Association of Gut Microbiome and Vitamin D in Knee Osteoarthritis Patients

Background: A Few preclinical studies have shown that Knee osteoarthritis (KOA) is linked to gut microbiome dysbiosis and chronic inammation. This pilot study was designed to look at the gut microbiome composition in KOA patients and normal individuals with or without vitamin D deciency (VDD, serum vitamin D <30 ng/ml). Methods: This pilot study was conducted prospectively in 24 participants. The faecal samples of all the participants were taken for DNA extraction. The V3-V4 region of 16s rRNA was amplied and the library was prepared and sequenced on the Illumina Miseq platform. Results: The mean (±SD) age was 45.5 (±10.2) years with no dened co-morbidities. Of 447 total Operational Taxonomic Units (OTUs), a differential abundance of 16 nominally signicant OTUs between the groups were observed. Linear discriminate analysis (LEfSe) revealed a signicant difference in bacteria among the study groups. Pseudobutyrivibrio and Odoribacter were specic for VDD while Parabacteroides, Butyricimonas, and Gordonibacter were abundant in the KOA_VDD group and Peptococcus, Intestimonas, Delftia, and Oribacterium were abundant in the KOA group. About 80% of bacterial species were common among different groups and hence labeled as core bacterial species. However, the core microbiome of KOA and VDD groups were not seen in the KOA_VDD group, suggesting that these bacterial groups were affected by the interaction of the KOA and VDD factors. Conclusion: Parabacteroides, Butyricimonas, Pseudobutyrivibrio, Odoribacter, and Gordonibacter are the predominant bacteria in vitamin D decient patients with or without KOA. Together these results indicate an association between the gut microbiome, vitamin D, and knee osteoarthritis.


Introduction
Osteoarthritis (OA) is a degenerative disease consisting of a progressive deterioration of the articular cartilage, subchondral bone sclerosis, and synovial in ammation, resulting in joint pain and disability 1 .
While various factors such as genetic factors, sex, ethnicity, age, and obesity were considered as central contributing factors associated with disease progression, the etiology of this disease remains still a "mystery". Although obesity increases OA risk through increased mechanical loading on weight-bearing joints, obesity also increases the risk for OA in non-weight-bearing joints 2,3 that might be caused by a lowgrade systemic in ammation 4,5 .
Besides, various studies showed that the gastrointestinal microbiome contributes to a triggering of this low-grade systemic in ammation in obesity 5,6 , through the bacterial endotoxin lipopolysaccharide (LPS) 7 . Moreover, not only LPS has been retrieved at an elevated concentration in the blood and the synovium of OA patients, but also it was found to be associated with a more severe knee OA, knee pain, and in ammation 8 . Other studies have also associated the gut-microbiome composition with a low-grade systemic and local in ammation observed in OA 9 . Together these ndings support that the intestinal microbiome contributes to OA pathogenesis although this interaction has still not been de ned and needs to be further explored.
On the other hand, observational studies found that low levels of vitamin D was associated with a higher prevalence of knee OA along with an increased risk of disease progression 10,11,12 . In effect, vitamin D could be considered as an important factor in the OA pathogenesis since vitamin D was shown to reduce cartilage degradation, which characterizes osteoarthritis 13,14 . So, it appears that vitamin D insu ciency or de ciency is closely associated with OA and its associated health implications 15 . It is also known that dietary nutrients shape the gut microbiome and Vitamin D supplementation is found to in uence microbiome 16 . Particularly, vitamin D improves barrier function in the gut by inducing the expression of Ecadherin and improving the epithelial cell junctions in the gut 15 . The de ciency of vitamin D can cause breakdown of these epithelial cell junctions resulting in a leaky gut 7 . It was known that Vitamin D affects the composition of the intestinal microbiome through mainly the expression of cathelicidin antimicrobial peptide (CAMP) gene 17 . So, the dysbiosis induced by the vitamin D de ciency might contribute to the severity of the OA by increasing intestinal barrier permeability and then producing LPS-mediated intestinal in ammation.
To investigate this interrelation between vitamin D de ciency, microbiome and KOA, we conducted a clinical pilot study in which we compared the gut microbial composition of KOA patients with or without vitamin D de ciency with that of normal individuals with low and high serum vitamin D levels.

Material And Methods
Study participants: This observational study was conducted in a tertiary care teaching hospital between June 2015 and December 2016 to investigate the association of gut microbiome and serum vitamin D status in knee osteoarthritis (KOA) patients and in normal healthy individuals. The diagnosis of KOA was based on the American College of Rheumatology Criteria. All patients of primary KOA between 30 to 60 years of age and willing to participate in the study with a follow-up of six months were included.
