Oral microbiota mediates the association of healthy dietary patterns with lower risk of advanced cardiovascular-kidney-metabolic syndrome

Abstract

Objectives

Existing evidence links diet and the oral microbiota to individual cardiovascular, kidney, and metabolic conditions, but these relationships have not been extended to the progressive stages of cardiovascular-kidney-metabolic (CKM) syndrome. We therefore investigated the associations among dietary patterns, the oral microbiome, and CKM stages.

Methods

This cross-sectional study included 2051 representative US adults. Oral microbiota was characterized via 16S rRNA gene sequencing of oral rinse samples. Dietary patterns were assessed using multiple dietary indexes derived from 24-h dietary recalls. CKM stages were defined according to American Heart Association criteria. MaAsLin3 analysis and ordinal logistic regression were used to evaluate the associations of dietary indexes and oral microbiota with CKM stages. Mediation analysis was used to assess the mediating role of oral microbiota in the associations between various dietary indexes and CKM stages.

Results

As CKM stages advanced, the oral microbial community tended to exhibit gradient shifts, primarily characterized by a potentially progressive reduction in the relative abundance of Pseudomonadota. Neisseria and Lautropia within Pseudomonadota, together with Gemella and Bergeyella, were negatively associated with CKM stages, whereas Lactobacillus was positively associated. Pseudomonadota-dominated salivatypes were associated with the lowest risk of advanced CKM stages. Higher healthy dietary indexes were associated with lower likelihood of advanced CKM stages and with higher abundance of Pseudomonadota taxa, particularly Neisseria. Notably, Neisseria and its higher taxonomies predominantly mediated the associations between dietary indexes and CKM stages.

Conclusion

This study suggests that Neisseria-related taxa are predominant mediators linking dietary patterns to CKM stages. Modulation of the diet–microbiome axis may represent a complementary strategy for promoting health and potentially influencing CKM-associated outcomes, but our cross-sectional data should not be interpreted as establishing a causal protective role.

Introduction

Cardiovascular disease (CVD), diabetes and chronic kidney disease (CKD) are leading causes of death in the United States and together accounted for approximately one-quarter of all deaths in 2020 [1]. Over the past two decades, cardiovascular, kidney, and metabolic multimorbidity has increased steadily [2]. These conditions are closely interconnected and share pathogenetic mechanisms, including obesity and insulin resistance, which create a vicious cycle of mutual exacerbation and ultimately lead to adverse clinical outcomes [3, 4]. The recently proposed cardiovascular-kidney-metabolic (CKM) syndrome describes a systemic disorder in which interactions among metabolic risk factors, CKD, and the cardiovascular system cause multiorgan dysfunction [5]. This framework defines progressive stages of disease: stage 0 (no CKM risk factors), stage 1 (excess or dysfunctional adiposity), stage 2 (metabolic risk factors or CKD), stage 3 (subclinical CVD), and stage 4 (clinical CVD) [5]. However, the mechanisms underlying CKM progression remain poorly understood, highlighting the need to identify mechanisms that may inform clinical diagnosis and intervention.

The oral microbiota interacts with the host immune system and plays a vital role in both oral and systemic health [6]. The oral microbiome can also contribute to host homeostasis by maintaining ecological balance, interacting with mucosal immunity, and participating in metabolic pathways relevant to systemic physiology [6]. Increasing evidence has linked oral microbial dysbiosis to type 2 diabetes mellitus (T2DM) [7], CKD [8], and CVD [9]. Nevertheless, its role in the coexistence of these conditions remains insufficiently studied [10]. Common microbial alterations reported in these disorders include reduced microbial diversity [11, 12] and increased abundance of periodontal pathogens [13]. In addition, genera such as Neisseria, Streptococcus, and Haemophilus have shown discriminatory potential for these diseases [8, 14, 15]. Several mechanisms may underlie these associations. Oral microbial invasion into the bloodstream can cause bacteremia, and oral-gut microbiota translocation can alter gut microbial composition [16]. Moreover, oral microbiota can induce the release of pro-inflammatory cytokines [17], thereby aggravating systemic inflammation and contributing to CKM-related diseases [9, 18]. Together, these findings suggest that progressive alterations in the oral microbiota may accompany CKM progression. However, no study has investigated shifts in distinct clusters within the oral microbiome, known as salivatypes [19], across CKM stages.

Oral microbiota is shaped by both host and environmental factors, particularly diet [20]. Diet is also relevant to systemic homeostasis because vegetable-derived dietary nitrate can enter the enterosalivary circulation, where oral bacterial communities reduce nitrate to nitrite, increasing nitric oxide bioavailability and potentially supporting vascular and metabolic regulation [21, 22]. Dietary patterns, which capture overall food intake, have been shown to influence the oral microbiota [23]. For instance, the Mediterranean Diet Index(MED) [24] is associated with reduced abundance of oral periodontal pathogens [25] and favorable effects on cardiometabolic health [26, 27]. Other healthy dietary indexes, including plant-based diets and the Dietary Approaches to Stop Hypertension Index (DASH), have also been linked to improvements in components of CKM syndrome [26]. Because commonly used dietary indexes may show somewhat different associations with cardiometabolic health [28], multiple dietary indexes are needed to comprehensively assess dietary quality. However, dietary quality assessed by diverse dietary indexes across CKM stages has not been well characterized. Furthermore, the oral microbiota may mediate the relationship between dietary patterns and cardiometabolic health [29], but its potential role in linking diverse dietary patterns to CKM stages remains unclear. Clarifying this issue may provide a theoretical basis for dietary strategies aimed at preventing or slowing CKM progression.

Therefore, using recently released oral microbiome data from a nationally representative US population, this study aimed to: (i) identify oral microbial taxa and salivatypes associated with CKM stages, (ii) investigate the associations of diverse dietary indexes with both the oral microbiota and CKM stages, (iii) explore the mediating role of the oral microbiota in the relationship between diet and CKM.

Materials and methods

Study design and participants

Data were obtained from the National Health and Nutrition Examination Survey (NHANES), which evaluates the health and nutritional conditions of a nationally representative US population. Detailed study information is available on the Centers for Disease Control and Prevention (CDC) website at https://www.cdc.gov/nchs/nhanes/. This study was approved by the National Center for Health Statistics (NCHS) Ethics Review Board (The IRB for each survey cycle was available at: https://www.cdc.gov/nchs/nhanes/irba98.htm), and all participants provided written informed consent. We used the 2009–2012 cycles because oral microbiome data were recently released for these years. Participants were excluded if they (i) were aged under 20 years old or pregnant; (ii) had missing information on oral microbiota, CKM stage, dietary indexes, or covariates; (iii) used antibiotics known to affect the oral microbiota (e.g., amoxicillin, macrolides) within 30 days (determined based on prescription drug information) [30]; (iv) had a sequencing depth of <10,000 reads during amplicon sequence variant (ASV) rarefaction. The final analysis included 2051 individuals (Fig. S1). As this observational study used a pre-existing nationally representative database, no a priori sample size calculation was performed; all eligible NHANES 2009–2012 participants were included, exceeding most oral microbiome studies in sample size.

