OA susceptibility in mice is partially mediated by the gut microbiome, is transferrable via microbiome transplantation, and is associated with immunophenotype changes

Abstract

Objectives:

The MRL/MpJ ‘superhealer’ mouse strain is protected from post-traumatic OA, although no studies have evaluated the microbiome in the context of this protection. This study characterized microbiome differences between MRL and wild-type mice, evaluated microbiome transplantation and OA, and investigated microbiome-associated immunophenotypes.

Methods:

Cecal material from mixed-sex C57BL6/J (B6) or female MRL/MpJ (MRL) was transplanted into B6 and MRL mice, then OA induced by disruption of the medial meniscus surgery (DMM). In other experiments, transplantation was performed after DMM and transplantation was performed into germ-free mice. Transplanted mice were bred through F2. OARSI, synovitis, and osteophyte scores were determined blindly 8 weeks after DMM. 16S microbiome sequencing was performed and metagenomic function imputed. Immunophenotypes were determined using mass cytometry.

Results:

MRL-into-B6 transplant prior to DMM showed reduced OA histopathology (OARSI score 70% lower transplant vs. B6 control), synovitis (60% reduction), and osteophyte scores (30% reduction) 8 weeks after DMM. When performed 48h after DMM, MRL-into-B6 transplant improved OA outcomes, but not when performed 1–2 weeks after DMM. Protection was seen in F1 (60% reduction) and F2 progeny (30% reduction). Several cecal microbiome clades were correlated with either better (e.g. Lactobacillus, R=−0.32, P=0.02) or worse (e.g. Rikenellaceae, R=0.43, P=0.001) OA outcomes. Baseline immunophenotypes associated with MRL-into-B6 transplants and MRL included reduced double-negative T cells and increased CD25+CD4+ T cells.

Conclusion:

The gut microbiome is responsible in part for OA protection in MRL mice and is transferrable by microbiome transplantation. Transplantation induces resting systemic immunophenotyping changes that correlate with OA protection.

Introduction

Osteoarthritis (OA) is a chronic musculoskeletal disease characterized by progressive loss of joint function leading to pain, reduced mobility, increased morbidity, and early mortality [1,2]. Both the most common musculoskeletal disease worldwide and the leading cause of disability in the US, it is associated with several comorbidities including stroke, metabolic syndrome, anxiety/depression, and heart disease [3–5]. Although the etiology of OA is still incompletely understood, it is clear that several biological processes are at play, including chronic low-level inflammation [6]. One potential driver of this inflammation, as has been seen in a variety of other chronic conditions, is perturbation of the gut microbiome [7]. Several studies have demonstrated alterations in the gut microbiome of OA patients [8–11], and fecal microbiome transplantation from human OA patients with metabolic syndrome worsens OA outcomes in mice [12].

These findings raise the possibility of microbiome-targeted OA therapeutics. Supplementation of dietary fiber has been shown to alter the gut microbiome composition and reduce OA severity in a post-traumatic destabilization of the medial meniscus (DMM) mouse model [13], although no improvement was seen in high-fat/high-sucrose diet rat OA models following prebiotic fiber supplementation [14]. Pretreatment of mice with antibiotics reduces histologic severity following ACL rupture [15] and DMM surgery [16], and germ-free mice have reduced OA pathology following DMM surgery compared to conventionally housed animals [17]. Oral probiotic supplementation with various Lactobacillus strains have also shown benefit in OA animal models, including Lactobacillus M5 (used in the production of the Mongolian fermented dairy product kumiss) in a high-fat diet-induced mouse OA model [18] and Lactobacillus acidophilus and Lactobacillus rhamnosus in an inflammatory monoiodiacetate (MIA) injection rat OA model [19,20].

These previous studies have focused on microbiome differences following OA induction in susceptible animals; however, no studies have yet evaluated the contribution of the microbiome to OA-resistant mouse strains. We have previously demonstrated that the gut microbiome plays a role in ear cartilage healing in Murphy Roths Large (MRL) mice [21]. This curious mouse strain was first identified in the late-1990s as a “superhealer” strain owing to its unusual ability to close 2.0mm earhole puncture wounds spontaneously after 4 weeks [22], and has been subsequently found to have a generalized healing phenotype, including neonatal digital tip regrowth [23], peripheral nerve regeneration [24], and cardiac wound healing [25,26]. Importantly, MRL mice also regenerate knee cartilage in a full-thickness cartilage injury OA model [27], a trait strongly correlated with ear hole closure [28].

In the present study, we set out to perform a comprehensive investigation of the gut microbiome in OA-resistant MRL and OA-susceptible B6 mice, including microbiome differences, evaluation of OA following microbiome transplantation into germ-free (GF) and non-germ-free mice, microbiome-associated changes in systemic inflammation via mass cytometry immunophenotyping, estimation of gut permeability via measurement of serum lipopolysaccharide (LPS) levels, and imputation of microbiome metagenomic function.