Patients with various deformity > 5 degrees, valgus deformity, exion deformity > 5 degrees, on anticoagulants, BMI > 30 kg/m 2 , history of hospitalization in last 8 weeks, on antibiotics for last 12 weeks, gastrointestinal co-morbidities (including in ammatory bowel disease, irritable bowel syndrome, gastrointestinal malignancies, surgical resection) or on pharmacologic doses (> 3 x recommended daily allowance) of vitamins or minerals were excluded. The Institutional Ethics Committee clearance was obtained (IRB Min No:9432-29/04/2015) and the patients were recruited after getting their written informed consent. The KOA patients were categorized into one of the following two groups; Group 1 with vitamin D levels < 30 ng/ml were classi ed as Vitamin D de cient (KOA_VDD, n = 7) and Group 2 with vitamin D levels > 30 ng/ml were classi ed as normal vitamin D status (KOA, n = 4). The asymptomatic individuals (patient attendant or hospital staff) were categorized into Group 3 (Vitamin D levels < 30 ng/ml, VDD, n = 7), and Group 4 (Vitamin D levels > 30 ng/ml, normal/ NVD, n = 6). There were no differences in age, geography, and gender between the four groups.
Data / Specimen collection: The demographic pro le, socioeconomic status, and medical history were entered into a predesigned proforma. The dietary intakes of food groups and macronutrients were calculated from 24 h dietary recall and food frequency. The daily ingestion of macro and micronutrients were matched with recommended dietary allowances for Indians using the DIGEST programme by ICMR 18 . The KOA disability was assessed using the WOMAC score 19 .
For KOA_VDD and KOA groups, plain radiographs of both Knees (Anteroposterior and lateral views) in the standing position were taken. The severity of KOA was classi ed as per Kellgren-Lawrence grade. The blood samples were collected in all 4 groups and the following parameters were checked: serum hemoglobin level, erythrocyte sedimentation rate, C-reactive protein, serum creatinine, serum uric acid, serum calcium/phosphate, serum alkaline phosphatase, parathyroid hormone, and thyroid function test and vitamin D levels. The faecal samples were collected from all participants in a plastic container on ice and then aliquoted in the lab, within 2 hours, into a 2 ml Eppendorf tube for storage at -80•C.
DNA isolation, V3-V4 region 16 s rRNA sequencing and library preparation: A 200 mg of faecal sample from the Eppendorf tube was taken for DNA isolation with a modi ed DNeasy PowerLyzer Power Soil kit (Cat No: 12855-100, Qiagen, Qiagen GmbH, Germany). The eluted DNA was quantitated using Nanodrop and checked using 1% agarose gel electrophoresis. The extracted faecal DNA was used to amplify the V3-V4 region of the 16 s rRNA gene using region-speci c primers (Table 1)  The quality-ltered sequence reads were imported into Mothur, and the read pairs were aligned with each other with minimum 30 bp overlap to form contigs. Reads with ambiguous base calls were rejected. The high-quality contigs were checked for identical sequences and duplicates were merged. The qualityltered contigs were classi ed into taxonomical outlines based on the Silva NR v123 database. The classi ed contigs were ltered for any undesirable lineage like Mitochondria, Chloroplast, Fungi or Archaea. The contigs were then clustered into OTUs (Operational Taxonomic Unit) based on the phylotype. The abundance of each OTU in the samples was estimated. The alpha and beta diversity of OTUs between the groups was tested by Kruskal-Wallis test after normalizing the OTU abundance. The relative abundance of bacterial groups was analysed using LEfSe (Linear discriminant analysis effect size) 21 . Canonical correspondence analysis (CCA) was performed using R version 3.1.0 on a Bray-Curtis dissimilarity matrix. Adonis was used to test the association between the distinct groups and the overall microbiota composition based on distance matrices using the "Adonis" function in "vegan" R package.

Results
A total of 24 participants, aged 45.5 ± 10.2 (mean ± SD) years were recruited into four groups, with seven, four, seven and six participants in KOA_VDD, KOA, VDD and Normal (NVD) groups respectively ( Table 2).
These participants were predominantly rural or semi-urban dwellers belonging to middle-to lower-middle socioeconomic status. None of them were on Vitamin D supplementation and they did not consume milk.