Sample collection and DNA extraction

Oral rinse samples were collected during 2009–2012. Participants were instructed to rinse their mouths with mouthwash for 5 s and then gargle for 5 s, repeating this procedure three times before expectorating into a collection cup. Oral rinse samples were stored and transported to the Ohio State University Gillison Laboratory for processing. DNA was extracted from the cell pellet using the Puregene DNA purification kit protocol following centrifugation, cell lysis, and digestion with RNase A and Proteinase K [31–33].

16S rRNA gene sequencing and bioinformatics procedures

Details of sequencing and bioinformatics are available in the NHANES documentation [31]. Briefly, purified DNA was sent to the Knight laboratory at the University of California, San Diego. Polymerase chain reaction (PCR) amplification and sequencing were performed using the Illumina HiSeq 2500 according to the Earth Microbiome Project protocol [34]. Raw forward reads were processed with DADA2 to generate an ASV feature table because paired-end reads did not overlap. Taxonomic annotations were provided in the NHANES dataset using the SILVA v123 database. Phylum-level nomenclature was updated to align with current International Committee on Systematics of Prokaryotes (ICSP) consensus (e.g., Bacillota replacing Firmicutes; Table S1) [35]. Genus-level assignments remain limited by short-read 16S rRNA sequencing of oral rinse samples and the older SILVA v123 database; thus, some taxa may reflect outdated nomenclature or low taxonomic resolution, as detailed in Table S2. Alpha- and beta-diversity indexes, and abundance tables were generated using QIIME1.

We used the NHANES feature tables with non-bacterial taxa excluded. For consistency, diversity data and feature tables were rarefied to 10,000 reads/sample. Alpha-diversity was calculated as the mean of 10 random sampling repetitions for each index (Observed ASVs, Faith’s phylogenetic diversity (Faith’s PD), Shannon, Simpson). Beta-diversity was assessed using Bray-Curtis dissimilarity. For downstream analyses, we retained taxa present in at least 10% of samples with abundance >0.01%, yielding 11 phyla, 21 classes, 30 orders, 48 families, and 95 genera.

Because NHANES provides processed ASV tables rather than raw sequence reads or representative ASV sequences, phylogeny-based tools such as Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2) could not be applied. As an exploratory supplementary analysis, we used Functional Annotation of Prokaryotic Taxa (FAPROTAX) [36], a taxonomy-to-function annotation approach, to estimate nitrate-reduction potential. In brief, taxonomic assignments were matched to a literature-curated prokaryotic functional-trait database, and the sample-level value was calculated from the abundance-weighted contribution of taxa annotated with nitrate reduction. This variable therefore represents taxonomy-matched functional potential rather than direct measurement of nitrate-reductase genes, nitrite production, or enzymatic nitrate-reducing activity.

Definition of CKM stages

The NHANES component includes questionnaires (demographics, disease and medication history, and lifestyles), laboratory testing (e.g., blood biochemistry and complete blood count), and physical examination (e.g., body mass index (BMI) and blood pressure). Details on the definition of CKM stages in NHANES have been reported previously [37]. CKM stages were defined as the coexistence of subclinical or clinical CVD, CKD, and metabolic disorders based on the 2023 American Heart Association (AHA) Presidential Advisory on CKM Health [5], with detailed definitions for each component shown in Table S3. Clinical CVD was determined by a history of heart failure, coronary heart disease, heart attack, or stroke. Subclinical CVD was defined as either (i) very high-risk CKD according to the Kidney Disease: Improving Global Outcomes (KDIGO) classification, or (ii) a high predicted 10-year CVD risk (≥ 20%) calculated using the AHA Predicting Risk of CVD EVENTs (PREVENT) equations (algorithms are detailed in Table S4) [38]. Estimated glomerular filtration rate (eGFR) was calculated using the 2021 race-free CKD Epidemiology Collaboration creatinine equation [39]. CKD risk was stratified by cross-classification of eGFR and urinary albumin to creatinine ratio (UACR) according to KDIGO (Table S3). In this study, CKD was defined as a KDIGO classification of moderate-to-high-risk CKD, excluding the low-risk category (eGFR ≥60 mL/min/1.73 m² and UACR < 30 mg/g). Metabolic disorders were defined as overweight/obesity, abdominal obesity, prediabetes, diabetes, hypertension, hypertriglyceridemia, or metabolic syndrome [5].

Briefly, participants were classified into five CKM stages: stage 0 (no CKM risk factors), stage 1 (excess or dysfunctional adiposity), stage 2 (metabolic risk factors or moderate-to-high-risk CKD), stage 3 (subclinical CVD with CKM risk factors), and stage 4 (clinical CVD with CKM risk factors) [5].

Dietary index

Daily dietary intake was assessed using 2-day 24 h dietary recalls. Energy and nutrients for foods and beverages were assessed using the United States Department of Agriculture (USDA) Food and Nutrient Database for Dietary Studies (FNDDS). Dietary recall data were further linked to the Food Patterns Equivalent Database (FPED), which classifies foods into 37 USDA Food Pattern Components. Dietary intakes of energy, nutrients, and food groups were averaged across 2 days when both recalls were available; otherwise, data from the first day were used.

To comprehensively assess the overall dietary quality, we applied several widely researched dietary indexes, including Dietary Inflammatory Index (DII) [40], MED [41], Healthy Eating Index-2020 (HEI-2020) [42], The Alternative Healthy Eating Index-2010 (AHEI-2010) [43], and DASH [44], based on daily intakes of nutrients and food groups. DII was derived from literature-based inflammatory effect scores for 45 dietary parameters; higher scores indicated a more pro-inflammatory, less healthy diet. The components of MED, HEI-2020, AHEI-2010, and DASH are shown in Table S5. Briefly, these indexes included adequacy components, mainly vegetables, fruits, whole grains, legumes, nuts, and seafood, and moderation components, mainly red and processed meat and sugars. Higher scores indicate better diet quality. All dietary indexes were calculated using the R package “Dietaryindex” [45].

