Abstract

Introduction:

We previously showed that vitamin D deficiency leads to gingival inflammation and alveolar bone loss in mice, and that topical vitamin D3 administration prevents that bone loss and inflammation and fosters a health-associated oral microbiota in a murine ligature model of periodontal disease. To understand the relationship between vitamin D, the oral microbiome, and host factors, we performed taxonomic profiling of the oral microbiome from C57Bl/6 mice fed either a vitamin D-deficient diet or a standard diet.

Methods:

This was a 13-week study, with a group crossover period at week 7. Oral microbiomes were sampled weekly. At the end of the 13 weeks, single-cell analysis was performed on the gingival and buccal tissues.

Results:

During the first 6 weeks, the vitamin D3-deficient group 1 showed higher diversity at the start of the experiments but was more volatile in alpha-diversity values, with a notable dip in diversity at week 8. Group 2 showed lower initial diversity but was more stable by mid-study and remained relatively higher during the period where group 1 diversity crashes (weeks 6-8). The most striking feature occurs around weeks 6-8, coinciding with the change in vitamin D diet, group 1 plummets while group 2 either remained stable or rose.

Discussion:

This showed that elimination of vitamin D3 in the diet altered the diversification of bacterial species in favor of an oral microbiome associated with inflammation and bone loss. This persistent dysbiosis contrasts with the transcriptomic changes, which showed mice on a vitamin D deficient diet displayed an overall enrichment of gene sets involved in epithelial development, suggesting that re-introduction of vitamin D into the diet may help improve mucosal barrier health in the face of persistent microbiome dysbiosis.

1 Introduction

Vitamin D deficiency has been associated with periodontal disease in several epidemiological studies (Ziada et al., 2025), as well as mental health disorders, such as late-life clinical depression (Gao et al., 2025), and autoimmune diseases, such as inflammatory bowel disease (IBD) (Triantos et al., 2022). In addition, when severe vitamin D deficiency was measured in elderly and middle-aged Americans, it was linked to a higher mortality compared with moderately deficient and sufficient cohorts in a prospective study (Hu et al., 2025).

When taken orally as cholecalciferol, vitamin D is inactive and needs to be enzymatically converted to 25-hydroxy-vitamin D3 (25OHD3) and then to the active form of 1,25-dihydroxy-vitamin D3 (1,25(OH)2D3). The canonical sites for these reactions are the liver and then the kidney, respectively (Pike and Christakos, 2017). However, other local sites for these conversions have been reported by our group and others, such as in airway epithelial cells of the lung (Rigo et al., 2012) and at other sites in the body (Hewison et al., 2007; Fritz et al., 2019; Christakos, 2021). Previously, we have shown that cholecalciferol can be converted to the active form of vitamin D3, both in vitro in human gingival epithelial cells (McMahon et al., 2011; Menzel et al., 2019) and in vivo in mice when applied topically to the gums (Menzel et al., 2019; Kirkwood et al., 2024).

Active vitamin D has multiple effects, including immunomodulatory and anti-inflammatory properties, and has been shown to enhance bacterial killing (Yim et al., 2007; Charoenngam and Holick, 2020; Figgins et al., 2024). Recently we demonstrated that topical application of cholecalciferol to the gums of mice can reverse the oral alveolar bone loss and gum inflammation in a ligature model of periodontal disease (Kirkwood et al., 2024). These improvements were associated with shifts in the oral microbiome. That is, Enterococcus faecalis appeared as the bone loss occurred and disappeared with the application of cholecalciferol.

In similar fashion to the oral microbiome, studies have also demonstrated shifts in the gut microbiome upon vitamin D deficiency (Li et al., 2022; Aggeletopoulou et al., 2023). However, no study has examined vitamin D deficiency and its effect on the oral microbiome over time. Here, we expanded upon our published study wherein alveolar bone loss and inflammation were observed in mice fed a vitamin D-deficient diet for six weeks (Menzel et al., 2019) to examine the effect of vitamin D deficiency on their oral microbiome over time. In addition, we sampled the microbiome for seven extra weeks after vitamin D deficient mice were returned to a regular, vitamin D sufficient diet to determine the effect of supplementation following vitamin D deficiency. Finally, at the end of the study, we excised gingival and buccal tissues to examine the transcriptomic effects of a vitamin D-deficient diet in these key tissues at the single cell resolution.