Methods

Ethics Statement:

The institutional review board and institutional animal care and use committee (IACUC) of the Oklahoma Medical Research Foundation (OMRF) approved this study. OMRF IACUC#19–56, 22–68, and 22–54.

Mouse husbandry:

C57BL6/J and MRL/MpJ mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA) and housed at OMRF. All animals were permitted ad libitum access to food and water (NIH31) in a 12-hour light-dark cycle. Germ-free animals (C57BL6/n) were maintained within the OMRF gnotobiotic animal facility. All animal husbandry procedures adhered to the NIH Guide for the Care and Use of Laboratory Animals. Our experiments utilized 117 total mice (Table 1), including 12 B6 controls, 10 MRL controls, 12 MRL-into-B6 transplants, 6 B6-into-MRL transplants, 8 B6-into-GF, 8 MRL-into-GF, 8 MRL-into-B6 transplant F1 progeny, 8 B6-into-MRL transplant F2 progeny, 5 48-hour transplant-after-DMM, 5 1-week transplant-after-DMM, and 5 2-weeks transplant-after-DMM mice. As a control we also performed DMM on GF mice without prior transplantation, and kept these animals GF after DMM. Immunophenotyping was performed on a separate cohort of mice without OA being induced, including 6 B6 controls, 6 MRL-into-B6 transplants, 6 B6-into-MRL transplants, and 6 MRL controls.

Table 1:

Mouse groups and number of mice per group.

Mouse Group	Number of mice	Sex	Age at euthanasia
Transplant before DMM
B6 control	12	M	20 weeks
MRL-into-B6 transplant	12	M	20 weeks
B6-into-MRL transplant	6	M	20 weeks
MRL control	10	M	20 weeks
Progeny of MRL-into-B6 transplant
F1	8	M	20 weeks
F2	8	M	20 weeks
MRL-into-B6 transplant after DMM
48h post-DMM	5	M	20 weeks
1wk post-DMM	5	M	20 weeks
2wks post-DMM	5	M	20 weeks
Germ-free transplant experiments
MRL-into-GF	8	M	20 weeks
B6-into-GF	8	M	20 weeks
GF controls	6	M	20 weeks
Immunophenotyping analysis (no DMM performed)
B6 control	6	M	13 weeks
MRL-into-B6 transplant	6	M	13 weeks
B6-into-MRL transplant	6	M	13 weeks
MRL control	6	M	13 weeks
Total mice, all experiments	117

Mouse cecal microbiome transplantation, OA induction:

Cecal donor mice (adult male and female C57BL6/J or female MRL/MpJ, 10–14 weeks of age) were euthanized and immediately dissected under sterile conditions. Female MRL donors and mixed male/female B6 donors were used based on data from our previous earhole closure study [21], where female MRL mice showed significantly better earhole closure but no sex differences were seen among B6 animals. Only male mice were included as recipients in our study, as female animals do not reliably exhibit an OA phenotype following DMM [29]; the microbiome contributions of sex differences in DMM outcomes are the subject of another study in our laboratory. Although MRL mice are significantly larger than B6 mice, transplantation did not result in significant weight change in either MRL-into-B6 nor B6-into-MRL animals. In transgenerational experiments, both sire and dam were transplanted at 12 weeks of age, no DMM performed, and pups were used either for DMM experiments or for further breeding (for F2 generation experiments) OA surgical, and germ-free mouse, and histology details are provided in the Supplementary Methods.

16S ribosomal RNA (rRNA) gene sequencing and data analysis:

Microbiome profiles were determined by sequencing a ~460bp region including the V3 and V4 region of bacterial 16S rRNA genes, amplified from ~30ng of DNA in each sample using a high-fidelity polymerase (NEB Q5, New England Biolabs). PCR master mixes were decontaminated with double-stranded DNAse treatment (PCR decontamination kit, Arcticzymes, Tromsø, Norway). Illumina Nextera XT indices were ligated, then libraries were pooled in equimolar amounts and sequenced on an Illumina miSeq sequencer using a 300bp paired-end sequencing protocol by the Clinical Genomics Center at OMRF. Microbial analyses were performed using the Quantitative Insights into Microbial Ecology (QIIME) software package, version 2022.2.0. Demultiplexed raw sequences were quality filtered, denoised, and chimeras removed using deblur [30]. Taxonomy was assigned via a custom Naïve Bayes taxonomic classifier trained on the Silva 138 99% OTUs full-length sequence dataset.

Diversity analyses:

Alpha diversity was computed with both a phylogenetic index (Faith”s Phylogenetic Diversity), group differences were compared with a non-parametric Mann-Whitney test, as the MRL-into-B6 transplant group was found to be non-normally distributed (D”Agostino & Pearson P=0.04). Beta diversity was characterized using an unweighted unifrac model. Group differences were calculated using a permuted analysis of variance (PERMANOVA) test with 999 permutations. In each case, significance was computed using a Kruskal-Wallis test and corrected for multiple comparisons using a Benjamini-Hochberg procedure.

