Meta-analysis of oral microbiome reveals sex-based diversity in biofilms during periodontitis

Abstract

Sex is an often overlooked, yet compulsory, biological variable when performing biomedical research. Periodontitis is a common yet progressively debilitating chronic inflammatory disorder affecting the tissues supporting teeth that ultimately leads to tooth loss if left untreated. The incidence of periodontitis is sex biased, with increased prevalence in males compared with females but with unknown etiology. We performed a sex-specific meta-analysis using publicly available oral microbiome data from different sampling sites of patients with periodontitis and periodontally healthy controls; sex balance was established for each periodontal health condition. Our results show sex-based diversity in oral biofilms of individuals with periodontitis but not in their saliva, with increased abundance of several periodontal pathogens in subgingival plaques from females compared with males. We devised a quantitative measure, uniquely defined as the Microsexome Index (MSI), which indicates that sexual dimorphism in subgingival bacterial composition is a distinct feature of reduced microbial diversity during periodontitis but not under healthy conditions. In addition, we found that smoking exacerbates microsexome diversity in supragingival biofilms, particularly during periodontitis. Taken together, we provide insights regarding sex-based diversity in periodontitis, a disease with multiorgan associations, and provide the rationale for further mechanistic, diagnostic, and therapeutic studies.

Keywords: Epidemiology; Inflammation; Microbiology.

Figure 1. Overview of main determinants driving differential enrichment of microbial composition between sexes.

(A) Variable correlation plot, eigenvalues bar plot, and contribution of variables to PCA dimensions 1 and 2 (Dim1 and Dim2, respectively) (left to right). Arrows (left panel) show color gradient from blue to red, representing low to high importance. Red dashed lines (far right panels) indicate expected average contribution of variables to PCA Dim1 and Dim2. Smoking, periodontitis, (sampling) site, and BioProject are identified variables contributing to the most substantial variance in data set. (B) Sample size for each site and periodontal condition (left). Studies on y axis grouped by disease status (periodontitis [P], violet; healthy controls [H], green) and ordered by decreasing sample size (top to bottom) within each site. AUC-ROC for genus-level random forest (RF) classifiers (middle). Classification of saliva samples was poor for both periodontal conditions (periodontitis: AUC = 0.60; healthy periodontium: AUC = 0.63). Plaque microbiome distinguishes females (Fs) from males (Ms) with healthy periodontium for 1 data set (AUC = 0.70), while biofilm microbiome (dental, subgingival plaques) distinguishes Fs from Ms during periodontitis with high to very high ability to classify in 4 data sets. Number of genera enriched in Fs (pink) and Ms (blue) for each study; *q < 0.05 by Welch’s test, Benjamini-Hochberg FDR correction (right). Only biofilm-associated microorganisms during periodontitis show differences between sexes. (C) Qualitative details of sex-based differential enrichment in genera during periodontitis; *q < 0.05 by Welch’s test, Benjamini-Hochberg FDR correction. (D) Relative abundance plot of phyla by sampling site in Fs and Ms during periodontitis (left) and in healthy individuals (right).

Figure 2. Meta-analysis reveals disease-associated sex-specific enrichment of biofilm composition from dental/subgingival plaques and reliably predicts sex during periodontitis.

