Oral microbiome modulation mitigates hyperglycemia exacerbation in gestational diabetes mellitus

Shengtao Gao; Nanlin Yin; Rujun Wei; Xiaoqing Li; Qiuhan Cheng; Alateng Zhula; Wen Zhou; Yifei Zhang; Sihan Li; Wei Zhou; Xiaoya Wang; Ruiqi Zhang; Qiannan Wang; Huimin Fan; Sijia Peng; Hongping Zhang; Kefeng Li; Yongfei Hu; Yuan Gao; Wenyu Shi; Hongbo Qi; Jinfeng Wang

doi:10.1038/s41467-026-74917-w

. 2026 Jul 3;17:8274. doi: 10.1038/s41467-026-74917-w

Oral microbiome modulation mitigates hyperglycemia exacerbation in gestational diabetes mellitus

Shengtao Gao ^1,^2,^3,^#, Nanlin Yin ^4,^5,^6,^#, Rujun Wei ^1,^#, Xiaoqing Li ^7,^#, Qiuhan Cheng ¹, Alateng Zhula ^8,⁹, Wen Zhou ¹⁰, Yifei Zhang ¹¹, Sihan Li ¹, Wei Zhou ¹, Xiaoya Wang ¹, Ruiqi Zhang ¹, Qiannan Wang ¹, Huimin Fan ¹, Sijia Peng ¹, Hongping Zhang ⁷, Kefeng Li ¹², Yongfei Hu ⁸, Yuan Gao ¹³, Wenyu Shi ^14,^✉, Hongbo Qi ^4,^5,^6,^✉, Jinfeng Wang ^1,^✉

PMCID: PMC13469150 PMID: 42399225

Abstract

Dysglycaemia and periodontal inflammation frequently co-occur during pregnancy, but the microbial mechanisms linking these conditions and their potential for intervention remain incompletely understood. Here, we establish prospective pregnancy cohorts including more than 2500 volunteers and longitudinally profile oral microbiome dynamics in 534 pregnant women. We show that gestational diabetes mellitus (GDM) is associated with a progressive shift from Streptococcus-dominated oral microbiota to Prevotella/Porphyromonas-enriched dysbiosis. In mouse and cellular models, this dysbiotic oral microbiota induces periodontal inflammation, systemic IL-17 and IL-1β responses, suppression of glucagon-like peptide-1 and insulin, and exacerbation of hyperglycemia. Conversely, oral microbiota remodeling through transplantation of Streptococcus-dominated bacteria attenuates periodontal inflammation, restores glucagon-like peptide-1 and insulin levels, and improves glycaemic status in mice. Salivary metabolomics identifies docosahexaenoic acid (DHA) depletion in GDM, and in vitro assays show selective suppression of dysbiosis-associated oral pathogens by DHA. We therefore test topical gingival DHA in a double-blind randomized controlled trial of 40 pregnant women with GDM (ChiCTR2400080741), with probing depth and fasting blood glucose as primary endpoints and gingival index, attachment loss and plaque index as secondary endpoints. Daily gingival DHA application for six weeks improves probing depth and attenuates fasting glucose increase compared with placebo, with median fasting glucose changes from baseline of 0.10 versus 0.27 mmol/L. Together, these findings identify oral dysbiosis as a microbial driver of periodontal and glycaemic deterioration during pregnancy and support oral microbiome modulation as a potential adjunctive strategy for pregnancy care, although the clinical findings remain preliminary and require validation in larger trials with broader glycaemic endpoints.

Subject terms: Microbiome, Metagenomics

Here, the authors associate changes in the oral microbiome during gestational diabetes with gum inflammation and blood sugar control, and show in animal models and a human trial that restoring healthier oral bacteria or topical docosahexaenoic acid, reduce these effects.

Introduction

Metabolic imbalances and oral diseases pose major challenges to global medical and public health, with diabetes and periodontitis being of particular concern^1–3. Diabetes frequently coexists with periodontitis, and the bidirectional relationship between these two conditions accelerates disease progression in both^4–6. Gestational diabetes mellitus (GDM), a specific type of diabetes that develops during pregnancy, is also highly comorbid with periodontitis^7,8.

In addition to sharing similar risks as other forms of diabetes^9,10, GDM exacerbates the likelihood of adverse pregnancy outcomes, including hypertensive disorders, excessive fetal growth, and long-term metabolic complications for both mother and child^11–13. Despite its high global average prevalence (~14.5%)¹⁴, many treatment options for diabetes are avoided during pregnancy due to concerns about fetal safety, leaving GDM patients with fewer therapeutic options¹⁵. Breaking the link between GDM and periodontitis as well as protecting periodontal health is important for maintaining glycemic balance during pregnancy and intergenerational health of mother and baby. However, the causal relationships and triggering modalities between GDM and periodontitis remain incompletely elucidated, making it a challenge to control their mutually reinforcing progression and deterioration.

Disruption of the oral microbial community has been recognized as the initiating factor in the etiology of periodontitis. This is primarily attributed to an increase in the number of opportunistic pathogens, such as Porphyromonas and Prevotella, which are classified as “red and orange complex bacteria”^16–18. These bacteria cause elevated levels of specific inflammatory factors and local periodontal inflammation^19,20. This local periodontal inflammation is often dysregulated and inadequately controlled, especially in susceptible individuals, resulting in sustained inflammatory destruction of the periodontium and the establishment of a nutritionally favorable environment for the proliferation of periodontitis-associated microorganisms, which represents an early initiating phase of periodontitis²¹. Notably, similar alterations in oral microbiota have also been documented in patients with blood glucose dysregulation^22–24. Although periodontitis was considered to be an important risk factor for advancing the GDM process and exacerbating the hyperglycemia^25–27, whether these microbial changes occur independently or play a crucial role in mediating the pathogenesis of periodontitis in patients with impaired glucose metabolism is largely unknown.

In this work, we investigate the dynamic interplay between the oral microbiome and glycemic regulation by longitudinally tracking oral microbiota changes during pregnancy and investigating microbial contributions to GDM progression. By integrating multi-omics analyses, in vitro microbial cultures and animal model experiments, we show that oral dysbiosis associated with imbalanced glucose metabolism is a key factor linking GDM and periodontitis, and we explore avenues for curbing glycemic deterioration. Based on the elucidation of the oral microbe-mediated cascade effects, we further verify the efficacy of oral microbiota reprogramming in improving periodontal status and glycemic levels through clinical intervention.

Results

Succession of oral dysbiosis coincides with the onset and progression of GDM

To investigate whether the oral microbiome serves as a critical mediator between glycemic imbalance and periodontitis, we primitively enrolled 2523 volunteers and ultimately established three prospective cohorts (Discovery Cohort: C_Dis; Intensive Cohort: C_Int; and Validation Cohort: C_Val) involving 534 pregnant women across two independent research centers in east and west China. The cohort of C_Dis was designed to explore the potential association between glycemic imbalance and periodontitis across a large sample size and over a broad time span (from the first to third trimester). C_Int was a densely sampled, high-resolution cohort designed to capture dynamic changes in the oral microbiome surrounding the oral glucose tolerance test (OGTT) time point, both before and after GDM diagnosis. C_Val was an independent external cohort intended to validate the key findings from C_Dis and C_Int, thereby enhancing the robustness and generalizability of the results.

Based on the OGTT diagnosis at 24 weeks of pregnancy, the volunteers were divided into a GDM group and a healthy (CON) group (Fig. 1A). Totally 747 saliva samples were collected from 534 women at 7 time points throughout the pregnancy (Table S1, S2). All samples were used for 16S rRNA amplicon sequencing, which yielded 77,793,328 sequences, generating the largest human oral microbiome dataset to date on the relationship between dysglycaemia and oral microbiota during pregnancy. Cross-cohort core microbiome analysis showed that the oral microbiome was underpinned by a stable set of highly prevalent core genera (Fig. S1A). Principal coordinates analysis (PCoA) based on Bray-Curtis distances across all three cohorts showed substantial overlap between GDM and CON. Nevertheless, PERMANOVA detected a significant overall difference between GDM and CON (Fig. S1B, R² = 0.011, P < 0.001), suggesting that gestational dysglycaemia is associated with a subtle but reproducible ecological shift in the oral microbiome.

To identify the microbial features associated with GDM and to explore the etiology of GDM susceptibility to periodontitis, we first performed oral microbiome clustering in the discovery cohort C_Dis, yielding two distinct consortia (Fig. 1B, Fig. S1C, D). For the first consortium, Streptococcus was the dominant bacterium with an average relative abundance greater than 30% (Fig. 1B, Fig. S1E). This consortium was termed as Streptococcus-dominated Oral Microbiota (SOM). In contrast to the presence of an absolute dominant genus in SOM, the relative abundance of the bacteria enriched in the second consortium was more evenly distributed. Genera represented by Prevotella and Porphyromonas had a relatively high abundance in this consortium (Fig. 1B, Fig. S1C, Fig. S1E), which was named Prevotella-enriched Oral Microbiota (POM). There was a strong preference for GDM over CON in the ratio of SOM to POM. More individuals with GDM belonged to the POM than to the SOM (Fig. 1B, Pearson’s Chi-squared test P = 0.002). Specifically, less Streptococcus and more Prevotella and Porphyromonas harbored in the oral cavity of GDM patients compared to CON (Fig. 1B, one-tailed Mann-Whitney test P: <0.001, <0.001 and 0.034, respectively), demonstrating that the absence of SOM bacteria and the enrichment of POM bacteria are critical microbial features associated with GDM.

To unveil the connections between GDM and oral microbiota, we analyzed the dynamic responses of SOM and POM bacteria to the onset and progression of GDM longitudinally across the entire pregnancy. Compared with CON, SOM bacteria in GDM gradually declined throughout pregnancy with follow-up time, while POM bacteria showed a progressive increase (P < 0.05, Fig. 1B, C). In the third trimester, the representative bacteria of SOM (Streptococcus) and POM (Prevotella) significantly decreased and increased (P < 0.05) in GDM, respectively (Fig. 1C), indicating that the dysbiosis of oral microbiota associated with glycemic imbalance persists and worsens throughout the pregnancy. We combined the cohorts of C_Dis and C_Int and analyzed the transformations between POM and SOM of paired samples from early, mid, and late pregnancy revealed that in GDM, more individuals were classified as SOM in early pregnancy but switched to POM in late pregnancy compared to CON (0.81 vs 0.50, Fig. 1D). Such finding suggests that GDM may induce a shift in the oral microbiota from SOM, where Streptococcus dominates, to POM in which Prevotella and Porphyromonas were enriched. This resulted in a significant increase in the diversity and dysbiosis index of oral microbiome of GDM in the middle and late pregnancy (Fig. S1F, G). Consistent with this progressive SOM-to-POM transition, correlation network analysis of representative SOM- and POM-associated genera revealed stage-dependent rewiring of oral microbial interactions, with reduced network connectivity and increased modularity in GDM, particularly in late pregnancy (Fig. S1H, I). To exclude the interference of subject age, we performed the above analyses in the age-matched GDM and CON subsets, which also confirmed this result (Fig. S2A–C). In addition, multivariable-adjusted analyses controlling for age, BMI, and gestational week showed a similar pattern, further supporting the robustness of the observed association between progressive oral dysbiosis and GDM (Fig. S2D). If the OGTT results were taken into account, POM was more prevalent than SOM in GDM cases with 2 and 3 abnormal indexes in the OGTT diagnosis (Fig. 1E). These findings showed a strong relationship between changes in the ratio of SOM/POM bacteria and GDM onset.

To validate the positive relationship between POM and GDM, based on the multi-center principle, we introduced another cohort (C_Val) for oral microbiome sequencing and analyzed the proportion of POM in 104 pregnant women. In them, 24 out of 27 individuals who were later diagnosed with GDM were classified to POM (Fig. 1F). These testing results at around 12 weeks of gestation imply that oral dysbiosis in patients with GDM occurred prior to the OGTT diagnosis and predisposed to POM. Further analysis of Potential of Heat-diffusion for Affinity-based Trajectory Embedding (PHATE, a dimensionality reduction method designed to preserve both local and global structure in high-dimensional data) and dissimilarity between SOM and POM subjects of C_Dis and C_Val together validated robustness of the relationship between POM and GDM (Fig. 1F), and supported that POM was a stable phenotype associated with gestational glycemic imbalance, characterized by enrichment of bacteria such as Prevotella and Porphyromonas in oral microbiome.

Subsequently, we performed functional analysis based on metagenomic data of human saliva samples collected from Trimester 3 of the CDis cohort (a total of 33 metagenomic samples [GDM: 16; CON: 17]) to evaluate possible outcomes associated with oral microbiota disorder in women with GDM. Repeated balanced subsampling and cross-platform comparisons using matched 16S rRNA profiles consistently supported the robustness of the metagenomic signals and their overall concordance with the 16S rRNA-defined dysbiosis pattern, as reflected by similar abundance structure and concordant GDM-versus-CON changes across shared genera. (Fig. S3A–E). The oral microbiome of GDM exhibited a significant enrichment in metabolic pathways associated with LPS synthesis (Fig. S3F), a well-established proinflammatory pathway, as well as multiple biological processes involved in the regulation and activation of inflammation (Fig. S3G, Table S3). The findings collectively demonstrated that oral dysbiosis in women with GDM may amplify the proinflammatory effects of oral microbiota.

POM triggers in situ and systemic inflammation and aggravates hyperglycemia

Since some members of the POM bacteria, mainly Porphyromonas gingivalis and Prevotella intermedia are well-established periodontal pathogens, upon observing the oral dysbiosis and enriched proinflammation effect of oral microbiota of pregnant women with GDM, we naturally realized that the formation of POM may contribute to periodontitis onset. We thus transplanted oral microbiota by orally administering saliva samples obtained from patients with and without GDM to the oral cavities of mice (Fig. 2A). After 3 weeks of continuous perfusion, MicroCT scanning and three-dimensional reconstruction were performed on the maxillary bone tissue to assess the periodontal health status. The transplantation of GDM oral microbiota into mice resulted in significant alveolar bone loss, as indicated by an increased cementoenamel junction to alveolar bone crest (CEJ–ABC) distance and a reduced bone volume/total volume (BV/TV) ratio (Fig. 2B, Fig. S4A). It was accompanied by a general increase in blood inflammation levels, especially the inflammatory factors interleukin-17 (IL-17), which was elevated significantly after GDM oral microbiota transplantation (Fig. 2C, Fig. S4B–D). This denotes that oral dysbiosis not only caused periodontal damage, but also stimulated a systemic inflammatory response together.

During routine blood glucose monitoring, we incidentally observed that mice receiving oral microbiota from GDM patients exhibited significantly higher random blood glucose (RBG) compared to mice receiving oral microbiota from healthy pregnant women at week 2 of treatment (Fig. 2D). The finding was confirmed by following OGTT, wherein fasting plasma glucose showed abnormal elevation in mice received GDM microbiota (Fig. 2D). Additionally, transplantation of GDM-derived oral microbiota significantly decreased the levels of insulin (INS) and glucagon-like peptide-1 (GLP-1) (Fig. 2E), which directly contribute to the regulation of blood glucose. This suggested that oral microbiota may influence blood glucose homeostasis by modulating hormone secretion. To further investigate the relationships between oral dysbiosis and blood glucose dysregulation, we isolated SOM and POM bacteria from human saliva, cultured six representative strains in vitro, mixed them proportionally to construct two oral microbiota consortia simulating SOM and POM (Streptococcus salivarius and Gemella morbillorum for SOM; and P. gingivalis, P. intermedia, Neisseria flava, and Rothia dentocariosa for POM). All these species were isolated from oral cavities and matched the dominant genera identified in our stratified SOM/POM microbiome profiles. While the full SOM and POM communities comprise broader taxonomic diversity, these six species were chosen based on their relative abundance, biological relevance, and our ability to culture and preserve them under laboratory conditions. This selection strategy allowed us to construct experimentally tractable, ecologically grounded consortia that captured the key compositional and functional differences between SOM and POM microbiota.

We transplanted a simulated POM consortium into the oral cavity of mice, with PBS and antibiotic cocktail as control (Fig. 2F). As a result, oral transplantation of POM bacteria elicited periodontal damage in mice (Fig. 2F, Fig. S4E–H) and induced significant elevations in both RBG and OGTT 0 h blood glucose at the 3rd week (Fig. 2G). In terms of blood glucose-regulating hormones and inflammatory factors, it was also observed that oral transplantation of POM bacteria significantly reduced the levels of INS and GLP-1, but significantly increased the levels of IL-17 and IL-1β in blood (Fig. 2H, I). These findings replicated the outcomes of the salivary microbiota transplantation experiment successfully, thereby suggesting that once the oral microbiota develops GDM-like disorder, it will induce periodontal and systemic inflammatory response and trigger a further exacerbation of hyperglycemia symptoms.

Because there is the possibility that oral bacteria travel downstream into the gastrointestinal tract to regulate blood glucose, we designed a supplementary experiment to test whether the observed systemic inflammation and hyperglycemia phenotypes were a result of bacterial entry into the gastrointestinal tract during oral saliva perfusion. We selected S. salivarius from SOM, and P. intermedia and P. gingivalis from POM for gavage administration. The results indicated that a 2-week bacterial gavage with POM bacteria enriched in the oral cavity of GDM women, did not significantly increase blood glucose levels compared to the control and SOM bacteria gavage groups (Fig. 2J). This is preliminary evidence that in the absence of oral microbiota transplantation and periodontal inflammation, periodontal pathogens entering the gastrointestinal tract alone will not result in an elevation of blood glucose. These experiments elucidate that hyperglycemia only occurs when POM bacteria in situ induce periodontal inflammatory responses in the oral cavity.