Group analyses:

Group differences were determined using the linear discriminant analysis effect size (LEfSe) pipeline [31]. LEfSe performs a non-parametric Kruskal-Wallis sum-rank test to detect features with significant differential abundance between groups then uses a linear discriminant analysis (LDA) to estimate the effect size of each differentially abundant feature. An LDA threshold of ≥2 was considered significant, corresponding to P≤0.01. Correlations were determined by comparing mean histologic scores of individual animals with microbiome clades, P≤0.05 was considered significant.

Mass cytometry immunophenotyping:

To further investigate underlying microbiome-mediated baseline systemic immunological changes that might correlate with OA susceptibility and/or resistance, we performed deep immunophenotyping via mass cytometry. Transplantation was performed on a separate cohort of 12-week-old male mice (6 B6 controls, 6 MRL-into-B6 transplants, 6 MRL controls, and 6 B6-into-MRL transplants) without OA induction. One week after transplantation, animals were sacrificed and splenocytes collected, stained, and imaged by the Human Phenotyping Core facility at OMRF. Additional details in Supplementary Methods.

Serum LPS analysis:

A Pierce chromogenic endotoxin quantification kit was used to quantify LPS (Thermo Fisher, Waltham, MA, USA) using an amebocyte lysate method and has a sensitivity of 0.01 EU/mL and an assay range of 0.01–0.1 EU/mL. LPS-free plasticware was utilized. Endotoxin-free water was used to dilute standards and samples were diluted 1:10. All analyses were performed using 2 technical replicates. The coefficient of determination (R²) of the standard curve was 0.95. Statistical significance was defined as P≤0.05. Inadequate serum was available for evaluation in 7 mice; final group sizes for LPS measurements were: B6 control 9, MRL control 8, B6-into-MRL transplant 6, MRL-into-B6 transplant 10.

Prediction of metagenome content and imputed bacterial functional classification:

The Phylogenetic Investigation of Communities by Reconstruction of Unobserved States version 2 (PICRUSt2) software package was used to impute bacterial metagenomes from our 16S deep sequencing microbial DNA data and functional annotation applied using the Kyoto Encyclopedia of Gene and Genomes (KEGG) catalog [32]. Statistical significance was calculated using a Student T-test, Benjamini-Hochberg FDR-corrected q≤0.05 considered significant.

Results:

Gut microbiome transplantation from MRL into B6 mice prior to OA induction prevents OA development. Transplantation from B6 into MRL worsens OA in recipient animals.

MRL mice had significantly reduced OA histologic OARSI scores (MRL: 0.64±0.15 vs. B6: 3.0±0.23, mean±SEM, P=6E-7), synovitis scores (0.54±0.10 vs. 1.2±0.14, P=0.002), and osteophyte scores (0.50±0.088 vs. 0.93±0.13, P=0.02) compared to B6 controls after DMM (Figure 1A, Figure 2). Transplantation of MRL cecal microbiota into B6 mice 48 hours prior to DMM reduced OA histologic severity (MRL-into-B6: 0.92±0.19 vs. B6 3.0±0.23, P=6E-7), synovitis (0.71±0.10 vs. 1.2±0.14, P=9E-3) and osteophytes (0.29±0.060 vs. 0.93±0.13, P=1E-4), reducing each of these measures to a level equivalent to MRL controls (Figure 1A). Transplantation of B6 cecal material into MRL animals resulted in mild worsening of OA histology (B6-into-MRL: 1.5±0.28 vs. MRL: 0.64±0.15, P=0.04) and synovitis (1.2±0.073 vs. 0.54±0.10, P=0.001) scores but did not result in worsened osteophyte score compared to MRL controls (0.71±0.087 vs. 0.50±0.088, P=0.2).

Figure 1:

OA outcomes in mice under various transplantation conditions, including OARSI score, histologic synovitis and histologic osteophyte scores. Horizontal line represents mean, bars represent SEM. (A) Results of mice transplanted 48 hours prior to DMM and B6/MRL controls; (B) Results of MRL-into-B6 mice transplanted at various times after DMM (48h, 1 week, and 2 weeks), along with transplant-before-DMM and B6 controls for comparison; (C) Results of progeny of MRL-into-B6 transplanted mice, with primary transplanted (F0) mice and B6 controls for comparison; (D) Results of MRL- and B6-transplantation into C57BL6/n germ-free mice, transplanted 48h prior to DMM, with B6 and MRL controls for comparison. P values were calculated using a non-FDR-corrected Student t-test (data confirmed to be normally distributed).

Figure 2:

Example histology images from various transplantation groups, modified coronal sections stained with Safranin-O. Medial knee compartment shown.

