Dysbiotic Microbiome–Metabolome Axis in Childhood Obesity and Metabolic Syndrome

Abstract

The prevalence of childhood metabolic syndrome (MetS) and obesity is rising, with emerging evidence suggesting these conditions negatively affect oral health. However, the underlying molecular determinants are unclear. This study investigated the oral microbiome, inflammatory markers, and metabolites in children with obesity and MetS to explore the interrelationships between systemic disease and oral health. We recruited 76 periodontally healthy, caries-free individuals aged 10 to 17 y into 3 groups: MetS (29), metabolically healthy obesity (MHO) (30), and normal-weight healthy (NWH) controls (17). Unstimulated saliva was collected. Bacterial DNA was isolated, V3–V4 regions amplified, and 16S sequencing performed on the Illumina MiSeq platform. Sequences were annotated against the HOMD database. Multiplex assays quantified adipokines and cytokines, with significance determined by Tukey honestly significant difference. Gas chromatography/mass spectrometry identified metabolite peaks that were annotated against the Small Molecule Pathway Database, with enrichment analysis determining significance. Integrated multiomics analysis was performed using multiblock sparse partial least squares regression discriminant analysis. The MHO and MetS groups demonstrated lower abundances of Streptococcus, Actinomyces, and Schaalia and higher levels of Aggregatibacter, Campylobacter, Alloprevotella, Prevotella, Leptotrichia, and Porphyromonas compared with NWH, despite similar clinical oral status in all cohorts. MetS and MHO also had increased leptin, tumor necrosis factor–α, interleukin (IL)–1β, and IL-15 and decreased adiponectin levels versus NWH. Disease-associated metabolites, including glutamate, cholesterol, isoleucine, tyrosine, phenylalanine, serine, and indoleacetic acid, were significantly enriched in the MetS and MHO groups. Integrated multiomic analysis identified key correlations in the saliva of subjects with metabolic health or disease. Decreases in health-associated species and increases in proinflammatory cytokines and disease-associated metabolites in the saliva of MetS and obese adolescents with clinical oral health indicate an “at-risk” environment, potentially explaining their elevated risk for oral diseases. Increased salivary leptin and decreased adiponectin levels highlight the potential of saliva as a noninvasive biomarker source for childhood MetS.

Introduction

Seventeen percent of all children and adolescents suffer from obesity in the United States, and the prevalence of obesity has tripled over the past 4 decades, resulting in significant morbidity and mortality (Weiss et al. 2004). Obese individuals are at 4 times higher risk for developing metabolic syndrome (MetS) (Rahmawati et al. 2024). MetS is a cluster of several biochemical and metabolic abnormalities including insulin resistance, glucose intolerance, dyslipidemia, and hypertension, in addition to increased waist circumference (Janem et al. 2017).

Childhood MetS and obesity are strongly linked with metabolic disease in adulthood (Magnussen et al. 2010). Children diagnosed with MetS are more prone to developing type 2 diabetes as adults and have an increased risk for cardiovascular disease, obesity, and dyslipidemia (Morrison et al. 2008). Emerging evidence shows that MetS in children and adolescents is associated with an increased risk for caries and periodontal diseases (Zeigler et al. 2015; Martens et al. 2017; Manohar et al. 2020). Recently, childhood oral infections were shown to be associated with MetS and adverse metabolic parameters in adulthood, suggesting a possible bidirectional relationship (Pussinen et al. 2020).

While these studies link oral conditions and childhood MetS, the molecular underpinnings underlying these associations are lacking. The oral cavity is an open ecosystem hosting biofilm that consists of hundreds of microbial species continually interacting with the host immune system at the gingival barrier. Homeostasis in these interactions defines health, while disruptions in this homeostasis (dysbiosis and dysregulated inflammation) lead to disease (Lamont et al. 2018). Small molecules (metabolites) in this environment produced by host cells and microbes may be useful for determining the interrelationships between the resident microbiome and immune cells and may provide evidence for mechanistic relationships between MetS and oral health.