2 Materials and methods2.1 Mice

C57Bl/6 female mice were purchased from Charles River Laboratories at 6–8 weeks old and housed in a barrier facility with a standard diet containing vitamin D for another 2 weeks to acclimatize the mice to the animal facility. All experimental protocols were approved by the University of Louisville’s Institutional Animal Care and Use Committee (IACUC protocol #20804). All methods were carried out as specified by the relevant guidelines and regulations. Mice were housed in groups during acclimatization, then all mice were separated into individual cages and sampled on Day 0. Group size (n=8 per group) was based on prior unpublished studies demonstrating n=8 as sufficient for statistical significance for both microbiome and scRNA sequence analysis. In addition, we were able to demonstrate significant alveolar bone loss and inflammation with 8 mice (Menzel et al., 2019).

2.2 Vitamin D sufficient and deficient diets and crossover study

After acclimating the mice for two weeks, on Day 0 all mice were housed in individual cages. Eight mice were switched to a standard, autoclaved diet (Labdiet 5010, purchased from Cincinnati Lab and Pet Supply, Cincinnati, OH, given ad libitum (Group 2). This was the vitamin D3 sufficient (control) diet that contained 4.2 IU vitamin D3/g and 1.0% calcium (standard levels for all ages of mice). The full composition is depicted in Supplementary Figure 1. To introduce vitamin D deficiency in another group of 8 mice, also housed in individual cages, an irradiated diet free of vitamin D3 with 0.02% calcium and 0.8% strontium (Inotiv) was given ad libitum. Strontium is included to inhibit the activity of 1-α-hydroxylase, which activates 25OHD3 to 1,25(OH)2D3 (Li et al., 2002). Calcium is included to prevent hypocalcemia. The full composition of this diet is given in Supplementary Figure 2. These 8 mice represented the vitamin D3 deficient group (Group 1). Group 1 was given a vitamin D3 deficient diet for 6 weeks, then switched to a vitamin D3 sufficient diet for weeks 7-13. Group 2 was vitamin D3 sufficient for 6 weeks, then switched to a vitamin D3 deficient diet for weeks 7-13. The rationale for making the crossover study 13 weeks is because mice cannot go longer than 6–7 weeks of vitamin D deficiency without having detrimental effects on their overall health (S. Christakos, personal communication). An extra week after 6 weeks was added to assure vitamin D deficiency disappeared in Group 1 at the crossover point (week 7) and then 6 more weeks was continued, totaling 13 weeks for the entire length of the study. In addition, the previous study relating to this study used a 6-week time course of vitamin D deficiency (Menzel et al., 2019). To clarify, vitamin D sufficiency (Vit +) means that the mice were on a regular diet for the last 6 of the 13 weeks. Vitamin D deficiency (Vit -) means that the mice were on the special diet devoid of vitamin D for the last 6 of the 13 weeks. No vitamin D was added to the diet other than what was present in the regular rodent diet. A timeline is shown in Supplementary Figure 3.

2.3 Sampling for microbiome analysis

Mice were swabbed before the diet started on (indicated as Week 0) and at the end of each week (Weeks 1-12) with a sterile cotton swab around the cheeks and gums for 30 seconds. The swabs were swirled in an Eppendorf tube (1.5 ml) with 1.0 ml of 95% ethanol in deionized H2O and frozen at -20°C until analysis. The air in the mouse barrier facility was also sampled as the negative control.