OA protection following gut microbiota transplantation wanes when transplantation is performed later after DMM

We next sought to determine whether microbiome modification via transplantation was effective as a therapeutic rather than simply a preventative; that is, whether it was effective when performed after OA induction had already occurred. To that end, we transplanted MRL cecal contents into B6 mice 48 hours after, 1 week after, and 2 weeks after DMM. Using the same endpoint as our previous prior-to-DMM experiments (8 weeks), we found improvement in OA histologic score at the 48h-post-DMM timepoint (48h: 1.6±0.12 vs. control: 3.0±0.23 mean OARSI score, P=0.02), although this was inferior to transplantation prior to DMM (P=0.003). There was a nonsignificant improvement in histologic outcomes at the 1 week- (2.2±0.32, P=0.1) and 2 weeks- (2.8±0.44, P=0.7) post-DMM transplantation timepoints. No improvement was seen in synovitis nor osteophyte scores when transplantation occurred after DMM (Figure 1B).

OA protection associated with gut microbiome transplantation is transgenerationally heritable

In a previous study of gut microbiome transplantation and earhole closure outcomes [21], we found transplantation-mediated improvements in wound healing to be transgenerationally heritable in the first progeny generation (F1). Therefore, we next analyzed the heritability of OA protection in progeny of transplanted mice. First-generation (F1) progeny of transplanted mice demonstrated similar OA scores compared to primary transplantation mice (F0): histopathologic OARSI (F0: 0.92±0.19 F1: 1.28±0.11, mean±SEM, P=0.2), synovitis (F0: 0.71±0.10 vs. 0.43±0.060 F1: P=0.06), and osteophyte (F0: 0.29±0.060 vs. F1: 0.46±0.085, P=0.1) outcomes. Second-generation (F2) progeny demonstrated mild protection in OA histopathology (F2: 2.03±0.34 vs. B6: 3.0±0.23, P=0.03) although mild worsening in synovitis (F2: 1.7±0.043 vs. B6: 1.2±0.14, P=0.04) and no change in osteophytes (F2: 1.2±0.071 vs. B6: 0.93±0.12, P=0.2) compared to B6 controls (Figure 1C).

MRL microbiota prevents OA whereas B6 microbiota transplantation is permissive for OA development when transplanted into Germ-free mice

As an additional validation of our microbiome transplantation findings, we then transplanted B6 and MRL microbiota into adult Germ-free (GF) B6 male mice, then performed DMM one week later. In agreement with our previous experiments, B6-into-GF mice exhibited similar OA histologic (B6-into-GF: 2.78±0.42 vs. B6: 3.0±0.23, mean±SEM, P=0.8), synovitis (B6-into-GF: 1.16±0.11 vs. 1.2±0.14, P=0.8), and osteophyte (B6-into-GF: 0.96±0.12 vs. B6: P=0.7) outcomes compared to B6 controls, whereas MRL-into-GF mice had similar OA histologic (P=0.4), synovitis (P=0.9), and osteophyte (P=0.3) outcomes compared to MRL controls. In all three measures, B6-into-GF mice exhibited worse outcomes than MRL-into-GF comparators (histologic P=0.01, synovitis P=0.002, osteophyte P=0.002) (Figure 1D). As a control, we also performed DMM on GF male mice without prior microbiome transplantation and confirmed they were mostly protected from OA, although not statistically different than MRL-into-GF mice (n=6 male GF OARSI score: 1.49±0.26, P=0.08 vs. MRL-into-GF).

Gut microbiome composition differs between OA-resistant MRL and OA-susceptible B6 mouse strains. Specific gut microbiota clades are associated with OA protection following transplantation and are implicated in transgenerational heritability

Next, we compared cecal microbial profiles generated by 16S rRNA gene deep sequencing among the various microbiome transplantation groups. A total of 66 gut microbiome clades were different in at least one comparison (Figure 3A, Supplementary Table 2). Of these, 25 clades were altered with MRL-into-B6 transplantation vs. B6, 20 in MRL control vs. B6 control comparison, 24 in MRL vs. B6-into-MRL transplantation, 30 in progeny (F1 and F2) of MRL-into-B6 vs. B6 controls, and 13 in B6-into-GF vs. MRL-into-GF comparisons. Four clades demonstrated consistent changes in 4 of the 5 group comparisons, including increases in family Coriobacteriaceae and decreases in Bacteroidetes ovatus, family Rikenellaceae, family Bacteroidaceae, and genus Bacteroidetes in association with MRL- or MRL-into-B6 transplantation, with reductions in these clades in B6-into-MRL transplants. We next performed an individual animal-level correlation analysis comparing microbiome clades with OARSI histopathologic scores in B6 controls, MRL controls, MRL-into-B6 transplants, and F1 and F2 progeny of MRL-into-B6 transplants. We identified 9 correlated clades (Pearson P≤0.05), including 5 negative correlations; i.e., increased presence of these clades associated with better OA outcomes (genus Lactobacillus: R=−0.32, P=0.02, Akkermansia muciniphila: R=−0.31, P=0.03, genus Oscillospira: R=−0.28, P=0.04, and genus Coprobacillus: R=−0.28, P=0.05). We found 5 clades with positive correlation; i.e., increased presence associated with worse OA outcomes (family Rikenellaceae: R=0.43, P=0.001, family Christensenellaceae: R=0.37, P=0.007, genus Anaerostipes: R=0.35, P=0.01, Oscillospira guilliermondii: R=0.33, P=0.02, and Bacteroidetes ovatus: R=0.28, P=0.04) (Figure 3B).