They had no other co-morbidities. The mean BMI was > 30 with the highest BMI 28.9 ± 2.6 (mean ± SD) in VDD group and the lowest BMI 22.9 ± 2.5 (mean ± SD) in the Normal group. Normal group was signi cantly different (p < 0.05) from all the three-study group, however, BMI among the three-study groups was not different. Their hemoglobin levels were 13 ± 1.7 (mean ± SD) gm/dl and none of them were on any treatment. KOA and KOA_VDD groups had a chronic history of pain ranging from 4 weeks to  Figure S1). However, when OTUs with counts less than 50 were removed, all 4 groups shared only 4 OTUs, whereas 14, 2,1 and 0 OTUs were speci c to the KOA, KOA_VDD, VDD and normal groups respectively (Fig. 1B). We also evaluated the overall differences in beta diversity between faecal microbiome of KOA_VDD, KOA, VDD and Normal Participants using canonical correspondence analysis (CCA), based on Bray-Curtis distances (Fig. 1C). All study groups emerge to cluster differently although the ADONIS test does not explain statistically this difference (p = 0.12).
Knee osteoarthritis-associated dysbiosis: To assess the effect of the Knee osteoarthritis disease on the gut microbiome, we compared the gut microbiome composition of individuals with normal vitamin D status, diagnosed with Knee osteoarthritis disease (KOA), and those considered as "healthy" (NVD). We rst evaluated the overall differences in beta diversity between faecal microbiome samples of both groups using canonical correspondence analysis (CCA), based on Bray-Curtis distances (Fig. 2A). The two groups appear to cluster differently although the ADONIS test does not explain statistically this difference (p = 0.13).
We estimated α-diversity using the observed OTU and Shannon index (using the Shannon index measuring how evenly OTUs are distributed in a sample) in both groups (Fig. 2B) To investigate differentially abundant taxa between both groups, we then performed LEfSe analysis to compare the abundance of bacterial taxa in KOA patients and healthy subjects. A histogram of the Linear Discriminate Analysis scores was computed for features that showed differential abundance between healthy (NVD) subjects and KOA patients (Fig. 2C). The LDA scores indicated that the relative abundances of Peptococcus, Shimwellia, Propionibacterium, Intestinimonas, and Pavimonas were enriched in patients with KOA patients than in healthy (NVD) subjects. The most differentially abundant bacterial taxon in patients with KOA was Peptococcus and for the NVD group, it was Anaero lum with a 3-fold difference in LDA score.
Knee osteoarthritis-associated dysbiosis in patients with vitamin D de ciency: To assess the effect of Vitamin D de ciency on the Knee osteoarthritis disease on the gut microbiome, we compared the gut microbiome composition of individuals with Knee osteoarthritis patient with vitamin D de ciency (KOA_VDD) and those with vitamin D de cient (VDD). We rst evaluated the overall differences in beta diversity between faecal microbiome samples of both groups using canonical correspondence analysis (CCA), based on Bray-Curtis distances (Fig. 3A). The two groups appear to cluster differently although the ADONIS test does not explain statistically this difference (p = 0.19).
We estimated α-diversity using the observed OTU and Shannon index (using the Shannon index measuring how evenly OTUs are distributed in a sample) in both groups (Fig. 3B) and they were not statistically signi cant (observed OTU index, P = 0.7; and Shannon index, P = 0.42).
Vitamin D de ciency impacts Knee osteoarthritis-associated dysbiosis: We assessed the effect of the Vitamin D de ciency on the gut microbiome of KOA patients, by comparing the gut microbiome of KOA patients with and without Vitamin D de ciency (KOA vs KOA_VDD). Although the CCA analysis presents two distinct clusters representing the KOA and KOA_VDD groups, this difference is not signi cantly different (Adonis, p = 0.055) (Fig. 4A). We observed also a decreased alpha diversity in the KOA_VDD group compared to the KOA (observed OTU index, 158.3 ± 28.4 vs. 174.2 ± 14.8; and Shannon index, 3.0 ± 0.5 vs. 3.2 ± 0.19), but this difference was not statistically different (Fig. 4B). However, as revealed by the LEfSe analysis, the gut microbiome of KOA_VDD are enriched by bacterial taxa, including Bacteroides, Parabacteroides, Pseudobutyrivibrio, Odoribacter, and Butyricimonas. On the other hand, the KOA microbiome was characterized by OTUs including Sphingomonas, Hydrogenoanaerobacterium, Rickenellaceae, Luteimonas, Selenomonas, Oxalobacteraceae, Ruminococcaceae, and Neisseriaceae (Fig. 4C).
