, IFN-γ, and IL-1β.
   Bifidobacterium and Ruminococcus are recognized as anti-inflammatory
   bacteria known for producing short-chain fatty acids ([74]20, [75]21).
   This implied that the increased abundances of certain anti-inflammatory
   bacteria during pregnancy may be linked to a reduction in
   pro-inflammatory cytokines levels.

Alterations in the fecal and plasma metabolites in pregnant women and their
relationships with cytokines

   Metabolomics could reflect biochemical process alterations of host
   influenced by genetics, environment, and microbiome ([76]22).
   Therefore, we employed untargeted liquid chromatography-mass
   spectrometry (LC-MS)-based metabolomics to analyze fecal and plasma
   samples in order to investigate metabolic changes during pregnancy. In
   the fecal samples, a total of 401 metabolites were identified in
   positive ion mode (ES+), while 280 metabolites were identified in
   negative ion mode (ES−). In the plasma samples, 140 metabolites at ES+
   and 198 metabolites at ES− were identified. The overall metabolic
   variations between pregnant women and non-pregnant women were further
   revealed through orthogonal partial least-squares discriminant analysis
   (OPLS-DA). A distinct separation of metabolome between the two groups
   was observed both in ES+ and ES− ([77]Fig. 3A and B; Fig. S3),
   indicating substantial metabolic alterations in feces and plasma during
   pregnancy. The permutation test was used to validate the robustness of
   OPLS-DA model. With the gradual decrease of replacement retention, both
   R^2 and Q^2 of the random model decreased gradually, suggesting no
   overfitting issue arise, and the model was robust (Fig. S4A through D).
   To identify significantly differential metabolites, we calculated
   variable importance in the projection (VIP) scores from the OPLS-DA
   models. Metabolites, whose VIP score exceeded 1 and P value was <0.05,
   were regarded as differential metabolites between groups. In summary, a
   total of 44 differential fecal metabolites and 53 differential plasma
   metabolites were detected in pregnant women compared to non-pregnant
   women ([78]Fig. 3C and D; [79]Fig. 4A and B). The majority of
   differential fecal metabolites are lipids and lipid-like molecules,
   such as arachidonic acid and chenodeoxycholate, both of which were
   downregulated in the pregnant group ([80]Fig. 3C and D). In plasma
   metabolites, a similar downregulation of arachidonic acid was observed
   ([81]Fig. 4B). Additionally, substances related to bile acid metabolism
   were decreased in pregnant females, including glycochenodeoxycholate,
   glycodeoxycholic acid, and deoxycholic acid ([82]Fig. 4A and B). The
   Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment pathway
   further revealed that bile acid secretion is the most significantly
   altered pathway, involving the highest number of metabolites ([83]Fig.
   4C). Arachidonic acid and bile acids have both been reported to be
   associated with inflammation ([84]23[85]–[86]25). They may be potential
   factors influencing immune responses during pregnancy.

Fig 3.

   [87]Fig 3
   [88]Open in a new tab

   Fecal metabolite profiles of the pregnancy and control group. (A and B)
   OPLS-DA score plot of fecal samples from pregnancy group and control
   group in ES+ and ES−, respectively. Samples from the same group are
   represented by points of the same color. The distribution of points
   reflects differences between and within groups. (C and D) Differential
   fecal metabolites in ES+ and ES− between pregnant and non-pregnant
   women. The x-axis represents the values of fold change in differential
   expression between pregnancy group samples and control group samples.
   Red color represents fold change >1, indicating upregulated
   differential metabolites in the pregnancy group, while green color
   represents fold change <1, indicating downregulated differential
   metabolites in the pregnancy group.

Fig 4.

