Vitamin D metabolites and the gut microbiome in older men

Prior studies suggest that gut microbes can influence intestinal vitamin D metabolism and that probiotics may alter circulating vitamin D levels. Low vitamin D status has been associated with multiple adverse health outcomes in epidemiologic studies, yet large trials of vitamin D supplementation have shown limited benefits for outcomes like cardiovascular disease and cancer. While 25-hydroxyvitamin D (25(OH)D) reflects vitamin D stores and is the clinical marker of sufficiency, the active hormone 1,25-dihydroxyvitamin D (1,25(OH)2D) binds the vitamin D receptor and regulates gene expression, also driving catabolism via 24-hydroxylase. Ratios reflecting activation (1,25(OH)2D/25(OH)D) and catabolism (24,25(OH)2D/25(OH)D) may better capture vitamin D metabolic flux and predict clinical outcomes. The study investigates whether circulating vitamin D metabolites and metabolic ratios are associated with gut microbiome diversity and composition in older men, hypothesizing that active vitamin D and vitamin D flux metrics, rather than storage 25(OH)D alone, relate to more favorable gut microbial profiles.

Evidence indicates bidirectional interactions between vitamin D signaling and the gut microbiome. Animal studies show that disrupting vitamin D signaling or VDR reduces beneficial Firmicutes and alters microbial composition, while probiotic interventions can increase VDR expression and protect against colitis. Human observational studies link low 25(OH)D to various diseases, though MrOS and large RCTs have not consistently shown benefits of supplementation for cardiovascular disease, cancer, or diabetes. Prior work suggests that vitamin D metabolic ratios (activation and catabolism) may better predict fractures, mortality, and CKD outcomes than 25(OH)D alone. Small human trials indicate vitamin D supplementation may modify gut microbiota primarily in deficient states. Collectively, literature supports exploring associations of active vitamin D and metabolic flux with microbiome diversity and taxa, beyond 25(OH)D levels.

Design and population: Cross-sectional analysis within the Osteoporotic Fractures in Men (MrOS) Study. From 1841 survivors at visit 4 (2014–2016), 567 men with fasting serum vitamin D metabolites and stool microbiome data were included. Participants were community-dwelling older men (mean age ~84 years) from six US sites (Birmingham, Minneapolis, Palo Alto, Pittsburgh, Portland, San Diego). IRB approvals and informed consent were obtained.

Stool collection and 16S rRNA sequencing: Participants collected fecal samples using OMNIgene-GUT OMR-200 kits and mailed to the Portland site; the first 599 samples were sequenced at Baylor College of Medicine following the Earth Microbiome Project protocol. The V4 region of 16S rRNA was amplified with primers 515F/806R and sequenced on Illumina MiSeq (2×250). Demultiplexing used bcl2fastq; primers trimmed with cutadapt 1.18. Data were uploaded to Qiita and quality controlled; forward reads trimmed to 150 nt. Features (sOTUs) were inferred with Deblur 1.1.0; phylogenetic placement used SEPP 4.3.5 against Greengenes 13.8 to minimize sequencing errors and phylogenetic bias. Taxonomic labels were assigned by traversing the insertion tree from sOTU toward the root to a known OTU.

Vitamin D metabolite measurements: Fasting serum 25(OH)D, 1,25(OH)2D, and 24,25(OH)2D3 were quantified by LC-MSMS at University Hospitals Leuven/KU Leuven. 1,25(OH)2D and 24,25(OH)2D were measured using a two-dimensional LC-MSMS method (based on Cassetta et al., modified by Vanderschueren and Cools), monitoring lithium adduct transitions on an AB Sciex 5500 QTRAP (ESI positive). Between-run imprecision: 6.7% at 40 pg/mL for 1,25(OH)2D; 7.6% at 2.0 ng/mL for 24,25(OH)2D; LLOQ: 10 pg/mL and 0.2 ng/mL respectively. 25(OH)D was measured with APCI LC-MSMS using online cleanup and Kinetex F5 analytical column; linear ranges 2–80 ng/mL (25(OH)D3) and 0.7–47 ng/mL (25(OH)D2); between-run imprecision 5.6% at 28.9 ng/mL. Laboratory performance was monitored via DEQAS participation.