Vitamin D de ciency affects the gut bacterial communities: To investigate the effect of the Vitamin D de ciency on the gut microbiome composition, we analyzed the gut microbiome composition of participants with levels of Vitamin D considered as de cient (< 30 ng/ml, VDD group) and normal (> 30 ng/ml, NVD group). Globally, as shown, the CCA analyzed (Fig. 5A) the VDD and NVD groups clustered separately, displaying an effect of the Vitamin D de ciency on the gut microbiome, although the ADONIS test did not con rm it (ADONIS, p = 0.43). We observed that the Vitamin D de ciency tended to increase the alpha diversity ( On the other hand, when LEfSe Analysis was performed on all four study groups, the distinct gut microbiome patterns for each of the groups were identi ed. The distinct OTUs for KOA were Peptococcus, Delftia, and Oribacterium while it was Gordonibacter, Butyricimonas, and Parabacteroides for KOA_VDD (supplementary information Figure S4). While Pseudobutyrivibrio and Odoribacter are speci c for VDD and Normal group (NVD) was enriched with Faecalibacterium and Anaero lum. These distinctions are signi cant and show a strong association between Vitamin D status, microbiome, and KOA.

Discussion
This study reveals for the rst time interaction of Vitamin D and the KOA status on the gut microbiome. For the rst time, in a well-de ned KOA phenotype, we observed an effect of the vitamin status on the gut microbiome in KOA patients. Signi cant enrichment of speci c OTUs was observed between KOA patients with and without vitamin D de ciency.
Previously, Jackson et al. 2018 reported that the abundance of speci c gut microbes (Lentispherea) was negatively associated with the prevalence of OA and rheumatoid arthritis 22 . As expected, the gut microbiome of KOA patients included in our study (KOA) was characterized by a dysbiosis compared with that of healthy individuals (NVD groups). it is suggested that the gut microbiome interacts with risk factors of OA and modulates the disease process. The local and systemic in ammation have an association with KOA. The release of pro-in ammatory mediators (TNF-α, IL-6) is augmented by numerous mechanisms such as epigenetic alterations, mitochondrial dysregulation, or cellular senescence. Previous literature showed that there are increased pro-in ammatory anaerobes and decreased anti-in ammatory microbes (i.e, Faecalibacterium prauznitzii) in the gut with aging. Besides our study showed Faecalibacterium is enriched in the NVD group when compared to the KOA_VDD group, suggesting that the KOA-associated microbiome could participate in the in ammation process and the degeneration process in the joint.
A previous study based on large-scale population-cohorts of Caucasian adults found a microbiome-wide association with knee WOMAC pain and Streptococcus spp 22 . However, they have not considered Vitamin D status in their study although they considered the BMI status of individuals in their analysis.
To our knowledge, this is the rst clinical observation study exploring the association between Vitamin D status and gut microbiome in KOA patients. Our study observed adequately an effect of the vitamin de ciency on the gut microbiome. Compared to healthy individuals (NVD), Pseudobutyrivibrio, and Odoribacter are retrieved speci cally in VDD patients. Interestingly we also observed an effect of the vitamin status on the KOA-associated dysbiosis. The gut microbiome of KOA patients with de cient vitamin D (KOA_VDD) was characterized by an increased abundance of Parabacteroides, Butyricimonas, Gordonibacter, while Intestimonas, Delftia, Peptococcus speci c for KOA. These results suggest that vitamin de ciency shapes the KOA linked gut microbiome. In effect, distinct signatures were observed when KOA (KOA) was compared with healthy individuals and Vitamin de cient KOA (KOA_VDD) with Vitamin de cient patients (VDD). This reveals that vitamin de ciency affects the gut microbiome but these alterations are dependent on the KOA status of individuals. It was probable that the unfavorable effects of vitamin D de ciency in the KOA disease could be mediated in part through the gut microbiome. conversely, vitamin D supplementation probably exerts a bene cial role on the KOA-altered gut microbiome, thus modulating the disease process of OA. Nevertheless, the determination of the role of this interaction needs to be determined in the future.
In support of this observation, the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) analysis of our study revealed that the gut microbiome involved in lipopolysaccharide biosynthesis and lipopolysaccharide biosynthesis protein is seen in the KOA_VDD group (found in Supplementary data). Along with these functions, we observed that cell mobility, secretions, cellular antigens, transcription-related proteins, and glycan synthesis and metabolism are also controlled by the gut microbiome found in KOA_VDD patients. On the other hand, amino acid metabolism (taurine and hypotaurine metabolism, cyanoaminoacid metabolism), glycolysis, gluconeogenesis, pyruvate metabolism, and phosphatidylinositol signaling system are controlled by the Gut microbiome of the KOA group. These observations indicate that vitamin D probably has control on the functionality of the gut microbiome responsible for the production of lipopolysaccharide along with cell mobility, secretion, and cellular antigen expression.