2.4 Bacterial DNA sequencing

DNA was isolated from the swabbed material in 95% ethanol as described previously (Duran-Pinedo et al., 2014). PicoGreen was used for DNA quantification. 50pg of gDNA was used for multiplex PCR in a 20ul reaction mix. PCR amplification of DNA (10-50ng) was performed using universal primers targeting the V3-V4 region of 16S genes (F341, R806). The products were purified using AMPure purification kit (Beckman Coulter, Brea CA USA). Amplicons were pooled in libraries (100ng) that were gel-purified and quantified by qPCR before being sequenced. The PCR protocol consisted of a 30-second incubation at 98 °C followed by five cycles of 98 °C, 10 sec; 63 °C, 5min; and 65 °C, 1 min; then 26 cycles of 98 °C, 10 sec; 64 °C, 1min; and 65 °C, 1 min. The multiplex PCR product was purified with AMPure® XP Beads, and after two 80% ethanol (approximately) washes, it was eluted with 5µl of i5 index, 10µl of i5 index, and 35µl of Indexing Reaction Mix. The indexing PCR was performed by incubating at 37 °C for 20min. The indexing PCR was cleaned with an adding ratio of 0.85 PEG NaCl into the Indexing PCR. The individual library was quantified using the KAPA library quantification kit (Kapa Biosystems) and monitored on the BioRad CFX 96 real-time PCR system (BioRad, Hercules, CA USA). Barcoded samples were pooled equimolarly for sequencing one MiSeq 2x250 cycle run (Illumina, San Diego, CA USA). The library was prepared at the Gene Expression & Genotyping of the Interdisciplinary Center for Biotechnology Research (University of Florida). The MiSeq run was performed at NextGen of the Interdisciplinary Center for Biotechnology Research (University of Florida).

2.5 Taxonomic profiling

In all bioinformatics analyses, GNU parallel was used (Tange, 2018) when possible. Sequences were filtered for quality using Trimmomatic (Bolger et al., 2014). Once filtered, raw paired-end reads were filtered, trimmed, and denoised using the DADA2 (Callahan et al., 2016). Amplicon sequence variants (ASVs) were inferred, merged, and chimeras were removed. Taxonomy was assigned against a custom oral mouse microbiome reference database. We generated a custom Kraken2 library with the oral microbiome genomes indicated in Stashenko et al (Stashenko et al., 2019). Phylogenetic assignment and relative quantification were performed using Kraken2 (Lu and Salzberg, 2020) and Bracken (Lu et al., 2017) against our custom 16S rRNA database for the oral microbiome extracted from the HOMD database (Chen et al., 2010). Sequences were aligned using DECIPHER, and a maximum-likelihood phylogenetic tree was built using phangorn with the GTR+G+I model. Samples with standardized residuals > 3 from an initial linear model were excluded. Group differences were tested per-week using Wilcoxon tests with FDR correction, and overall using a two-way ANOVA (Time × Group interaction) with a linear model. A detailed protocol is attached in the Supplementary Material as a jupyter notebook.

2.6 Bar-plot species

Species composition of the different groups and time points were represented as bar plots using the package ‘phyloseq’ (McMurdie and Holmes, 2013). Kruskal-Wallis test for multiple pairwise comparisons was performed using the function ‘comparison’ from the R packages ‘agricolae’ (de Mendihuru Delgado, 2023).

2.7 Tissue processing into single cell suspension

Gingiva and oral mucosa tissue from vitamin D sufficient (Group 1; vitamin D sufficient at 13 weeks) and deficient (Group 2; vitamin D deficient at 13 weeks) mice were harvested at the endpoint of the experiment (see dietary details above). Tissues were placed in 10% DMSO/90% fetal bovine serum (FBS) and slowly frozen in a Mr. Frosty and preserved at -80C until ready for processing. Tissues were placed in 2 ml of Advanced RPMI media supplemented with 10% (FBS) (Both from Gibco, Billings, MT), 100 U/ml of heparin (Sigma-Aldrich, St. Louis, MO) and 100 U/ml of DNAse I (StemCell Technologies, Vancouver, BC). After mincing the tissue, Liberase TH was added at 75 mg/ml (Roche, Basel, Switzerland). The tissue was incubated at 37 °C in an elliptical shaker. After incubation, cells passed through a 70 mm filter and washed before final resuspension in PBS with 2% FBS. Single cell suspensions were loaded into a HIVEs capture device (Honeycomb BioTechnologies, Waltham, MA) as per manufacturer’s instructions.

2.8 Honeycomb HIVE sequencing

Cells from gingiva and oral mucosa were loaded into HIVEs capture devices (Honeycomb Biotechnologies, Waltham, MA) as described in the manufacturer’s protocol. Loaded devices were frozen at -80°C and shipped in dry ice to the Honeycomb headquarters for processing and sequencing.

2.9 Data alignment

Read processing was performed using the Honeycomb HIVE recommended workflow. Briefly, fastq files were aligned using the Honeycomb Beenet software package (v1.1.3) and the mouse reference genome mm10 (20210714-mm10.104) with default settings. The generated transcript count matrices (TCM) were used for subsequent analysis.