(A) Relative abundance plot of global phyla by (sampling) site in females (Fs) and males (Ms) without (top) and with (bottom) periodontitis from 7 included studies. (B) Heatmap showing log₁₀ mean difference between sexes at phylum level across sites during periodontitis, with significant sex-specific enrichment indicated by asterisks (*q < 0.05 by Welch’s test, Benjamini-Hochberg FDR correction). Mean difference of phyla shown by sex, with pink/blue indicating enrichment in Fs and Ms, respectively. Opacity ranges from mean difference of 0.05 (M) and 0.025 (F) to 0 (white), wherein white indicates no difference between sexes. (C) α-Diversity (observed ASVs) by sex in healthy controls (left) and during periodontitis (right) across sites show reduced diversity in Fs at level of dental (healthy) and subgingival (periodontitis) biofilms; *q < 0.05, **q < 0.01 between sexes by Wilcoxon rank-sum test. (D) β-Diversity of samples from all 7 studies based on Bray–Curtis distance indicate strong influence of sampling. Box plots show samples projected onto first 2 principal components broken down by site; P values by Kruskal-Wallis test. (E) α-Diversity of biofilms does not correlate with age in both sexes. Salivary α-diversity shows weak, negative correlation with age in periodontally healthy Ms after calculation of Pearson’s correlation coefficient. (F) AUC-ROC curves built with RF and gradient boosting machine (GBM) show predictive power of microbiome to predict sex for each sampling site and periodontal condition. Different from salivary microbiome, biofilm-associated microbiome correctly classifies Ms versus Fs during periodontitis, while in periodontally healthy individuals, classification of sex was poor for both salivary and subgingival microbiomes.

Figure 3. Smoking exacerbates sexual dimorphism in dental plaque composition but, in subgingival plaque, is independent of smoking and unique to periodontitis.

Individuals with unknown smoking status were excluded. (A) Alluvial plot showing sexually dimorphic genera and their position across reported axes strata. Splines correspond to genera identified as consistently different between males (Ms) and females (Fs) using multiple differential abundance (DA) analysis methods (see Supplemental Figure 4 for DA outputs). Axes correspond to periodontal condition, smoking habits, and (sampling) site; each genus is indicated by a specific color. (B) Multiple response analysis, reporting number of diverse genera by sex across combinations of periodontal condition, smoking habits, and site (black dots, connecting bars). More diverse genera based on sex are found in supragingival plaque of smokers during periodontitis. Summary of number of different genera by sex in corresponding conditions. (C) Venn diagrams show only 1 genus (i.e., Desulphovibrio) as consistently different between sexes in 2 different sites, saliva and supragingival plaque, of nonsmokers with periodontitis. Differential enrichment of subgingival plaque by sex is only evident during periodontitis, independent of smoking. (D) α-Diversity stratified by sex across periodontal condition and smoking habits show sex-based differences in α-diversity in saliva and subgingival biofilms from nonsmokers with periodontitis. *P<0.05, **P<0.01 by Wilcoxon rank-sum test. (E) β-Diversity by smoking status in each disease condition. Segregation patterns are affected by site and smoking habits, determined by Kruskal-Wallis test.

Figure 4. The Microsexome Index (MSI) represents a quantitative measure of sex-specific diversity.

In order to reduce bias from smoking, MSI was calculated based on subgingival microbiome composition of nonsmokers, excluding individuals with either unknown smoking history or who reported being smokers. (A) Genera differentially abundant between sexes in periodontally healthy individuals and during periodontitis, after applying FDR to 2-tailed t test (31), were used to compute MSI, a summary statistic of observed sex-specific microbial diversity wherein values different from zero indicate increased sex-specific enrichment (left). Violin plots show statistical difference in MSI between sexes by periodontal condition, indicating MSI as a summary measure of sex-predictive power of subgingival microbiome (right). ***P < 0.001; ****P < 0.0001 by Wilcoxon rank-sum test. TM7x, Saccharibacteria. (B) Relationship of MSI with subgingival microbial richness overall (left) and by sex (right), calculated by applying Pearson’s correlation coefficient. Different from periodontally healthy controls (top left), MSI shows significant, inverse relationship with subgingival microbial richness during periodontitis (bottom left). When dissecting information by sex (right), periodontally healthy males uniquely show a significant, direct correlation of MSI with richness. (C) Relationship of MSI and subgingival microbial richness with age by periodontal condition using Pearson’s correlation coefficient, with evidence of a trend toward increased MSI and parallel decrease in richness with age, not observed in healthy controls. (D) Consistent with this, MSI reliably predicts sex in both periodontal conditions, but addition of age increases predictive accuracy only during periodontitis. Data visualization refers to healthy Fs (N = 21), healthy Ms (N = 18), periodontitis Fs (N = 15), and periodontitis Ms (N = 11). AUC-ROC with RF is applied.