Assessment of covariates

Questionnaires collected demographic information, including age, sex, race (Mexican American, other-Hispanic, non-Hispanic White, non-Hispanic Black, and other race), marital status (never married, married, living with partner, widowed/divorced/separated), poverty-to-income ratio (PIR), and education level (less than high school, high school graduate, some college or AA degree, college graduate or above). PIR, an indicator of socioeconomic status based on household income relative to federal poverty thresholds by household size, year, and state, was categorized as <1.3, 1.3–3.5, and >3.5. Smoking status was divided into never smoker (those who had never smoked at least 100 cigarettes in life), former smoker (those who had smoked at least 100 cigarettes in life and now quit smoking), and current smoker (those who had smoked at least 100 cigarettes in life and now smoked cigarettes every day or sometimes). Drinking status was divided into never drinker (those who had never drunk at least 12 alcohol drinks in life), former drinker (those who had drunk at least 12 alcohol drinks in life and quit drinking in the past year), and current drinker (those who had drunk at least 12 alcohol drinks in life and still drank alcohol in the past year). Physical activity was measured as metabolic equivalent of task (MET)-minutes/week, calculated by multiplying activity duration by the corresponding MET value. Because antidiabetic medications may affect the oral microbiota [46], antidiabetic drug use was included as a covariate and identified from prescription records (e.g., metformin, insulin, acarbose) within the past month [30]. Oral health was assessed by the self-reported question, “Rate the health of your teeth and gums,” and was dichotomized as excellent/very good/good or fair/poor.

Statistical analysis

Characteristics of study participants are presented as median (Q25, Q75) for continuous variables with skewed distribution, and as frequency (percentage) for categorical variables. Differences across CKM stages were assessed using the Kruskal-Wallis test for skewed continuous variables, with Dunn’s test for pairwise comparisons, and the Pearson Chi-squared test for categorical variables.

Microbial alpha-diversity (Observed ASVs, Faith’s PD, Shannon, Simpson) and beta-diversity (Bray-Curtis dissimilarity) were analyzed. Beta-diversity was assessed by principal coordinate analysis (PCoA). Differences in microbial composition were tested using permutational multivariate analysis of variance (PERMANOVA; adonis) with 999 permutations. Post-hoc pairwise PERMANOVA comparison across CKM stages was conducted using the R package “pairwiseAdonis”.

The two-part MaAsLin3 (microbiome multivariable associations with linear models) framework was used to evaluate the independent associations of predictors with microbial prevalence (logistic regression) and non-zero abundance (linear regression) [47]. In the abundance model, the regression coefficient represents the change in log-transformed abundance, whereas in the prevalence model, the coefficient denotes the log-odds ratio (Log-OR) for taxon presence. Covariates were consistent with subsequent analyses. This approach avoids pseudo-count biases and mitigates false-positives common in compositional microbiome data, outperforming traditional differential abundance methods [47].

Microbial communities were clustered into “salivatypes” using Dirichlet multinomial mixtures (DMM) [19, 29, 48]. To determine the optimal number of clusters, DMM models with increasing Dirichlet components (k) were fitted, and the Akaike information criterion (AIC) and Bayesian information criterion (BIC) were used to balance goodness-of-fit and complexity. The optimal number of salivatypes was defined as the k with the lowest AIC or BIC, while also considering cluster numbers reported in previous studies.

Because CKM stages represent an ordered disease-severity hierarchy, ordinal logistic regression was used to assess the associations of oral microbial taxa, salivatypes, and dietary indexes with CKM stages. Linear regression was used to assess the associations between dietary indexes and oral microbial taxa. Covariates included age (years), sex, BMI (kg/m²), race (Mexican American, other-Hispanic, non-Hispanic White, non-Hispanic Black, and other race), marital status (never married, married/living with partner, widowed/divorced/separated), PIR (<1.3, 1.3–3.5, and >3.5), education level (less than high school, high school graduate, some college or AA degree, college graduate or above), smoking status (never, former, current), drinking status (never, former, current), physical activity (MET-minutes/week), total energy intake (kcal), antidiabetic drug use (yes/no), and oral health (excellent/very good/good and fair/poor). Before association analysis, diversity and dietary indexes were standardized as Z-scores, and microbial relative abundances were inverse normal transformed.

Significant associations among dietary indexes, microbial taxa, and CKM stages (P < 0.05) were included in mediation analyses. Mediation analyses were conducted to assess the mediating role of microbial taxa in the associations between dietary indexes and CKM stages, with adjustment for the same covariates described above. A significant mediation effect was defined as a false discovery rate (FDR)-corrected P-value for the average causal mediation effects (ACME) < 0.05. The proportion mediated was calculated as ACME divided by the total effect.

All statistical analyses were performed using R version 4.4.2, primarily with the packages including “vegan”, “maaslin3”, “MASS”, “DirichletMultinomial”, “mediation”, and “ggplot2”. All statistical tests were two-sided. For microbial taxa at each taxonomic level, P-values were adjusted using the Benjamini-Hochberg (BH) method to control the FDR, and FDR-corrected P-value (P-FDR) < 0.05 was considered statistically significant. For all other analyses, P-value < 0.05 was considered statistically significant.

Results

Characteristics of study participants by CKM stages

The median participant age was 47 (25th-75th percentiles: 35, 58) years, and 52.2% were male (Table S6). There were significant between-group differences in all demographic characteristics and lifestyles at different CKM stages (P < 0.05). Participants with higher CKM stages were more likely to be older, male, non-Hispanic Black, and current smokers, as well as have a lower education level, a lower PIR, higher BMI, reduced physical activity, a higher rate of antidiabetic drug use, and poorer oral health.

The five most dominant phyla were Bacillota (formerly Firmicutes, 48.66%), Bacteroidota (formerly Bacteroidetes, 18.25%), Actinomycetota (formerly Actinobacteria, 15.91%), Pseudomonadota (formerly Proteobacteria, 11.18%), and Fusobacteriota (formerly Fusobacteria, 5.13%), accounting for 99.13%. The ten dominant genera were Streptococcus (29.36%), Rothia (11.66%), Prevotella_7 (9.28%), Veillonella (6.66%), Haemophilus (5.96%), Gemella (4.19%), Fusobacterium (2.92%), Prevotella (2.87%), Neisseria (2.85%), and Actinomyces (2.64%), accounting for 78.39% (Table S7). The top five contributors to the explained variance of oral microbiome variability were smoking status, race, age, education level, and CKM stage (Figure S2).

Alterations of oral microbiota associated with CKM stages

Preliminary exploratory analyses revealed significant differences in three alpha-diversity indexes (Observed ASVs, Shannon, and Simpson) across CKM stages (Fig. 1A and Table S8). These indexes tended to decrease with higher CKM stages; median Shannon values for stages 0-4 were 4.78, 4.77, 4.78, 4.62, and 4.57, respectively. Significant pairwise differences occurred only between stages 0–2 and 3–4. We observed significant differences in microbial composition for CKM stages (P_adonis = 0.001, Fig. 1B). Notably, PCoA1—capturing the primary microbial variation—differed significantly across stages (P < 0.001), showing a progressive trend with higher CKM stages. This suggests a directional shift in microbial composition paralleling CKM severity. Furthermore, pairwise comparisons revealed stage 4 was significantly distinct from all others, exhibiting the most profound microbial divergence. However, covariate-adjusted ordinal logistic regression revealed no significant associations for diversity indices (Table S9).

Fig. 1. Alterations of the oral microbiota across CKM stages.