Figure 3:

Microbiome composition analysis by 16S deep sequencing. (A) Microbiome group comparisons. Numbers represent linear discriminant analysis effect size (LDA-ES); negative values indicate enrichment in mouse groups associated with worse OA outcome (B6, B6-into-MRL, B6-into-GF), whereas positive values indicate enrichment in mouse groups associated with better OA outcomes (MRL, MRL-into-B6, progeny of MRL-into-B6, MRL-into-GF). Only statistically significant clade difference (Kruskal-Wallis P≤0.05 followed by linear discriminant analysis) values are shown. (B) OARSI score correlation with individual-level gut microbiome data, top 3 negatively- and top 3 positively correlated microbiome clades presented. This correlation includes B6 controls, MRL controls, MRL-into-B6 transplants, and F1 and F2 progeny of MRL-into-B6 mice. Blue line indicates best-fit linear regression among all groups, red curved lines are 95% confidence intervals of linear regression. R values represent Pearson correlation coefficient, P values presented significance level of Pearson correlation coefficient.

Microbiome transplantation-mediated OA protection correlates with specific splenocyte immunophenotypes

Next, we explored microbiome-mediated changes in circulating immune cell subsets and intracellular cytokine levels that correlate with gut microbiome transplantation. To do this, we performed mass cytometry (CyTOF) on baseline (no DMM) splenocytes in a reduced set of mice (n=5 B6 controls, n=6 MRL-into-B6, n=6 MRL controls, n=6 B6-into-MRL). Two cellular subsets were significantly different in all 3 group comparisons: naïve B cells and double-negative T cells. In MRL-into-B6 transplanted mice we found 7 differences compared to B6 controls (Figures 4A and 4B, Supplementary Table 3); all 7 of these were also different when MRL were compared to B6 mice. Four of these cell subsets had same-direction changes in both MRL-into-B6 transplanted and MRL (double-negative T cells reduced in transplant and MRL, naïve B cells reduced in transplant and MRL, CD62L+ monocytes reduced in transplant and MRL, and CD25+CD4+ T cells increased in transplant and MRL). In MRL controls vs. B6 controls, we identified an additional 9 cellular subset differences that were not significantly changed in MRL-into-B6 transplants. In our B6-into-MRL transplanted mice, we found 11 group differences, of which 6 were also present in the MRL control vs. B6 control comparisons.

Figure 4:

Mass cytometry immunophenotyping of a second group of animals post-transplantation but without DMM surgery, reflecting baseline immunophenotype changes induced by microbiome transplantation. (A) t-distributed stochastic neighbor embedding plot (tSNE) under various microbiome conditions (B) individual peripheral immune cell subset differences, only statistically significant cell subsets presented; (C) intracellular cytokine differences, only selected statistically significant cytokines presented; (D) serum lipopolysaccharide (LPS) levels among various transplant groups. P values calculated using a non-FDR-corrected Student t-test (data confirmed to be normally distributed).

We further identified 21 intracellular cytokine differences in MRL-into-B6 transplants vs. B6 controls (Figure 4C, Supplementary Table 4), 7 differences among MRL controls vs. B6 controls (3 shared with the MRL-into-B6 transplants vs. B6 controls comparison including reduction of TNF in dendritic cells in both MRL-into-B6 transplant and MRL mice, similar reduction in IL-10 in monocytes, and similar reduction in IL-4 in double negative T cells). We also found 6 differences among MRL controls vs. B6-into-MRL transplants, one of which was shared with previous comparisons (increased TNF in DCs in B6-into-MRLs vs. MRL controls).

Gut permeability is decreased with MRL-into-B6 microbiome transplants and increased in B6-into-MRL transplants

Next, we indirectly evaluated intestinal permeability via measuring serum LPS levels. MRL mice had higher baseline LPS compared to B6 mice (B6: 0.41±0.01 Endotoxin units [EU]/mL vs. MRL: 0.48±0.02, P=0.004, Figure 4D). MRL-into-B6 transplantation reduced LPS levels (MRL-into-B6: 0.38±0.01 vs. B6: 0.41±0.01, P=0.02), whereas B6-into-MRL transplantation increased LPS levels (0.70±0.04 vs. 0.48±0.02, P=1E-5).