Vitamin D is known to induce the expression of E-cadherin and improve the epithelial cell junction in the gut. In contrast, vitamin D de ciency causes the breakdown of the junction resulting in a leaky gut and favoring the passage of bacterial components, as well as products from the bacterial metabolism. This way, KOA patients with a de ciency in vitamin D could have increased permeability of the gut barrier.
Bacterial peptidoglycans were detected in the synovial uid of patients with reactive arthritis and OA, suggesting that these antigens result in in ammation of the joints. In a study, rats injected with Propionibacterium acnes into the shoulder joint developed erosive arthritis, showing the involvement of the microorganisms and/or bacterial components in the in ammation process of the joint. In general, the common bacterial pathogens detected in joint in ammation correspond to bacteria retrieved in the gut microbiome such as Pseudomonas sp, Shigella sp, and E. coli species.
Osteoarthritis requires lifelong treatment either for a cure or for arresting the progression of the disease. Because of its plasticity, the gut microbiome is an interesting target for preventive or therapeutic interventions. The data generated by our study show promising leads for effectively relieving indications of KOA. The microbes observed in VDD without KOA (Lachnospiraceae, Xylanivorans, pseudobutyrivibrio, Dialister) may offer protection from KOA, despite a de ciency of vitamin D as these patients did not develop KOA. Speci cally, efforts in the Identi cation of bacterial species belonging to these taxa could lead to identifying newer probiotics with potential bene ts for the KOA. As shown in the PICRUSt analysis from our study, the modulation of metabolic process, cellular mobility, cellular secretion, and antigen expression with LPS formation are predominant actions on the KOA-associated gut microbiome. Earlier studies on oral ingestion of Lactobacillus casei (2 X 10 10 CFU, 500 mg/kg of body weight) with type II collagen (250 mg/ kg body weight) and glucosamine (250 mg/kg body weight) reduced pain, cartilage destruction, and pannus formation compared to Glucosamine or L. Casei alone in rat RA model. L casei is a probiotic that suppresses the proin ammatory cytokines (IL-1b, IL-6, etc.) and increases the antiin ammatory cytokines (IL-10 and IL-4) in the rat OA model. To date, there are no human studies that have investigated the role of probiotics in osteoarthritis. Although these ndings are based on the analysis of few individuals and they need to be con rmed on a larger cohort. However, the patient groups included in our study were homogenous in terms of dietary intake, smoking, alcohol intake, oral medication usage, physical activity, sun exposure, or BMI, which con rms that the observed interaction between KOA and Vitamin D status is not produced by other confounding factors and strengthens our data.

Conclusion
The gut microbiome in knee osteoarthritis patients with or without vitamin D de ciency presents different bacterial abundances. From these data, we suggest that the impact of vitamin D de ciency in the KOA disease could be mediated through the gut microbiome. These observations recommend a larger Vitamin D interventional study in KOA patients to con rm the bene cial effects of the observed dysbiosis.

Declarations
Con ict-of-interest statement: The authors of this manuscript declare that they have no con icts of interest to disclose.

Figure 2
Knee osteoarthritis-associated dysbiosis. (A) canonical correspondence analysis (CCA) was performed to study the beta Diversity between stool microbiome in KOA and Normal participants. (B) Alpha-diversity using the observed OTU and Shannon index based on Bray-Curtis distances. (C) A histogram of the log 10 transformed LDA scores was computed for features that showed differential abundance between healthy subjects and KOA patients. Knee osteoarthritis-associated dysbiosis in patients with D vitamin de ciency (A) canonical correspondence analysis (CCA) was performed to study the beta Diversity between stool microbiome in KOA_VDD and VDD patients. (B) Alpha-diversity using the observed OTU and Shannon index based on Bray-Curtis distances and (C) Histogram of the log 10 transformed LDA scores were computed for features that showed differential abundance between KOA_VDD and VDD patients. Effect of Vitamin D de ciency on Knee osteoarthritis-associated dysbiosis. (A) canonical correspondence analysis (CCA) was performed to study the beta Diversity between stool microbiome in KOA_VDD and KOA patients. (B) Alpha-diversity using the observed OTU and Shannon index based on Bray-Curtis distances. (C) A histogram of the log 10 transformed LDA scores was computed for features that showed differential abundance between KOA_VDD and KOA_NVD patients.