POM exacerbates hyperglycemia via a dual pathway of hormones and stress

Given the concomitant decrease in INS and GLP-1 levels, along with elevated blood levels of IL-17 and IL-1β observed in mice transplanted with oral microbiota from GDM patients (Fig. 2C, E, H, I), we hypothesized that oral microbes may disrupt glycemic homeostasis by inducing an inflammatory response. The relationship between blood glucose and IL-17 and IL-1β was then investigated by intraperitoneally injecting three doses of these two substances into mice (Fig. 3A). After intraperitoneal injection of IL-17, the fasting blood glucose levels of mice were increased in a dose-dependent manner and were significantly elevated compared with the non-injected group once the dosage of IL-17 reached 0.5 μg (Fig. 3B). Notably, intraperitoneal injection of IL-1β resulted in a significant rise in fasting plasma glucose levels, even at a very low dose (0.1 μg) (Fig. 3C). Further assays showed that intraperitoneal injection of either IL-17 or IL-1β significantly reduced the levels of INS and GLP-1 in the murine bloodstream (Fig. S5A, B), and decreased the expression of GLP-1 protein in ileal tissues (Fig. 3D). In other words, both IL-17 and L-1β can raise blood glucose levels by suppressing the secretion of INS and GLP-1, albeit at different dosages.

Fig. 3 — A Experimental design of IL-17 and IL-1β intraperitoneal injection. At 24 h after injection, fasting blood glucose was measured, and blood and ileal tissues were collected. B, C Fasting blood glucose 24 h after intraperitoneal injection of different dosage of IL-17 and IL-1β. D Relative protein expression of GLP-1 in ileum of mice treated with 1 μg IL-17 and IL-1β. β-actin was housekeeping gene. E Experimental design of blockage of IL-17 and IL-1β using DIM intraperitoneal injection. Twelve periodontitis mice were constructed using *P. intermedia*. After 13 days of *P. intermedia* transplantation, half of them were intraperitoneally injected with DIM. F RGBT of CON, *P. intermedia*, and DIM mice. From 7th experimental day, the RGB of mice with transplantation of *P. intermedia* were persistently higher than the mice treated with PBS + ABX (P < 0.05). From 15th experimental day, 2 days after the introduction of DIM, RGB of *P. intermedia* + DIM mice were significantly lower than *P. intermedia* mice (P < 0.05) and were similar with PBS + ABX mice (P > 0.05). G, H Blood content of GLP-1 and INS of CON, *P. intermedia*, and DIM mice. I Experimental design of cell culture added with IL-17 and IL-1β. Both of these 2 substances were added to the medium at 0, 80, 160 ng/ml. STC-1 and NCL-H716 are the mouse and human enteroendocrine cell lines, respectively. J, K Relative RNA expression level of GLP-1 in NCL-H716 cell lines after treated with different concentration of IL-17 and IL-1β. L Relative protein expression level of GLP-1 in NCL-H716 cell lines after treated with 160 ng/ml IL-17 and IL-1β. β-actin was housekeeping gene. For (D, L) GLP-1 and β-actin were analyzed using the same protein lysates loaded in the same lane order on gels processed in parallel. For box plots, the center line indicates the median, box limits indicate the 25th and 75th percentiles, and whiskers extend to the smallest and largest values within 1.5× the interquartile range; individual points represent biologically independent mice or independently treated cell-culture wells. For bar plots, bars show mean values and error bars show SEM; overlaid points represent individual biological replicates. For (B, C) CON n = 14 mice and each IL-17 or IL-1β dose group n = 10 mice. For (D) n = 3 biologically independent mice per group. For (F–H) CON n = 7 mice, *P. intermedia n* = 6 mice after group splitting, and *P. intermedia* + DIM n = 6 mice. For (J, K) n = 3 independently treated cell-culture wells per group. For (L) n = 3 independent protein samples per group. Statistical comparisons were performed using one-tailed Welch’s t-tests. Exact t statistics, degrees of freedom, Cohen’s d values, 95% confidence intervals and exact P values are provided in the Source Data statistical reporting summary. Source data and exact statistical reporting are provided as Source Data files. ****: P < 0.0001; ***: P < 0.001; **: P < 0.01; *: P < 0.05; n.s.: P > 0.05.

In addition to INS and GLP-1, other possible glucose-elevating pathways involving IL-17 and IL-1β were investigated. Levels of cortisol, epinephrine, and oxidative stress, which also possess hyperglycemic effects, were quantified. The results showed that intraperitoneal injection of IL-1β significantly augmented the levels of stress-related hormones cortisol and EPI (Fig. S5C, D). Concurrently, there was a significant reduction in the activities of the oxidative stress-associated enzymes catalase (CAT), superoxide dismutase (SOD), and glutathione peroxidase (GPx) in the blood (Fig. S5E–G). These findings underscore the critical role played by IL-1β in activating the systemic stress response. In contrast, the impact of IL-17 on these aspects was not discernible. Taken together, apart from suppressing INS and GLP-1 secretion, IL-1β can also elevate blood glucose levels by eliciting systemic stress responses, thereby leading to an imbalance in blood glucose regulation even at lower concentrations. This partially explains why blood glucose is more responsive to IL-1β compared to IL-17.

To validate the role of IL-17 and IL-1β in the induction of hyperglycemia, we next examined whether joint inhibition of these inflammatory axes could restore GLP-1 signaling and glycemic homeostasis in periodontitis-induced hyperglycemia models (Fig. 3E). We employed 3,3’-diindolylmethane (DIM), a naturally occurring compound that downregulates IL-1β expression through inhibition of NF-κB signaling and interferes with IL-17 signaling by suppressing IL-17 receptor (IL-17RA) activation and downstream transcriptional responses. Following our established protocol, P. intermedia was used to induce periodontal inflammation in mice. From day 7 after P. intermedia transplantation, blood glucose levels in transplanted mice were consistently and significantly higher than those in non-transplanted mice. As expected, after administering DIM injections to decrease the abundance of IL-1β and inhibit the binding of IL-17 to its receptor, a significant reduction in blood glucose levels was observed (Fig. 3F). Even during continuous oral infusion of P. intermedia, DIM effectively regulated blood glucose as well as GLP-1 and INS levels (Fig. 5G, H). The evidence provided by this experiment supports the mediating role of IL-17 and IL-1β in periodontal inflammation and hyperglycemia caused by oral bacteria.

A Metabolomic assays of compounds in the saliva of GDM and CON pregnant women in late pregnancy. Left panel shows the top five most upregulated and downregulated compounds in GDM saliva compared with CON. For fructose, the values > 15 were replaced with 15. For glycerate, the values > 3 were replaced with 3. For ketovaline, the values > 10 were replaced with 10. The asterisk (*) represents the significant differences between GDM and CON (OPLSDA VIP > 1, FDR adjusted Mann-Whitney P < 0.1, |log2foldchange| > 0.5). The shades behind the compounds shows the class of these compounds. In the middle panel, the average concentration of each class in GDM and CON. The pie chart in the right panel illustrates the distribution of compounds across different classes identified in this study. See also Table S5. B PUFAs and carbohydrates were most influenced by GDM compared with CON. The raw abundance data of each metabolite utilized for this figure was standardized across samples. The red or green circles represent the |log2foldchange| of metabolic class in GDM/CON were higher than 0.5, and gray circles represent the |log2foldchange| of metabolic class in GDM/CON were lower than 0.5. C Concentration of fructose, glycerate, glutamine, and DHA in saliva from early pregnancy were measured using UPLC-MS/MS. D Experimental design of oral perfusion of GDM-enriched metabolites (fructose and oxoglutarate). The bottom panel shows the representative pictures of MicroCT analysis of alveolar bone tissues. Red circles highlight the extent of alveolar bone resorption of different groups. E Effect of oral cavity perfusion of fructose and oxoglutarate on the blood content of IL-17 and IL-1β. See also Fig. S9. F We selected top five up-regulated or down-regulated metabolites in GDM saliva and prepared three concentration gradients (10×, 100×, 1000×) based on their concentrations in saliva. The impacts of these metabolites on oral microbiota were estimated using the changes of the area under the growth curves (AUC) of the POM and SOM species. This picture shows the impacts of 100× concentration of the metabolites on the growth of POM and SOM species. The color of each square indicates the variation in AUC for species cultured with different metabolites. The growth curves of the control group (without additional metabolites) are represented by black lines, while those of each metabolite are shown in red. The X axis range of each subgraph is set from 0 to 50 h, with the culture time being fixed at 48 h, except for *Streptococcus* which has a culture time of 32 h. The Y axis represents the Optical Density of the cultures at 600 nm wavelength. The range of Y axis is 0–2. The clustering of rows and columns is based on the hierarchical clustering method with complete linkage distance. ΔAUC = AUC of bacteria with metabolites added—AUC of bacteria without metabolites added. For box plots, the center line indicates the median, box limits indicate the 25th and 75th percentiles, and whiskers extend to the smallest and largest values within 1.5× the interquartile range; individual points represent independent saliva samples or biologically independent mice. For Fig. 5A, targeted salivary metabolomics included CON n = 9 and GDM n = 12 saliva samples in late pregnancy. For (C) sample sizes were DHA, fructose and glycerate: CON n = 21 and GDM n = 30 saliva samples; glutamine: CON n = 14 and GDM n = 17 saliva samples. C was analyzed using one-tailed Welch’s t-tests. For (E) n = 10 biologically independent mice per group, and comparisons among PBS + ABX, fructose and oxoglutarate groups were performed using one-way ANOVA. For (F) bacterial growth assays were performed with n = 3 independent cultures per condition. Exact test statistics, degrees of freedom where applicable, effect sizes, 95% confidence intervals, FDR-adjusted P values where applicable, and exact P values are provided in the Source Data statistical reporting summary. Source data and exact statistical reporting are provided as Source Data files.

To further elucidate the inhibitory effects of IL-17 and IL-1β on GLP-1 expression in enteroendocrine cells and the potential pathways by which they exacerbate glucose imbalance, we then used human (NCL-H716) and mouse (STC-1) enteroendocrine cell lines for in vitro treatment with IL-17 and IL-1β (Fig. 3I). The results of qPCR and western blot demonstrated that added IL-17 and IL-1β significantly downregulated GLP-1 expression in enteroendocrine cells at both mRNA and protein levels (Fig. 3J–L). This verifies a suppressive effect of both IL-17 and IL-1β on GLP-1 expression in human and murine cell lines. We also conducted a transcriptomic sequencing to compare the gene expression profiles of the two cell lines before and after treatment with IL-17 and IL-1β. After IL-1β treatment, 327 and 104 differentially expressed genes we identified in NCL-H716 and STC-1, respectively. These numbers were significantly higher than the 12 and 20 differential genes observed in NCL-H716 and STC-1 after IL-17 treatment (Fig. S5H). By KEGG functional annotation analysis of the intersected differentially expressed genes in NCL-H716 and STC-1 (Fig. S5I), we found that signaling pathways of NF-kappa B, IL-17, and TNF, which are involved in the regulation of inflammatory responses and stress, were activated by IL-1β (Table S4). Our analyses proved that IL-1β had a more pronounced effect on both enteroendocrine cells compared to IL-17, and may downregulate GLP-1 expression through inflammatory response and stress pathways (Table S4, Fig. S5J).

Remodeling of the oral microbiota improves periodontal and glycemic status

Given the predominance of SOM bacteria in the oral cavity of healthy pregnant women, we proceeded to investigate whether SOM bacteria, comprising S. salivarius and G. morbillorum, could ameliorate or suppress periodontal inflammation and glycemic imbalance mediated by oral dysbiosis (Fig. 4A). The addition of SOM bacteria significantly ameliorated alveolar bone loss compared to oral perfusion with POM bacteria alone (Fig. 4A, Fig. S6A–C). Both immunohistochemical and immunofluorescence analyses of periodontal tissue revealed that the introduction of SOM bacteria reduced the accumulation of IL-17 and IL-1β in periodontal tissue (Fig. 4B–D). The levels of IL-17 and IL-1β in the murine bloodstream were significantly lower in the presence of SOM bacteria than their absence (Fig. S6D, E). These findings reveal that SOM consisting of S. salivarius and G. morbillorum, is effectively in alleviating POM-induced local and systemic periodontal inflammation.

Fig. 4 — A Experimental design of oral microbiota remodeling with SOM consortium. The POM mice only received POM bacteria, the POM + SOM mice received POM and SOM bacteria. The bacteria oral infusion lasted for 21 days and then the blood, alveolar bone, and ileal tissues were collected. Pictures on the bottom panel were the representative results of MicroCT analysis of alveolar bone tissue of POM and POM + SOM mice. Red circles highlight the extent of alveolar bone resorption in POM and POM + SOM group. B Representative results of immunohistochemical analysis of IL-17 and IL-1β levels in periodontal tissues of POM and POM + SOM mice. C Quantitative statistics of immunohistochemical results showed that IL-17 and IL-1β levels decreased after the introduction of SOM. D Immunofluorescence of IL-17 and IL-1β levels in periodontal tissues. Cytokines were stained with green fluorescence; alveolar bone tissues were stained with blue fluorescence. E Blood GLP-1 levels of POM and POM + SOM mice. F Relative protein expression of GLP-1 in ileum tissues of POM and POM + SOM mice. See also Fig. S6. G Experimental design of the oral microbiota remodeling with SOM bacteria for female, male, and pregnant mice. The 14 days RBGT and OGTT 0 h blood glucose elevation induced by POM consortium were ameliorated by SOM consortium. M, F, and P represent the male, female, and pregnant mice, respectively. See also Fig. S7. H Experimental design of oral transplantation with single bacterium. Except for the 10 CON mice who were treated with PBS and antibiotics, 20/30 mice received *P. intermedia* and 10/30 mice received *P. gingivalis* in the first 10 days. For the close following 11 days, 10/20 of the *P. intermedia* mice and all the *P. gingivalis* mice received *S. salivarius*. The remaining 10/20 *P. intermedia* mice continue received *P. intermedia*. See also Fig. S6. I Successive RBG surveillance after bacterial transplantation. First 10 days after oral transplantation with *P. intermedia*, RBG of three *P. intermedia* groups were all higher than PBS + ABX group, while from 17 to the end of this study, the RBG of *P. intermedia* group was higher than all the other three groups. The P-values displayed in the bottom panel represent the highest P-values observed within each comparison group in RBG surveillance. J Blood content of IL-17 and IL-1β at the end of oral transplantation with single bacterium experiment. For (F) GLP-1 and β-actin were analyzed using the same protein lysates loaded in the same lane order on gels processed in parallel. For box plots, the center line indicates the median, box limits indicate the 25th and 75th percentiles, and whiskers extend to the smallest and largest values within 1.5× the interquartile range; individual points represent biologically independent mice or independently quantified tissue images. For bar plots, bars show mean values and error bars show SEM; overlaid points represent individual biological replicates. For (C) n = 8 independently quantified periodontal tissue images per group. For (E) POM n = 38 and POM + SOM n = 37 mice. For (F) n = 3 biologically independent mice per group. For (G) sample sizes were male mice: POM n = 14 and POM + SOM n = 13; female mice: POM n = 12 and POM + SOM n = 12; pregnant mice: POM n = 10 and POM + SOM n = 10. For (I–J) n = 10 biologically independent mice per group. Pairwise comparisons were performed using one-tailed Welch’s t-tests. Exact t statistics, degrees of freedom, Cohen’s d values, 95% confidence intervals and exact P values are provided in the Source Data statistical reporting summary. Source data and exact statistical reporting are provided as Source Data files. ****: P < 0.0001; ***: P < 0.001; **: P < 0.01; *: P < 0.05; n.s.: P > 0.05.

We subsequently assessed the efficacy of SOM bacteria in preventing hyperglycemia exacerbation. The introduction of SOM bacteria significantly attenuated the POM-induced elevation in RBG and 0 h and 1 h blood glucose during the OGTT, while concomitantly elevating the levels of hormones implicated in blood glucose regulation, viz., GLP-1 and INS (Fig. 4E, Fig. S6F, G). SOM bacteria likewise significantly up-regulated GLP-1 secretion in mice with glucose imbalance when the protein level of GLP-1 in the ileum tissue of mice was quantified by western blot (Fig. 4F). Transcriptomic sequencing and functional enrichment analyses of mouse pancreatic tissue further revealed that SOM bacteria up-regulated signal transduction-related processes and INS expression in pancreas (Fig. S6H, I). On this basis, by using metabolomics and qPCR in colonic contents and liver tissues respectively, we evaluated whether oral bacteria affected glucose homeostasis through regulating intestinal metabolism and hepatic gluconeogenesis. The results showed that there were neither significant differences in the metabolome of colon content (Fig. S6J) nor in the mRNA expression levels of rate-limiting enzyme genes (PCK and G6PC) for gluconeogenesis in the liver between POM and POM + SOM bacteria transplantation mice (Fig. S6K). Additionally, to ascertain whether the effect of SOM bacteria on blood glucose levels was influenced by sex or pregnancy, we conducted oral bacterial infusion and periodontitis modeling on male, female, and pregnant mice, respectively (Fig. 4G). All mice exhibited consistent responses in terms of inflammatory factors, glucose, INS, and GLP-1, as well as hepatic PCK and G6PC gene transcript levels (Fig. S7A–H). Collectively, the improvement of glycemic imbalance by SOM bacteria was not attributable to intestinal metabolism and hepatic gluconeogenesis; rather, it likely resulted from the mitigation of periodontal inflammation and restoration of GLP-1 expression in intestinal endocrine L cells. This enhanced the signal stimulation received by pancreas B cells and subsequently increased INS secretion. Notably, these findings were consistent regardless gender and pregnancy, indicating that glycemia-raising effect of periodontitis is not restricted to pregnancy. Glycemic imbalance of other type of diabetes will also be partially attributed by periodontitis.