Comparison of A Alpha-diversity indexes (Observed ASVs, Faith’s PD, Shannon, and Simpson), B beta-diversity assessed by PCoA based on Bray-Curtis distance, and the relative abundances of oral microbiota at the C phylum, D family, and E genus levels across CKM stages. Differences in diversity indexes and relative abundances across CKM stages were tested using the Kruskal–Wallis test. Pairwise comparisons between groups were performed by the Dunn’s test. Asterisks in A indicate statistically significant differences in pairwise comparisons (**P < 0.01, *P < 0.05). Differences in microbial composition across CKM stages were compared using PERMANOVA with 999 permutations, followed by pairwise PERMANOVA comparisons. Asterisks next to the taxa names in the legends of (C–E) denote taxa with significantly different relative abundances across all CKM stages based on unadjusted P-values. Abbreviations: ASVs amplicon sequence variants, Faith’s PD Faith’s phylogenetic diversity, CKM cardiovascular, kidney, and metabolic syndrome, PCoA principal coordinate analysis.

Relative abundances of oral microbiota at phylum, family, and genus levels across CKM stages are illustrated in Fig. 1C–E. A total of 77 taxa differed significantly across CKM stages (P-FDR < 0.05, Table S10), predominantly within Pseudomonadota (n = 23), Bacillota (n = 17), and Actinomycetota (n = 12). Most Pseudomonadota taxa declined with advancing CKM stages (mean abundances: 12.79%, 11.62%, 10.90%, 11.49%, 9.94% for stages 0–4, respectively). Conversely, Actinomycetota exhibited an increasing trend.

Covariate-adjusted MaAsLin3 analysis identified 57 taxa significantly associated with CKM stages (using an exploratory P-FDR < 0.25 to capture a broader range of potential associations) (Fig. 2A and Table S11). Of these, 20 belonged to the phylum Pseudomonadota. Notably, both the abundance and prevalence of Neisseria were significantly associated with CKM stages, exhibiting a 0.129 decrease in log-relative abundance per 1-stage increment in CKM. The previously observed significant differences for taxa within Actinomycetota were no longer evident. Only a minority of taxa exhibited significant positive correlations, including the legacy broad taxon Lactobacillus sensu lato (hereafter referred to as Lactobacillus), as well as taxa within Bacteroidota and Mycoplasmatota. Furthermore, covariate-adjusted ordinal logistic regression models revealed that 15 taxa were significantly associated with CKM stages (P-FDR < 0.1, Fig. 2B). Fourteen of these overlapped with the taxa identified by MaAsLin3 (with the exception of Burkholderiaceae), underscoring the robustness of their associations with CKM progression (Fig. 2C). With the exception of Lactobacillus and Lactobacillaceae, which correlated positively with CKM stages, all other identified taxa displayed negative correlations. The odds ratios (ORs) (95% confidence intervals (CIs)) for significant genus-level taxa were 0.851 (0.767, 0.946) for Lautropia, 0.861 (0.778, 0.953) for Neisseria, 0.863 (0.786, 0.948) for Gemella, 1.200 (1.062, 1.356) for Lactobacillus, and 0.857 (0.780, 0.943) for Bergeyella (currently classified within the family Weeksellaceae) (Fig. 2B).

Fig. 2. Significant associations between oral microbial taxa and CKM stages in covariate-adjusted models.

Significant associations between oral microbial taxa and CKM stages from A MaAsLin3 analysis (P-FDR < 0.25) and B ordinal logistic regression (P-FDR < 0.1). C Venn diagram showing the overlap of significant taxa identified by the MaAsLin3 analysis and ordinal logistic regression model. All models were adjusted for age, sex, BMI, race, marital status, education level, PIR, smoking status, drinking status, physical activity (MET-minutes/week), total energy intake (kcal), antidiabetic drug use, and oral health. P-values were adjusted using FDR correction for microbial taxa at each taxonomic level. Abbreviations: CKM cardiovascular, kidney, and metabolic syndrome, 95% CI 95% confidence interval, OR odds ratio.

Associations between salivatypes and CKM stages

We further identified distinct clusters of the oral microbial community by DMM modelling. Based on prior studies [29, 49] and model fit (lowest BIC at three clusters; Fig. S3), we defined three salivatypes: type Pseudomonadota, type high diversity, and type Actinomycetota, according to the diversity and differentially abundant taxa (Fig. 3A–D). Microbial structure differed significantly across the three salivatypes (Fig. 3A), indicating robust clustering. Type high diversity exhibited the highest alpha-diversity, followed by type Pseudomonadota and type Actinomycetota (Fig. 3B). The three salivatypes were dominated by Pseudomonadota (Haemophilus and Neisseria), Fusobacteriota (Fusobacterium), and Actinomycetota (Rothia), respectively (Fig. 3C, D). With advancing CKM stages, the proportion of type Pseudomonadota decreased, whereas type Actinomycetota increased; type high diversity remained relatively stable (Fig. 3E). In covariate-adjusted ordinal logistic regression models, compared to type Pseudomonadota, type Actinomycetota (OR = 1.456, 95% CI: 1.148, 1.845), but not type high diversity (OR = 1.173, 95% CI: 0.935, 1.471), was significantly associated with higher CKM stages. A significant positive association was observed between Actinomycetota and Pseudomonadota ratio (A/P ratio) and CKM stages (OR per SD = 1.103, 95% CI: 1.003, 1.213), and the third tertile showed a borderline significant association with higher CKM stages compared to the first tertile (OR = 1.238, 95% CI: 0.975, 1.572) (Fig. 3F).

Fig. 3. Associations between salivatypes and CKM stages.

A Beta-diversity assessed by PCoA based on Bray-Curtis distance, B alpha-diversity indexes, and the relative abundances of oral microbiota at the C phylum and D genus levels across the three salivatypes. E Proportion of the three salivatypes across CKM stages. F Associations of salivatypes and Actinomycetota-to-Pseudomonadota ratio with CKM stages identified by ordinal logistic regression. The three salivatypes were identified using DMM. Differences in microbial composition across salivatypes were compared using PERMANOVA with 999 permutations. The differences in alpha-diversity indexes across salivatypes were tested using the Kruskal–Wallis test, and pairwise comparisons were performed by the Dunn’s test. ****P < 0.0001. Asterisks next to the taxa names in the legends of (C–E) denote taxa with significantly different relative abundances across salivatypes based on unadjusted P-values. Differences in proportion of the three salivatypes across CKM stages were tested using the Chi-square tests. Models were adjusted for age, sex, BMI, race, marital status, education level, PIR, smoking status, drinking status, physical activity (MET-minutes/week), total energy intake (kcal), antidiabetic drug use, and oral health. Abbreviations: DMM Dirichlet multinomial mixtures, ASVs amplicon sequence variants, Faith’s PD Faith’s phylogenetic diversity, CKM cardiovascular, kidney, and metabolic syndrome, 95% CI 95% confidence interval, OR odds ratio, T1 the first tertile, T2 the second tertile, T3 the third tertile, A/P ratio, Actinomycetota to Pseudomonadota ratio.