Hence, we aimed to characterize the oral microbiome–metabolome–immune axis in children and adolescents with obesity and MetS. Inadequate control for host-confounding variables, such as presence of oral clinical disease and inflammation, can obscure true associations in microbiome studies while limiting the likelihood of identifying potential causal relationships. Thus, we investigated these microbiome, metabolome, and immune parameters in subjects in the absence of clinical oral disease to identify the impact of obesity and MetS on oral health within the disease-naïve oral environment. We hypothesize that altered metabolic parameters will significantly disrupt the composition of the oral microbiome and its associated ecosystem, shifting it toward a more pathogenic and immunogenic environment.

Materials and Methods

Clinical Study Design

Patient recruitment

Study approval was obtained from the University of Iowa Institutional Review Board under protocol #201907808. Seventy-six periodontally healthy, caries-free children and adolescents between the ages of 10 and 17 y were enrolled in 3 groups: normal weight healthy controls (NWH), metabolically benign obese (MHO), and metabolic syndrome (MetS). Subjects were recruited for this study at the Pediatric Cardiometabolic Disorders Clinics at Iowa River Landing Clinics and the Pediatric Dentistry Clinic at the College of Dentistry. The study and reporting conform to STROBE (Strengthening the Reporting of Observational studies in Epidemiology) guidelines.

Inclusion criteria

Diagnosis of MetS and MHO

Two board-certified pediatric endocrinologists diagnosed pediatric MetS and MHO based on the International Diabetes Federation Criteria (Zimmet et al. 2007). These criteria were chosen as they account for ethnicity-specific measurements. Pediatric MetS was diagnosed in the presence of a waist circumference greater than 90th percentile along with 2 additional factors from the following 4: high triglycerides, low high-density lipoprotein (HDL) cholesterol, high blood pressure, and increased plasma glucose. Subjects who presented with only central obesity without satisfying the diagnostic criteria for pediatric MetS were diagnosed with MHO. Subjects presenting with an absence of all the above parameters were placed into the NWH.

Periodontal health

Gingival and periodontal health were determined using the 2017 American Academy of Periodontology classification (Caton et al. 2018). All teeth exhibited attachment levels <2 mm, probing depths <3 mm, a plaque index <2, and a gingival index <2 (Löe 1967). Drs. Salman, Reyes, and Ganesan performed the oral examination (97% interrater agreeability). All participants were frequency matched for age, gender, and ethnicity. Demographic data, dental and oral hygiene history, and clinical parameters including glycosylated hemoglobin (HbA1c), insulin resistance, lipid levels, blood pressure, medications, and other comorbidities were recorded. Laboratory measurements were collected within 2 wk of sample collection from the University of Iowa Hospitals and Clinics or affiliate labs, with all samples obtained in a fasting state.

Exclusion criteria

Subjects who had received antibiotic therapy or dental cleaning within the previous 3 mo; required antibiotics before dental treatment; used immunosuppressant medications, bisphosphonates, or steroids; used nonsteroidal anti-inflammatory drugs in the past 48 h; had a prior diagnosis of type 1 or 2 diabetes; had a history of HIV; or had a history of periodontal disease were excluded. Subjects who were smokers/e-cigarette users or pregnant were also excluded. All subjects diagnosed with gingivitis (biofilm and nonbiofilm induced) or periodontitis (of any form) based on the American Academy of Periodontology/European Federation of Periodontology 2017 classification were excluded (Caton et al. 2018). Patients with active or recurrent dental caries (International Caries Detection and Assessment System score > 0) (Ai et al. 2007) were also excluded.

Sample collection

Two millilters of unstimulated saliva was collected by having subjects allow saliva to pool in their mouth for 3 min and then drool into a tube continuously for 1 min (Navazesh 1993). Saliva was immediately transferred to dry ice and stored at −80 °C. Subjects were instructed to refrain from brushing, eating, drinking, or rinsing in the 30 min prior to sample collection. All of our sample collections were collected before 1 pm.

DNA isolation and sequencing

Bacterial DNA was isolated using a Qiagen DNA MiniAmp kit (Qiagen). 16S sequencing (amplification of the V3–V4 region) was performed using an Illumina MiSeq 2*300bp sequencing platform. Sequence quality filtering, paired-end merging, Amplicon Sequence Variant inference, and chimera removal were performed using DADA2 v1.20 (Callahan et al. 2016). A naïve Bayes classifier as implemented in QIIME2 (qiime feature-classifier fit-classifier-naive-bayes, default parameters) was trained on the Human Oral Microbiome Database (16S rRNA RefSeq v15.22) for taxonomic classification of the DADA2-derived amplicon sequence variants (ASVs). Details of the parameters are listed in the supplemental text.