   [89]Fig 4
   [90]Open in a new tab

   Plasma metabolites profiles of the pregnancy and control group. (A and
   B) Differential plasma metabolites in ES+ and ES− between pregnant and
   non-pregnant women. The x-axis represents the values of fold change in
   differential expression between pregnancy group samples and control
   group samples. Red color represents fold change >1, indicating
   upregulated differential metabolites in the pregnancy group, while
   green color represents fold change <1, indicating downregulated
   differential metabolites in the pregnancy group. (C) KEGG pathway
   enrichment analysis of differentially expressed plasma metabolites. The
   y-axis represents various KEGG metabolic pathways, while the x-axis
   represents the number of differentially expressed metabolites within
   each KEGG pathway. Colors denote the P values from the enrichment
   analysis, with darker colors indicating smaller P values and greater
   enrichment significance. (D) Venn diagram of fecal metabolites and
   plasma metabolites in ES+. (E) Venn diagram of fecal metabolites and
   plasma metabolites in ES−. (F) Heatmap of correlations between
   metabolites and cytokines. The colors on the heatmap represent the
   magnitude of the correlation coefficients, with red indicating a
   positive correlation and blue indicating a negative correlation. *P <
   0.05, **P < 0.01, and ***P < 0.001.

   In the process of digestion, food undergoes a complex series of
   biochemical reactions and produces various metabolites. The gut
   microbiota plays a role in metabolizing specific nutrients and
   regulating the levels of intestinal metabolites. Metabolites
   originating from dietary sources and those produced by the gut
   microbiota can enter the blood circulation through the enterohepatic
   circulation, exerting biological effects. As a result, we intersected
   fecal metabolites with plasma metabolites, identifying 81 common
   metabolites in ES+ and 121 common metabolites in ES− ([91]Fig. 4D and
   E). Among the 202 common metabolites, 36 exhibited intergroup
   differences in plasma, forming the focus of our further investigations
   (Fig. S5). We further evaluated the correlations between these shared
   metabolites and cytokines. Our findings revealed that certain enriched
   metabolites in pregnant women were negatively correlated with
   pro-inflammatory cytokines, while depleted metabolites in pregnant
   women were positively correlated with pro-inflammatory cytokines
   ([92]Fig. 4F). For instance, we observed a notable negative correlation
   between taurocholate, which is enriched in the pregnant group, and
   pro-inflammatory cytokines such as IFN-γ, IL-1β, and TNF-α. Conversely,
   metabolites like glycochenodeoxycholate, deoxycholic acid, and
   arachidonic acid, which are enriched in the control group, exhibited a
   positive correlation with several pro-inflammatory cytokines.

Association of gut microbiota, metabolites, and cytokines

   The gut microbiota is considered to play a key role in the host
   metabolism, capable of both regulating the host’s metabolites and
   producing its own metabolites which have a significant impact on the
   host ([93]26). In this study, cytokines were associated with both gut
   microbiota and their metabolites. Moreover, the correlations between
   cytokines and metabolites were more numerous and stronger than those
   between cytokines and microbiota. Several studies have reported on the
   mediating role of metabolites in the interaction between microbiota and
   immune functions or disease phenotypes ([94]19, [95]27, [96]28).
   Therefore, we hypothesize that metabolites may mediate the association
   between gut microbiota and cytokines during pregnancy. We performed
   directional mediation analysis to further explore which metabolites may
   mediate the impact of gut microbiome on immunity. Finally, a total of
   46 linkages among microbes, metabolites, and cytokines were identified
   ([97]Fig. 5A; Table S4). We found that some microbes may alleviate the
   systemic inflammatory response by inhibiting pro-inflammatory
   metabolites. For instance, OTU_82, enriched in pregnant women and
   belongs to Ruminococcus callidus, decreased the TNF-α and IL-2 levels
   by reducing deoxycholic acid levels ([98]Fig. 5B and C; Fig. S6A and
   B). Moreover, OTU_943, a bacteria enriched in the pregnancy group,
   contributes to lower TNF-α and IL-12 levels by decreasing arachidonic
   acid ([99]Fig. 5D and E; Fig. S6C and D). Conversely, several
   metabolites mediate the impact of gut microbiota on pro-inflammatory
   cytokines. OTU_2033 depleted in pregnant women, which may increase the
   IFN-γ and IL-1β by decreasing the bisindolylmaleimide I level
   ([100]Fig. 5F and G). Taken together, our results suggest that gut
   microbes may alter plasma cytokines by interacting with host
   metabolites.

Fig 5.