2.10 Data filtering and Seurat object generation

The TCMs from Beenet alignment were further processed using R (version 4.5.1) and Seurat (version 5.3.0) (Butler et al., 2018). Seurat objects were generated from these TCMs and merged into either a buccal or gingival Seurat object. Next, a priori filtering was applied to each object to remove cells that had fewer than 50 genes or greater than 2500 genes, and to remove cells whose mitochondrial content was >20%. A lower limit of 50 genes was used to increase the sensitivity for detecting neutrophils (Wigerblad et al., 2022). These filtered Seurat objects were then processed using the standard Seurat workflow. ‘NormalizeData’ and ‘FindVariableFeatures’ functions were run with default arguments. The ‘ScaleData’ function was ran with the percent of mitochondrial genes as a regression variable.

2.11 Dimensionality reduction, cell clustering, and annotation

After Seurat object generation and processing, linear dimensionality reduction was performed using the ‘RunPCA’ function with default arguments. PCA outputs were then passed into the ‘FindNeighbors’, ‘FindClusters’, and ‘RunUMAP’ functions to generate non-linear dimensionality reductions for clustering in uniform manifold approximation projection (UMAP) space. Clustering was performed using the Leiden method. To identify cell types for annotation, differentially expressed genes for each Seurat cluster were determined using the ‘FindAllMarkers’ function with a minimal cell percent threshold of 0.25. A combination of these DEGs and canonical markers were used to determine the cell type of each cluster. After initial clustering with assigning of coarse annotations (e.g., epithelial, endothelial), these clusters were subset into individual objects for further analysis and clustering to examine for any subpopulations of interest. After subsetting, each Seurat object was re-analyzed through the standard Seurat workflow, clustering methods, and DEG determination as above.

2.12 Pathway analysis

The R package clusterProfiler (version 4.16.0) (Yu et al., 2012) was utilized to perform gene set enrichment analysis (GSEA) on the total cell objects of buccal and gingival tissues and for over-representation analysis (ORA) on the individual coarse cell clusters after subsetting into separate Seurat objects. DEGs were determined using the ‘FindMarkers’ function from Seurat. For both methods of pathway analysis, all ontologies within the Gene Ontology database were included and the Benjamini-Hochberg method of p-value adjustment was implemented with a cutoff of 0.05.

2.13 Figure generation

To generate publication-ready figures, a combination of the R packages Seurat, SeuratExtend (Hua et al., 2025), and clusterProfiler and ggplot2 (Wickham, 2016) were utilized for the UMAP, dot plot, and bar graphs.

3 Results3.1 Crossover experiment design – 13 weeks

In the group fed the vitamin D-deficient diet over the first 6-week period, we observed a time-dependent diversification of the sampled microbiome in the oral cavity (Figures 1, 2). Although Figures 1 and 2 showed taxa with more than 5% abundance in at least 2 samples, the entire taxonomic profile was used for analysis. At week 7, the diets were switched to examine if microbiome diversity can be rescued/normalized by simple re-introduction of vitamin D into the diet as opposed to treatment doses per se.

Effect of vitamin D deficiency over time - relative abundance of bacterial genera in the oral microbiome of mice under different dietary conditions. After week 6, the diets were switched (indicated by a line). Top bar graph (Group 1) indicates a start on a vitamin D-deficient diet, with a switch to a regular vitamin D sufficient diet (week 7–13 samples). Bottom bar graph (Group 2) indicates a start on a regular vitamin D sufficient diet, with a switch to a vitamin D deficient diet (week 7–13 samples). Bar plots display genus-level taxonomic profiles of the oral microbiome in mice fed a vitamin D3-deficient diet (Group 1, top) or a standard diet (Group 2, bottom). Only genera representing more than 5% relative abundance were shown. 16S rRNA gene sequences were processed using the DADA2 pipeline, with taxonomy assigned via a custom mouse oral microbiome reference database. Amplicon Sequence Variants (ASVs) were agglomerated at the genus level, transformed to relative abundances, and visualized with ggplot2. Colors represent distinct genera, and each bar reflects the mean composition for each group. *Weeks with significant differences between Group 1 vs. Group 2, PERMANOVA analysis with p-value < 0.05.