Comparative meta-taxonomics

Alpha diversity estimates and hypothesis testing were determined with the R packages breakaway and DivNet. Beta diversity distances and permutational multivariate analysis of variance (PERMANOVA) were calculated with QIIME2 v2023.9. Linear discriminant analysis dimensionality reduction plots were visualized using PhyloToAST v1.4 (Dabdoub et al. 2016). Both phylogenetic (weighted normalized UniFrac) and abundance-based (Bray–Curtis) distance matrices were used to estimate beta diversity. The significance of the overall groupwise clustering was determined using PERMANOVA. Differential abundance of the species-condensed data between the groups (NWH as reference) was calculated by negative binomial generalized linear modeling (GLM) on cumulative sum scaling–normalized counts (Paulson et al. 2013) using MaAsLin2 v1.18.0 (Mallick et al. 2021). False discovery rate (FDR) correction (Benjamini–Hochberg) was performed with a cutoff of q < 0.1.

Immunoinflammatory marker analysis

Interleukins and tumor necrosis factor–α (TNF-α) were assayed using the Bio-Plex Pro Cytokine 10-plex Human Panel (Bio-Rad Corporation). Levels of adipokines leptin, resistin, and adiponectin (Quantikine; R&D Systems Inc.), matrix metalloproteases (MMPs), and Th17 pathway markers (Millipore, Sigma-Aldrich) were determined according to the manufacturer’s instructions. A standard curve was generated for each assay before quantification of the samples. The Steel–Dwass all pairs test was used to determine the statistical significance of groupwise concentration differences.

Metabolomic analysis

Sample preparation for gas chromatography/mass spectrometry (GC/MS) and analysis were performed at the Iowa Metabolomics Core as previously described (Tompkins et al. 2019). Metabolite peaks identified from GC/MS were annotated against the Small Molecule Pathway Database, and the significance was determined by enrichment analysis. Barycentric visualization of metabolite relative abundance was performed using klaR v1.7.0. Differential metabolite abundance between the groups, using NWH as reference, was calculated through GLM analysis on centered log ratio (CLR)–transformed data with the R package MaAsLin2 v1.18.0 (Mallick et al. 2021). FDR correction (Benjamini–Hochberg) was performed with a cutoff of q < 0.1.

Multiomic data integration and analysis

Supervised data integration and variable selection incorporating the species-condensed/genus-condensed counts and metabolomic and immunoinflammatory marker data were performed with multiblock sparse partial least squares regression discriminant analysis (DIABLO) implemented in the R package mixOmics v6.22.0. The circosPlot function was used to depict correlations between variables across measurement types with a minimum of Rho (ρ) ≥ 0.5. Prior to analysis, entries in the counts tables below a cumulative (across all subjects) relative abundance of 10⁻⁴ were removed. To handle the compositional nature of all data types, a CLR transform was applied with Bayesian-multiplicative replacement of count zeros handled through the cMultRepl function of the R library zCompositions (v1.14.1) as an alternative to using 1 or a small fractional number (Palarea-Albaladejo and Martín-Fernández 2015).

Rigor and reproducibility

Our study used stringent inclusion criteria, controls for microbiome sequencing and metabolomics, rigorous data filtering, and robust statistical analysis to minimize confounding factors, ensure accuracy, and maintain consistency, avoiding batch effects for reliable and reproducible results. All R scripts used for the analysis are available on GitHub (https://github.com/smdabdoub/PediatricMetSPaper). Detailed measures are listed in the supplemental text.

Results

Demographics

Demographic parameters such as age, race, sex assigned at birth, and the systemic parameters HbA1c, height, and blood pressure were not different between the groups (Table). A significant difference (P = 0.043, Fisher’s exact test) was found in the prevalence of mixed and permanent dentitions. Weight (P = 4.65 × 10⁻⁹, Kruskal–Wallis), body mass index (BMI) (P = 6.56 × 10⁻¹¹, Kruskal–Wallis), and waist circumference (P = 7.39 × 10⁻¹¹, ANOVA) were significantly different between the control and the MHO/MetS groups. No statistically significant differences were observed in BMI and waist circumference between the MHO and MetS groups. Triglycerides (P < 0.001, Mann–Whitney) and glucose titers (P = 0.008, Welch’s t test) were significantly higher in the MetS group compared with the MHO group, while HDL titers (P = 1.87 × 10⁻¹¹, Welch’s t test) were significantly higher in the MHO group compared with the MetS group (Table).