   [101]Fig 5
   [102]Open in a new tab

   Mediation linkages between gut microbiota, metabolites, and cytokines.
   (A) The Sankey diagram displays mediation linkages among gut
   microbiota, metabolites, and cytokines. Columns from left to right show
   the OTUs, metabolites, and cytokines, respectively. The curved lines
   across the columns indicate the mediation effects. Different colored
   curves represent the subclasses to which the metabolites belong. (B)
   Analysis of the effect of OTU_82 on TNF-α as mediated by deoxycholic
   acid [indirect effect: −0.113, 95% boot confidence interval (CI):
   −0.298, −0.038]. The proportion of the mediation effect is labeled at
   the center of the ternary diagrams. (C) Analysis of the effect of
   OTU_82 on IL-2 as mediated by deoxycholic acid (indirect effect:
   −0.120, 95% boot CI: −0.324, −0.025). (D) Analysis of the effect of
   OTU_943 on TNF-α as mediated by arachidonic acid (indirect effect:
   −0.085, 95% boot CI: −0.220, −0.008). (E) Analysis of the effect of
   OTU_943 on IL-12 as mediated by arachidonic acid (indirect effect:
   −0.094, 95% boot CI: −0.241, −0.008). (F) Analysis of the effect of
   OTU_2033 on IFN-γ as mediated by bisindolylmaleimide I. (G) Analysis of
   the effect of OTU_2033 on IL-1β as mediated by bisindolylmaleimide I.

DISCUSSION

   To investigate the interaction between host and microbiota during
   pregnancy and offer preliminary insights into this aspect, we explored
   the associations among gut microbiota, metabolites, and plasma
   cytokines of pregnant women and non-pregnant women. We observed
   alterations in the gut microbiome, fecal metabolite, and plasma
   metabolite characteristics in pregnant women. Further analysis
   indicated a correlation between these changes and variations in
   cytokine levels. Mediation analysis revealed the potential role of
   metabolites in mediating the connection between gut microbiota and
   cytokines. To the best of our knowledge, the associations among altered
   gut microbiota, metabolites, and plasma cytokines during normal
   pregnancy were first explored in this study.

   To ensure the normal progression of pregnancy, the maternal immune
   system shifts toward a Th2-type response in the mid to late stages of
   gestation, exhibiting more anti-inflammatory characteristics. Th1 cells
   secrete pro-inflammatory cytokines, such as IL-2, IFN-γ, and TNF-α
   ([103]29), and are involved in specific autoimmune responses and
   cellular immunity. Th2 cells, which produce IL-4 and IL-10, facilitate
   the proliferation of B cells, antibody production, and tolerance to
   allogeneic rejection ([104]30). In our study, pro-inflammatory
   cytokines such as IL-1β, IL-6, IFN-γ, and TNF-α exhibited a
   downregulation in the plasma of pregnant women, while the
   anti-inflammatory cytokine IL-4 showed an upregulation, largely
   consistent with the result of a previous study ([105]31). However, the
   IL-10 levels in pregnant women were found to be lower than those in
   non-pregnant women, contradicting the findings of prior research
   ([106]31, [107]32). This discrepancy may be due to the inconsistency
   between the characteristics of cytokines in whole blood and those
   isolated in vitro. Furthermore, we did not observe a marked increase in
   anti-inflammatory cytokines in the late stages of pregnancy. One
   possible explanation for this observation is that our control group
   consisted of non-pregnant women. Compared to them, pregnant women
   typically exhibit a state of relative immunosuppression throughout
   their pregnancy, a conclusion that is supported by the work of Abu-Raya
   and colleagues ([108]33). Additionally, the timing of blood sample
   collection from the pregnant participants was prior to the onset of
   labor. During this phase, the fetus remains in a rapid growth and
   development stage, with the maternal local immune response
   predominantly skewed toward Th2-type dominance.