Imputed microbiome metagenomes suggest increases in several metabolic pathways are associated with microbiome-mediated OA protection

Finally, we imputed metagenomic function from our 16S analysis using PICRUSt2 [32]. For this analysis, we focused on B6 control vs. MRL control and B6 control vs. MRL-into-B6 transplant comparisons, which demonstrated the greatest degree of OA histology, synovitis, and osteophyte protection. We identified 41 significantly altered metagenomic pathways (FDR-corrected q<0.05) in MRL vs. B6 controls and 186 pathways different in MRL-into-B6 transplants vs. B6 controls (Supplementary Table 5). We then reasoned that pathways important for microbiome-mediated OA protection would likely be shared between the two comparisons; 28 such overlaps were identified (Table 2). Of these, 26 had concordant changes among the two comparisons; that is, increased (or decreased) pathway activity was present in both MRL and MRL-into-B6 animals. The most strongly induced pathways in MRL and MRL-into-B6 transplants included TCA cycle VII (acetate producers) (MRL-into-B6 vs. B6 ratio 25.7, q=0.02, MRL vs. B6 ratio 37.3, q=5E-4), ADP-L-glycero/beta-D-manno-heptose biosynthesis (MRL-into-B6 vs. B6 ratio 11.3, q=0.04, MRL vs. B6 ratio 3.0, q=0.02), and glycolysis/Entner-Doudoroff superpathway (MRL-into-B6 vs. B6 ratio 4.0, q=0.02, MRL vs. B6 ratio 1.9, q=0.03). The two discordant pathways were GDP-D-glycero/alpha-D-manno-heptose biosynthesis (increased in MRL-into-B6 transplants but decreased in MRL vs. B6 controls) and superpathway of glucose and xylose degradation (also increased in MRL-into-B6 transplants but decreased in MRL vs. B6 controls).

Table 2:

Imputed metagenomic pathways shared among (MRL vs. B6) and (MRL-into-B6 vs. B6) comparisons. FDR correction by Benjamini-Hochberg procedure (q values) presented.

Metagenomic Pathway	q transpl vs b6	q mrl vs b6	B6 mean±SEM	MRL mean±SEM	MRL-into-B6 transpl mean±SEM	MRL-into-B6:B6 ratio	MRL:B6 ratio	Concordance between group comparisons
TCA cycle VII (acetate-producers)	0.02	0.0005	140±21	5100±900	3500±1200	25.7	37.3	Y
ADP-L-glycero-β-D-manno-heptose biosynthesis	0.04	0.02	250±30	740±130	2800±1000	11.3	3.0	Y
superpathway of glycolysis and Entner-Doudoroff	0.02	0.03	10000±1300	19000±2400	41000±11000	4.0	1.9	Y
GDP-D-glycero-α-D-manno-heptose biosynthesis	0.04	0.03	3800±440	1920±310	13000±3800	3.5	0.5	N
preQ0 biosynthesis	0.008	0.005	14000±810	20000±1200	44000±8200	3.2	1.5	Y
tetrapyrrole biosynthesis II (from glycine)	0.007	0.0005	13000±1000	22000±1100	42000±6500	3.2	1.7	Y
S-adenosyl-L-methionine cycle I	0.007	0.02	34000±2600	50000±3700	61000±6200	3.1	1.7	Y
thiazole biosynthesis I (E. coli)	0.007	0.001	15000±735	23000±1500	41000±6500	2.8	1.6	Y
adenosylcobalamin salvage from cobinamide I	0.0001	0.02	12000±790	21000±2600	33000±2900	2.7	1.8	Y
adenosylcobalamin salvage from cobinamide II	0.0001	0.03	11000±740	20000±2700	31000±2800	2.7	1.8	Y
adenosylcobalamin biosynthesis from cobyrinate a,c-diamide I	0.0001	0.03	11500±731	20000±2700	31000±2800	2.7	1.7	Y
superpathway of arginine and polyamine biosynthesis	0.02	0.05	4100±460	7000±870	11000±2300	2.6	1.7	Y
peptidoglycan biosynthesis IV (Enterococcus faecium)	0.02	0.0003	18000±1600	35000±2000	46000±9200	2.6	2.0	Y
superpathway of tetrahydrofolate biosynthesis	0.02	0.008	22000±1700	31000±1400	54000±11000	2.5	1.4	Y
superpathway of tetrahydrofolate biosynthesis and salvage	0.02	0.02	27000±2100	37000±1700	64000±12000	2.4	1.4	Y
purine nucleobases degradation I (anaerobic)	0.02	0.002	5300±570	20000±3200	13000±2500	2.4	3.8	Y
peptidoglycan maturation (meso-diaminopimelate containing)	0.004	0.04	29000±2500	41000±2800	65000±6900	2.3	1.4	Y
purine nucleotides degradation II (aerobic)	0.02	0.03	15000±1200	26000±2900	35000±6200	2.3	1.7	Y
glycolysis II (from fructose 6-phosphate)	0.01	0.05	42000±4600	60000±3300	94000±14000	2.2	1.4	Y
glycerol degradation to butanol	0.05	0.02	3700±740	11000±1900	8200±1700	2.2	2.9	Y
homolactic fermentation	0.008	0.02	42000±4000	61000±3400	92000±14000	2.2	1.4	Y
glycolysis I (from glucose 6-phosphate)	0.008	0.04	46000±4000	64000±3700	100000±14000	2.2	1.4	Y
methanogenesis from acetate	0.02	0.004	1700±360	8100±1400	3600±590	2.2	4.8	Y
acetylene degradation	0.01	0.02	29000±2900	47000±3900	58000±9000	2.0	1.6	Y
pyruvate fermentation to acetate and lactate II	0.008	0.01	49000±3200	72000±4800	99000±13000	2.0	1.5	Y
superpathway of glucose and xylose degradation	0.02	0.02	23000±4200	6300±1100	44000±6200	1.9	0.3	N
4-aminobutanoate degradation V	0.04	0.01	1400±220	4900±890	2600±430	1.8	3.5	Y
tetrapyrrole biosynthesis I (from glutamate)	0.007	0.0005	14000±1000	23000±1100	43000±7300	1.8	1.5	Y