Associations between dietary indexes and CKM stages

Dietary indexes differed significantly across CKM stages (Fig. 4A and Table S12). With increasing CKM stage, DII showed a significant increasing trend, while other healthy dietary indexes showed a significant decreasing trend. CKM stage distributions also differed significantly across tertiles of dietary indexes (Fig. 4B). For DII, the proportion of stage 0 decreased, whereas stages 3 and 4 increased from T1 to T3. For the four healthy dietary indexes, T3 showed the highest proportion of stage 0-1 and the lowest proportion of stages 2–4. In covariate-adjusted ordinal logistic regression, higher DII was significantly associated with higher CKM stages (OR per SD = 1.206, 95% CI: 1.070, 1.359), and both T2 (OR = 1.379, 95% CI: 1.090, 1.746) and T3 (OR = 1.479, 95% CI: 1.123, 1.946) were significantly associated with higher CKM stages compared to T1 (Fig. 4C). In contrast, all four healthy dietary indexes were inversely associated with CKM stages (MED: OR = 0.847, 95% CI: 0.766, 0.937; HEI-2020: OR = 0.848, 95% CI: 0.768, 0.936; AHEI-2010: OR = 0.801, 95% CI: 0.722, 0.887; DASH: OR = 0.828, 95% CI: 0.746, 0.919). Compared to T1, T3 of all four healthy dietary indexes was significantly associated with lower CKM stages.

Fig. 4. Associations between dietary indexes and CKM stages.

A Boxplots of dietary indexes (DII, MED, HEI-2020, AHEI-2010, and DASH) across CKM stages. B Proportion of CKM stages across tertiles of different dietary indexes. C Associations between dietary indexes and CKM stages identified by ordinal logistic regression. Differences in dietary indexes across CKM stages were tested using the Kruskal–Wallis test, and pairwise comparisons were performed by the Dunn’s test. ****P < 0.0001, ***P < 0.001, **P < 0.01, and *P < 0.05. Differences in proportion of CKM stages across tertiles of different dietary indexes were tested using the chi-square test. Models were adjusted for age, sex, BMI, race, marital status, education level, PIR, smoking status, drinking status, physical activity (MET-minutes/week), total energy intake (kcal), antidiabetic drug use, and oral health. Abbreviations: CKM cardiovascular, kidney, and metabolic syndrome, 95% CI 95% confidence interval, OR odds ratio, T1 the first tertile, T2 the second tertile, T3 the third tertile, DII Dietary Inflammatory Index, MED Mediterranean Diet Index, HEI-2020 Healthy Eating Index-2020, AHEI-2010 Alternative Healthy Eating Index-2010, DASH Dietary Approaches to Stop Hypertension Index.

Alterations of oral microbiota associated with dietary indexes

Alpha-diversity indexes significantly differed across tertiles of all dietary indexes (Fig. S4 and Table S13). Specifically, T3 in DII showed increased alpha-diversity compared to T1/T2, whereas T3 in other healthy indexes showed decreased diversity. There were significant differences in microbial composition across tertiles of all dietary indexes (Fig. S5). After covariate adjustment, dietary indexes were no longer associated with alpha-diversity indexes, while DII and AHEI-2010 remained significantly associated with beta-diversity indexes (Table S14).

Non-parametric tests across the tertiles of various dietary indices identified numerous differentially abundant taxa (Table S15–S19). A substantial overlap of 55 taxa—predominantly Pseudomonadota (n = 12), Bacteroidota (n = 12), and Bacillota (n = 11)—indicated consistent microbial alterations across dietary patterns (Fig. S6A, S6B). Notably, Pseudomonadota abundance was lower in the highest DII tertile (T3) than T1-T2, but higher in T3 for healthy indices; other phyla exhibited reverse trends (Fig. S6C). Furthermore, salivatype proportions differed significantly across tertiles of all five indices (Fig. 5A). From T1 to T3, type Pseudomonadota decreased in DII but increased in healthy indices, while type high diversity showed opposite trends. The Actinomycetota type remained stable.

Fig. 5. Associations between dietary indexes and oral microbial taxa.

A Proportion of the three salivatypes by tertiles of different dietary indexes. Significant associations between dietary indexes and oral microbial taxa from B MaAsLin3 analysis (P-FDR < 0.05) and C linear regression (P-FDR < 0.05). D Heatmap summarizing the significant associations between dietary indexes and oral microbial taxa from covariate-adjusted models. Only taxa significantly associated with at least two dietary indexes are displayed. Differences in proportion of the three salivatypes by tertiles of different dietary indexes were tested using the chi-square tests. Models were adjusted for age, sex, BMI, race, marital status, education level, PIR, smoking status, drinking status, physical activity (MET-minutes/week), total energy intake (kcal), antidiabetic drug use, and oral health. The color gradient indicates beta values from the linear regression, as shown in the color key. P-values were adjusted using FDR correction for microbial taxa at each taxonomic level. Abbreviations: T1 the first tertile, T2, the second tertile, T3 the third tertile, DII Dietary Inflammatory Index, MED Mediterranean Diet Index, HEI-2020 Healthy Eating Index-2020, AHEI-2010 Alternative Healthy Eating Index-2010, DASH Dietary Approaches to Stop Hypertension Index.

After adjusting for covariates, both the MaAsLin3 analysis and linear regression consistently showed that most significant taxa belonged to Pseudomonadota (Fig. 5B–D). DII was inversely associated with these taxa, whereas the other healthy dietary indices were positively associated. In the MaAsLin3 models, Neisseria and Kingella (Pseudomonadota) were associated with four dietary indices (Fig. 5B and Table S20). In the linear regression models, higher taxonomic levels of Neisseria (Betaproteobacteria, Neisseriales, and Neisseriaceae) were significantly associated with all five dietary indices, showing negative correlations with DII and positive correlations with the other indices (Fig. 5C). In contrast, taxa from Bacillota and Actinomycetota generally showed associations in the opposite direction. In MaAsLin3, Bifidobacterium and its higher taxonomic levels (Bifidobacteriales, Bifidobacteriaceae) were associated with at least three dietary indices (Fig. 5B and Table S20). Similarly, linear regression indicated that Bifidobacterium, Coriobacteriaceae and its higher taxonomic levels (Coriobacteriia, Coriobacteriales) were significantly associated with at least four dietary indices (Fig. 5C).

The significant taxa identified by the two models were highly consistent, supporting the robustness of our findings (Fig. 5D). Taxa significantly associated with all five dietary indices included Neisseria and its higher taxonomies (Betaproteobacteria, Neisseriales, Neisseriaceae), Pseudomonadaceae, and Anaeroglobus from Bacillota.