   Gut microbiota is closely related to the host immune system, which may
   directly or indirectly influence mucosal immunity and ultimately affect
   systemic immune function ([109]13, [110]34). To explore immune
   alteration during pregnancy whether affected by the intestinal
   microbiota, we analyzed the characteristics of the gut microbiota and
   immune status in pregnant women and non-pregnant women and conducted a
   correlation analysis between these factors. Compared to non-pregnant
   women, pregnant women showed an increased abundance of Actinobacteriota
   and Proteobacteria, which is largely consistent with the findings of
   Koren et al. ([111]11). Gut microbiota does not exist in isolation;
   they work as guilds, collectively thriving or declining in response to
   environmental changes ([112]35, [113]36). In ecological and biological
   research, identifying microbial modules helps researchers understand
   how complex microbial ecosystems operate. Therefore, we categorized the
   gut microbiota into clusters, dividing them into multiple modules to
   reduce the dimensionality of the microbial data. In the pregnant group,
   modules 1, 2, and 4 demonstrated significantly higher abundance and
   were negatively correlated with certain pro-inflammatory cytokines.
   This seems to indicate that microbial members belonging to the same
   module have the same or similar functions ([114]37). Additionally,
   Bifidobacterium are known as probiotics in the human intestine. Some
   members of Bifidobacterium have been shown to exert an immunological
   effect ([115]38, [116]39). They induce high levels of anti-inflammatory
   cytokines or reduce pro-inflammatory cytokines through pili ([117]40),
   extracellular polysaccharides ([118]41, [119]42), etc. OTUs belonging
   to Bifidobacterium demonstrated a negative correlation with
   pro-inflammatory cytokines, indicating a relationship between
   Bifidobacterium and the reduction of pro-inflammatory cytokines during
   pregnancy.

   Metabolites play roles in the immune system that are independent of
   their conventional functions as sources or intermediates in
   biosynthesis and bioenergetics ([120]43). Specific metabolites have the
   capacity to either regulate cytokine production or demonstrate
   cytokine-like effects ([121]43). Hence, we analyzed the metabolic
   characteristics of pregnant women and the associations between
   differential metabolites and cytokines. The most significant metabolic
   pathway that differed between pregnant and non-pregnant women was the
   bile acid secretion pathway, which also involved the largest number of
   differential metabolites. The metabolism of bile acids is a complex
   process. Primary bile acids such as cholic acid and taurocholic acid
   are synthesized in the liver and secreted into the bile, subsequently
   entering the intestine ([122]44). Enzymes produced by intestinal
   microbiota further metabolize primary bile acids into secondary bile
   acids ([123]45), including deoxycholic acid and glycodeoxycholic acid.
   Subsequently, approximately 95% of bile acids in the intestine are
   reabsorbed into the portal vein before reaching the distal ileum,
   completing one cycle of enterohepatic circulation ([124]46). In our
   study, primary bile acids such as taurocholate in the plasma of
   pregnant women were upregulated, while secondary bile acids such as
   deoxycholic acid, glycochenodeoxycholate, and glycodeoxycholic acid
   were downregulated. This may be attributed to reduced farnesoid X
   receptor (FXR) activity in the intestines during pregnancy ([125]47).
   FXR can enhance the expression of organic solute transporter α and β,
   promoting the reabsorption of bile acids into the portal vein
   ([126]48). Consequently, the levels of secondary bile acids in the
   plasma during pregnancy were reduced. Studies have reported that serum
   levels of conjugated bile acids, especially taurine-conjugated ones,
   dominate the bile acid profile in late pregnancy ([127]49). As
   gestational weeks increase, taurocholic acid levels rise, while
   deoxycholic acid levels decrease ([128]49), consistent with our
   research findings. Bile acids play an essential role in regulating
   immunity ([129]50, [130]51). Recently, bile acids have been reported to
   influence host immunity by modulating adaptive immune cells ([131]52).
   In the present study, we observed negative associations between
   taurocholate, which was significantly increased in the pregnant group,
   with pro-inflammatory cytokines such as TNF-α and IL-1β. A fundamental
   research confirmed that high concentrations of taurocholic acid can
   inhibit the production and gene expression of TNF-α and IL-1β
   ([132]53). Deoxycholic acid and glycochenodeoxycholate, which were
   depleted in pregnant women, were positively correlated with
   pro-inflammatory cytokines. Glycochenodeoxycholate and deoxycholic acid
   have both been demonstrated to possess inflammatory activation
   properties ([133]24, [134]25). Our findings indicated a strong
   correlation between bile acid metabolism and the variations in plasma
   cytokine levels during pregnancy. Furthermore, a significant portion of
   metabolites present in the blood originates from the intestinal tract
   ([135]54, [136]55), suggesting that metabolites may play a pivotal role
   in the intricate interplay between gut microbiota and host responses.
   The relationship between metabolites and cytokines was found to be
   closer than that with microbiota, indicating that metabolites may exert
   a more direct influence on pregnancy immunity than gut microbes. In our
   research, certain microbes may indirectly influence immune function by
   modulating the upregulation or downregulation of metabolites. For
   example, OTU_82, which belongs to Ruminococcus callidus, may modulate
   the levels of TNF-α and IL-2 by decreasing the level of deoxycholic
   acid. Ruminococcus participates in the oxidation and epimerization of
   hydroxyl groups at the C3, C7, and C12 positions in bile acid
   metabolism ([137]56). A previous study believed that a high level of
   deoxycholic acid may activate inflammatory ([138]25). An in vitro
   experiment demonstrated that deoxycholic acid promotes the release of
   TNF-α ([139]57). Our results are consistent with these findings.
   Regrettably, this study was observational in nature. Although we
   conducted directional mediation effect analysis to infer the
   associations among microbiota, metabolism, and immunity, it does not
   prove a causal relationship between them. The findings still require
   experimental validation.