Table.

Demographics of recruited subjects. ^a

Parameter	Normal-Weight Healthy (NWH)	Metabolically Healthy Obese (MHO)	Metabolic Syndrome (MetS)	P-value
Number of subjects	29	30	17
Sex assigned at birth				P = 0.47 b
Female (%)	15 (51.7%)	17 (56.5%)	7 (41.1%)
Male (%)	14 (48.2%)	13 (43.3%)	10 (58.8%)
Race				P = 0.45 b
Black/African American	8 (27.6%)	6 (20.7%)	2 (11.8%)
White/Caucasian	20 (68.9%)	21 (72.4%)	15 (88.2%)
Alaska Native	—	1 (3.4%)	—
Multiple	1 (3.5)%	1 (3.4%)	—
Age (y) (median ± SD)	13 ± 1.8	13 ± 2.2	14 ± 2.2	P = 0.61 c
Weight (kg) (median ± SD)	50.8 ± 12	93 ± 29.7	97.5 ± 27	P = 4.65 × 10−9c MHO-MetS: p = 0.19 d
Height (cm) (median ± SD)	160 ± 12.8	160 ± 30.3	166.9 ± 8.1	P ± 0.21 e
BMI (kg/m2) (median ± SD)	18.7 ± 2.9	34.5 ± 9.4	36.07 ± 7.6	P = 6.56 × 10−11c MHO-MetS: P = 0.43 d
Dentition				P = 0.043 b
Mixed dentition	8 (27.5 %)	16 (53.3%)	9 (52.9%)
Permanent dentition	21 (72.4%)	14 (46.6%)	8 (47%)
Waist circumference (cm) (median ± SD)	78.7 ± 13.2	104 ± 19.3	106 ± 217.6	P = 7.39 × 10−11e MHO-MetS: P = 0.38 f
Triglycerides (mg/dL) (median ± SD)	—	91.5 ± 87.6	161 ± 121	P < 0.001 g
High-density lipoprotein (mg/dL) (median ± SD)	—	44.5 ± 8.6	37 ± 6.9	P = 1.87 × 10−11h
Glucose (mg/dL) (median ± SD)	—	91 ± 6.6	103 ± 9.2	P = 0.008 h
HbA1c (%) (median ± SD)	—	5.3 ± 0.4	5.3 ± 0.36	P < 0.39 g
Systolic blood pressure (mm Hg) (median ± SD)	116 ± 5.32	112 ± 9.2	116.5 ± 8.3	P < 0.18 g
Diastolic blood pressure (mm Hg) (median ± SD)	70 ± 5.12	70 ± 8.0	71 ± 78.8	P = 0.36 h

SD, standard deviation.

^aNormality assessed by Q-Q plot and Shapiro–Wilk test.

^bFisher’s exact test.

^cKruskal–Wallis.

^dDunn’s test, Benjamini–Hochberg correction.

^eOne-way analysis of variance.

^fTukey honestly significant difference.

^gMann–Whitney.

^hWelch 2-sample t test.

Sequence Statistics

A total of 12,317,852 reads were sequenced from the recruited subjects from which 181,494 ASVs were recovered after quality filtering, paired-end merging, and chimera removal through the DADA2 workflow.