Discussion:

In the present study, we demonstrate that direct augmentation of the gut microbiome via transplantation is sufficient to affect PTOA outcomes in the DMM model as measured by OARSI, synovitis, and osteophyte scores. PTOA severity can be reduced in B6 mice via transplantation from MRL mice, whereas MRL outcomes can be worsened via transplantation from B6 mice. We go on to show that the improvements seen in MRL-into-B6 mice persist through second-generation progeny, although at a diminished level. We also find that MRL or B6 gut microbiota, when introduced into germ-free B6 mice, is sufficient to protect or render susceptible, respectively, recipient animals to OA. Finally, we identify a number of gut microbiome clades and imputed metagenomic pathways that correlate with OA protection or susceptibility, as well as identify several systemic inflammatory cell population and gut permeability changes that are altered with transplantation.

Our findings mirror those we previously published regarding gut microbiome and the earhole healing phenotype of MRL mice to B6 recipients [21]; indeed, earhole closure rates have been shown to strongly correlate with OA outcomes in various mouse strains [28]. Several gut microbial clades we previously found correlated with earhole healing were also correlated with OA protection in the current study, including genus Lactobacillus (Pearson R=+0.65 for earhole closure [higher Lactobacillus associated with improved earhole closure], R=−0.32 for DMM histology [higher Lactobacillus associated with lower OA histology score and, thus, improved OA outcome]), family Christensenellaceae (R=−0.65 for earhole closure, R=+0.37 for DMM histology [more Christensenellaceae associated with poorer earhole and OA outcomes]), as well as one clade with differing directionality in ear and joint, genus Akkermansia, R=−0.31 for DMM histology and R=−0.80 for earhole closure [associated with improved OA but worse earhole outcomes].

The premise for the current study was our expectation that the MRL genetic background might induce (previously described) differences in immune responses not directly, but rather via influencing the gut microbiome which, has a subsequent effect of immunoregulation. If this were indeed the case, this immunoregulatory effect would be transferrable via microbiome transplantation, which we show is indeed the case. Previous studies of MRL OA resistance have focused on full-thickness cartilage injury and have not shown that partial-thickness cartilage injury results in tissue regeneration [27], although the DMM model we used in this study has not been previously evaluated in MRL mice. Indeed, our longitudinal transplantation findings (transplant after DMM) demonstrate that an OA-protective effect is lost at later timepoints, suggesting that the effects we see are likely related to the acute inflammatory process involved in OA rather than cartilage regeneration seen in the previous full-thickness MRL injury studies. Indeed, previous genetic studies have demonstrated strong positive genetic correlation between MRL cartilage regeneration and ear-wound and cartilage hole repair, suggesting that regeneration in MRL mice may be a strictly genetic feature whereas immune-mediated outcomes may be partially microbiome-dependent. This hypothesis is reinforced by our previous finding that earhole closure in MRL animals cannot be made worse via gut microbiome transplantation from B6 animals [21], where earhole closure represents a more “pure” regeneration phenotype rather than a prevention-of-damage-accrual phenotype that we find in the current OA study.