We also explored the potential of oral microbiota remodeling as a therapeutic approach for managing or ameliorating glycemic imbalance in patients with periodontal inflammation and concurrent glucose metabolism disturbance. For this purpose, we first established two mouse models of periodontitis complicated by glucose imbalance induced by P. intermedia and P. gingivalis, respectively (Fig. 4H). Then all P. gingivalis treated mice and half of P. intermedia treated mice were administrated with S. salivarius for oral microbiota remodeling. On day 7 after remodeling of the oral microbiota, periodontal inflammation-induced blood glucose elevation was significantly attenuated (Fig. 4I), and hormones associated with blood glucose regulation and inflammatory response were also markedly restored (Fig. 4J, Fig. S8A–E). This discovery identified novel therapeutic targets for the management of diabetes, and proposed that consideration should be given to increasing the proportion of S. salivarius in oral cavities for diabetic patients with concomitant periodontitis.

Polyunsaturated fatty acids differentially modulate oral microbes to form POM

To develop a more feasible scheme for clinical intervention rather than oral microbial transplantation, we intended to look for active ingredients from oral metabolites that have a regulatory effect on microbes. We performed metabolomic assays of compounds in the saliva of volunteers in late pregnancy to measure the extent of oral metabolomic changes in GDM patients (Fig. 5A, B). Of the detected 174 compounds in 18 categories, including carbohydrates, amino acids, and organic acids, the greatest differences were in polyunsaturated fatty acids (PUFAs) (Fig. 5B). The amount of 22 compounds differed significantly between GDM and CON, and 17 (five were PUFAs and their derivatives) of them were deficient in GDM (Table S5). To track the onset of imbalances in compounds of GDM saliva, we then moved the assay forward to quantify the compounds in saliva collected in early pregnancy. We specifically detected the content of the most upregulated (fructose and glycerate) and the most downregulated compounds (glutamine and docosahexaenoic acid [DHA]) in GDM saliva of late pregnancy. The results showed that the amount of these compounds was already significantly different between GDM and CON at this stage (Fig. 5C), and the direction of changes were the same as in late pregnancy. Such a pattern was consistent with that observed in the oral microbiome, where dysbiosis emerges with GDM in early pregnancy and persists throughout.

To test whether these compounds have independent effect on the glycemia, we administered an antibiotic cocktail to decrease the resident oral bacterial load in mice, followed by infusion of two carbohydrates (fructose and oxoglutarate) that are significantly enriched in the oral cavity of women with GDM, respectively. We then assessed both periodontal and glycemic status (Fig. 5D). It was observed that although fructose and oxoglutarate exhibited significant enrichment in the oral cavity of women with GDM and carbohydrate displayed a high abundance in their mouth (Table S5), without oral microbiota, GDM-enriched metabolites alone could not induce any substantial alterations in alveolar bone volume fraction, lingual or buccal CEJ-ABC distance (Fig. 5D, Fig. S9A–C). In other words, these metabolites did not independently induce significant periodontal damage. The absence of significant changes in glucose values and blood levels of IL-17, IL-1β, GLP-1, INS, GHb, and CRP (Fig. 5E, Fig. S9D–H) also effectively suggested that compounds present in the oral cavity cannot directly contribute to periodontal inflammation and hyperglycemia without microbial involvement.

To further investigate the relationship between these differential compounds and periodontal inflammation and glycemic imbalance as well as to analyze the causality between dysbiosis and metabolic disorder of GDM, we first testify whether the compounds were produced by oral bacteria. We cultured six representative bacterial strains isolated from human saliva in vitro, mixed proportionally to construct two oral microbiota consortia simulating SOM and POM, and detected the metabolites of these two consortia. Based on the results of the metabolomic assay, only 5 of the 22 compounds that differed significantly between GDM and CON could be detected in cultures (Fig. S10A), suggesting that the vast majority of compounds were not produced by these oral bacteria. Among the compounds detected, glutamine showed the greatest change before and after incubation and was enriched in the simulated culture system of SOM (Fig. S10A). Accordingly, SOM bacteria may have an advantage in glutamine synthesis. Metagenomic sequencing from human saliva samples and comparative genome analysis provided compelling evidence supporting this hypothesis. The gene abundance of enzymes involved in the catabolism of fructose and biosynthesis of glutamine, such as phosphotransferase system and glutamine biosynthesis enzymes, were decreased in the oral metagenome of GDM as compared to CON (Fig. S10B). We also conducted a comprehensive quantification of the copy number of enzyme genes related to glutamine biosynthesis across 1602 completed genomes of oral bacteria. The findings revealed an expansion of these enzyme genes within the Streptococcus bacterial genomes, but they were found to be partially or completely absent in typical POM bacteria such as Prevotella and Porphyromonas (Fig. S10C). This disparity may not only serve as a fundamental explanation for the variations in glutamine synthesis and fructose catabolic capacity of SOM and POM bacteria, but also explain the cause of the lower fructose but higher glutamine content observed in the simulated culture system of SOM as well as in the oral metabolome of the CON.

We then carried out an integration analysis of the microbiome and metabolome to estimate the co-occurrence probability of oral microbes and compounds in addition to infer the potential interactions between them. Five compounds enriched in GDM were found to have high probabilities of co-occurrence with POM bacteria, whereas Streptococcus from SOM strongly co-occurred with DHA which was deficient in GDM (Fig. S10D). Correlation calculations also verified that DHA were positively correlated with SOM bacteria (Fig. S10E), raising the possibility of their co-occurrence. These GDM-associated co-variations between oral microbes and compounds may serve as key regulation to concurrence of periodontitis and glycemic imbalance.

In view of the majority of differential compounds were not derived from SOM or POM bacteria and they cannot directly contribute to periodontal inflammation and hyperglycemia without microbial involvement; we then hypothesized that these compounds may regulate periodontal and glycemic status through oral microbes. To test this hypothesis, we added each of the 10 compounds most affected by GDM into the mono-culture of each of the six SOM or POM bacteria to ascertain whether these compounds promoted or hindered the growth of these bacteria (Fig. 5F, Fig. S11A). The growth curves of each bacterial strain in the presence of each compound were recorded. Compared to blank control without additional metabolites added to the medium, substances which are notably deficient in GDM patients, namely DHA and eicosapentaenoic acid (EPA), exhibited the most significantly suppressive effects on POM bacteria including P. intermedia and P. gingivalis (Fig. 5F). We then collected fresh human saliva, mixed thoroughly, and inoculated into medium supplemented with DHA. After 24 h of incubation, the regulatory effect of DHA on oral microbiota was verified by 16S rRNA gene amplicon sequencing. The results corroborated that DHA supplementation significantly suppressed the growth of Prevotella, while not significantly influenced the growth of Streptococcus (Fig. S11B, C). Based on these results, we conducted an in vitro experiment with six bacterial species isolated from human saliva to validate the suppressive effect of DHA on POM bacteria. We observed a reduction in the total number of bacterial cells in the simulated mixed culture system upon DHA supplementation (Fig. S11D). The relative abundance of P. intermedia decreased significantly, while that of S. salivarius remained unchanged (Fig. S11E, F). It strongly indicated that DHA consolidated the dominance of S. salivarius. To dissect the way in which DHA modulates the oral microbiota, we quantified DHA concentrations in the culture media of S. salivarius, P. intermedia, and P. gingivalis at 24 h post-inoculation. A significantly more pronounced reduction in DHA content was observed in media inoculated with S. salivarius than in media inoculated with P. intermedia and P. gingivalis (Fig. S11G), suggesting that S. salivarius may actively consume DHA, thereby attenuating the inhibitory effect of DHA on its growth. These findings mutually support the differential regulatory effects of DHA on the growth of S. salivarius and P. intermedia, where DHA tends to decrease the proportion of POM bacteria. GDM-related DHA deficiency may be a contributing factor to the imbalance of oral microbiota and its metabolism.

DHA improves periodontal and glycemic status of pregnant women with GDM

According to the specific inhibitory effect of DHA on periodontal pathogens, we finally applied for a randomized clinical intervention trial to validate the improvement effect of DHA on the periodontal inflammation and subsequently the glycemic status of pregnant women with GDM. We recruited 40 pregnant women with a clinical diagnosis of GDM by OGTT (Fig. 6A, Fig. S12A). Subjects were randomly assigned to the GDM and GDM + DHA groups, with comparable baseline blood glucose levels between the two groups (one-tailed t-test P = 0.3212). Women in the GDM + DHA group received daily gingival application of 1 ml of 100 mg/ml DHA, while those in the placebo group received 1 ml of water. Due to the relatively low compliance and limited feasibility of directly assessing periodontal inflammation-specific biomarkers (e.g., local inflammatory cytokines in gingival crevicular fluid) in pregnant women, we instead selected clinically accessible and standardized periodontal parameters including probing depth, attachment loss, gingival index, and plaque index, to reflect the improvement of periodontal inflammation. Through up to 6 weeks of continuous monitoring, we found that the DHA intervention significantly improved periodontal pocket probing depth (Fig. 6B, Fig. S12B, C). Although no significant effects were observed on plaque index, attachment loss, or gingival index, improvements were noted in the GDM + DHA group (Fig. S12D–F). To verify whether DHA improved periodontal inflammation status by remodeling the oral microbiota, we collected 116 GDM and GDM + DHA saliva samples over a 7-week period. Copy number of P. gingivalis, P. intermedia, and S. salivarius was quantified using qPCR (Fig. S12G–I, Table S6). The results showed a significant inhibition in the abundance of P. gingivalis and P. intermedia after gingival application of DHA (Fig. S12J, K), accompanied by a notable increase in the ratio of S. salivarius to P. gingivalis and P. intermedia (Fig. 6C, Fig. S12L). Such an intervention restored the population of beneficial bacteria in the mouth, specifically the dominant presence of S. salivarius. Continuous monitoring revealed a median increase in fasting blood glucose of 0.27 mmol/L (5.2% increase compared with baseline) in the GDM group versus 0.1 mmol/L (0.1% increase compared with baseline) in the GDM + DHA group compared to baseline, suggesting that DHA may attenuate fasting glucose elevation in this pilot RCT (Fig. 6D–F). These results suggest that topical DHA may help improve periodontal status and fasting glucose trajectory in pregnant women with GDM, although larger trials are required for confirmation.

Fig. 6 — A Experimental design of the clinical trial. 40 GDM pregnant women were recruited, with 20 of whom daily treated with DHA. Participants were randomly assigned in a 1:1 ratio to the placebo group or the DHA group, and received daily gingival application of 1 ml DHA solution (100 mg/ml) or 1 ml water for six consecutive weeks. Baseline measurements were obtained before intervention, and periodontal parameters, fasting blood glucose, and saliva samples were collected weekly throughout the trial. Periodontal probing depth and fasting blood glucose were the primary outcomes of this trial, whereas gingival index, attachment loss, and plaque index were secondary outcomes. B Integrated change in periodontal probing depth relative to baseline across follow-up visits in GDM and GDM + DHA pregnant women. The actual periodontal probing depth values at baseline and follow-up visits used to derive these baseline-relative changes are shown in Fig. S12B, C. C Change of the ratio of *S. salivarius* to *P. gingivalis* and *P. intermedia* after application of DHA compared with baselines. D Fasting blood glucose values of GDM and +DHA pregnant women from baseline through weekly follow-up visits. The height of the shade in each box represents the fasting glucose value. The red or green triangles in boxes represent the fasting glucose increased or decreased more than 0.1 compared with the primary value. Darker the triangles’ color, greater the variation of the fasting blood glucose compared with previously time points. E Baseline fasting blood glucose values before DHA application. F Overall change in fasting blood glucose during the intervention period relative to baseline, calculated by averaging the available follow-up measurements obtained after DHA application for each participant. For clinical plots, individual points represent participant-level measurements or participant-level changes from baseline with available follow-up data. For box plots, the center line indicates the median, box limits indicate the 25th and 75th percentiles, and whiskers extend to the smallest and largest values within 1.5× the interquartile range. For (B) GDM n = 27 and GDM + DHA n = 26 values were included. For (C) sample sizes were *S. salivarius*/*P. intermedia* ratio: GDM n = 35 and GDM + DHA n = 37; *S. salivarius/P. gingivalis* ratio: GDM n = 26 and GDM + DHA n = 30. For (E) GDM n = 17 and GDM + DHA n = 19 values were included. For (F) GDM n = 23 and GDM + DHA n = 23 values were included. Between-group comparisons were performed using one-tailed Welch’s t-tests unless otherwise specified. Exact t statistics, degrees of freedom, Cohen’s d values, 95% confidence intervals and exact P values are provided in the Source Data statistical reporting summary. Source data and exact statistical reporting are provided as Source Data files. **: P < 0.01; *: P < 0.05; n.s.: P > 0.05.

Discussion

In this study, we established prospective cohorts comprising over 2500 volunteers and longitudinally monitored the oral microbiome dynamics of 534 participants throughout pregnancy to investigate the etiology and reciprocal mechanisms underlying the comorbidity of GDM and periodontitis. We found that the decrease of the ratio of SOM/POM bacteria performed crucial role in mediating the GDM onset. Periodontal and systemic inflammatory response induced by POM bacteria triggered a further exacerbation of hyperglycemia symptoms via a dual pathway of hormones and stress. Remodeling of the oral microbiota and recovering the predominance of SOM bacteria in the oral cavity, especially S. salivarius, improved periodontal and glycemic status. This reveals a novel manner for microbe-mediated glycemic regulation. In addition to oral transplantation with SOM bacteria, our findings indicate that DHA, which is notably deficient in the GDM oral cavity, exhibits the most significant inhibitory effects on POM bacteria. The randomized clinical intervention trial showed that gingival application of DHA restored the dominant presence of S. salivarius and hindered the abundance of P. gingivalis and P. intermedia and effectively controlled the glycemia of GDM pregnant women.

Previous studies examining the oral microbiota of GDM patients have primarily relied on cross-sectional analyses conducted during mid-to-late pregnancy, after GDM diagnosis^28–30. This approach limits the ability to trace microbiota perturbations preceding disease onset. Based on a prospective cohort of a large population, in the present study, we found that GDM-associated shifts in the composition of the human oral microbiota actually happened in early pregnancy. Considering that lagging OGTT testing at 24–28 weeks of gestation negatively affects the prevention and timely treatment of GDM, the discovery of these signals from the oral microbiota may be of practical value for early screening, monitoring and diagnosis of glycemic imbalances during pregnancy, as is the case with the use of intestinal microbiome for prediction of glycemic responses^31,32. With regard to the outcomes of the oral dysbiosis, a common perception is that the oral cavity of patients with GDM is enriched in some periodontitis-associated bacteria^33,34. Although a wide variety of bacteria were mentioned, the most important periodontal pathogen, P. gingivalis, was absent^35,36. Our study points to a shift in the composition of the oral microbiota of patients with GDM, marked by the flourishing of the well-defined periodontal pathogens (e.g., Prevotella and Porphyromonas) and the wilting of the resident bacterium Streptococcus. Such a shift stimulates periodontal and systemic inflammations and further exacerbates hyperglycemia. The phenotype that hyperglycemia alters the oral microbiota to increase its periodontal pathogenicity has also been reported in mouse models of diabetes³⁵. Confusingly, the high levels of Enterococcus, Staphylococcus, and Aerococcus observed therein are all not typical periodontal pathogens. Differences from our findings may depend on the type of diabetes⁶, or may reflect differences in oral microbiota composition or response mechanisms between mice and humans.

Due to the unpredictable onset of diabetes and periodontitis, the underlying causes and pathways of their high concurrency remain confusing²¹. Pro-inflammatory cytokines such as tumor necrosis factor (TNF)-α, IL-1β and IL-6 impair insulin action^37,38, just like an increase in pro-inflammatory cytokines induced by periodontal pathogens enhances insulin resistance^39,40. One of the most striking inflammatory factors is IL-17, which is elevated in human periodontitis lesions and in arthritis and obesity caused by periodontitis^41,42. We found that inoculation of oral microbiota of GDM patient origin or periodontal pathogens into the oral cavity of mice elevated periodontal and blood levels of IL-17, which in turn inhibited GLP-1 expression in enteroendocrine cells and ultimately led to impaired pancreatic insulin secretion. In this way, insulin resistance seems to be a unidirectional process secondary to the inflammation. But there is another reverse route, i.e., diabetes-enhanced IL-17 alters the oral microbiota and renders it more pathogenic³⁵. The two-sided role of IL-17 further emphasizes the two-way relationship between periodontitis and hyperglycemia⁴³, implying that once one of the two diseases attacks, there is the potential for a vicious cycle of bidirectional aggravation. In addition to IL-17 mediating the dysfunction of GLP-1 secretion in pancreatic β-cells, the present study reveals a potent role of IL-1β in inducing blood glucose elevation through the stress pathway. IL-1β transcription is responsible for NF-κB, and local and systemic inflammation induced by periodontal pathogenic bacteria will lead to activation of the latter⁴⁴. This pathway is supported by our transcriptomic data, in which NF-κB-related genes were enriched in human and mouse cells under inflammatory factor stimulation. IL-1β can directly increase the levels of stress-related hormones such as cortisol and EPI, and these hormones were documented to have significant glucose-elevating effects⁴⁵. The elevation of blood glucose levels induced by inflammation through upregulation of EPI⁴⁶ and cortisol⁴⁷ in circulation often poses a challenge for dietary interventions to achieve significant improvement in blood glucose managements. Therefore, future research on the glucose-elevating response mediated by IL-1β may prioritize strategies aimed at inhibiting inflammation and reducing circulating levels of IL-1β.

