Metagenomic analysis reveals associations between salivary microbiota and body composition in early childhood

Abstract

Several studies have shown that body mass index is strongly associated with differences in gut microbiota, but the relationship between body weight and oral microbiota is less clear especially in young children. We aimed to evaluate if there is an association between child growth and the saliva microbiome. We hypothesized that associations between growth and the saliva microbiome would be moderate, similarly to the association between growth and the gut microbiome. For 236 toddlers participating in the New Hampshire Birth Cohort Study, we characterized the association between multiple longitudinal anthropometric measures of body height, body weight and body mass. Body Mass Index (BMI) z-scores were calculated, and dual-energy x-ray absorptiometry (DXA) was used to estimate body composition. Shotgun metagenomic sequencing of saliva samples was performed to taxonomically and functionally profile the oral microbiome. We found that within-sample diversity was inversely related to body mass measurements while community composition was not associated. Although the magnitude of associations were small, some taxa were consistently associated with growth and modified by sex. Certain taxa were associated with decreased weight or growth (including Actinomyces odontolyticus and Prevotella melaninogenica) or increased growth (such as Streptococcus mitis and Corynebacterium matruchotii) across anthropometric measures. Further exploration of the functional significance of this relationship will enhance our understanding of the intersection between weight gain, microbiota, and energy metabolism and the potential role of these relationships on the onset of obesity-associated diseases in later life.

Figure 1

Associations between body mass index (BMI) measured at 3 or 4 years of age versus child growth metrics measured between 0 and 2 years of age. (A) Age and sex adjusted BMI z-score versus rapid weight gain stratified by sex for 157 children. Wilcoxon p-value indicates difference in BMI z-score by rapid weight gain group for females and males. (B) Growth charts using the weight-for-length (weight-for-height) ratio plotted for 195 children between the ages of 0 and 2 stratified by BMI percentile groupings at 3 or 4 years of age. Growth indices are colored by the sex of the child.

Figure 2

Saliva microbiome composition among 236 children grouped by body mass index percentile groupings. (A) Phylum-specific composition by mean relative abundance. (B) The composition of the top 10 genera based on the highest mean relative abundance across all samples. (C) The composition of the top 10 species based on the highest mean relative abundance across all samples. Color ranges (i.e., spectrum of blue for Streptococcus) for species are used to delineate species from the same genus. UW = underweight.

Figure 3

Comparison of associations between growth metrics and microbes. (A) Venn diagram depicting the top 10 genera by lowest p-value produced from adjusted MaAsLin2 regressions. (B) Venn diagram depicting the top 10 species by lowest p-value. For both (A) and (B), blue and red are indicative of positive and negative coefficients respectively. (C) Dot and whisker plots to represent the relative abundance change attributable to the child growth metric. Each row represents the coefficient estimate from a different linear regression model. These adjusted regression models included the exposure (growth variable) and the following covariates: delivery mode (vaginal or cesarean), sex (male or female), sample age in days, maternal BMI, gestational age in weeks, and solid foods start age in months. The sample size of the adjusted models for the growth metrics measured at the time of saliva sample collection was 202. The sample size for the rapid weight gain model was 138. Species were selected for univariate linear regression analysis due to their overlap in the species-level Venn diagram.

Figure 4

Assessing the joint effects of body mass and sex on saliva microbiota at 3 or 4 years of age. MaAsLin2 models included in addition to the interaction term: the growth variable, sex (male or female), delivery mode (vaginal or cesarean), sample age in days, maternal BMI, gestational age in weeks, and solid foods start age in months. Black circles indicate a p-value < 0.1. (A) Coefficient for interaction between female sex and child growth metric (age and sex adjusted BMI z-score or total fat mass in grams using a DXA scan) on the relative abundance of genera. (B) Coefficient for the interaction between female sex and child growth metrics on the relative abundance of species. (C) and (D) use the same data but the models for the interaction term represents joint effects with males instead of females. For species, only associations with an effect size > 0.005 or < − 0.005 are included.

Figure 5

Exploratory analysis of body mass metrics and saliva microbiome functional profiles from pathway analysis. (A) Tile plot demonstrating the effect size and statistical significance of associations between child growth metrics (age and sex-adjusted BMI and DXA measured total fat mass in kilograms) and functional pathways. Effect sizes are derived from MaAsLin2 analyses. The models were adjusted for delivery mode (vaginal or cesarean), sex (male or female), sample age in days, maternal BMI, gestational age in weeks, and solid foods start age in months. Black circles represent a p-value < 0.15. Only associations with an effect size absolute value greater than 0.00001 in at least one of the two models were included in the plot. (B) and (C) Species-specific contributions of select pathways and KEGG gene families by overweight status. Pathways shown were selected from results from (A) and had the highest effect size in the two models.