Covariates: Age, race, BMI, smoking, alcohol, self-rated health, Physical Activity Scale for the Elderly score, season, medication use (antibiotics in past 30 days, vitamin D supplements, antidepressants, probiotics, laxatives, statins, antihistamines, proton pump inhibitors), and estimated dietary resistant starch intake (>1 g/100 g from 80 FFQ food sources).

Outcomes and diversity metrics: Alpha-diversity quantified by Faith’s phylogenetic diversity (PD). Beta-diversity computed using unweighted UniFrac with principal coordinates analysis (PCoA).

Statistical analyses: Non-redundant covariates explaining alpha- and beta-diversity were identified using forward stepwise redundancy analysis (RDA) with permutation testing (1000 permutations) and adjusted R2 stopping criteria (vegan package in R 3.6.1). For beta-diversity, the first 10 PCoA coordinates from unweighted UniFrac were used in RDA; sensitivity analyses using all coordinates yielded the same set of significant covariates. PERMANOVA with Benjamini–Hochberg FDR correction (q-values) tested associations of covariates with beta-diversity. Multiple linear regression models assessed associations between PD and each vitamin D measure (25(OH)D, 1,25(OH)2D, 24,25(OH)2D, activation ratio 1,25(OH)2D/25(OH)D, catabolism ratio 24,25(OH)2D/25(OH)D), adjusting for selected confounders via backward selection (liberal p<0.20), then including the vitamin D measure; diagnostics included residual plots, QQ plots, DFBETAs and sensitivity analyses. 25(OH)D was analyzed as continuous and dichotomous (<20 ng/mL deficiency) variables.

Taxa identification: Random forest classification with 5-fold-within-5-fold nested cross-validation (Python 3.6.10, sklearn.ensemble.RandomForestClassifier) was applied to samples in the top and bottom deciles for each vitamin D measure. Models with mean AUC >0.70 (kept for 1,25(OH)2D and activation ratio only) were used to identify candidate sOTUs with feature importance >0.2% (Gini importance). Directions of association were assessed with Spearman rank correlation and BH-FDR correction; sOTUs with adjusted p≤0.05 were considered significant.

Data and code: Sequencing data available at EBI (ERP107984). Analysis code available at https://github.com/knightlab-analyses/vitamin-d and Zenodo DOI 10.5281/zenodo.4123576.