The use of inflammatory inhibitors may be one approach to break the link between periodontal inflammation and hyperglycemia. For example, treatment with IL-17 antibody reduced periodontal pathogenicity in the oral microbiota of diabetic mice³⁵. Similarly, in our study, blood glucose levels were significantly downregulated immediately after we used DIM to reduce IL-1β abundance and block IL-17 binding to its receptor^48,49. However, the practical feasibility of such an intervention is uncertain and is particularly unsuitable for pregnancy. We confirm in clinical trials that DHA differentially regulates the growth of oral bacteria, thereby exerting anti-inflammatory and anti-hyperglycemic effects. Previously, omega-3 PUFAs such as DHA and EPA have been used to ameliorate complications associated with type II diabetes^50–52. A clinical double-blind, randomized controlled trial has demonstrated that supplementation of 4 g of omega-3 fatty acids to prescription medication on a daily basis can reduce levels of triglycerides and very low-density lipoprotein cholesterol in the bloodstream, while also significantly increasing high-density lipoprotein cholesterol levels⁵¹. The consumption of DHA and EPA is also significantly inversely associated with fasting blood glucose and HbA1c levels in the bloodstream. Taking 520 mg of DHA- and EPA-rich fish oil daily can effectively alleviate related symptoms of type 2 diabetes^53,54. Here we propose a temperate approach that distally prevents hyperglycemia from worsening through modulation of the oral microbiota. Glucose imbalance is ameliorated by using small amounts of DHA smearing on the gingiva, without the necessity for a large oral intake. For pregnant women, this is a friendlier approach than the one that relies on conventional protocols for treating periodontitis to achieve glycemic control⁵⁵. Although, a larger scale validation is required, it expands the application scenarios of omega-3 PUFAs. DHA and EPA are commonly used as daily supplements during pregnancy to maintain maternal and neonatal health because of their important effects including regulation of neural transmission, neural plasticity, and signal transduction impacting brain development and function^56,57, and maintenance of normal cognitive function⁵⁸. It should be noted that the recommended daily intake of DHA for pregnant women is 100–200 mg/day, and doses up to 1000 mg/day are considered safe⁵⁹. For pregnant women smeared of DHA on their gingiva combining of oral intake, the total intake of DHA should be quantified to avoid overdose.

Our findings reveal that reduced salivary DHA levels are associated with a shift toward a dysbiotic oral microbiome in GDM. DHA has known antimicrobial and anti-biofilm properties against pathogens such as P. gingivalis and F. nucleatum, likely via membrane disruption and oxidative stress^60,61. Differential susceptibility to DHA may drive the compositional shifts observed in GDM. We observed that DHA inhibited the growth of P. intermedia and P. gingivalis in vitro, but had less inhibitory effect on S. salivarius. Further, we quantified DHA concentrations in the culture media after 24-h incubation and found that DHA was significantly more depleted in S. salivarius cultures compared to P. intermedia and P. gingivalis (Fig. S9G), suggesting that S. salivarius may actively utilize DHA. This selective metabolism could attenuate DHA’s antimicrobial effect on S. salivarius while maintaining inhibitory pressure on POM-associated pathogens, contributing to community restructuring in favor of symbiotic taxa.

The observed reduction in salivary DHA concentrations among GDM patients is likely multifactorial and may reflect broader metabolic disturbances associated with gestational insulin resistance. It is well established that GDM is associated with impaired transport and altered metabolism of PUFAs, including DHA, particularly with respect to fetal delivery^62,63. Such dysregulation may also compromise the availability of circulating DHA for transfer to the oral cavity. Additionally, structural and functional alterations of the salivary glands have been reported in GDM, including reduced secretory output and disrupted acinar organization⁶⁴, which may further contribute to diminished DHA concentrations in saliva. Together, these factors, reduced systemic DHA mobilization and impaired glandular secretion, may mechanistically explain the observed depletion of salivary DHA in GDM and its downstream effects on the oral microbiota.

GDM may not be exactly the same as other types of diabetes due to the influence of factors such as hormones⁶. Similarly, there may be differences between men and women. Nevertheless, our intervention trials in male/female and non-pregnant/pregnant mice clearly demonstrated that oral microbiota remodeling effectively ameliorates glycemic imbalance induced by periodontal inflammation, independent of sex or pregnancy status. This suggests that such an intervention strategy of downregulating local and systemic IL-17 and/or IL-1β levels by targeting oral microbiota may also be applicable to dysglycaemia in the non-pregnant phase. One of the regrets of this study is that only DHA was used in the clinical intervention. DHA not only has antibacterial properties^60,61,65, but also inhibits the inflammatory response through activation of the G protein-coupled receptor GPR120 on the plasma membrane of inflammatory cells, leading to a reduction in NF-κB activation and inflammatory cytokine production^66,67. So even though we detected a higher proportion of Streptococci and fewer periodontal pathogens in the mouths of pregnant women with GDM whose glycaemia was controlled after the DHA intervention, we could not determine the order in which inflammation and periodontal pathogens were suppressed on this basis. Limited by the riskiness and ethics of conducting microbial transplantation experiments in humans, we did not use live S. salivarius bacteria for clinical interventions. Alternatively, we provide evidence in animal experiments that the use of S. salivarius in the oral cavity also ameliorates glycemic imbalances induced by periodontal inflammation, thus certifying the dominant role of reshaping the oral microbiota. A further limitation of the antibiotic-based transplantation experiments is that we did not directly quantify the extent of oral microbiota depletion induced by antibiotic pretreatment, nor resolve which oral niches were most affected. Therefore, our antibiotic-based preconditioning should be interpreted as a microbiota-reduction strategy rather than a niche-specific depletion model. Another limitation of this study is the relatively small sample size and the lack of long-term follow-up, which restricts the ability to more firmly confirm the effects of glycemic control and to assess potential side effects of DHA on pregnant women and their infants. The sample size was planned to detect a 0.5 mmol/L between-group difference in fasting blood glucose based on preliminary cohort-derived glucose data; however, the observed difference in this independent RCT was smaller than this prespecified value, indicating reduced power for detecting modest glycaemic effects. The clinical trial should therefore be interpreted as a preliminary proof-of-concept study. Although the analysis followed the prespecified SAP and focused on baseline-relative delta values, this approach did not fully model time-by-treatment interactions. In addition, fasting blood glucose was the only glycemic endpoint assessed. Further larger trials using more comprehensive glycemic outcomes and prespecified longitudinal treatment-by-time analyses are needed to validate these findings.

Although the primary objective of our clinical trial was to assess DHA efficacy on periodontal inflammation and subsequently on glycemic status, we elected to use routine, widely accepted clinical periodontal indices as the trial endpoints because directly measuring molecular biomarkers of periodontal inflammation proved less feasible in pregnant women (poor compliance and limited bedside operability). We acknowledge that this choice may appear to conflate two related but distinct concepts (inflammation vs irreversible tissue destruction). However, probing depth and attachment loss are established, consensus clinical measures used to define and stage periodontitis in epidemiological and clinical research, and therefore capture the structural manifestations of disease that follow from sustained periodontal inflammation⁶⁸. In our randomized trial, a statistically significant improvement in probing depth was observed following gingival DHA application, indicating that the subclinical chronic inflammation commonly present in pregnant women with GDM was significantly ameliorated by DHA. For gingival index and plaque index, which are commonly used as validated indices to assess gingival inflammatory status and plaque accumulation, improvements were observed in the same direction. However, these changes did not reach statistical significance, likely due to the relatively short duration and small sample size of the trial (N = 40), as well as the fact that all participants with GDM were free from overt periodontitis at baseline.

Benefiting from the fact that the hyperglycemic symptoms of GDM debut during pregnancy, in this study, the resolution of the oral microbiome and metabolome from hundreds of women throughout pregnancy allows us to recognize the importance of oral microbiota in the bidirectional relationship between glucose metabolism disorders and periodontal inflammation. We elucidate the mechanism of cross-radicalization between periodontal inflammation and glucose metabolism imbalance. Responses and feedback from the oral microbiota are linked to this matter. A specific oral microbial consortium consisting of several periodontal pathogens, which we define as POM, is labeled as a key mediator. We also demonstrate that the overgrowth of POM bacteria and oral dysbiosis backward stimulated the rise of blood glucose. Finally, a complete chain starting from oral dysbiosis, to regional and systemic inflammatory activation, and then stepwise to hyperglycemia exacerbation is established. By restoring oral microbiota through microbial transplantation or targeted regulation, hereby we mark a target and new way to break the link between complications to alleviate hyperglycemia and prevent further deterioration of the disease.

Methods

Ethics statement

All research conducted in this study complied with all relevant ethical regulations. The human cohort study was approved by the Ethics Committee of Wenzhou People’s Hospital (IRB approval No. WRY2021-215) and the Women and Children’s Hospital of Chongqing Medical University (IRB approval No. 20172601), and was conducted in accordance with the World Medical Association Declaration of Helsinki. All participants provided written informed consent for participation in the study and for the use of their samples and data in this paper. The randomized controlled trial was approved by the Ethics Committee of Wenzhou People’s Hospital and was prospectively registered in the Chinese Clinical Trial Registry (ChiCTR2400080741). All animal experiments were approved by the Laboratory Animal Welfare and Animal Experimental Ethics Committee of China Agricultural University.

Human participants, sex and gender reporting

All human participants included in this study were pregnant women. Sex and pregnancy status were determined based on obstetric medical records and participant self-report at recruitment. Because GDM occurs during pregnancy, the human cohort and randomized controlled trial were designed specifically in pregnant women; therefore, no male participants were enrolled and no sex-disaggregated comparison was performed for the human clinical analyses. Participant age and clinical characteristics for the cohort study and randomized controlled trial are provided in Tables S2, S7 and S8 respectively.

In the prospective cohort study, 2523 pregnant women were initially recruited from two hospitals, and 534 pregnant women were included in the longitudinal oral microbiome analyses after eligibility assessment, consent, sample availability and sequencing quality control. These included 379 participants in C_Dis, 29 participants in C_Int and 104 participants in C_Val. In the randomized controlled trial, 40 pregnant women with GDM were enrolled and randomly assigned to the placebo group or DHA group, with 20 participants per group. Written informed consent was obtained from all participants for study participation and for the use of their samples and data in this paper. Participants did not receive financial compensation, but study-related sample collection and clinical assessments were provided at no cost.

Prospective cohorts

Totally 2523 pregnant women in three cohorts were recruited from two hospitals, Wenzhou People’s Hospital/Wenzhou Maternal and Child Health Care Hospital and the Women and Children’s Hospital of Chongqing Medical University. Recruitment, clinical data collection and saliva sampling were performed during routine prenatal visits according to predefined protocols. All participants recruited in these cohorts were of Han Chinese ethnicity. Pregnant women with GDM were diagnosed according to the diagnostic guideline of International Association of Diabetes and Pregnancy Study Groups (IADPSG). Fasting plasma glucose ≥ 5.1 mmol/L or 1-h blood glucose ≥ 10.0 mmol/L or 2-h blood glucose ≥ 8.5 mmol/L after 75 g OGTT are the diagnostic threshold⁶⁹.

Participants enrolled in the cohort study met the following inclusion criteria: pregnant women aged between 25 and 45 years, with a body weight of at least 45 kg and a pre-pregnancy body mass index (BMI) between 18.0 and 28.0 kg/m² (inclusive). All participants were in good physical and mental health, with no history of major diseases, drug allergies, systemic or infectious diseases (including periodontitis, pneumonia, sexually transmitted infections, or urinary tract infections), mental disorders, or pregnancy-related complications other than gestational diabetes (e.g., preeclampsia). Additionally, none had recently used hypoglycemic medications. Participants were excluded if they:¹ had participated in other drug clinical trials within 1 month prior to enrollment;² had pregnancy complications other than gestational diabetes or major systemic diseases;³ reported alcohol consumption or smoking during the first 6 months of pregnancy; or⁴ had a history of illegal drug use, or had taken probiotics or antibiotics within 1 month prior to enrollment.

Salivary specimens of the human participants were sampled by trained professionals under strict aseptic conditions and a uniform protocol. All specimens were placed in sterile tubes or vials, immediately frozen upon collection at −20 °C, and then transported to the laboratory and stored at −80 °C until further analysis.

Cohort of C_Dis. In this cohort, we recruited 379 pregnant women and collected salivary specimens in Trimester 1 (~12 weeks of gestational ages), from August 2014 to May 2015. GDM screening according to the results of OGTT was performed in the Trimester 2 (~24 weeks of gestational ages) with 181 subjects were classified to GDM and 198 classified to healthy control (CON) group. Meanwhile, 31 GDM and 29 CON subjects were picked out for Trimester 2 saliva specimens’ collection respectively.

Cohort of C_Int. Salivary specimens of 29 pregnant women were serially collected in gestational age of 12, 14, 16, 18, and 22/24 (16 samples in week 24, 13 samples in week 22) weeks, from April 2022 to November 2022. OGTT was conducted in week 24. Among of the 29 pregnant women, 10 women were diagnosed with GDM and 19 women were without GDM.

Cohort of C_Val. We established a prospective cohort including of 104 pregnant women in their early gestational ages (~12 weeks) with the saliva specimens collected, from November 2021 to February 2022. These included 27 pregnant women were diagnosed with GDM and 77 pregnant women without GDM.

Randomized controlled trial

Our clinical study was prospectively registered in the Chinese Clinical Trial Registry on February 6, 2024 (ChiCTR2400080741; https://www.chictr.org.cn/showproj.html?proj=215373), before participant recruitment began. The randomized controlled trial was conducted at Wenzhou People’s Hospital according to the preregistered study protocol (version 3.0, December 28, 2023) and statistical analysis plan (version 2.0, January 10, 2024), which are provided as Supplementary Methods 1 and 2, respectively. No deviations were made from the preregistered intervention, randomization procedure, primary and secondary outcomes, sample size plan, or prespecified delta-value analysis.

Power analysis based on a two-sample t-test (α = 0.05, β = 0.20) and preliminary glucose data (mean = 4.95 mmol/L, SD = 0.61, N = 248) indicated that 20 participants per group (N = 40 total) were needed to detect a 0.5 mmol/L difference (Cohen’s d = 0.78) with 80% power. Forty pregnant women diagnosed with GDM based on OGTT in Trimester 2 was enrolled in this study. The inclusion and exclusion criteria for participants are detailed in Supplementary Methods 1. The recruitment began on February 8, 2024. After enrollment, the trial for the participants commenced the following morning. Subsequently, 40 pregnant women were randomly assigned to either the placebo group or DHA group (n = 20 per group) based on the envelopes they selected. Thereafter, participants received daily applications of either DHA or a placebo on the gingiva for a duration of 6 weeks. The study was conducted in a double-blind manner, with both participants and research staff responsible for data collection blinded to group allocation. The primary outcomes of this trial were periodontal probing depth and fasting blood glucose, whereas the secondary outcomes included gingival index, attachment loss, and plaque index. Periodontal parameters including probing depth, attachment loss, gingival index, and plaque index were measured as indicators of periodontal inflammation status. Both periodontal parameters and blood glucose levels were measured prior to treatment initiation and on a weekly basis throughout the trial period. Salivary samples were also collected concurrently. All these data and samples collection were performed by professional nursing staff of Wenzhou People’s Hospital. All specimens were placed in sterile tubes or vials, immediately frozen upon collection at −20 °C, and then transported to the laboratory and deposited at −80 °C until further use. For statistical analysis of the trial outcomes, greater emphasis was placed on the changes at each follow-up visit relative to baseline. One-tailed Student’s t-test was used for between-group comparisons, with P < 0.05 considered statistically significant. More clinical trial details and statistical analysis plan are in Supplementary Methods 1 and Supplementary Methods 2. Baseline characteristics of the randomized trial arms are summarized in Table S8, and participant-level baseline data are provided in Table S7.

Animal experiments

All animal experiments were approved by China Agricultural University and were performed in accordance with the approved institutional animal care and use protocols. Female, male and pregnant specific-pathogen-free C57BL/6J mice were used as indicated below. Unless otherwise stated, female SPF C57BL/6J mice aged 4 weeks and weighing 22 ± 2 g were used in the primary murine mechanistic experiments. Mice were purchased from HFK BIOSCIENCE (Beijing, China; strain code 11001A). Animals were maintained under controlled conditions with a 12 h light/12 h dark cycle, temperature of 22 ± 3 °C and humidity of 50 ± 20%, and were fed an AIN-93G diet.

Sex was considered in the study design. Because this study focused on GDM and pregnancy-associated oral dysbiosis, the primary mechanistic mouse experiments were performed mainly in female mice. To assess whether the glycaemia-modulating effect of oral microbiota remodeling was influenced by sex or pregnancy status, an additional validation experiment was performed in male, female and pregnant mice, with data analyzed and reported separately by sex/pregnancy status. Disaggregated data for individual animal experiments are provided in the Source Data files.

In the experiments using antibiotics, a combination of antibiotics (ABX) containing 100 µg/ml neomycin, 100 µg/ml penicillin, 50 µg/ml vancomycin, and 100 µg/mL metronidazole (Sigma, St. Louis, MO) were dissolved in distilled water and administered to C57BL/6J mice via ad libitum drinking water to reduce the amount of endogenous bacteria⁵⁴. Before salivary, metabolic or microbial transfer, ABX drinking was provided to the animals for 7 consecutive days then change to clean water without ABX and microbes.

In the salivary transfer experiment, 40 female SPF C57BL/6J mice aged 4 weeks were used and randomly assigned to receive saliva from CON subjects or GDM subjects, with 20 mice per group. 10 ml fresh saliva was collected from 10 GDM and 10 CON subjects respectively. After vigorously vortex, CON and GDM saliva were diluted with equal volume of saline and subpackaged to sterile 1.5 ml tubes with 500 μl of each. Diluted saliva stored in −80 °C until transfer to mice. One day before the salivary transfer, periodontitis model was constructed as described in the “Periodontitis model” section⁷⁰. Forty mice were used in this experiment with 20 mice received saliva from CON subjects and the others received saliva from GDM subjects. Salivary transfer was conducted every morning with 20 μl diluted saliva for each mouse using pipette to oral cavity and lasted for 21 days. RBG was detected in 7 days and 14 days. OGTT was performed in 21 days. Then the mice were sacrificed, and blood and maxillary tissues were collected.