Oral microbiota mediates associations between dietary indexes and CKM stages

Covariate-adjusted mediation analysis showed that 11 taxa significantly mediated 36 associations between dietary indexes and CKM stages (P-FDR < 0.05), with mediation proportions ranging from 3.97% to 10.67% (Fig. 6A and Table S21). Notably, five Pseudomonadota taxa significantly mediated 22 of these associations. Neisseria and its higher taxonomic levels (Betaproteobacteria, Neisseriales, Neisseriaceae), together with Lactobacillus and Lactobacillaceae, were significant mediators for all dietary indexes. Other significant mediators included Lautropia (Pseudomonadota), Scardovia (Actinomycetota), and Flavobacteriaceae and its higher taxonomic levels (Flavobacteriia and Flavobacteriales). To identify specific dietary components related to these mediator taxa, we further examined associations with a comprehensive panel of dietary factors, comprising 37 food groups and 41 nutrients (macronutrients, vitamins, and minerals). Most significant associations were concentrated in six taxa: Betaproteobacteria, Neisseriales, Neisseriaceae, Neisseria, Lactobacillaceae, and Lactobacillus (Fig. 6B). Higher intakes of dark green vegetables, protein, dietary fiber, total fat (total polyunsaturated fatty acids), total cholesterol, lutein, food folate, total choline, phosphorus, magnesium, iron, zinc, sodium, potassium, and selenium were associated with enrichment of taxa negatively related to CKM stages (Betaproteobacteria, Neisseriales, Neisseriaceae, and Neisseria) and depletion of taxa positively related to CKM stages (Lactobacillaceae and Lactobacillus). On the other hand, carbohydrates and sugars showed the opposite pattern. These findings support the existence of complex interactions among diet, oral microbiota, and host health. Within this multidirectional framework, modulation of the diet–microbiome axis may represent a complementary strategy for promoting health and potentially influencing CKM-associated outcomes.

Fig. 6. The mediating role of oral microbial taxa in the associations between dietary indices and CKM stages, and their specific dietary associations.

A The significant mediating role of oral microbial taxa in the associations between dietary indexes and CKM stages (P-FDR < 0.05). B Associations of dietary factors with the significant mediator taxa identified in (A). Associations between dietary indexes and the oral microbial taxa were identified by linear regression, and associations between dietary indexes and CKM stages were identified by ordinal logistic regression. Significant mediations were defined by an FDR-corrected P-value of the ACME < 0.05. P-values were adjusted using FDR correction for all significant diet–oral microbial taxa–CKM associations. The figure shows the proportion of mediation (%) for each corresponding mediation analysis. Associations between dietary factors and the significant mediator taxa were identified by linear regression. P-values were adjusted using FDR correction separately across the 37 food groups and 41 nutrients. Models were adjusted for age, sex, BMI, race, marital status, education level, PIR, smoking status, drinking status, physical activity (MET-minutes/week), total energy intake (kcal), antidiabetic drug use, and oral health. Abbreviations: DII Dietary Inflammatory Index, MED Mediterranean Diet Index, HEI-2020 Healthy Eating Index-2020, AHEI-2010 Alternative Healthy Eating Index-2010, DASH Dietary Approaches to Stop Hypertension Index, F_TOTAL total intact or cut fruits and fruit juices (cup eq.), F_CITMLB intact fruits (whole or cut) of citrus, melons, and berries (cup eq.), F_OTHER intact fruits (whole or cut); excluding citrus, melons, and berries (cup eq.), F_JUICE fruit juices, citrus and non citrus (cup eq.), V_TOTAL total dark green, red and orange, starchy, and other vegetables; excludes legumes (cup eq.), V_DRKGR dark green vegetables (cup eq.), V_REDOR_TOTAL total red and orange vegetables (tomatoes + other red and orange) (cup eq.), V_REDOR_TOMATO tomatoes and tomato products (cup eq.), V_REDOR_OTHER other red and orange vegetables, excluding tomatoes and tomato products (cup eq.), V_STARCHY_TOTAL total starchy vegetables (white potatoes + other starchy) (cup eq.), V_STARCHY_POTATO white potatoes (cup eq.), V_STARCHY_OTHER other starchy vegetables, excluding white potatoes (cup eq.), V_OTHER other vegetables not in the vegetable components listed above (cup eq.), V_LEGUMES legumes computed as vegetables (cup eq.), G_TOTAL total whole and refined grains (oz. eq.), G_WHOLE whole grains (oz. eq.), G_REFINED refined or non-whole grains (oz. eq.), PF_TOTAL total meat, poultry, seafood, organ meats, cured meat, eggs, soy, and nuts and seeds; excludes legumes (oz. eq.), PF_MPS_TOTAL total meat, poultry, seafood, organ meats, and cured meat (oz. eq.), PF_MEAT, beef, veal, pork, lamb, game meat excludes organ meats and cured meat (oz. eq.), PF_CUREDMEAT cured/luncheon meat made from beef, pork, or poultry (oz. eq.), PF_ORGAN organ meat from beef, veal, pork, lamb, game, and poultry (oz. eq.), PF_POULT chicken, turkey, Cornish hens, and game birds; excludes organ meats and cured meat (oz. eq.), PF_SEAFD_HI seafood (finfish, shellfish and other seafood) high in n-3 fatty acids (oz. eq.), PF_SEAFD_LOW seafood (finfish, shellfish and other seafood) low in n-3 fatty acids (oz. eq.), PF_EGGS, eggs (chicken, duck, goose, quail) and egg substitutes (oz. eq.), PF_SOY soy products excluding calcium fortified soy milk and immature soybeans (oz. eq.), PF_NUTSDS peanuts, tree nuts, and seeds, excludes coconut (oz. eq.), PF_LEGUMES legumes computed as protein foods (oz. eq.), D_TOTAL total milk, yogurt, cheese, and whey (cup eq.), D_MILK fluid milk and calcium fortified soy milk (cup eq.), D_YOGURT yogurt (cup eq.), D_CHEESE cheese (cup eq.), OILS oils (grams), SOLID_FATS solid fats (grams), ADD_SUGARS foods defined as added sugars (tsp. eq.), A_DRINKS alcoholic beverages (no. of drinks).

Given that certain Neisseria species are confirmed nitrate reducers [22], we further investigated associations of predicted nitrate-reduction potential with CKM stages and dietary indexes. Most oral microbial taxa assigned to the FAPROTAX-predicted nitrate-reduction function belonged to the Pseudomonadota (Fig. S7A), and Neisseria abundance was strongly correlated with predicted nitrate-reduction potential (Fig. S7B). Furthermore, predicted nitrate-reduction potential decreased significantly with advancing CKM stages (Fig. S7C and Table S22). Predicted nitrate-reduction potential was negatively correlated with DII but positively correlated with other healthy dietary indices (Fig. S7D). Collectively, these findings suggest that FAPROTAX-predicted nitrate-reduction potential, potentially driven by Neisseria, may represent a functional link between dietary patterns and CKM progression.