Children with MetS and Obesity exhibit distinct oral microbiome in the absence of clinical oral disease

Although alpha diversity demonstrated no differences between the groups, analysis of the Bray–Curtis pairwise distances revealed significant differences between the group centroids (P = 0.04, PERMANOVA) in salivary microbial composition (Fig. 1A, Supplemental Fig. 1). While dentition status differed significantly between the NWH and MetS/MHO groups, it did not influence group centroids (P = 0.912, PERMANOVA) (Supplemental Fig. 2). Notably, bacterial genera previously associated with oral disease and infections, including Porphyromonas, Prevotella, Moraxella, Leptotrichia, and Aggregatibacter, were significantly enriched in the MetS and MHO groups (FDR-adjusted P value <0.05) (Fig. 1B and Supplemental Fig. 3). Furthermore, specific species, Aggregatibacter HMT 513, Alloprevotella HMT 914, and Campylobacter HMT 044, were enriched only in the MetS cohorts (FDR-adjusted P value <0.1). Several commensal members of Streptococcus, such as Streptococcus salivarius and Streptococcus anginosus, and members of Actinomyces, such as Actinomyces graevenitzii and Veillonella dispar, and Schaalia genera were increased in the NWH cohort and decreased in the MetS and MHO cohorts (FDR-adjusted P value <0.05), demonstrating microbial disruptions even in the absence of dental caries and periodontal diseases (Fig. 1B and Supplemental Fig. 3). When microbial data were integrated with clinical metrics, relative abundances of several bacterial members, including Saccharibacteria, Leptotrichia HMT 498, Leptotrichia buccalis, and Aggregatibacter HMT 898, increased with increasing waist circumference and triglyceride levels. In contrast, levels of Streptococcus anginosus, Streptococcus constellatus, Streptococcus salivarius, Eikenella corrodens, Lautropia mirabilis, and Selenomonas decreased with increasing waist circumference and triglyceride levels.

Figure 1.

(A) Linear discriminant analysis of Bray–Curtis distances of the genera composition of periodontally healthy and normal-weight healthy (NWH), patients with metabolic syndrome (MetS), and metabolically healthy obesity (MHO). Microbial profiles of subjects clustered by metabolic health, creating 3 statistically significant clusters (P = 0.04, permutational multivariate analysis of variance). (B) Differential abundance of species-minimum taxonomic assignments performed with MaAsLin2 using a negative binomial generalized linear modeling approach. All comparisons are relative to the NWH group. Significantly increased/decreased abundance relative to NWH is indicated by red up arrows or blue down arrows, respectively, next to the group name (q < 0.1, Benjamini–Hochberg false discovery rate).

Metabolic status alters the inflammatory milieu

While the proinflammatory adipokine leptin was significantly increased in the MetS and the MHO groups compared with NWH (Steel–Dwass, P < 0.05), the cardioprotective adipokine adiponectin was decreased in the MetS and MHO groups when compared with the NWH group (Steel–Dwass, P < 0.05). Saliva from the MetS and MHO groups contained significantly greater levels of several proinflammatory cytokines, including TNF-α and interleukin (IL)–1β, when compared with NWH (Steel–Dwass, P < 0.05). In addition, MMPs such as MMP-3 were elevated in the MetS and MHO groups (Fig. 2).

Figure 2.

Levels of select statistically significant inflammatory mediators and adipokines matrix metalloproteases (MMPs) in periodontally and dentally healthy normal-weight healthy (NWH) patients, patients with metabolic syndrome (MetS), and metabolically healthy obesity (MHO). The y-axis represents the pg/mL concentration. P < 0.05, Steel–Dwass.

Specific metabolites correlate with childhood obesity and MetS

Amino acids valine, isoleucine, leucine, and tyrosine and tryptophan were significantly enriched in MetS and MHO saliva, whereas metabolites associated with carbohydrate metabolism such as pyruvate, fumarate, citrate, malate, and glycerol were significantly enriched in the NWH cohorts (FDR-adjusted P value <0.1) (Fig. 3A). When metabolomic data were integrated with clinical metrics, several metabolites, including indole acetic acid, 2-hydroxyglutarate, beta-alanine, methylmalonate, tryptophan, valine, thymine, alanine, aminoadipate, isoleucine, succinate, methionine, threonine, xanthine, cholesterol, leucine, and glutamate, demonstrated increased levels with increasing waist circumference and triglyceride levels (Fig. 4A and Supplemental Figs. 4 and 5). Levels of fumarate, citrulline, malate, pyruvate, glycerol, hypotaurine, stearate, phosphoenolpyruvate, inositol, 3-hydroxypropionate, cysteine, serotonin, nonanoic acid, ribose-5-phosphate, palmitate, asparagine, 3-hydroxypropionate, and hydroxybutyrate decreased with increasing waist circumference and triglycerides (Fig. 2B and Supplemental Figs. 4 and 5). Glucose metabolite levels remained consistent across all samples regardless of the metabolic diagnosis.