In the microbial gastric infusion experiment, 40 female SPF C57BL/6J mice aged 4 weeks were randomly assigned to four groups: PBS control, Streptococcus salivarius, Prevotella intermedia and Porphyromonas gingivalis, with 10 mice per group. Representative species of Streptococcus, Prevotella, and Porphyromonas (Streptococcus salivarius, Prevotella intermedia, and Porphyromonas gingivalis) were administered intragastrically to mice. These three bacteria were incubated in an anaerobic incubator at 37 °C until OD600 reaches 0.6–0.8. The cultures were then centrifuged (3500 × g for 10 min), and the bacteria in the pellet were washed three times in PBS with centrifugation at 3500 × g for 5 min each time. The bacteria were then resuspended in PBS to prepare a fresh solution for gavage. Forty mice were randomly assigned to four groups with three microbial gavage groups and 1 control group. Mice in the microbial gavage group were gavaged with 0.2 ml fresh bacterial solution (1 × 10⁹ microbial cells) daily. Mice in the control group received equal volume PBS in the same manner and consecutive ABX drinking for the whole experiment. This experiment lasted for 14 days with RBG of each mouse conducted at 7 days and 14 days.

In the oral metabolite infusion experiment, 30 female SPF C57BL/6J mice aged 4 weeks were randomly assigned to receive fructose, oxoglutarate or water control, with 10 mice per group. Infusion of metabolites was conducted every morning with 20 μl of each mouse using pipette to oral cavity and lasted for 21 days. Concentrations of these metabolites were set according to the results of targeted metabolomics analysis of saliva and were 100× of the concentrations in saliva. RBG was detected in 7 days and 14 days. OGTT was performed in 21 days. Then the mice were sacrificed and blood and maxillary tissues were collected.

In the oral microbiota consortium infusion experiment, 120 female SPF C57BL/6J mice aged 4 weeks were randomly assigned to CON, POM and POM + SOM groups, with 40 mice per group. For CON group, the mice received PBS gargle and consecutive ABX drinking. For POM group, the mice received bacteria cocktail gargle consisting of representative species of POM (P. intermedia, Neisseria flava, Rothia dentocariosa, and P. gingivalis). For POM + SOM group, the mice received bacteria cocktail gargle consisting of representative species of SOM (S. salivarius and Gemella morbillorum) and POM (P. intermedia, N. flava, R. dentocariosa, and P. gingivalis). Total number of microbial cells in the bacteria cocktail gargle of POM and POM + SOM were similar (~1 × 10⁹). The ratio of the species in the bacteria cocktail gargle was set based on the results of 16S rRNA amplicon sequence analysis of C_Dis cohort. One day before the oral perfusion, periodontitis model was constructed as described in the “Periodontitis model” section. RBG was detected in 7 days and 14 days. OGTT was performed in 21 days. Then the mice were sacrificed and blood and maxilla were collected. For the mice of POM and POM + SOM, tissue of ileum and pancreas and colonic contents were also collected.

In the intraperitoneal IL-17 and IL-1β injection experiment, 70 female SPF C57BL/6J mice aged 4 weeks were randomly assigned to seven groups: vehicle control, three IL-17 dose groups and three IL-1β dose groups, with 10 mice per group. After acclimatization for 1 week in cages, six groups were intraperitoneally injected with three dosages of IL-17 and IL-1β (0.1, 0.5, and 1 μg) respectively. The CON group was intraperitoneally injected with equal volume of vehicle of the cytokines (50 μl, sterile water). After injection, the diets were removed immediately and fasting blood glucose was detected 24 h later. Then the mice of CON, 1 μg IL-17, and 1 μg IL-1β were sacrificed and blood and tissue of ileum were collected.

In the DIM intervention experiment, 19 female SPF C57BL/6J mice aged 4 weeks were used. Seven mice received PBS as controls and 12 mice received Prevotella intermedia. In the second phase, the 12 Prevotella intermedia-treated mice were split into Prevotella intermedia and Prevotella intermedia + DIM groups, with six mice per group. The experiment of intraperitoneal injection with 3,3’-diindolylmethane (DIM) including two phases. In phase I, 12 mice received Prevotella intermedia (1 × 10⁹ microbial cells in 20 μl PBS) gargle daily, seven CON mice received equal volume of PBS gargle with consecutive ABX drinking. In phase II, seven CON mice received same conditions of that in phase I, and 12 mice inoculated with Prevotella were randomly split to two group. One group was intraperitoneally injected with 25 mg/kg BW DIM. The other group was intraperitoneally injected with equal volume vehicle (100 μl, including 5% DMSO, 40% PEG300, 5% Tween-80, 50% Saline). Phase one lasted for 13 days, phase two lasted for 4 days. One day before phase I, periodontitis model was constructed as described in the “Periodontitis model” section. RGB was detected every 2 days and all the mice were sacrificed and blood samples were collected.

In the oral S. salivarius remodeling experiment, 40 female SPF C57BL/6J mice aged 4 weeks were used. Ten mice received PBS control, 20 mice received Prevotella intermedia and 10 mice received Porphyromonas gingivalis during the first phase. During the second phase, 10 Prevotella intermedia-treated mice and all 10 Porphyromonas gingivalis-treated mice received Streptococcus salivarius for oral microbiota remodeling. Phase I lasted for 10 d and phase II lasted for 11 days. One day before phase I, periodontitis model was constructed as described in the “Periodontitis model” section. Two groups received P. intermedia gargle, one group received P. gingivalis gargle, and one group received equal volume of PBS gargle with consecutive ABX drinking. In phase II, the Porphyromonas group and one of the Prevotella group were changed to S. salivarius gargle. The other two groups received same conditions of that in phase I. All the bacterial gargle consisting of 1 × 10⁹ microbial cells in 20 μl PBS. RBG was evaluated in the 7 days, 10 days in phase 1 and 7 days, 9 days, and 11 days in phase 2. Then all the mice were sacrificed and blood samples were collected.

In the sex- and pregnancy-status validation experiment, 60 SPF C57BL/6J mice were used, including male, female and pregnant mice. Mice were assigned to POM or POM + SOM groups within each sex/pregnancy-status category, with 10 mice per group. The animals were treated with POM or POM + SOM with six groups and ten mice in each group. The content of the gargle of POM and POM + SOM and the construction of periodontitis model as described in the “Periodontitis model” section. RBG was detected in 7 days and 14 days. OGTT was performed in 21 days. The mice then were sacrificed and blood was collected. For the mice of POM and POM + SOM, tissue of ileum and liver were also collected.

Periodontitis model

The mouse periodontitis model was established using sterile silk ligatures. Mice were anesthetized, and sterile silk sutures were gently placed bilaterally around the maxillary second molars to promote local plaque accumulation and periodontal inflammation. The ligatures were checked daily throughout the indicated experimental period. If a ligature became detached, it was replaced promptly to maintain continuous modeling. At the endpoint, maxillary tissues from both sides were collected for MicroCT, immunohistochemistry or immunofluorescence analyses.

Bacterial cultivations

Five in vitro bacterial culture experiments were conducted. All the cultivations were performed with Gifu Anaerobic Medium (GAM, Qingdao Hi-Tech Industrial Park Hope Bio-Technology Co., Ltd.) at 37 °C. Six representative species isolated from saliva were used to mimic the consortium of SOM (S. salivarius, G. morbillorum) and POM (P. intermedia, N. flava, R. dentocariosa, and P. gingivalis). For the bacterial consortium cultivation, the ratio of the species inoculated was set based on the results of 16S rRNA amplicon sequence analysis of C_Dis cohort. In SOM group, the ratio of S. salivarius and G. morbillorum was around 90:10. In POM group, the ratio of P. intermedia, N. flava, R. dentocariosa, and P. gingivalis was round 31:30:24:15. Both the fermentation of SOM and POM were repeated in six tubes. After anaerobic fermentation at 37 °C for 24 h, the broth of SOM and POM were collected. Three tubes of germ-free medium were synchronously put in 37 °C incubator and collected after 24 h as control. All these samples were frozen in −80 °C immediately after collection for untargeted metabolomics analysis.

Effect of the five most significantly increased (fructose, glycerate, ketovaline, oxoglutarate, (S)-OMV) and five most significantly decreased (glutamine, DHA, pyruvate, EPA, DPAn-6) salivary metabolites in GDM compared with CON on the growth of six bacterial species (S. salivarius, G. morbillorum, P. intermedia, N. flava, R. dentocariosa, and P. gingivalis) was evaluated using bacterial growth curves. Four concentration gradients (CON/0, 10×, 100×, and 1000×) were set according to the concentration of each metabolite in saliva detected using targeted metabolomics analysis. The single and pure colonies of these six species were pre-cultured into GAM for growth until OD600 reaches 0.6–0.8. One hundred microliters of each pure bacterial suspension were cultured to 10 ml GAM with different metabolites until the variations of OD600 reached plateau. All experimental conditions were tested triple. During the microbial cultivation, tubes were inverted to resuspend cultures and small aliquots were transferred multiple times to clean microplate. OD600 measurements were taken using a microplate spectrophotometer (Epoch 2, BioTek Instruments, Inc., Winooski, Vermont, USA) to assess the growth of each species.

The impact of DHA on the growth of these six representative species was evaluated using salivary cultivation and mixed bacterial cultivation, respectively. For salivary cultivation, six healthy people were enrolled for salivary collection with around 5 ml saliva of each people. The saliva was pooled together and vortexed vigorously for 30 s. Then 200 μl fresh saliva was inoculated to 10 ml GAM with or without DHA. For the mixed cultivation, all the six bacterial species were mixed inoculated into 10 ml GAM included or not included DHA with a ratio of 57:6:12:11:9:6. Both salivary cultivation and mixed bacterial cultivation were conducted at 37 °C in anaerobic conditions for 24 h and then the impacts of metabolites on the bacterial growth were evaluated using 16S rRNA sequencing. In the mixed bacterial cultivation, the number of bacterial cells per unit volume of broth was counted using optical microscope.

In the last bacterial cultivation, we inoculated S. salivarius, P. intermedia, and P. gingivalis into the medium with DHA added to test the metabolic potential of DHA by these bacteria. The cultivation lasted for 48 h with three replications for each species. The control was set with same dose DHA but without bacteria. Then the change of DHA content in the broth was measured using UPLC-MS/MS.

Cell cultures

STC-1 and NCL-H716 cell lines were used for the validation of the negative regulating effect of IL-17 and IL-1β on the expression of GLP-1 in enteroendocrine cells. STC-1 (BNCC342403) was obtained from Benai Chuanglian Biology Research Institute (Beijing, China). NCL-H716 (TCHu210) was obtained from CAS Center for Excellence in Molecular Cell Science (CEMCS, Shanghai, China). All cell lines were used after confirming that they were Mycoplasma (-) after Mycoplasma testing with the PCR Mycoplasma Detection Kit (TaKaRa, Shiga, Japan) according to the manufacturer’s instructions. The medium for STC-1 was 90% DMEM-H + 10% fetal bovine serum (FBS) (NEWZERUM, Christchurch, New Zealand), for NCL-H716 was 90% RPMI1640 + 10% FBS. Both these two mediums were purchased from Solarbio (Beijing, China). STC-1 and NCL-H716 cells were seeded in 6-well culture plates (5 × 10⁴ cells per well) and were maintained at 37 °C in a humidified atmosphere of 5% CO₂ until to 80% confluence. Cells were then treated with 0, 80, and 160 ng/ml IL-17 or IL-1β according to the purpose of experiment, with three replicates for each treatment. After treatment, the cells were washed with PBS and then collected. GLP-1 protein expression was detected by western blotting using the procedure described in the “Western blot” section.

Oral glucose tolerance test (OGTT) of mice

For OGTT, 16-h fasted mice were weighed and a baseline blood glucose concentration was measured. Then mice were administered 2 g/kg BW glucose in sterile water by gavage through a gastric needle inserted into the stomach. Blood glucose concentrations were measured at 60 min and 120 min after glucose administration using the blood from tail ends.

Micro-CT analysis

The maxillary tissues were fixed with 4% paraformaldehyde for micro-computed tomography (CT) scanning using an Inveon™ Micro-CT instrument (Siemens Medical Solutions) for which the tube voltage and rotation center had been calibrated before use. The parameters were set as follows: Voltage 60 kV, current 220 μA, exposure time 1500 ms, effective pixel value 8.89 μm 360 °rotation, every 1 ° exposure. The region of interest (ROI) was set from 0.8 to 1.3 mm below the growth plate. Radiographs and the ratio of bone volume to total volume (BV/TV) were reconstructed and analyzed with Inveon Research Workplace software. The bone loss height was measured from the CEJ to the ABC at the mesial and distal sites of the second molars.

Blood parameters assays

The mice blood was centrifuged at 3500 × g for 15 min, and the serum were harvested. The levels of blood parameters, including TNF-α, IL1β, IL-17, CRP, GLP-1, INS, GHb, Cortisol, and EPI, were determined using ELISA kits (Gene-Lab, Beijing, China) according to the manufacturers’ guidelines. The blood parameters of CAT, SOD, MDA, and GPx were detected using the relevant kits from Beyotime (Beyotime Biotech. Inc.).

Immunohistochemistry

Maxillary tissues were fixed in 4% paraformaldehyde. Tissue slices were dehydrated and embedded in paraffin. Endogenous peroxidase activity was blocked by 30-min incubation in 0.3% H₂O₂ in methanol. Sections were then blocked with 3% BSA for 1 h and incubated overnight at 4 °C with primary antibodies diluted in 3% BSA, including rabbit anti-IL-17A antibody (Abcam, ab302922, clone EPR26410-51; 1:500) and rabbit anti-IL-1β antibody (Abcam, ab2105; dilution 1:400). Sections were subsequently incubated with HRP-conjugated goat anti-rabbit IgG secondary antibody (Servicebio, GB23303; dilution 1:200). DAB (3,3’-diaminobenzidine) and hematoxylin was used to label the positive color and restain nuclei. The positive color was brown and yellow, nuclei was blue. The results are interpreted under a white light microscope. Image-Pro Plus (v.6.0) software was utilized for Integrated Optical Density (IOD) calculation by summing the pixel intensity values within a representative region of interest (ROI) in each image. The areal density was subsequently determined using the formula: areal density = integrated optical density value (IOD)/tissue area.

Immunofluorescence

Tissue fixation, dehydration, embeddedness, and the primary antibodies were the same with immunohistochemical analysis. A microwave in a 10 mM citrate antigen was adopted to repair solution for antigen retrieval. Sections were blocked with 3% BSA and incubated overnight at 4 °C with rabbit anti-IL-17A antibody (Abcam, ab302922, clone EPR26410-51; 1:500) or rabbit anti-IL-1β antibody (Abcam, ab2105; dilution 1:800). Detection and labeling were performed using secondary antibodies conjugated to Alexa Fluor-488 fluorophores (Servicebio, GB25303; dilution 1:400). Microscopy detection and collect images by Fluorescent Microscopy.

Western blot

Proteins from mouse ileum tissues were extracted using RIPA lysis buffer (Beyotime Institute of Biotechnology, Shanghai, China) supplemented with protease and phosphatase inhibitors. Protein concentrations were determined using a bicinchoninic acid assay kit (Solarbio, Beijing, China). Protein samples were mixed with loading buffer, denatured at 95 °C for 5 min, separated by 12% SDS-PAGE and transferred to 0.45-μm PVDF membranes. After blocking with 5% non-fat milk in TBST for 2 h at room temperature, membranes were incubated overnight at 4 °C with rabbit anti-GLP-1 antibody (Abcam, ab200474, clone 24H1L3; 1:250) or anti-β-actin antibody (Abcam, ab252556, clone BLR057F; 1:5000). β-actin was used as the loading control. After washing with TBST, membranes were incubated with HRP-conjugated goat anti-rabbit IgG secondary antibody (Bioss, bs-0295G-HRP; dilution 1:5000) for 1 h at room temperature. Signals were detected using enhanced chemiluminescence reagent and imaged with a chemiluminescence imaging system. Protein band intensities were quantified using ImageJ software (v.1.54 g; National Institutes of Health, Bethesda, MD, USA). Lot numbers were not recorded and are indicated as “not recorded” in the Reporting Summary.

Real-time quantitative PCR for mRNA

Regulating effect of IL-17 and IL-1β on the expression of GLP-1 in STC-1 and NCL-H716 were determined by RTPCR. GAPDH was used as the internal reference for both STC-1 and NCL-H716. The primers for real-time qPCR were synthesized by Sangon Biotech (Shanghai, China) and are listed in Table S6. Total RNA from the cultured cells was isolated by the TRIzol method and then reverse transcribed into cDNA using a PrimeScript RT kit with the gDNA Eraser (RR047A, TaKaRa, Japan). Real-time qPCR was performed using SYBR® Premix Ex Taq (RR420, TaKaRa, Japan) and the QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific, USA). Experimental data were analyzed using the 2^−ΔΔCt method.

DNA extraction

For each salivary specimen, approximately 0.2 ml saliva and equal volumes of PBS and Qiagen’s AL buffer were mixed and vortexed thoroughly. Total DNA was extracted from the suspension of each sample type using QIAamp DNA Mini Kit (Qiagen, Valencia, CA). DNA quality and concentrations were assessed with 1.0% agarose gel electrophoresis and a NanoDrop® ND-2000 spectrophotometer (Thermo Scientific Inc., USA) before downstream processing.

16S rRNA gene amplicon sequencing and data preprocesses

We amplified hypervariable region V3-V4 of the 16S rRNA gene using primer pairs 338 F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806 R (5′-GGACTACHVGGGTWTCTAAT-3′) by an ABI GeneAmp® 9700 PCR thermocycler (ABI, CA, USA). Purified amplicons with different index sequences were pooled in equimolar amounts and paired-end sequenced on an Illumina PE250 platform (Illumina, San Diego, USA) according to the standard protocols by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China).

Raw sequencing reads of the 16S rRNA gene sequences were quality filtered and trimmed using QIIME2 (v.2021.11). The trimmed paired reads were merged and denoised using the DADA2 algorithm. Taxonomy assignment was performed with the SILVA V138 ribosomal reference database⁷¹ using a previously trained classifier specifically for the V3-V4 region-specific primers (338F-806R) and clustered at 99% sequence similarity. Two samples from the C_Dis cohort with sequence counts lower than 10000 were discarded from the following analysis.