Additionally, we examined the associations between covariates and Neisseria, revealing significant correlations with race and sex. Notably, Black individuals and smokers exhibited the significantly lowest abundance of this genus (Table S23).

Discussion

In this nationally representative US population, our study provides novel insights into the interplay among dietary patterns, oral microbiota, and CKM syndrome. As CKM stages worsen, oral microbial composition tended to exhibit gradient alterations, particularly a decline in Pseudomonadota abundance. Neisseria and Lautropia within Pseudomonadota, as well as Gemella and Bergeyella, were negatively correlated with CKM stages, whereas Lactobacillus was positively correlated. Pseudomonadota-dominated salivatypes were associated with the lowest risk of advanced CKM stages. Healthier dietary patterns were associated with lower risk of advanced CKM and were consistently linked to higher abundance of Pseudomonadota taxa (e.g., Neisseria and Kingella). Moreover, Neisseria and its higher taxonomies predominantly mediated the association between dietary indexes and CKM stages.

The core oral microbiome in our study resembled that reported in healthy individuals [50], with Streptococcus, Veillonella, Neisseria, Haemophilus, Rothia, Actinomyces, Prevotella, and Fusobacterium being highly abundant. Our exploratory finding of reduced alpha-diversity and compositional shifts across CKM stages extends prior reports in obesity, T2DM, and CKD [10–12]. Although evidence linking oral microbiota directly to CVD remains inconsistent, likely due to small sample sizes [18, 51, 52], our findings suggest that increasing cardiovascular, kidney, and metabolic multimorbidity tends to parallel a gradient shift in the oral microbiome. Notably, the largest decline in alpha-diversity occurred between stages 0–2 (pre-CVD) and 3–4 (subclinical and clinical CVD), and only stage 4 showed a microbial structure distinct from all other stages, suggesting that cardiovascular damage may mark a key point in oral microbial shifts within the CKM framework.

With CKM progression, taxa within Pseudomonadota decreased while Actinomycetota increased. Consistent with this pattern, individuals with diabetes have shown lower Pseudomonadota and higher Actinomycetota abundance than healthy controls [53]. After covariate adjustment, the inverse association between Pseudomonadota and CKM stages remained robust. At the genus level, Neisseria and Lautropia (Pseudomonadota), as well as Gemella (Bacillota) and Bergeyella (Bacteroidota), were negatively correlated with CKM stages. Previous studies have reported Neisseria enrichment in CKD and T2DM [14, 15], whereas others found an inverse association with lipid profiles [52]; most were limited by small sample sizes (n < 500). The established role of certain oral Neisseria species as nitrate reducers provides a mechanistic basis for their beneficial effects on cardiometabolic health via the nitrate-nitrite-nitric oxide (NO) pathway [22, 54]. Lautropia, a beneficial oral commensal, is depleted with CKD [55], and some species within this genus also reduce nitrate and may similarly confer cardiovascular benefits. Although small studies have reported oral Gemella enrichment in CKD and T2DM [8, 56], the negative association observed here may reflect differences in study populations and our larger sample. Oral Bergeyella has shown a nominal inverse association with incident hypertension [57], and depletion of its newly classified family, Weeksellaceae, has been linked to gestational diabetes [58], broadly supporting our findings. Conversely, we found that Lactobacillus was positively correlated with CKM stages, consistent with reports of higher salivary Lactobacillus in atherosclerosis [59] and strong evidence for its enrichment in T2DM saliva from a systematic review [60]. As the first nationally representative study of CKM-related oral microbial changes, our findings provide relatively robust evidence.

We further defined salivatypes based on distinct clusters of the oral microbiota and investigated their associations with CKM stages. While enterotypes based on gut microbiota have been widely studied and are strongly linked to diet and lifestyle [61, 62], salivatypes remain less explored [19, 29, 49]. In our study, the three salivatypes closely resembled those reported in a European population, being dominated by Neisseria, Prevotella (with high alpha-diversity), and Rothia, respectively [29]. This suggests that, similar to enterotypes, salivatypes show robust clustering across populations [61]. With CKM progression, the proportion of Pseudomonadota-dominated (Neisseria) salivatypes decreased, whereas Actinomycetota-dominated (Rothia, a health-associated taxon around teeth or implants) salivatypes increased. This pattern is consistent with a prior study in diabetes [29], although that study did not adjust for covariates. In contrast, after covariate adjustment, we found that Pseudomonadota-dominated salivatypes were associated with the lowest risk of advanced CKM stages. These findings extended current understanding of salivatype compositional changes across CKM stages, particularly the reduction in Pseudomonadota-dominated salivatypes with higher CKM stages.

Our analysis reinforces the established benefits of healthy dietary patterns for cardiometabolic health [28]. Diets such as the MED, characterized by higher intakes of fruits, vegetables, legumes, seafood, and nuts, and lower intakes of red and processed meat and added sugars [24], are associated with improved outcomes in T2DM, CKD, and CVD [26, 27, 63]. We further showed that greater adherence to healthy dietary patterns was robustly associated with lower CKM stages after covariate adjustment. A proteomic study suggested that immune function and metabolic regulation may represent a shared pathway linking dietary patterns to cardiometabolic health [28], and pathways related to lipid metabolism and immune function are central to this process [64]. Notably, the inflammatory DII increased significantly only at advanced CVD stages (stages 3–4), whereas other healthy dietary indexes declined from the metabolic disorder stage (stage 2). This pattern suggests that different dietary dimensions may affect distinct phases of CKM progression, which warrants future investigation.

In this study, the oral microbial alterations associated with several dietary indexes were highly consistent. Healthier dietary patterns were primarily associated with enrichment of Pseudomonadota and Neisseria and depletion of Bifidobacterium and Lactobacillus. Consistent with these findings, salivary Neisseria has been linked to vegetarian diets and high dietary fiber intake [65], both characteristic of healthy diets. In contrast, high intake of sugar-sweetened beverages has been associated with increased salivary Bifidobacteriaceae and Lactobacillus [66], and oral Lactobacillus is more abundant in high-sugar intake groups [67]. These findings support our results, as healthy dietary patterns generally involve lower sugar intake. Notably, salivary pH after glucose loading has been negatively correlated with Neisseria abundance [68], whereas the acidogenic and aciduric properties of Bifidobacterium and Lactobacillus favor their proliferation in low-pH environments [69, 70]. Thus, unhealthy diets (e.g., high sugar intake) that promote oral acidification may plausibly deplete Neisseria while enriching Bifidobacterium and Lactobacillus.