Figure 3.

(A) Barycentric plot of relative abundances of metabolites detected in saliva in normal-weight healthy (NWH), metabolic syndrome (MetS), and metabolically healthy obesity (MHO) cohorts. Selected significantly enriched metabolites (from enrichment analysis) are highlighted in different colors. (B) Multiblock sparse partial least squares regression discriminant analysis (MB sPLS-DA/DIABLO) Circos plot showing strong (ρ ≥ 0.5) correlations between metabolites, microbial species, and immune mediators in the NWH, MetS, and MHO cohorts. Curves inside the circle represent correlations between predictors, with orange lines indicating positive correlations and black indicating negative correlations. Normalized groupwise abundance of each predictor is displayed by the line plots around the outside of the circle.

Figure 4.

Linear models built with MaAsLin2 on centered log ratio (CLR)–transformed metabolite concentration data with waist circumference as a covariate demonstrate (A) metabolites exhibiting a significant increase in abundance as waist circumference increases (P < 0.1 false discovery rate [FDR]) and (B) metabolites exhibiting a significant decrease in abundance as waist circumference increases (P < 0.1 FDR).

Multiomic analysis identified discriminant patterns of correlations in metabolic health and disease in saliva

Multiomic analysis using DIABLO (Data Integration Analysis for Biomarker discovery using Latent cOmponents) showed metabolically healthy NWH controls and MetS/MHO groups were distinguished by distinct patterns of associated microbiota, immune mediators, and metabolites (Fig. 3B). Adiponectin levels were positively correlated with oral commensals such as Actinomyces graevenitzii and Prevotella salivae and health-associated metabolites such as malate, citrate, and fumarate. Fumarate, a widely recognized antioxidant and anti-inflammatory metabolite, decreased with increasing waist circumference and triglyceride levels, negatively correlated with Porphyromonas sp. and Granulicatella elegans and several proinflammatory interleukins. In addition, fumarate positively correlated with commensal members such as Prevotella salivae and Corynebacterium matruchotii (Fig. 3B).

Discussion

MetS and obesity are chronic inflammatory conditions associated with significant health risks and systemic complications (Weiss et al. 2004; Weiss et al. 2013). MetS is linked to gut microbial dysbiosis, where disruptions in the gut barrier and host metabolism contribute to a low-grade inflammatory state (Kim et al. 2020). Previous work has also indicated that MetS is associated with oral inflammation (Lamster and Pagan 2017; Pirih et al. 2021). However, most prior studies have been performed in the adult population, and no controlled study has investigated molecular alterations in the oral environment of pediatric MetS without the confounding factor of existing oral diseases. This study addresses this gap by enrolling a well-defined cohort of pediatric subjects without active caries or periodontal inflammation and identifying several individual molecular indicators and interrelated multiomic markers that support an association between childhood MetS, obesity, and a disruption in oral homeostasis.

Microbial dysbiosis is strongly linked to oral inflammation (Abusleme et al. 2013). In this study, obesity and MetS were characterized by a loss of health-associated commensal species, such as Streptococcus, Veillonella dispar, and Schaalia, and an enrichment of pathogenic species, including Porphyromonas, Prevotella, Moraxella, Leptotrichia, and Aggregatibacter. Notably, the levels of these disease-associated species increased with waist circumference and triglyceride levels in these patient cohorts. Interestingly, animal studies have demonstrated that P. gingivalis administration leads to a higher proportion of fat mass, worsened glucose tolerance, increased adipose inflammation, and enhanced gut permeability (Yoshida et al. 2021). The significant increase in Porphyromonas in the MetS and MHO cohorts, along with the direct relationships between increased abundance with both waist circumference and triglyceride levels, provides additional evidence for a potential role of Porphyromonas in human metabolic disorders (Hajishengallis and Chavakis 2021).