The amplicon sequence variants (ASVs) with very small counts (lower than 5) in 20% samples and low variance based on inter-quantile range (IQR)⁷² were removed to improve downstream statistical analysis. The relative abundance of each ASV was normalized to the total abundance of all ASVs from each sample. Rarefaction and Good’s coverage evaluation was evaluated based on the normalized ASVs table, followed by alpha diversity (Shannon and Simpson) calculation using R package vegan (https://github.com/vegandevs/vegan).

Metagenomic sequencing and data preprocesses

Thirty-three salivary specimens collected from Trimester 3 of C_Dis (GDM: 16, CON: 17) with sufficient material and DNA quality for metagenomic library construction were included for metagenomic sequencing. Metagenomic DNA libraries was constructed using NEXTFLEX Rapid DNA-Seq (Bioo Scientific, Austin, TX, USA). Paired-end sequencing was performed on Illumina Novaseq 6000 (Illumina Inc., San Diego, CA, USA) according to the manufacturer’s instructions. KneadData (v. 0.7.7) software was used for quality control (based on Trimmomatic) and de-hosting (based on Bowtie2) of raw reads. The parameters for Trimmomatic were set as default. Abundance of KEGG Genes, KEGG Orthology, KEGG Enzyme, and KEGG Pathway was estimated using the BWA (v.0.7.17) algorithm for short reads mapping⁵ and SAMtools (v.1.15) for format transformation and alignment summary statistics⁶. eggNOG mapper (v.2.1.2) was used for Gene Ontology annotation.

Targeted metabolomics measurement

Targeted metabolomics was performed on 72 saliva samples, including 21 late-pregnancy T3 samples (GDM, n = 12; CON, n = 9) and 51 early-pregnancy T1 samples (GDM, n = 30; CON, n = 21). Each participant-derived saliva sample was treated as one biological replicate. Internal standards, calibrators, pooled quality-control samples and blank samples were included across the analytical run for quality control and metabolite quantification. Briefly, 120 μl of each saliva sample was lyophilized, reconstituted with 20 μl methanol, and mixed with 120 μl ice-cold internal standard solution using an Eppendorf epMotion Workstation (Eppendorf Inc., Hamburg, Germany). After centrifugation at 4000 × g for 30 min using an Allegra X-15R centrifuge (Beckman Coulter, Indianapolis, IN, USA), 30 μl of supernatant was mixed with 20 μl freshly prepared derivatization reagent. After derivatization, 330 μl ice-cold 50% methanol was added to dilute the sample. Targeted metabolites were quantified using an ultra-performance liquid chromatography-tandem mass spectrometry system (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA, USA). Peak integration, calibration and quantification were performed using MassLynx software (v.4.1, Waters). Metabolite concentrations were calculated based on calibration curves and internal-standard normalization.

Untargeted metabolomic measurement

Untargeted LC-MS/MS metabolomics was performed on in vitro bacterial culture samples, including six independent SOM culture replicates, six independent POM culture replicates and three germ-free medium controls. A pooled quality-control sample was prepared by mixing equal volumes from all extracted samples and was injected at regular intervals throughout the analytical sequence to monitor instrument stability and analytical reproducibility. Briefly, 100 μl of culture supernatant was mixed with 400 μl methanol/acetonitrile solution (1:1, v/v), vortexed and sonicated at 40 kHz for 30 min at 5 °C. Samples were incubated at −20 °C for 30 min and centrifuged at 13,000 × g for 15 min at 4 °C. The supernatant was evaporated to dryness under nitrogen and reconstituted in 100 μl acetonitrile/water solution (1:1, v/v), followed by brief sonication and centrifugation before LC-MS/MS analysis.

Untargeted metabolomic data were acquired using an ultra-high-performance liquid chromatography system coupled to a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific) equipped with an electrospray ionization source. Chromatographic separation was performed on a Waters ACQUITY HSS T3 column (100 mm × 2.1 mm, 1.8 μm). Mobile phase A was 0.1% formic acid in water/acetonitrile (95:5, v/v), and mobile phase B was 0.1% formic acid in acetonitrile/isopropanol/water (47.5:47.5:5, v/v). The gradient was: 0–3.5 min, 0–24.5% B; 3.5–5.0 min, 24.5–65% B; 5.0–5.5 min, 65–100% B; 5.5–7.4 min, 100% B; 7.4–7.6 min, 100–51.5% B; 7.6–7.8 min, 51.5–0% B; and 7.8–10.0 min, 0% B for column equilibration. The flow rate was 0.4 ml/min, injection volume was 2 μl, column temperature was 40 °C, and samples were maintained at 4 °C. Data were acquired in both positive and negative ion modes. The spray voltage was 3,500 V in positive mode and −3,500 V in negative mode, heater temperature was 425 °C, capillary temperature was 325 °C, sheath gas flow rate was 50 arb, and auxiliary gas flow rate was 13 arb. MS/MS data were acquired using data-dependent acquisition with stepped normalized collision energies of 20, 40 and 60 eV. The full MS resolution was 60,000, MS/MS resolution was 7500, and scan range was 70–1050 m/z. Raw LC-MS/MS data were processed using Progenesis QI (Waters Corporation) for peak detection, alignment, deconvolution and peak area extraction. Internal-standard peaks and known false-positive signals, including noise, column bleed and reagent-derived peaks, were removed. Metabolite annotation was performed by matching accurate mass, MS/MS spectra and retention information against HMDB, METLIN and the Majorbio in-house database.

Whole-tissue RNA-seq

Approximately 3 g pancreatic tissue was collected and was homogenized by bead-beating in tubes containing 2.8-mm beads and TRIzol reagent. RNA was extracted according to the TRIzol protocol. RNA purity and concentration were then examined using NanoDrop 2000. RNA integrity and quantity were measured using the Agilent 2100 system. mRNA was purified from total RNA using polyT and then fragmented into 300–350 bp fragments. The first strand cDNA was reverse-transcribed using fragmented RNA and dNTPs and second strand cDNA synthesis was subsequently performed. The template was enriched by PCR, and the PCR product was purified to obtain the final library. Sequencing was conducted using Illumina NovaSeq 6000 with mode PE150. Gene expression was measured using Kallisto (v.0.48.0) with the transcriptome reference of GRCh38 or GRCm39 (Release 108). Transcripts Per Million (TPM) was used for gene abundance quantification.

Bacterial quantification by real-time PCR (qPCR)

The DNA from the precipitate was extracted and purified using the QIAamp DNA Stool Mini Kit (Qiagen, West Sussex, UK). Specific primers for S. salivarius, P. intermedia and P. gingivalis was referred to the published studies^73–75, and the specificity of them was detected before formal experiment (Fig. S12G–I). All oligonucleotide primers used in this study were commercially synthesized by Sangon Biotech (Shanghai, China). Primer sequences are listed in Table S5. Purified strains of these three species were cultured in liquid medium to the third generation, and then the standard curves of them were constructed respectively. The DNA concentrations of target species was converted into copy number with the following formula:

C o p i e s = \frac{6.02 \times 10^{23} \times D N A c o n c e n t r a t i o n \times 10^{- 9}}{t a r g e t D N A l e n g t h \times 660}

Microbial dysbiosis indexes and clustering analysis of saliva samples

Microbial dysbiosis indexes (MD-index)⁷⁶ was calculated using the log of [total abundance of genus increased in GDM] over [total abundance of genus decreased in GDM] for all salivary specimens in different trimester of the C_Dis cohort. The enriched or depleted bacteria in GDM compared with CON were identified based on the Mann-Whitney test P < 0.05. Clustering analysis of the samples from C_Dis was conducted at the genus level according to the previous protocol⁷⁷. Samples were clustered using Jensen-Shannon distance and partitioning around medoid (PAM) clustering. Optimal number of clusters was obtained based on the Calinski-Harabasz (CH) index. Clustering analysis for the samples from C_Val and C_Int were assigned according to the samples from C_Dis with the closest Jensen-Shannon distance.

Gene copies analysis

We retrieved 1602 complete genomes of bacteria belong to ten genera (Streptococcus, Veillonella, Prevotella, Neisseria, Rothia, Gemella, Porphyromonas, Actinomyces, Granulicatella, TM7x) in saliva from GenBank (ftp://ftp.ncbi.nlm.nih.gov/genomes/). KEGG genes of the enzymes involved in the metabolism from fructose to glutamine (ec:2.4.1.5, ec:2.6.1.2, ec:2.7.1.121, ec:1.2.4.1, ec:2.7.1.202, ec:6.3.5.6, ec:6.3.5.7) were downloaded using KEGG API. Gene copies of these enzymes in the 1602 genomes were analyzed using BLAST (v.2.9.0+).

Statistics and reproducibility

This study included prospective observational cohorts, mechanistic mouse experiments, in vitro bacterial and cell-culture experiments, and a randomized controlled trial. For the observational cohort components, sample sizes were determined by participant eligibility, consent, successful sample collection and sequencing/metabolomics data availability; no statistical method was used to predetermine sample size for these cohort analyses. For the randomized controlled trial, sample size was predetermined according to the preregistered study protocol and statistical analysis plan. Power analysis based on a two-sample t-test (α = 0.05, β = 0.20) and preliminary glucose data indicated that 20 participants per group (N = 40 total) were required to detect a 0.5 mmol/L between-group difference with 80% power. No statistical method was used to predetermine sample size for the animal, bacterial culture or cell-culture experiments; sample sizes were based on prior experimental experience, feasibility and commonly used group sizes in related mechanistic studies. Exact sample sizes and replicate definitions are provided in the figure legends and Source Data files.

Participants in the randomized controlled trial were randomly assigned in a 1:1 ratio to the placebo or DHA group, and both participants and research staff responsible for data collection were blinded to group allocation. Human cohort analyses were observational and were not randomized. For mouse intervention experiments, animals were randomly allocated to treatment groups where group assignment was required. For in vitro bacterial culture and cell-culture experiments, treatment groups were assigned according to the experimental design. Except for the randomized controlled trial, the investigators were not blinded to allocation during experiments and outcome assessment.

Two saliva samples from the CDis cohort with fewer than 10,000 16S rRNA gene sequencing reads were excluded from downstream microbiome analyses before statistical comparisons. Samples or measurements that failed technical quality control, lacked valid group information, or lacked the specific outcome variable required for a given analysis were not included in that specific analysis. In the randomized controlled trial, participant-level measurements with available follow-up data were used for longitudinal and delta-value analyses; missing follow-up values were not imputed. No other data were excluded from the analyses.

For the cohort study, the main analysis was a microbiome-stratification framework rather than a single regression model. Genus-level oral microbiome profiles in the discovery cohort were clustered using Jensen-Shannon distance and PAM, with the optimal cluster number determined by the Calinski-Harabasz index, as described above. Associations between microbiome consortia and GDM status were tested using Pearson’s chi-squared test, while differences in dominant genera, MD-index, and alpha diversity were assessed using one-tailed Mann-Whitney tests. Community-level differences were evaluated by Bray-Curtis PCoA/PERMANOVA as supporting analyses, and multivariable-adjusted logistic regression adjusting for age, BMI, and gestational week was used as sensitivity analysis.

Discriminatory genus between SOM and POM in salivary specimens and differential KEGG Enzyme and KEGG Pathway were analyzed using Linear discriminant analysis Effect Size (LEfSe), with an absolute value of log LDA score ≥ 2.0 and P ≤ 0.05 considered a differential signature. For the differential analysis of the dominant genus in saliva in different trimester between GDM and CON, the changes of MD-index and α-diversity GDM compared with CON, differential regulation effect of DHA on the growth of Streptococcus and Prevotella, it was conducted using one-tailed Mann-Whitney test with P ≤ 0.05 considered a differential signature. Discriminatory metabolites in saliva of Trimester 3 between GOM and CON was identified using R package ropls (v 1.36.0) for OPLSDA analysis with adjusted Mann-Whitney test P < 0.1, OPLSDA VIP > 1, and |log2foldchange| > 0.5. The differences of the content of fructose, DHA, glycerate, and glutamine in saliva of Trimester 1 GDM and CON were analyzed using one-tailed Student’s t-test, with P ≤ 0.05 considered statistically significant. One-tailed student’s t-test was also used for the differential analysis of mice blood measurements, quantized data of MicroCT and immunohistochemistry, WB, and qPCR. Differential gene expression in pancreas in mice treated with POM compared with POM + SOM was analyzed using DESeq2 (v.1.44.0) Genes with an |log2 fold change| > 0.5 and an adjusted P ≤ 0.05 were considered differentially expressed. Significantly enriched Gene Ontology by the differentially expressed genes in GDM compared with CON was identified using the R package clusterProfiler (v.4.12.2) with the algorithm of GSEA and P ≤ 0.05 was adopted as the threshold. Community-level differences were assessed using Bray-Curtis distance-based PCoA and PERMANOVA. Multivariable-adjusted logistic regression analyses were performed with adjustment for age, BMI, and gestational week, with P ≤ 0.05 was adopted as the threshold. Spearman correlation networks were constructed for selected oral genera, and only correlations passing the predefined thresholds (|rho| > 0.3, adjusted P ≤ 0.05) were retained for network topology analysis. Unless otherwise specified, adjusted P ≤ 0.05 was considered statistically significant. To enhance the visual clarity of subtle yet statistically significant differences, certain bar plots and line graphs in this study are displayed with y-axes that do not start at zero. This approach was used solely to improve interpretability where absolute values are small and group differences would otherwise be difficult to distinguish visually.

Microbe-metabolite interactions analysis

Microbe-metabolite interactions between oral microbiota and saliva metabolites were estimated based their co-occurrence probabilities using a neural network algorithm MMvec (v.1.0.6). The parameters were set default. The microbe-metabolite interactions were visualized based conditional probability matrix and the log conditional probabilities between observations. Spearman correlation analysis between oral microbiota and saliva metabolites was employed using a R package corrplot (v.0.92) in R (v.4.4.1).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(11MB, pdf)}

41467_2026_74917_MOESM2_ESM.pdf^{(41.9KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(51.3KB, xlsx)}

Supplementary Data 2^{(10.8KB, xlsx)}

Supplementary Data 3^{(63.6KB, xlsx)}

Supplementary Data 4^{(53.6KB, xlsx)}

Supplementary Data 5^{(13.2KB, xlsx)}

Supplementary Data 6^{(13KB, xlsx)}

Reporting Summary^{(119.1KB, pdf)}

Transparent Peer Review file^{(1.7MB, pdf)}

Source data

Source data^{(32.6MB, zip)}

Acknowledgements

We are grateful to Jinyang Zhang, Peifeng Ji, Huajing Teng, Tongyu Zhang, Fengxiang Zhao, Suxin Shi, Depeng Li, Qian Zhang and Fangqing Zhao for their supports. We appreciate the support of High-performance Computing Platform of China Agricultural University for providing computational resources. This study was supported by the National Natural Science Foundation of China (T2341010, 32370053, U21A20346 and 82171662), Key R&D and Achievement Transformation Program in the Inner Mongolia Autonomous Region (2026YFSH0100), National Key Research and Development Program of China (2022YFA1304102, 2022YFC2704702, and 2023YFC2705900), Joint Funds of the Zhejiang Provincial Natural Science Foundation of China (LBY23H200008), Science and Technology Planning Project of Wenzhou (ZY2021025), Sichuan Science and Technology Program (2023YFQ0005), Key Research Program of Chongqing Science and Technology Bureau (CSTB2022TIAD-KPX0156), and 2115 Talent Development Program of China Agricultural University. The funders had no role in study design, data collection, data analysis, data interpretation, manuscript writing, or the decision to submit the manuscript for publication.

Author contributions

J.W. conceived the study. S.G., R.W., and Q.C. performed the animal experiments and related samples measurements. N.Y and H.Q. constructed the cohort of C_Dis and C_Int. X.L. and H. Z. performed the recruitment of C_Val and the clinical trial. Q.C., R.W., S.L., W.Z., and X.W. performed the experiments related to microbial cultivation. Y.Z., W.Z., Q.C., and S.G. conducted the assays and data analysis of MicroCT. A.Z. and S.G. conducted the cell culture experiments and related samples measurements. R.Z., H.F., Q.W., and X.W. took the key auxiliary role during the mice periodontitis model construction. S.P. and W.Z. analyzed the content of compounds in saliva samples from early pregnancy. The invaluable support of Y.H. greatly facilitated numerous in vivo and in vitro experiments. S.G. and W. S. took a comprehensive and overarching role in data analysis-related work. S.G. drafted and K.L., Y.G., W.S. and J.W. revised the manuscript. W.S., H.Q. and J.W. supervised the study.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source data are provided with this paper. The raw sequencing and metabolomics data generated in this study have been deposited in public repositories under the following accession codes: PRJNA1477772 [https://www.ncbi.nlm.nih.gov/bioproject/1477772] (raw 16S rRNA amplicon sequencing data of saliva specimens from the C_Dis, C_Int and C_Val cohorts, and metagenomic sequencing raw data of late-pregnancy saliva specimens from the CDis cohort). PRJNA1478085 [http://www.ncbi.nlm.nih.gov/bioproject/1478085] (metagenomic sequencing raw data of late-pregnancy saliva specimens from the C_Dis cohort).HRA008796 [https://ngdc.cncb.ac.cn/gsa-human/browse/HRA008796] (RNA-seq raw data of NCL-H716 cell lines). CRA019417 (RNA-seq raw data of STC-1 cell lines and mouse pancreas samples from POM and POM + SOM groups). CRA019414 (16S rRNA amplicon sequencing raw data of in vitro saliva cultures and POM/SOM bacterial cultures). ST004888 [10.21228/M8GP04] (targeted saliva metabolomics data deposited in Metabolomics Workbench). ST004891 [10.21228/M83G3S] (untargeted LC-MS/MS metabolomics data of simulated oral bacterial consortia deposited in Metabolomics Workbench). The GenBank assembly accession codes for the 1602 previously published complete oral bacterial genomes used for gene-copy analysis are provided in Source data of GenBank assembly accession for gene-copy analysis. Source data are provided with this paper.