• Participant characteristics: 567 older men, mean age 84.2 years (SD 4.1), mean BMI 27.0 kg/m2. Vitamin D deficiency (25(OH)D <20 ng/mL) in 7.2%; 74.8% reported vitamin D supplement use; 6.7% reported antibiotic use in past 30 days.
• Site and sun exposure: 25(OH)D varied by site consistent with sun exposure (ANOVA p=0.023), highest in San Diego; 1,25(OH)2D did not differ by site (ANOVA p=0.38). 25(OH)D correlated with 1,25(OH)2D (r=0.43, p<0.001) and 24,25(OH)2D (r=0.80, p<0.001).
• Alpha-diversity (Faith’s PD): In redundancy analysis, 1,25(OH)2D explained the largest proportion of variance in alpha-diversity (~5%). In adjusted multiple linear regression, higher PD was associated with higher 1,25(OH)2D (slope 1.39, p=7.23×10^-7), higher activation ratio (slope 1.12, p=0.0002), higher catabolism ratio (slope 0.85, p=0.003), and higher 24,25(OH)2D (slope 0.64, p=0.02). 25(OH)D was not associated (slope 0.04, p=0.88). Recent antibiotic use was associated with reduced alpha-diversity (t-test p=0.00017), but associations with 1,25(OH)2D persisted after adjustment.
• Beta-diversity (unweighted UniFrac): 1,25(OH)2D explained ~2% of variation and was the strongest assessed factor in redundancy analyses. PERMANOVA with BH-FDR correction showed significant clustering by 1,25(OH)2D (q=0.004), 24,25(OH)2D (q=0.011), activation ratio (q=0.004), and catabolism ratio (q=0.004). 25(OH)D, analyzed as continuous (q=0.32) or dichotomous deficiency variable (q=0.503), was not significant. Other significant covariates included clinical site (q=0.004), race (q=0.004), total starch intake (q=0.004), physical activity (q=0.013), alcohol intake (q=0.035), and medication use (antibiotics q=0.006; antidepressants q=0.006; statins q=0.013; PPI q=0.026).
• Specific taxa: Random forest identified 12 sOTUs associated with vitamin D metabolism (models retained for 1,25(OH)2D and activation ratio; AUC >0.70). Six sOTUs associated with 1,25(OH)2D were all Firmicutes, class Clostridia, order Clostridiales, and recognized butyrate producers. Eight sOTUs associated with activation ratio included seven Firmicutes (Clostridia/Clostridiales) and one Lentisphaerae (family Victivallaceae). Firmicutes sOTUs were generally positively associated with higher 1,25(OH)2D and activation ratio; negative associations were observed for Oscillospira, Blautia, and Anaerotruncus.
• Vitamin D metabolic ratios: In vitamin D deficient men, activation ratio was higher; in vitamin D adequate men, catabolism ratio was higher (significant differences shown in Fig. 6).

Findings demonstrate that the active vitamin D hormone (1,25(OH)2D) and measures of vitamin D metabolic flux (activation and catabolism ratios) are more strongly associated with gut microbiome diversity and composition than the storage form 25(OH)D. The positive associations with alpha-diversity and significant clustering in beta-diversity suggest that higher biologically active vitamin D status relates to a more diverse and distinct microbial community. Identification of butyrate-producing Firmicutes associated with higher 1,25(OH)2D and higher activation ratio links vitamin D signaling to taxa considered beneficial for gut health. These associations persisted after accounting for key confounders, including antibiotic use, medications, diet (starch intake), physical activity, and site. The lack of site differences for 1,25(OH)2D, despite clear site differences in 25(OH)D, supports the notion that the active hormone is under tighter physiological regulation and may be influenced by host–microbiome interactions beyond sun exposure. Integrating prior literature, results are consistent with animal and cellular studies showing VDR-dependent effects on microbial composition and butyrate-mediated modulation of VDR pathways. Overall, the data support a bidirectional relationship in which vitamin D metabolism and the gut microbiome influence each other, with implications for metabolic and inflammatory health in older adults.

In a large cohort of older men, circulating 1,25(OH)2D and vitamin D activation/catabolism ratios, but not 25(OH)D, were robustly associated with greater microbial alpha-diversity and with beta-diversity differences, including enrichment of butyrate-producing Firmicutes. These findings provide evidence for meaningful interactions between vitamin D signaling and gut microbial ecology and suggest that metrics of vitamin D metabolic flux capture biologically relevant aspects of vitamin D status linked to the microbiome. Future studies should evaluate causality using longitudinal and interventional designs, assess functional microbial profiles via shotgun metagenomics and metabolomics, and determine whether targeted dietary or vitamin D interventions in deficient or specific clinical populations can favorably modulate gut microbiota and health outcomes.

• Cross-sectional design precludes causal inference and directionality of associations.
• Microbiome profiling used 16S rRNA amplicon sequencing rather than shotgun metagenomics, limiting taxonomic and functional resolution.
• Study population comprised predominantly older, white men, which may limit generalizability to women, younger individuals, and more diverse populations.
• Random forest taxa associations were derived only for measures with adequate predictive performance (1,25(OH)2D and activation ratio), potentially missing taxa related to other vitamin D measures.