   There are several limitations in our research. Firstly, the small
   sample size may limit our ability to effectively account for errors
   caused by individual differences. Secondly, this study is a
   cross-sectional study, and due to its inherent limitations, we cannot
   control for potential influences of confounding factors such as diet.
   Lastly, despite proposing some hypotheses regarding biological
   mechanisms through the integration of microbiota, metabolism, and
   immune data sets, causal relationships cannot be confirmed. The dynamic
   associations among microbiota, metabolism, and immunity during
   pregnancy still require validation through multicenter clinical trials
   with large sample sizes and fundamental experiments. Despite these
   limitations, this study has offered new insights into our understanding
   of the connections between the gut microbiome, metabolome, and the
   immune system during pregnancy.

   In summary, our study revealed complicated associations among gut
   microbiota, metabolites, and immune during pregnancy, and identified
   some specific metabolites which may act as mediators between symbiotic
   microorganisms and immune homeostasis. This study provided a
   comprehensive landscape of the gut microbiota and metabolites in
   pregnant women and expanded our current knowledge regarding how immune
   function is affected by gut microbiota and metabolites.

MATERIALS AND METHODS

Study design and participants

   Thirty pregnant women and 15 non-pregnant women were recruited from the
   First Affiliated Hospital of Jinan University between February 2019 and
   August 2020. The inclusion criteria for pregnant participants were as
   follows: aged between 18 and 34 years, naturally conceived with a
   singleton pregnancy, and a pre-pregnancy BMI ranging from 18.5 to 21.9
   kg/m^2. Pregnant women with complications such as gestational
   hypertension, gestational diabetes, or any immune-related disorders
   were excluded. The control group comprised healthy females aged between
   18 and 34 years with a BMI between 18.5 and 21.9 kg/m^2. All
   participants reported no use of probiotics or antibiotics in the 6
   months prior to participating in the study. Additionally, we also
   recorded the basic physiological information such as age, weight, and
   height.

Sample collection

   Fecal and peripheral blood samples from pregnant women were collected
   during the period from the 37th week of pregnancy until the onset of
   labor. For the non-pregnant women, samples were collected on the 14th
   day of their menstrual cycle. Participants defecated onto a sterile
   pad, after which professionals collected approximately 1 g of feces
   using a sterile spoon and transferred it into a sterile tube. The fecal
   samples were transported under cryogenic conditions and stored at −80°C
   until examined. Additionally, 5 mL of peripheral blood was drawn from
   fasting participants by a nurse into an EDTA anticoagulant tube. The
   blood sample was allowed to sit at room temperature for 30 minutes and
   then centrifuged at 4°C at 2,000 rpm for 10 minutes. The supernatant
   serum was collected and stored in a refrigerator at −80°C.

16S rRNA gene sequencing and species annotation

   Total genome DNA of the gut microbiota from samples was extracted by
   the CTAB method, and then DNA purity and concentration were detected by
   agarose gel electrophoresis. DNA was placed in a centrifuge tube and
   diluted to 1 ng/µL with sterile water. The V3 to V4 region of the 16S
   rRNA gene was amplified using specific primer (341F/806R) with the
   barcode. Using TruSeq DNA PCR-Free Sample Preparation Kit (Illumina,
   USA) generated sequencing libraries. The library was sequenced on an
   Illumina NovaSeq platform. Raw tags were truncated and chimera-filtered
   according to the QIIME (V1.9.1,
   [140]http://qiime.org/scripts/split_libraries_fastq.html)
   quality-controlled process ([141]58). Ultimately, sequences with ≥97%
   similarity were assigned to the same OTUs using the Uparse software
   (Uparse v7.0.1001, [142]http://www.drive5.com/uparse/) ([143]59).
   Representative sequences for each OTU were screened for further
   annotation.