Code availability

All code used in this study is available from the GitHub repositories https://github.com/gaoshengtao/periodontitis-and-GDM. The version of the code used in this study has been archived in Zenodo under 10.5281/zenodo.20485523⁷⁸.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Shengtao Gao, Nanlin Yin, Rujun Wei, Xiaoqing Li.

Contributor Information

Wenyu Shi, Email: shiwy@cau.edu.cn.

Hongbo Qi, Email: qihongbo@cqmu.edu.cn.

Jinfeng Wang, Email: wangjf@cau.edu.cn.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74917-w.

References

1.Benzian, H., Guarnizo-Herreño, C. C., Kearns, C., Muriithi, M. W. & Watt, R. G. The WHO global strategy for oral health: an opportunity for bold action. Lancet398, 192–194 (2021). [DOI] [PubMed] [Google Scholar]
2.Benzian, H., Watt, R., Makino, Y., Stauf, N. & Varenne, B. WHO calls to end the global crisis of oral health. Lancet400, 1909–1910 (2022). [DOI] [PubMed] [Google Scholar]
3.Zimmet, P., Alberti, K. G., Magliano, D. J. & Bennett, P. H. Diabetes mellitus statistics on prevalence and mortality: facts and fallacies. Nat. Rev. Endocrinol.12, 616–622 (2016). [DOI] [PubMed] [Google Scholar]
4.Lalla, E. & Papapanou, P. N. Diabetes mellitus and periodontitis: a tale of two common interrelated diseases. Nat. Rev. Endocrinol.7, 738–748 (2011). [DOI] [PubMed] [Google Scholar]
5.Choi, Y. H. et al. Association between periodontitis and impaired fasting glucose and diabetes. Diab. Care34, 381–386 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Elliott, A. et al. Distinct and shared genetic architectures of gestational diabetes mellitus and type 2 diabetes. Nat. Genet.56, 377–382 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Esteves Lima, R. P. et al. Association between periodontitis and gestational diabetes mellitus: systematic review and meta-analysis. J. Periodontol.87, 48–57 (2016). [DOI] [PubMed] [Google Scholar]
8.Chokwiriyachit, A. et al. Periodontitis and gestational diabetes mellitus in non-smoking females. J. Periodontol.84, 857–862 (2013). [DOI] [PubMed] [Google Scholar]
9.McIntyre, H. D. et al. Gestational diabetes mellitus. Nat. Rev. Dis. Prim.5, 47 (2019). [DOI] [PubMed] [Google Scholar]
10.Buchanan, T. A., Xiang, A. H. & Page, K. A. Gestational diabetes mellitus: risks and management during and after pregnancy. Nat. Rev. Endocrinol.8, 639–649 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Kumar, A. et al. Association between periodontal disease and gestational diabetes mellitus—a prospective cohort study. J. Clin. Periodontol.45, 920–931 (2018). [DOI] [PubMed] [Google Scholar]
12.Ye, W. et al. Gestational diabetes mellitus and adverse pregnancy outcomes: systematic review and meta-analysis. BMJ377, e067946 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Simmons, D. et al. Treatment of gestational diabetes mellitus diagnosed early in pregnancy. N. Engl. J. Med.388, 2132–2144 (2023). [DOI] [PubMed] [Google Scholar]
14.IDF. Prevalence of gestational diabetes mellitus (GDM), %, https://diabetesatlas.org/data/en/indicators/14/ (2021).
15.Horvath, K. et al. Effects of treatment in women with gestational diabetes mellitus: systematic review and meta-analysis. BMJ340, c1395 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Darveau, R. P. Periodontitis: a polymicrobial disruption of host homeostasis. Nat. Rev. Microbiol.8, 481–490 (2010). [DOI] [PubMed] [Google Scholar]
17.Endo, A. et al. Comparative genome analysis and identification of competitive and cooperative interactions in a polymicrobial disease. ISME J.9, 629–642 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Holt, S. C. & Ebersole, J. L. Porphyromonas gingivalis, Treponema denticola, and Tannerella forsythia: the “red complex”, a prototype polybacterial pathogenic consortium in periodontitis. Periodontol 200038, 72–122 (2005). [DOI] [PubMed] [Google Scholar]
19.Baughn, A. D. & Malamy, M. H. The strict anaerobe Bacteroides fragilis grows in and benefits from nanomolar concentrations of oxygen. Nature427, 441–444 (2004). [DOI] [PubMed] [Google Scholar]
20.Hajishengallis, G. et al. Low-abundance biofilm species orchestrates inflammatory periodontal disease through the commensal microbiota and complement. Cell Host Microbe10, 497–506 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Hajishengallis, G. & Chavakis, T. Local and systemic mechanisms linking periodontal disease and inflammatory comorbidities. Nat. Rev. Immunol.21, 426–440 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Ganesan, S. M. et al. A tale of two risks: smoking, diabetes and the subgingival microbiome. ISME J.11, 2075–2089 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Li, Y. et al. Dysbiosis of oral microbiota and metabolite profiles associated with type 2 diabetes mellitus. Microbiol. Spectr.11, e0379622 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Kunath, B. J. et al. Alterations of oral microbiota and impact on the gut microbiome in type 1 diabetes mellitus revealed by integrated multi-omic analyses. Microbiome10, 243 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Genco, R. J., Graziani, F. & Hasturk, H. Effects of periodontal disease on glycemic control, complications, and incidence of diabetes mellitus. Periodontol 200083, 59–65 (2020). [DOI] [PubMed] [Google Scholar]
26.Genco, R. J. & Borgnakke, W. S. Diabetes as a potential risk for periodontitis: association studies. Periodontol 200083, 40–45 (2020). [DOI] [PubMed] [Google Scholar]
27.Kocher, T., König, J., Borgnakke, W. S., Pink, C. & Meisel, P. Periodontal complications of hyperglycemia/diabetes mellitus: epidemiologic complexity and clinical challenge. Periodontol 200078, 59–97 (2018). [DOI] [PubMed] [Google Scholar]
28.Wang, J. et al. Dysbiosis of maternal and neonatal microbiota associated with gestational diabetes mellitus. Gut67, 1614–1625 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Teles, F., Wang, Y., Hajishengallis, G., Hasturk, H. & Marchesan, J. T. Impact of systemic factors in shaping the periodontal microbiome. Periodontol 200085, 126–160 (2021). [DOI] [PubMed] [Google Scholar]
30.Jang, H., Patoine, A., Wu, T. T., Castillo, D. A. & Xiao, J. Oral microflora and pregnancy: a systematic review and meta-analysis. Sci. Rep.11, 10.1038/s41598-021-96495-1 (2021). [DOI] [PMC free article] [PubMed]
31.Zeevi, D. et al. Personalized nutrition by prediction of glycemic responses. Cell163, 1079–1094 (2015). [DOI] [PubMed] [Google Scholar]
32.Berry, S. E. et al. Human postprandial responses to food and potential for precision nutrition. Nat. Med.26, 964–973 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Hajishengallis, G., Lamont, R. J. & Koo, H. Oral polymicrobial communities: assembly, function, and impact on diseases. Cell Host Microbe31, 528–538 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Maekawa, T. et al. Porphyromonas gingivalis manipulates complement and TLR signaling to uncouple bacterial clearance from inflammation and promote dysbiosis. Cell Host Microbe15, 768–778 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Xiao, E. et al. Diabetes enhances IL-17 expression and alters the oral microbiome to increase its pathogenicity. Cell Host Microbe22, 120–128.e124 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Kuboniwa, M. et al. Metabolic crosstalk regulates Porphyromonas gingivalis colonization and virulence during oral polymicrobial infection. Nat. Microbiol.2, 1493–1499 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Kolb, H. & Eizirik, D. L. Resistance to type 2 diabetes mellitus: a matter of hormesis? Nat. Rev. Endocrinol.8, 183–192 (2011). [DOI] [PubMed] [Google Scholar]
38.Lambelet, M. et al. Dysfunctional autophagy following exposure to pro-inflammatory cytokines contributes to pancreatic β-cell apoptosis. Cell Host Microbe9, 96 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Blasco-Baque, V. et al. Periodontitis induced by Porphyromonas gingivalis drives periodontal microbiota dysbiosis and insulin resistance via an impaired adaptive immune response. Gut66, 872–885 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Shoer, S. et al. Impact of dietary interventions on pre-diabetic oral and gut microbiome, metabolites and cytokines. Nat. Commun.14, 5384 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Cheng, W. C., Hughes, F. J. & Taams, L. S. The presence, function and regulation of IL-17 and Th17 cells in periodontitis. J. Clin. Periodontol.41, 541–549 (2014). [DOI] [PubMed] [Google Scholar]
42.Zelkha, S. A., Freilich, R. W. & Amar, S. Periodontal innate immune mechanisms relevant to atherosclerosis and obesity. Periodontol. 200054, 207–221 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Preshaw, P. M. et al. Periodontitis and diabetes: a two-way relationship. Diabetologia55, 21–31 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Wang, P. L. & Ohura, K. Porphyromonas gingivalis lipopolysaccharide signaling in gingival fibroblasts-CD14 and Toll-like receptors. Crit. Rev. Oral. Biol. Med.13, 132–142 (2002). [DOI] [PubMed] [Google Scholar]
45.Schmidt, M. I. et al. Fasting hyperglycemia and associated free-insulin and cortisol changes in Somogyi-like patients. Diab. Care2, 457–464 (1979). [DOI] [PubMed] [Google Scholar]
46.Kox, M. et al. Voluntary activation of the sympathetic nervous system and attenuation of the innate immune response in humans. Proc. Natl. Acad. Sci. USA111, 7379–7384 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Straub, R. H. & Cutolo, M. Glucocorticoids and chronic inflammation. Rheumatology55, ii6–ii14 (2016). [DOI] [PubMed] [Google Scholar]
48.Dong, L. et al. 3,3’-Diindolylmethane attenuates experimental arthritis and osteoclastogenesis. Biochem. Pharmacol.79, 715–721 (2010). [DOI] [PubMed] [Google Scholar]
49.Srikanth, M. & Rasool, M. 3, 3’- diindolylmethane hinders IL-17A/IL-17RA interaction and mitigates imiquimod-induced psoriasiform in mice. Int. Immunopharmacol.109, 108795 (2022). [DOI] [PubMed] [Google Scholar]
50.Kumar, M. et al. Omega-3 fatty acids and their interaction with the gut microbiome in the prevention and amelioration of type-2 diabetes. Nutrients14, 1723 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Davidson, M. H. et al. Efficacy and tolerability of adding prescription omega-3 fatty acids 4 g/d to simvastatin 40 mg/d in hypertriglyceridemic patients: an 8-week, randomized, double-blind, placebo-controlled study. Clin. Ther.29, 1354–1367 (2007). [DOI] [PubMed] [Google Scholar]
52.Wu, J. H. Y. et al. Omega-6 fatty acid biomarkers and incident type 2 diabetes: pooled analysis of individual-level data for 39 740 adults from 20 prospective cohort studies. Lancet Diab. Endocrinol.5, 965–974 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Kosmalski, M., Pękala-Wojciechowska, A., Sut, A., Pietras, T. & Luzak, B. Dietary intake of polyphenols or polyunsaturated fatty acids and its relationship with metabolic and inflammatory state in patients with type 2 diabetes mellitus. Nutrients14, 1083 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Jacobo-Cejudo, M. G. et al. Effect of n-3 polyunsaturated fatty acid supplementation on metabolic and inflammatory biomarkers in type 2 diabetes mellitus patients. Nutrients9, 573 (2017). [DOI] [PMC free article] [PubMed]
55.D’Aiuto, F. et al. Systemic effects of periodontitis treatment in patients with type 2 diabetes: a 12 month, single-centre, investigator-masked, randomised trial. Lancet Diab. Endocrinol.6, 954–965 (2018). [DOI] [PubMed] [Google Scholar]
56.Lauritzen, L. et al. DHA effects in brain development and function. Nutrients8, 6 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Djuricic, I. & Calder, P. C. Beneficial outcomes of omega-6 and omega-3 polyunsaturated fatty acids on human health: an update for 2021. Nutrients13, 2421 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Burckhardt, M. et al. Omega-3 fatty acids for the treatment of dementia. Cochrane Database Syst. Rev. 1–54. 10.1002/14651858.CD009002.pub3 (2016). [DOI] [PMC free article] [PubMed]
59.Cetin, I. et al. Omega-3 fatty acid supply in pregnancy for risk reduction of preterm and early preterm birth. Am. J. Obstet. Gynecol. MFM6, 101251 (2024). [DOI] [PubMed] [Google Scholar]
60.Sun, M. et al. Antibacterial and antibiofilm activities of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) against periodontopathic bacteria. Microb. Pathog.99, 196–203 (2016). [DOI] [PubMed] [Google Scholar]
61.Ribeiro-Vidal, H. et al. Antimicrobial activity of EPA and DHA against oral pathogenic bacteria using an in vitro multi-species subgingival biofilm model. Nutrients12, 10.3390/nu12092812 (2020). [DOI] [PMC free article] [PubMed]
62.Hai-Tao, Y. et al. Gestational diabetes mellitus decreased umbilical cord blood polyunsaturated fatty acids: a meta-analysis of observational studies. Prostaglandins Leukot. Essent. Fat. Acids171, 102318 (2021). [DOI] [PubMed] [Google Scholar]
63.Araújo, J. R., Correia-Branco, A., Ramalho, C., Keating, E. & Martel, F. Gestational diabetes mellitus decreases placental uptake of long-chain polyunsaturated fatty acids: involvement of long-chain acyl-CoA synthetase. J. Nutr. Biochem.24, 1741–1750 (2013). [DOI] [PubMed] [Google Scholar]
64.Matczuk, J. et al. Effect of streptozotocin-induced diabetes on lipids metabolism in the salivary glands. Prostaglandins Other Lipid Mediat.126, 9–15 (2016). [DOI] [PubMed] [Google Scholar]
65.Zhuang, P. et al. Eicosapentaenoic and docosahexaenoic acids attenuate hyperglycemia through the microbiome-gut-organs axis in db/db mice. Microbiome9, 1–21 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Oh, D. Y. et al. GPR120 is an omega-3 fatty acid receptor mediating potent anti-inflammatory and insulin-sensitizing effects. Cell142, 687–698 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Yan, Y. et al. Omega-3 fatty acids prevent inflammation and metabolic disorder through inhibition of NLRP3 inflammasome activation. Immunity38, 1154–1163 (2013). [DOI] [PubMed] [Google Scholar]
68.Botelho, J., Machado, V., Proença, L. & Mendes, J. J. The 2018 periodontitis case definition improves accuracy performance of full-mouth partial diagnostic protocols. Sci. Rep.10, 7093 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.IADPSG. International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care33, 676–682 (2010). [DOI] [PMC free article] [PubMed]
70.Marchesan, J. et al. An experimental murine model to study periodontitis. Nat. Protoc.13, 2247–2267 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Quast, C. et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res.41, D590–D596 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Chong, J., Liu, P., Zhou, G. & Xia, J. Using MicrobiomeAnalyst for comprehensive statistical, functional, and meta-analysis of microbiome data. Nat. Protoc.15, 799–821 (2020). [DOI] [PubMed] [Google Scholar]
73.Boutaga, K., van Winkelhoff, A. J., Vandenbroucke-Grauls, C. M. & Savelkoul, P. H. Periodontal pathogens: a quantitative comparison of anaerobic culture and real-time PCR. FEMS Immunol. Med. Microbiol.45, 191–199 (2005). [DOI] [PubMed] [Google Scholar]
74.Sakaguchi, S. et al. Bacterial rRNA-targeted reverse transcription-PCR used to identify pathogens responsible for fever with neutropenia. J. Clin. Microbiol.48, 1624–1628 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Shelburne, C. E. et al. Quantitative reverse transcription polymerase chain reaction analysis of Porphyromonas gingivalis gene expression in vivo. J. Microbiol. Methods49, 147–156 (2002). [DOI] [PubMed] [Google Scholar]
76.Gevers, D. et al. The treatment-naive microbiome in new-onset Crohn’s disease. Cell Host Microbe15, 382–392 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Arumugam, M. et al. Enterotypes of the human gut microbiome. Nature473, 174–180 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Gao, S. periodontitis-and-GDM: code used for data analysis and visualization, 10.5281/zenodo.20485523 (2026).