In parallel, systemic diseases such as MetS affect the microbiome and the host both synergistically and independently (Dabke et al. 2019). MHO and MetS patients exhibited significantly increased levels of TNF-α, IL-1β, and IL-15 in saliva. IL-15, in particular, has been implicated in promoting inflammation in adipose tissue and obesity-associated MetS (Lacraz et al. 2016). This study also revealed significant increases in specific salivary MMPs in MetS and obese cohorts. MMPs are crucial for extracellular matrix remodeling during obesity-mediated adipose tissue formation and tissue breakdown in periodontitis and dental caries. MMP-12 and MMP-3, which play roles in adipose tissue expansion and insulin resistance (Chavey et al. 2003), were significantly elevated in the MetS cohort. Furthermore, an observed increase in the obesity-associated adipokine leptin (Obradovic et al. 2021) and a decrease in the cardioprotective adipokine adiponectin was observed in MetS and obese patients (Ouchi et al. 2006). These salivary changes mirror the systemic levels of these markers typically found in MetS and obesity disease states, suggesting that saliva has potential to serve as an indicator of overall metabolic health.

The final, understudied, piece of the puzzle we examined underscores the importance of considering the role of the metabolome in shaping the oral environment. Elevated levels of amino acids such as tryptophan and valine in MetS and MHO cohorts align with prior findings, which suggest roles for these metabolites in obesity and insulin resistance (Lischka et al. 2022). Similarly, there was a marked increase in tryptophan and indole acetic acid, metabolites associated with the tryptophan pathway that modulate mucosal–microbial homeostasis across multiple body sites (Dehhaghi et al. 2019), in the MetS and MHO groups. Glycerol, typically a health-associated metabolite (Gawron et al. 2019) was enriched in the NWH cohort but significantly decreased in the MetS and MHO cohorts, further supporting the utility of metabolomics to identify disruptions in the oral environment.

Fitting multiple puzzle pieces together with integrative multiomic analysis enables the characterization of relationships among study cohorts at a more comprehensive level than when considering microbiota, metabolome, or inflammatory signatures alone (Kremer and van Steenbergen 2000; Wlodarska et al. 2017). Alterations in microbial composition have been linked to changes in adiponectin and resistin gene expression along with several metabolites in the gut environment in mice (Yao et al. 2020). This study is the first to demonstrate such a relationship in the human oral environment. In addition, this study provides evidence that health-associated metabolites and adipokines, such as fumarate and adiponectin, are positively associated with commensal bacteria and exhibit decreases in abundance as waist circumference and triglyceride levels increase. Furthermore, we found these same metabolites negatively correlate with disease-associated species (Fig. 3B). Finally, both fumarate and adiponectin were significantly elevated in NWH individuals. These findings underscore the potential roles of these metabolites as health-associated biomarkers or determinants, warranting validation in future studies. Together, our multiomic analysis provides a template for distinguishing a healthy versus disrupted oral environment in response to systemic metabolic status.

Overall, these findings support the concept of an oral–systemic connection in MetS and obesity, suggesting that MetS and obesity disrupt the oral environment even in the absence of clinical oral disease. Although cross-sectional, this study identifies potential biomarkers and pathways that could enhance our understanding of MetS-associated oral disease risks. Longitudinal studies are needed to determine whether the observed salivary biomarkers and microbial shifts precede or follow metabolic dysregulation, which could establish causality between systemic metabolic disruptions and oral diseases. Future research should extend to diverse populations and age groups to assess the generalizability of these findings and identify population-specific risk factors. In addition, the small sample size for the MetS group and the significant hormonal variations occurring during this developmental stage underscore the need for larger cohorts to validate these findings and explore the potential impact of hormonal influences. Understanding these determinants is crucial for developing personalized treatment approaches for at-risk pediatric populations.

Author Contributions

U. Salman, contributed to data acquisition, analysis, and interpretation, drafted and critically revised the manuscript; S.M. Dabdoub, contributed to data analysis and interpretation, critically revised the manuscript; A. Reyes, A. Sidahmed, K. Weber-Gasperoni, R. Brown, I. A. Evans, E. Taylor, A. Mangalam, contributed to data acquisition, analysis, and interpretation, critically revised the manuscript; L. Kanner, V. Curtis, contributed to conception or design, data interpretation, critically revised the manuscript; S.M. Ganesan, contributed to conception or design, contributed to data acquisition, analysis, and interpretation, drafted and critically revised the manuscript. All authors gave their final approval and agree to be accountable for all aspects of the work.