Effect of vitamin D deficiency over time - relative abundance of bacterial species in the oral microbiome of mice under different dietary conditions. After week 6, the diets were switched (indicated by a line). Top bar graph (Group 1) indicates a start on a vitamin D-deficient diet, with a switch to a regular vitamin D sufficient diet (week 7–13 samples). Bottom bar graph (Group 2) indicates a start on a regular vitamin D sufficient diet, with a switch to a vitamin D deficient diet (week 7–13 samples). Bar plots display species-level taxonomic profiles of the oral microbiome in mice fed a vitamin D3-deficient diet (Group 1, top) or a standard diet (Group 2, bottom). Only species representing more than 5% relative abundance were shown. 16S rRNA gene sequences were processed using the DADA2 pipeline, with taxonomy assigned via a custom mouse oral microbiome reference database. Amplicon Sequence Variants (ASVs) were agglomerated at the species level, transformed to relative abundances, and visualized with ggplot2. Colors represent distinct species, and each bar reflects the mean composition for each group. *Weeks with significant differences between Group 1 vs. Group 2, PERMANOVA analysis with p-value < 0.05.

3.2 Microbiome at week 0

At week 0, Actinomyces, and Fusobacterium dominated the microbiome of both groups when analyzed at the genus level of >5% (Figure 1). When analyzed to the species level of >5%, Acinetobacter baumannii and Fusobacterium nucleatum dominated both groups (Figure 2). Other bacteria identified at the genus level present (>5%) were Prevotella, Limosilactobacillus, Streptococcus, and Solobacterium in Group 1, and Streptococcus, Staphylococcus, Prevotella, Kingella and Lactobacillus in Group 2. At the species level, Limosilactobacillus panis and Solobacterium sp. Oral Clone 5RH-44 were also present in Group 1. In Group 2, at the species level of analysis, Staphylococcus nepalensis, Lactobacillus gasseri, and Kingella oralis, were present at week 0 (Figure 2).

3.3 Vitamin D sufficiency control – regular diet, weeks 1-6

During the first 6 weeks, Group 2 was on a regular diet containing vitamin D (Supplementary Figure 1). Each week during weeks 1-6, at the genus analysis level (>5%), Actinomyces, Fusobacterium, and Staphylococcus were present in the highest abundance. Prevotella and Lactobacillus were also present in lower abundance at each week except they were undetectable for one week (week 1 for Prevotella and week 6 for Lactobacillus). During week 1, Lentilactobacillus, Abiotrophia, Clostridium, and Athrospira were detected, but were not detected in weeks 2-6. Weizmannia was also detected in week 1 and in week 2. Limosilactobacillus was detected in weeks 2 and 5. Jonquetella, Bacillus, Anaeroglobus, and Eikenella appeared in week 3 but in no other weeks in Group 2. Gemella sp. Oral Clone ASCF12 only appeared in week 4.

At the species level in Group 2, compared to week 0, Acinetobacter baumanii and Fusobacterium nucleatum were the dominant and only species present consecutively throughout weeks 1-6. On this background, the microbiome fluctuated week-to-week: In week 2, Staphylococcus neplanesis, Lactobacillus gasseri, Kingella oralis, Limosilactobacillus panis, and Solobacterium sp._oral_clone_6RH-44 were first detected; in week 3, Kingella oralis and Limosilactobacillus panis were lost to detection, and by week 4 Solobacterium sp._oral_clone_6RH-44 was also undetected; however, both Limsilactobaccilus and Solobacterium sp._oral_clone_6RH-44 re-appeared in the sample of week 5.

Other Group 2 species that were not detected in week 0 that subsequently appeared were Streptococcus intermedius (week 1), Staphylococcus schleiferi (week 1), Weizmannia coagulans (week 1), Streptococcus pneumoniae (weeks 1 and 3), Clostridium saccharolyticum (week 1), Prevotella sp. Oral Clone 376 (weeks 2, 3, and 5), Limosilactobacillus fermentum (week 2), Actinomyces sp. Oral Clone DR002 (weeks 2 and 3), Eggerthia catenaformis (week 2), Bacillus anthracis (week 3), Jonquetella anthropi (week 3), Lactobacillus crispatus (weeks 3 and 4), Staphylococcus warneri (week 4), and Eubacterium Oral Clone 20-24 (week 6).