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(11MB, pdf)}

41467_2026_74917_MOESM2_ESM.pdf^{(41.9KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(51.3KB, xlsx)}

Supplementary Data 2^{(10.8KB, xlsx)}

Supplementary Data 3^{(63.6KB, xlsx)}

Supplementary Data 4^{(53.6KB, xlsx)}

Supplementary Data 5^{(13.2KB, xlsx)}

Supplementary Data 6^{(13KB, xlsx)}

Reporting Summary^{(119.1KB, pdf)}

Transparent Peer Review file^{(1.7MB, pdf)}

Source data^{(32.6MB, zip)}

Data Availability Statement

[CR1] 1.Benzian, H., Guarnizo-Herreño, C. C., Kearns, C., Muriithi, M. W. & Watt, R. G. The WHO global strategy for oral health: an opportunity for bold action. Lancet398, 192–194 (2021). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Benzian, H., Watt, R., Makino, Y., Stauf, N. & Varenne, B. WHO calls to end the global crisis of oral health. Lancet400, 1909–1910 (2022). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Zimmet, P., Alberti, K. G., Magliano, D. J. & Bennett, P. H. Diabetes mellitus statistics on prevalence and mortality: facts and fallacies. Nat. Rev. Endocrinol.12, 616–622 (2016). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Lalla, E. & Papapanou, P. N. Diabetes mellitus and periodontitis: a tale of two common interrelated diseases. Nat. Rev. Endocrinol.7, 738–748 (2011). [DOI] [PubMed] [Google Scholar]

[CR5] 5.Choi, Y. H. et al. Association between periodontitis and impaired fasting glucose and diabetes. Diab. Care34, 381–386 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Elliott, A. et al. Distinct and shared genetic architectures of gestational diabetes mellitus and type 2 diabetes. Nat. Genet.56, 377–382 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Esteves Lima, R. P. et al. Association between periodontitis and gestational diabetes mellitus: systematic review and meta-analysis. J. Periodontol.87, 48–57 (2016). [DOI] [PubMed] [Google Scholar]

[CR8] 8.Chokwiriyachit, A. et al. Periodontitis and gestational diabetes mellitus in non-smoking females. J. Periodontol.84, 857–862 (2013). [DOI] [PubMed] [Google Scholar]

[CR9] 9.McIntyre, H. D. et al. Gestational diabetes mellitus. Nat. Rev. Dis. Prim.5, 47 (2019). [DOI] [PubMed] [Google Scholar]

[CR10] 10.Buchanan, T. A., Xiang, A. H. & Page, K. A. Gestational diabetes mellitus: risks and management during and after pregnancy. Nat. Rev. Endocrinol.8, 639–649 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Kumar, A. et al. Association between periodontal disease and gestational diabetes mellitus—a prospective cohort study. J. Clin. Periodontol.45, 920–931 (2018). [DOI] [PubMed] [Google Scholar]

[CR12] 12.Ye, W. et al. Gestational diabetes mellitus and adverse pregnancy outcomes: systematic review and meta-analysis. BMJ377, e067946 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Simmons, D. et al. Treatment of gestational diabetes mellitus diagnosed early in pregnancy. N. Engl. J. Med.388, 2132–2144 (2023). [DOI] [PubMed] [Google Scholar]

[CR14] 14.IDF. Prevalence of gestational diabetes mellitus (GDM), %, https://diabetesatlas.org/data/en/indicators/14/ (2021).

[CR15] 15.Horvath, K. et al. Effects of treatment in women with gestational diabetes mellitus: systematic review and meta-analysis. BMJ340, c1395 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Darveau, R. P. Periodontitis: a polymicrobial disruption of host homeostasis. Nat. Rev. Microbiol.8, 481–490 (2010). [DOI] [PubMed] [Google Scholar]

[CR17] 17.Endo, A. et al. Comparative genome analysis and identification of competitive and cooperative interactions in a polymicrobial disease. ISME J.9, 629–642 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Holt, S. C. & Ebersole, J. L. Porphyromonas gingivalis, Treponema denticola, and Tannerella forsythia: the “red complex”, a prototype polybacterial pathogenic consortium in periodontitis. Periodontol 200038, 72–122 (2005). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Baughn, A. D. & Malamy, M. H. The strict anaerobe Bacteroides fragilis grows in and benefits from nanomolar concentrations of oxygen. Nature427, 441–444 (2004). [DOI] [PubMed] [Google Scholar]

[CR20] 20.Hajishengallis, G. et al. Low-abundance biofilm species orchestrates inflammatory periodontal disease through the commensal microbiota and complement. Cell Host Microbe10, 497–506 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Hajishengallis, G. & Chavakis, T. Local and systemic mechanisms linking periodontal disease and inflammatory comorbidities. Nat. Rev. Immunol.21, 426–440 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Ganesan, S. M. et al. A tale of two risks: smoking, diabetes and the subgingival microbiome. ISME J.11, 2075–2089 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Li, Y. et al. Dysbiosis of oral microbiota and metabolite profiles associated with type 2 diabetes mellitus. Microbiol. Spectr.11, e0379622 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Kunath, B. J. et al. Alterations of oral microbiota and impact on the gut microbiome in type 1 diabetes mellitus revealed by integrated multi-omic analyses. Microbiome10, 243 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Genco, R. J., Graziani, F. & Hasturk, H. Effects of periodontal disease on glycemic control, complications, and incidence of diabetes mellitus. Periodontol 200083, 59–65 (2020). [DOI] [PubMed] [Google Scholar]

[CR26] 26.Genco, R. J. & Borgnakke, W. S. Diabetes as a potential risk for periodontitis: association studies. Periodontol 200083, 40–45 (2020). [DOI] [PubMed] [Google Scholar]

[CR27] 27.Kocher, T., König, J., Borgnakke, W. S., Pink, C. & Meisel, P. Periodontal complications of hyperglycemia/diabetes mellitus: epidemiologic complexity and clinical challenge. Periodontol 200078, 59–97 (2018). [DOI] [PubMed] [Google Scholar]

[CR28] 28.Wang, J. et al. Dysbiosis of maternal and neonatal microbiota associated with gestational diabetes mellitus. Gut67, 1614–1625 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Teles, F., Wang, Y., Hajishengallis, G., Hasturk, H. & Marchesan, J. T. Impact of systemic factors in shaping the periodontal microbiome. Periodontol 200085, 126–160 (2021). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Jang, H., Patoine, A., Wu, T. T., Castillo, D. A. & Xiao, J. Oral microflora and pregnancy: a systematic review and meta-analysis. Sci. Rep.11, 10.1038/s41598-021-96495-1 (2021). [DOI] [PMC free article] [PubMed]

[CR31] 31.Zeevi, D. et al. Personalized nutrition by prediction of glycemic responses. Cell163, 1079–1094 (2015). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Berry, S. E. et al. Human postprandial responses to food and potential for precision nutrition. Nat. Med.26, 964–973 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Hajishengallis, G., Lamont, R. J. & Koo, H. Oral polymicrobial communities: assembly, function, and impact on diseases. Cell Host Microbe31, 528–538 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Maekawa, T. et al. Porphyromonas gingivalis manipulates complement and TLR signaling to uncouple bacterial clearance from inflammation and promote dysbiosis. Cell Host Microbe15, 768–778 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Xiao, E. et al. Diabetes enhances IL-17 expression and alters the oral microbiome to increase its pathogenicity. Cell Host Microbe22, 120–128.e124 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Kuboniwa, M. et al. Metabolic crosstalk regulates Porphyromonas gingivalis colonization and virulence during oral polymicrobial infection. Nat. Microbiol.2, 1493–1499 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Kolb, H. & Eizirik, D. L. Resistance to type 2 diabetes mellitus: a matter of hormesis? Nat. Rev. Endocrinol.8, 183–192 (2011). [DOI] [PubMed] [Google Scholar]

[CR38] 38.Lambelet, M. et al. Dysfunctional autophagy following exposure to pro-inflammatory cytokines contributes to pancreatic β-cell apoptosis. Cell Host Microbe9, 96 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR39] 39.Blasco-Baque, V. et al. Periodontitis induced by Porphyromonas gingivalis drives periodontal microbiota dysbiosis and insulin resistance via an impaired adaptive immune response. Gut66, 872–885 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Shoer, S. et al. Impact of dietary interventions on pre-diabetic oral and gut microbiome, metabolites and cytokines. Nat. Commun.14, 5384 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Cheng, W. C., Hughes, F. J. & Taams, L. S. The presence, function and regulation of IL-17 and Th17 cells in periodontitis. J. Clin. Periodontol.41, 541–549 (2014). [DOI] [PubMed] [Google Scholar]

[CR42] 42.Zelkha, S. A., Freilich, R. W. & Amar, S. Periodontal innate immune mechanisms relevant to atherosclerosis and obesity. Periodontol. 200054, 207–221 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR43] 43.Preshaw, P. M. et al. Periodontitis and diabetes: a two-way relationship. Diabetologia55, 21–31 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Wang, P. L. & Ohura, K. Porphyromonas gingivalis lipopolysaccharide signaling in gingival fibroblasts-CD14 and Toll-like receptors. Crit. Rev. Oral. Biol. Med.13, 132–142 (2002). [DOI] [PubMed] [Google Scholar]

[CR45] 45.Schmidt, M. I. et al. Fasting hyperglycemia and associated free-insulin and cortisol changes in Somogyi-like patients. Diab. Care2, 457–464 (1979). [DOI] [PubMed] [Google Scholar]

[CR46] 46.Kox, M. et al. Voluntary activation of the sympathetic nervous system and attenuation of the innate immune response in humans. Proc. Natl. Acad. Sci. USA111, 7379–7384 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR47] 47.Straub, R. H. & Cutolo, M. Glucocorticoids and chronic inflammation. Rheumatology55, ii6–ii14 (2016). [DOI] [PubMed] [Google Scholar]

[CR48] 48.Dong, L. et al. 3,3’-Diindolylmethane attenuates experimental arthritis and osteoclastogenesis. Biochem. Pharmacol.79, 715–721 (2010). [DOI] [PubMed] [Google Scholar]

[CR49] 49.Srikanth, M. & Rasool, M. 3, 3’- diindolylmethane hinders IL-17A/IL-17RA interaction and mitigates imiquimod-induced psoriasiform in mice. Int. Immunopharmacol.109, 108795 (2022). [DOI] [PubMed] [Google Scholar]

[CR50] 50.Kumar, M. et al. Omega-3 fatty acids and their interaction with the gut microbiome in the prevention and amelioration of type-2 diabetes. Nutrients14, 1723 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Davidson, M. H. et al. Efficacy and tolerability of adding prescription omega-3 fatty acids 4 g/d to simvastatin 40 mg/d in hypertriglyceridemic patients: an 8-week, randomized, double-blind, placebo-controlled study. Clin. Ther.29, 1354–1367 (2007). [DOI] [PubMed] [Google Scholar]

[CR52] 52.Wu, J. H. Y. et al. Omega-6 fatty acid biomarkers and incident type 2 diabetes: pooled analysis of individual-level data for 39 740 adults from 20 prospective cohort studies. Lancet Diab. Endocrinol.5, 965–974 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.Kosmalski, M., Pękala-Wojciechowska, A., Sut, A., Pietras, T. & Luzak, B. Dietary intake of polyphenols or polyunsaturated fatty acids and its relationship with metabolic and inflammatory state in patients with type 2 diabetes mellitus. Nutrients14, 1083 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR54] 54.Jacobo-Cejudo, M. G. et al. Effect of n-3 polyunsaturated fatty acid supplementation on metabolic and inflammatory biomarkers in type 2 diabetes mellitus patients. Nutrients9, 573 (2017). [DOI] [PMC free article] [PubMed]

[CR55] 55.D’Aiuto, F. et al. Systemic effects of periodontitis treatment in patients with type 2 diabetes: a 12 month, single-centre, investigator-masked, randomised trial. Lancet Diab. Endocrinol.6, 954–965 (2018). [DOI] [PubMed] [Google Scholar]

[CR56] 56.Lauritzen, L. et al. DHA effects in brain development and function. Nutrients8, 6 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Djuricic, I. & Calder, P. C. Beneficial outcomes of omega-6 and omega-3 polyunsaturated fatty acids on human health: an update for 2021. Nutrients13, 2421 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Burckhardt, M. et al. Omega-3 fatty acids for the treatment of dementia. Cochrane Database Syst. Rev. 1–54. 10.1002/14651858.CD009002.pub3 (2016). [DOI] [PMC free article] [PubMed]

[CR59] 59.Cetin, I. et al. Omega-3 fatty acid supply in pregnancy for risk reduction of preterm and early preterm birth. Am. J. Obstet. Gynecol. MFM6, 101251 (2024). [DOI] [PubMed] [Google Scholar]

[CR60] 60.Sun, M. et al. Antibacterial and antibiofilm activities of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) against periodontopathic bacteria. Microb. Pathog.99, 196–203 (2016). [DOI] [PubMed] [Google Scholar]

[CR61] 61.Ribeiro-Vidal, H. et al. Antimicrobial activity of EPA and DHA against oral pathogenic bacteria using an in vitro multi-species subgingival biofilm model. Nutrients12, 10.3390/nu12092812 (2020). [DOI] [PMC free article] [PubMed]

[CR62] 62.Hai-Tao, Y. et al. Gestational diabetes mellitus decreased umbilical cord blood polyunsaturated fatty acids: a meta-analysis of observational studies. Prostaglandins Leukot. Essent. Fat. Acids171, 102318 (2021). [DOI] [PubMed] [Google Scholar]

[CR63] 63.Araújo, J. R., Correia-Branco, A., Ramalho, C., Keating, E. & Martel, F. Gestational diabetes mellitus decreases placental uptake of long-chain polyunsaturated fatty acids: involvement of long-chain acyl-CoA synthetase. J. Nutr. Biochem.24, 1741–1750 (2013). [DOI] [PubMed] [Google Scholar]

[CR64] 64.Matczuk, J. et al. Effect of streptozotocin-induced diabetes on lipids metabolism in the salivary glands. Prostaglandins Other Lipid Mediat.126, 9–15 (2016). [DOI] [PubMed] [Google Scholar]

[CR65] 65.Zhuang, P. et al. Eicosapentaenoic and docosahexaenoic acids attenuate hyperglycemia through the microbiome-gut-organs axis in db/db mice. Microbiome9, 1–21 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Oh, D. Y. et al. GPR120 is an omega-3 fatty acid receptor mediating potent anti-inflammatory and insulin-sensitizing effects. Cell142, 687–698 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR67] 67.Yan, Y. et al. Omega-3 fatty acids prevent inflammation and metabolic disorder through inhibition of NLRP3 inflammasome activation. Immunity38, 1154–1163 (2013). [DOI] [PubMed] [Google Scholar]

[CR68] 68.Botelho, J., Machado, V., Proença, L. & Mendes, J. J. The 2018 periodontitis case definition improves accuracy performance of full-mouth partial diagnostic protocols. Sci. Rep.10, 7093 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR69] 69.IADPSG. International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care33, 676–682 (2010). [DOI] [PMC free article] [PubMed]

[CR70] 70.Marchesan, J. et al. An experimental murine model to study periodontitis. Nat. Protoc.13, 2247–2267 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR71] 71.Quast, C. et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res.41, D590–D596 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR72] 72.Chong, J., Liu, P., Zhou, G. & Xia, J. Using MicrobiomeAnalyst for comprehensive statistical, functional, and meta-analysis of microbiome data. Nat. Protoc.15, 799–821 (2020). [DOI] [PubMed] [Google Scholar]

[CR73] 73.Boutaga, K., van Winkelhoff, A. J., Vandenbroucke-Grauls, C. M. & Savelkoul, P. H. Periodontal pathogens: a quantitative comparison of anaerobic culture and real-time PCR. FEMS Immunol. Med. Microbiol.45, 191–199 (2005). [DOI] [PubMed] [Google Scholar]

[CR74] 74.Sakaguchi, S. et al. Bacterial rRNA-targeted reverse transcription-PCR used to identify pathogens responsible for fever with neutropenia. J. Clin. Microbiol.48, 1624–1628 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR75] 75.Shelburne, C. E. et al. Quantitative reverse transcription polymerase chain reaction analysis of Porphyromonas gingivalis gene expression in vivo. J. Microbiol. Methods49, 147–156 (2002). [DOI] [PubMed] [Google Scholar]

[CR76] 76.Gevers, D. et al. The treatment-naive microbiome in new-onset Crohn’s disease. Cell Host Microbe15, 382–392 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR77] 77.Arumugam, M. et al. Enterotypes of the human gut microbiome. Nature473, 174–180 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR78] 78.Gao, S. periodontitis-and-GDM: code used for data analysis and visualization, 10.5281/zenodo.20485523 (2026).

PERMALINK

Oral microbiome modulation mitigates hyperglycemia exacerbation in gestational diabetes mellitus

Shengtao Gao

Nanlin Yin

Rujun Wei

Xiaoqing Li

Qiuhan Cheng

Alateng Zhula

Wen Zhou

Yifei Zhang

Sihan Li

Wei Zhou

Xiaoya Wang

Ruiqi Zhang

Qiannan Wang

Huimin Fan

Sijia Peng

Hongping Zhang

Kefeng Li

Yongfei Hu

Yuan Gao

Wenyu Shi

Hongbo Qi

Jinfeng Wang

Abstract

Introduction

Results

Succession of oral dysbiosis coincides with the onset and progression of GDM

Fig. 1. Succession of oral dysbiosis coincides with the onset and progression of GDM.

POM triggers in situ and systemic inflammation and aggravates hyperglycemia

Fig. 2. POM triggers in situ and systemic inflammation and aggravates hyperglycemia.

POM exacerbates hyperglycemia via a dual pathway of hormones and stress

Fig. 3. POM exacerbates hyperglycemia via a dual pathway of hormones and stress.

Fig. 5. Polyunsaturated fatty acids differentially modulate oral microbes to form POM.

Remodeling of the oral microbiota improves periodontal and glycemic status

Fig. 4. Remodeling of the oral microbiota improves periodontal and glycemic status.

Polyunsaturated fatty acids differentially modulate oral microbes to form POM

DHA improves periodontal and glycemic status of pregnant women with GDM

Fig. 6. DHA improves periodontal and glycemic status of pregnant women with GDM.

Discussion

Methods

Ethics statement

Human participants, sex and gender reporting

Prospective cohorts

Randomized controlled trial

Animal experiments

Periodontitis model

Bacterial cultivations

Cell cultures

Oral glucose tolerance test (OGTT) of mice

Micro-CT analysis

Blood parameters assays

Immunohistochemistry

Immunofluorescence

Western blot

Real-time quantitative PCR for mRNA

DNA extraction

16S rRNA gene amplicon sequencing and data preprocesses

Metagenomic sequencing and data preprocesses

Targeted metabolomics measurement

Untargeted metabolomic measurement

Whole-tissue RNA-seq

Bacterial quantification by real-time PCR (qPCR)

Microbial dysbiosis indexes and clustering analysis of saliva samples

Gene copies analysis

Statistics and reproducibility

Microbe-metabolite interactions analysis

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Data availability

Code availability

Competing interests

Footnotes

Contributor Information

Supplementary information