Microbial analysis

   To identify representative microbes for the two groups, we conducted
   random forest analysis using R package “randomForest” in RStudio
   software (version 1.4.1717). Subsequently, the top 100 OTUs of variable
   importance predicted by random forest were clustered using the Ward
   clustering algorithm via the R package “WGCNA.” The microbes contained
   in the gray module are microbes that do not belong to any other
   modules, and they were excluded from subsequent analysis. The average
   abundance of modules between the two groups was compared using the
   Wilcoxon rank-sum test method.

Untargeted LC-MS-based metabolomics

   Fecal samples and plasma samples were detected by liquid
   chromatography-tandem mass spectrometry (LC-MS/MS) analysis. The
   samples were thawed at 4°C and mixed with cold methanol/acetonitrile
   (2:2:1, vol/vol). The mixture was centrifuged for 15 minutes (14,000 ×
   g, 4°C). The supernatant was dried in a vacuum centrifuge. For LC-MS
   analysis, the samples were re-dissolved in 100 µL acetonitrile/water
   (1:1, vol/vol) solvent. To monitor the stability and repeatability of
   instrument analysis, quality control (QC) samples were prepared by
   pooling 10 µL of each sample and analyzed together with the other
   samples. The QC samples were inserted regularly and analyzed in every
   five samples.

   Analyses were performed using a ultra-high performance liquid
   chromatography (UHPLC) (1290 Infinity LC, Agilent Technologies) coupled
   to a quadrupole time-of-flight (AB Sciex TripleTOF 6600) in Shanghai
   Applied Protein Technology Co., Ltd. The metabolites in the samples
   were detected by UHPLC-Q-TOF MS. The raw data were converted into
   .mzxml format by ProteoWizard. XCMS software was used to correct the
   retention time of peak alignment and extract the peak area. Firstly,
   the data of metabolite structure identification were preprocessed for
   the extracted data, then the quality of the experimental data was
   evaluated and analyzed.

Metabolomics data analysis

   Metabolomic data analysis was performed using the APT-BioCloud platform
   ([144]http://cloud.aptbiotech.com/#/login). The OPLS-DA was used for
   multivariate data analysis. Additionally, the robustness of this model
   was evaluated by the sevenfold cross-validation and response
   permutation testing. The VIP value of each metabolite in the OPLS-DA
   model was calculated to indicate its contribution to the
   classification. Metabolites with VIP value >1 and P values <0.05 of
   Student’s t-test were considered statistically significant.
   Differentially metabolites were further mapped to the KEGG database to
   identify the altered biological pathways.

Detection of plasma cytokines

   Concentrations of IFN-γ, TNF-α, interleukins IL-1β, IL-2, IL-4, IL-6,
   IL-10, and IL-12 were measured by multiplex bead assay analysis
   according to the manufacturer’s instructions. Differences of cytokines
   between groups were calculated and visualized using GraphPad Prism
   8.0.2 software.

Multi-omics analysis

   We conducted correlation analysis of gut microbiota, metabolites, and
   cytokines based on the Spearman method. In addition, we also performed
   a directional mediation analysis to evaluate the potential mediation by
   metabolites. We calculated the mediation effect by the R package
   “bruceR.” To test the mediation effect, a bootstrap method was
   conducted with 1,000 bootstrap samples. The Sankey diagram was created
   to visualize microbes affect cytokines by metabolites and drawn using
   the R package “ggalluvial.”

Statistical analysis

   The measurement data conforming to the normal distribution were
   expressed by mean ± standard deviation, and Student’s t-test was used
   to evaluate differences between the two groups. The nonparametric
   Mann-Whitney test was conducted for data that did not conform to a
   normal distribution. Statistical analysis was performed using RStudio
   software (version 1.4.1717) or GraphPad Prism software (version 8.0.2).

ACKNOWLEDGMENTS