3.4 Vitamin D deficiency

In general, vitamin D-deficiency induced time-dependent changes in both the number of types and abundance of bacteria at the genus and species levels of analysis, which did not reverse upon re-introduction of vitamin D into the diet. The results of these changes in the microbiota are depicted in Figures 1 and 2 and detailed in Table 1.

WeekGenusDirectionLog2FoldChangePadjBaseMeanWeek6LimosilactobacillusEnriched in G1-8.6797087933.68887E-17119.5932153Week6EikenellaEnriched in G1-8.0493718436.69591E-1013.90145774Week6StaphylococcusEnriched in G1-5.6477055082.89922E-0945.71105062Week6LactobacillusEnriched in G1-7.1191890992.89922E-0940.11152647Week6StreptococcusEnriched in G1-6.3650241480.00017834413.0670461Week6FusobacteriumEnriched in G1-5.2849977070.000946611275.7979529Week6AbiotrophiaEnriched in G1-8.7523261130.0014917224.46905179Week6SolobacteriumEnriched in G1-8.4724046980.0178616145.680487942Week6BlautiaEnriched in G1-8.188792720.0213100214.679886255Week6HaemophilusEnriched in G1-3.8271445210.0278379021.373160313Week6ParageobacillusEnriched in G1-7.5988828730.0309129913.91135681Week8PrevotellaEnriched in G1-7.4953713862.01618E-06106.4207086Week10PrevotellaEnriched in G1-9.1981045739.44837E-09172.2164796Week10LentilactobacillusEnriched in G1-6.6059703970.0008228567.355302212WeekSpeciesDirectionLog2FoldChangePadjBaseMeanWeek6Limosilactobacillus_panisEnriched in G1-8.7058322753.60924E-13115.7750416Week6Lactobacillus_gasseriEnriched in G1-7.5752064544.74741E-0732.18583809Week6Staphylococcus_warneriEnriched in G1-7.1419495414.65855E-0517.24452818Week6Staphylococcus_nepalensisEnriched in G1-5.5514470050.00016337823.47343916Week6Eikenella_sp._canine_oral_taxon_049Enriched in G1-8.4761949470.0001692045.597282849Week6Abiotrophia_defectivaEnriched in G1-8.6746336250.00078463222.88175455Week6Fusobacterium_nucleatumEnriched in G1-5.437129930.001019998285.8183073Week6Lactobacillus_crispatusEnriched in G1-7.1543383950.0010199988.493262618Week6Staphylococcus_aureusEnriched in G1-6.6725461090.0058993933.262586295Week6Solobacterium_sp._oral_clone_5RH-44Enriched in G1-8.4432802940.0058993935.368025151Week6Parageobacillus_genomosp._1Enriched in G1-7.9982344820.0138779085.380989587Week6Streptococcus_intermediusEnriched in G1-7.853185270.0144574146.061192481Week6Streptococcus_gordoniiEnriched in G1-7.8016402320.0302176283.221834658Week8Prevotella_sp._oral_taxon_376Enriched in G1-7.8985037992.05845E-06112.7388639Week8Lentilactobacillus_kisonensisEnriched in G1-6.3869384220.0476178494.765986242

Significant changes in microbiota.

When comparing Groups 1 and 2 during the period before the diet change at week 6, differences were not statistically significant either at the genus or species levels. However, at week 6, when mice are vitamin D-deficient (Menzel et al., 2019) we observed statistically significant differences in communities (Table 1). After that, Prevotella was enriched in Group 1 [switched to a vitamin D-sufficient regular diet (VitD+)] at weeks 8 and 10, and Lentilactobacillus at week 10.

In the abundance analysis of genus (> 5%), Actinomyces and Fusobacterium were abundant during all weeks in mice from both Groups 1 and 2, albeit in differing proportions (Figure 1). The abundance of Actinomyces decreased over time when mice were fed a vitamin D-deficient diet (Group 1, Weeks 1–6 and Group 2, Weeks 7-13) compared with mice on the vitamin D sufficient diet (Group 2, Weeks 1-6) in both the genus and species (Actinomyces_sp._oral_clone_DR002) abundance analyses. Interestingly, when the mice in Group 1 were switched to the regular diet at the cross over period, the Actinomyces genus and species abundance did not increase (Figures 1, 2, Group 1, Weeks 7-13).