Our mediation analysis further identified Neisseria-related taxa and Lactobacillus as mediators of the associations between healthy dietary patterns and CKM stages. In supplementary exploratory analyses based on taxonomy-to-function annotation, predicted nitrate-reduction potential was positively correlated with Neisseria abundance and healthy dietary indices, but declined with advancing CKM stages. These findings should not be interpreted as evidence that nitrate reduction was directly measured or that nitrate-reducing activity itself mediated the diet–CKM association. While certain oral Neisseria species are established nitrate-reducing bacteria [22], this trait is not universal within the genus. Rather than acting in isolation, Neisseria likely participates within a broader oral denitrification network where reduced nitrogen intermediates generated by one taxon can serve as substrates for others. Furthermore, specific dietary components, including dark green vegetables, protein, and fiber, were associated with higher Neisseria and lower Lactobacillus abundance, whereas higher carbohydrate and sugar intake was associated with lower Neisseria and higher Lactobacillus. Consistently, enrichment of salivary Neisseria species is linked to vegetarian diets and high dietary fiber intake [65], and individuals with Neisseria-dominated salivatypes reportedly consumed less sugar [29], whereas high sugar intake is associated with oral Lactobacillus enrichment [66, 67]. Rothia—another well-established oral nitrate reducer—was enriched in the Actinomycetota-dominated salivatype in our study, and this Rothia-enriched salivatype was associated with higher carbohydrate intake [29]. We therefore hypothesize that healthy diets, partly through lower sugar intake and higher protein availability, help maintain oral pH homeostasis and favor Neisseria over acidogenic and aciduric Lactobacillus [70] and carbohydrate-preferring taxa like Rothia, whereas poor diets may induce acidification, drive these microbial shifts, and contribute to CKM progression.

The mechanism may center on the nitrate-nitrite-NO pathway, coordinated by this complex oral denitrification network. Nitrate-reducing bacteria such as Neisseria convert dietary nitrate into nitrite, which is further metabolized to NO in the circulation [21]. NO may confer cardiometabolic benefits by reducing oxidative stress, lowering blood pressure, and improving endothelial function [6, 21]. Green vegetables are the major dietary nitrate source, and several studies have shown Neisseria enrichment following intake of nitrate-rich foods [22, 71], consistent with our findings. Dietary nitrate has also been associated with benefits for glucose metabolism, blood pressure, visceral fat accumulation, and cardiovascular and renal complications of T2DM [21]. Mechanistically, the nitrate-nitrite-NO pathway acts largely through mitochondria, with nitrate-derived reactive nitrogen species (RNS) interacting with multiple mitochondrial targets [54]. NO can bind cytochrome c oxidase and inhibit respiration, thereby affecting bioenergetics and cell signaling, while RNS can also interact with respiratory chain complex I [72] and reduce mitochondrial ROS. Other proposed mechanisms include AMPK activation, GLUT4 translocation, and modulation of NADPH oxidases [54]. We therefore hypothesize that healthy dietary patterns rich in green vegetables may increase Neisseria abundance, enhance beneficial NO production, and reduce the risk of CKM. Since epidemiological studies have not shown adverse health effects of dietary nitrate intake, increasing daily vegetable consumption may represent a safe strategy for CKM prevention. Our findings should be viewed as hypothesis-generating support for future studies testing whether vegetable-rich dietary patterns, oral microbial ecology, and nitrate–nitrite–NO metabolism jointly influence CKM-related outcomes.

Our study has several strengths. First, it is the first to report potential gradient alterations in oral microbial communities associated with CKM stages, advancing understanding of cardiovascular, kidney, and metabolic multimorbidity. Second, it is based on oral microbiome data from a nationally representative US population, enhancing the robustness and representativeness of our findings. Third, to reduce confounding, we excluded individuals with recent antibiotic use and adjusted for glucose-lowering medications using detailed prescription information [30]. However, several limitations warrant consideration. First, the cross-sectional design precludes causal inference; longitudinal studies are essential to clarify these associations. Second, although models with ordinal outcomes showed significant associations, they do not definitively establish a strictly progressive shift in diet or the microbiome across CKM stages, but rather broader directional trends. Third, nitrate-reduction potential was inferred using a taxonomy-to-function annotation approach rather than directly measured. Given the limitations of literature-curated taxon–function annotations, these nitrate-related findings should be interpreted as exploratory and require validation in future metagenomic, metabolomic, and experimental studies. Fourth, 16S rRNA gene sequencing limits species-level and functional resolution, including identification of species directly responsible for nitrate reduction, and certain genus-level assignments may reflect outdated nomenclature because the older SILVA v123 database was used. Future metagenomic studies will be needed to address these gaps. Fifth, oral rinse samples inherently capture mixed microbial communities from multiple habitats rather than site-specific niches. Sixth, despite extensive adjustments, the potential for residual confounding from unmeasured factors and recall bias from dietary data cannot be fully excluded. Finally, the findings are based on a US adult population, which may limit their generalizability to other populations.

Conclusions

In conclusion, analyses of oral microbiome data from a nationally representative US population demonstrated that CKM progression was accompanied by a trend toward gradient shifts in the microbial community, characterized primarily by a decline in Pseudomonadota. Within this phylum, Neisseria and Lautropia were inversely associated with CKM severity. Healthier dietary patterns were associated with lower risk of advanced CKM stages and were primarily linked to enrichment of Pseudomonadota taxa, particularly Neisseria. Neisseria-related taxa predominantly mediated the associations between dietary indexes and CKM stages. These findings provide association-based, hypothesis-generating evidence for complex diet–oral microbiota–CKM interactions and support future longitudinal and mechanistic studies rather than establishing a causal dietary prevention strategy.

Author contributions

All authors were responsible for the study concept and design. Weitao Su, Yayuan Mei, and Ang Li obtained funding. Weitao Su, Yayuan Mei, Ming Yang, and Ang Li conducted the literature review and designed the statistical methodology. Weitao Su, Yayuan Mei, and Jingtao Wu participated in the pre-processing of the datasets. Weitao Su and Yayuan Mei did the statistical analysis and drafted the manuscript. Weitao Su, Yayuan Mei, Ming Yang, and Ang Li critically reviewed and improved the drafts of the manuscript. All authors read and approved the final manuscript.

Funding

The authors disclose support for the research of this work from the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0513200), the National Natural Science Foundation of China (82404365, 82404278), Young Elite Scientists Sponsorship Program of the Beijing High Innovation Plan (20250761), the Joint Funds for the Innovation of Science and Technology, Fujian Province (2024Y9367), the Natural Science Foundation of Fujian Province (2025J01820), and the Institutional Fund of The Second Affiliated Hospital of Fujian Medical University (BS202503).

Data availability

The datasets supporting the conclusions of this article are publicly available from the NHANES repository: sample information is available at https://wwwn.cdc.gov/nchs/nhanes/, and oral microbiome diversity/taxonomic data are available at https://wwwn.cdc.gov/Nchs/Nhanes/Omp/. NHANES only provides diversity and taxonomic data, but does not provide raw sequencing data.

Code availability

The codes used for statistical analysis in this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethical approval

This study was sanctioned by the National Center for Health Statistics (NCHS) Ethics Review Board, and all participants provided written informed consent. All methods were performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki and applicable NCHS/NHANES protocols.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41387-026-00455-5.