Analysis of the Associations Between the Human Fecal Microbiome and Bone Density, Structure, and Strength: The Osteoporotic Fractures in Men (MrOS) Cohort

Abstract

In preclinical models, the composition and function of the gut microbiota have been linked to bone growth and homeostasis, but there are few available data from studies of human populations. In a hypothesis-generating experiment in a large cohort of community-dwelling older men (n = 831; age range, 78-98 years), we explored the associations between fecal microbial profiles and bone density, microarchitecture, and strength measured with total hip dual-energy X-ray absorptiometry (DXA) and high-resolution peripheral quantitative computed tomography (HRpQCT) (distal radius, distal and diaphyseal tibia). Fecal samples were collected and the 16S rRNA gene V4 hypervariable region sequenced. Sequences were bioinformatically processed through the DADA2 pipeline and then taxonomically assigned using SILVA. Generalized linear models as implemented in microbiome multivariable association with linear models (MaAsLin 2) were used to test for associations between skeletal measures and specific microbial genera. The abundances of four bacterial genera were weakly associated with bone density, structure, or strength (false discovery rate [FDR] ≤ 0.05), and the measured directions of associations of genera were generally consistent across multiple bone measures, supporting a role for microbiota on skeletal homeostasis. However, the associated effect sizes were small (log2 fold change < ±0.35), limiting power to confidently identify these associations even with high resolution skeletal imaging phenotypes, and we assessed the resulting implications for the design of future cohort-based studies. As in analogous examples from genomewide association studies, we find that larger cohort sizes will likely be needed to confidently identify associations between the fecal microbiota and skeletal health relying on 16S sequencing. Our findings bolster the view that the gut microbiome is associated with clinically important measures of bone health, while also indicating the challenges in the design of cohort-based microbiome studies. © 2022 American Society for Bone and Mineral Research (ASBMR).

Keywords: AGING; ANALYSIS/QUANTITATION OF BONE; EPIDEMIOLOGY; SYSTEMS BIOLOGY - BONE INTERACTIONS.

Figure 1.

Selection of participants; fully described in Methods.

Figure 2.

Associations between gut microbial taxa and bone measures. Differences in microbial composition per standard deviation of each bone measure were assessed using generalized linear models in MaAsLin 2, and the log₂ fold change and FDR-adjusted p-value for each taxon were used to create the heatmap. Only those associations with FDR <0.25 are included. A hierarchical clustering was performed to group associations across the variety of skeletal measures. *p< 0.05

Figure 3.

Finite mixture analysis of association z-score distribution. The smooth black curve is an estimate of the hypothetical normal distribution of null associations, based on the observed range and density of z-scores near zero. The jagged light-blue line shows the distribution (in bins of 1/14 ~ 0.0714 z-score units) of all actual, observed z-scores of the taxa-skeletal measure associations. The colored areas represent the number of taxa-skeletal measure associations not expected to occur within the hypothetical normal distribution of null effects, and thus representing potentially true associations: red the number of potentially true negative effects, and green the potentially true positive effects. The false discovery rates at each location on the z scale can be visually estimated by comparing the height of the red or green areas at that location to the height of the overall histogram (light-blue line) at the same location. See Supplemental Statistical Methods for a fuller description of the finite mixture model.