The Fusobacterium (>5% genus abundance) were present in larger proportions in the vitamin D-deficient mice when compared to vitamin D-sufficient mice within Group 1 and within Group 2, increasing in proportion as the mice were fed longer on the vitamin D deficient diet (Figure 1). A similar pattern occurred with Fusobacterium nucleatum in the species abundance (>5%) analysis (Figure 2).

The genus abundance (>5%) analysis in Figure 1 also revealed that Streptococcus decreased and disappeared entirely by Week 6 with vitamin D-deficiency (Group 1, Weeks 1-6) and did not return with the return to a vitamin D-sufficient diet (Group 1, Weeks 7-13), while in Group 2, Streptococcus decreased upon continuing on the regular diet (Weeks 1-6) but returned with vitamin D-deficiency (Weeks 7-12) but disappeared by Week 13 of vitamin D-deficiency.

At the species level, we observed a pattern similar to that at the genus level. Most of the statistically significant differences occurred at week 6, coinciding with documented vitamin D deficiency (Menzel) followed by the dietary change (Table 1). After that, only 2 species, Lentilactobacillus kisonensis and Prevotella sp. oral taxon 376, were significantly more abundant in Group 1 at week 8.

However, Streptococcus cristatus (Figure 2, species abundance >5%) was only found in the vitamin D-deficient mice (Group 1, Week4 and Group 2, Weeks 9, and 11).

Vitamin D-deficiency also decreased the abundance of genus Lactobacillus. Lactobacillus was present (genus abundance of >5%) in vitamin D-sufficient mice (Group 2, Weeks 1-5) but was present in lower abundance in vitamin D-deficient mice at Weeks 4 and 6 (Figure 1, Group 1). Lactobacillus was present in lesser abundance at Weeks 10 and 12 of vitamin D-deficient mice (Figure 1, Group 2) following the switch from 8 weeks of a vitamin D-sufficient diet that also had a greater abundance of Lactobacillus at Weeks 7 and 12 (Figure 1, Group 1). Lactobacillus gasseri was observed in greater abundance at Weeks 7 and 12 in mice fed the vitamin D-sufficient diet (Figure 2, Group 1), at Weeks 2 and 3 of the vitamin D-sufficient diet (Figure 2, Group 2) and in mice fed the vitamin D-deficient diet at Weeks 7 and 12 (Figure 2, Group 2). Lactobacillus gasseri was not detected in the vitamin D-deficient mice in Group 1 (Figure 2). Lactobacillus crispatus was present in low abundance in vitamin D-sufficient mice (Group 2, Weeks 3 and 4) and in Group 1, Week 12 once the mice were switched to the vitamin D-sufficient diet at Week 7 (Figure 2). Lactobacillus crispatus was not detected in any group of vitamin D-deficient mice (Figure 2).

However, genus Prevotella was present in mice receiving the vitamin D-sufficient diet but was not present in the vitamin D-deficient mice (Figure 1, Group 1, Weeks 1–6 and Group 2, Weeks 7-13). The disappearance of Prevotella_sp._oral_taxon_376 in the vitamin D-deficient mice appeared to occur at the species level as well (Figure 1, Group 1, Weeks 1–6 and Group 2, Weeks 7-13).

In Group 2, Enterococcus appeared in the vitamin D-deficient mice (Weeks 9 and 11), but not in the vitamin D-deficient mice in Group 1 (Figure 1, Weeks 1-6). Enterococcus was not detected (>5% abundance) at the species level (Figure 2). Enterococcus was not detected in vitamin D-sufficient mice in Group 1, Weeks 1–6 nor in Group 2, Weeks 7–12 at the genus or species level.

Staphylococcus appeared to increase in abundance with vitamin D-deficiency in Group 2, Weeks 7-13 (Figure 1). Staphylococcus also appeared with vitamin D-deficiency in Group 1 at Weeks 1, 2, 4, 5, and 6 and remained present in slightly greater proportions after the mice were returned to the regular diet at Weeks 7, 8, 9, 11, 12 and 13 (Figure 1).

The diversity of species that occurred as mice became vitamin D-deficient in the first 6 weeks (Figure 3, Group 1 versus Group 2, Weeks