Our data add to the growing literature on gut microbiome susceptibility in knee OA. By far the largest previous human study was performed in the Dutch Rotterdam RS(III) and LifeLines-DEEP OA cohorts in the Netherlands, including 2294 patients[8]. Four bacterial clades were associated with knee OA pain; these included class Bacilli, order Lactobacillales, family Streptococcaceae, and genus Streptococcus following adjustment for age, sex, and body mass index (BMI). This agrees with our finding of increased Lactobacillales as characteristic of worse DMM outcomes in B6 vs. MRL comparisons. Of note was our finding that an individual genus with Lactobacillales, Lactobacillus, was strongly correlated with improved DMM outcomes (see discussion later). We must assume that certain genera within this order (e.g. members of genus Streptococcus) may have pro-OA effects whereas others (e.g. members of genus Lactobacillus) may have OA protective effects. Future work will be needed to develop more detailed microbiome compositional data, particularly, whole-microbial-genome sequencing (shotgun sequencing) that may allow for full species-level taxonomical analysis. to help further clarify these disparate findings. We did not find evidence for Streptococcus in our microbiome data from any mouse group.

Christensenellaceae has been associated with both hand [9] and any-joint OA [33]; this agrees with our finding of depletion of this family in MRL-into-B6 transplants and enrichment in B6-into-GF vs. MRL-into-GF mice. We also found class Erysipelotrichi associated with OA in mice, in agreement with a 2021 study of the effects of vitamin D deficiency and gut microbiome in knee OA patients [34]; indeed, this class has also been associated with obesity- and load-induced OA in a mouse model [35]. Genus Clostridium has been associated with OA in rat models [36,37]. A large number of papers have also associated members of phylum Bacteroidetes with murine OA[16,17,38]; we found members of this phylum depleted in multiple group comparisons, including MRL vs. B6, MRL vs. B6-into-MRL, and B6 vs. transplant progeny. Within this phylum, we found family Rikenellaceae to be very strongly associated with worse OA outcomes; this family has been associated previously with OA in our human cartilage work [39], as well as others” work on murine OA [40]. Members of phylum Verrucomicrobia, particularly genus Akkermansia, have been associated with OA in human patients [10], rat models of OA [37], and mouse models of OA [41]. Notably, our results were discordant regarding Akkermansia muciniphila; in group analysis, this species was associated with worse outcomes in the progeny of MRL-into-B6 transplants vs. B6 controls, whereas in individual-level correlation analysis, it was strongly correlated with OA protection. Further evaluation of this species will be needed to determine its effects on OA outcomes. Among microbiome clades associated with improved OA outcomes, our data agree with several previous studies on Lactobacillus in animal models [36,42,43] and human patients [44,45], as does our association of Lachnospiraceae [9] and family Mogibacteriaceae (from a dog OA study) [46]. A final bacterial clade of note is family Coriobacteriaceae which was increased among mouse transplant groups that were protected from OA; this family has been shown to be increased in long-term runners [47], an activity associated with OA protection in humans.

Serum and synovial LPS is elevated in OA patients [48–50], likely from increased gut permeability, and has been shown to exacerbate OA in animal models [51]. In the current study, we found that MRL-into-B6 transplantation reduced serum LPS levels compared to B6 controls, and B6-into-MRL transplantation increased serum LPS levels compared to MRL controls (Figure 4D). Curiously, however, we found that the baseline levels of LPS in MRL mice were significantly higher than B6 controls, we can only speculate that MRL mice may be less reactive to LPS than B6 control animals. An important limitation to our LPS quantitation is that the initial blood collection in our animals was performed in standard blood collection tubes, rather than specific LPS-free collection tubes, although subsequent processing was performed using strictly LPS-free plasticware and reagents.

Our microbial metabolomics and splenocyte immunophenotyping identified several pathways that could help explain the phenotypic differences we observed. For example, the most upregulated microbial metabolomic pathway we identified was acetate production. Acetate has been shown to promote regulatory T cell (Treg) generation [52]. Indeed, our immunophenotyping results indicate that mouse groups protected from OA have expansions of peripheral CD25+CD4+ T cells, which are likely Tregs, although definitive identification will require further characterization including intracellular Foxp3 expression in future studies.. Previous studies on Tregs in OA have been somewhat contradictory; on the one hand, increases in joint-infiltrating Tregs correlate positively with OA symptoms [53], although this study noted a significantly increased number of Tregs in joint tissues compared to peripheral blood, indicating that recruitment of Tregs may be a reactive process related to ongoing inflammation rather than a primary change. Furthermore, adoptive transfer of Col II-specific anti-inflammatory Tregs protect mice from OA [54], and peripheral blood T-regs are decreased in knee OA patients [55].

Our CyTOF analysis also identified double-negative T cells (DNTs, lacking CD4 and CD8 co-receptors) as significantly reduced upon MRL-into-B6 and increased after B6-into-MRL transplantation. No studies have yet reported on this cell population in OA, although extensive evidence links their expansion in autoimmune and other inflammatory conditions. We also saw a reduction in CD62L+ monocytes in MRL-into-B6 and an increase in B6-into-MRL; presumably, these represent “classical” inflammatory monocytes that are rapidly recruited to infection/inflammation sites [56]; future work should include macrophage subset analysis in transplanted mice. One cellular subset (Naïve B cells) demonstrated a paradoxical decrease in B6-into-MRL mice compared to MRL controls, and CD62- B cells were increased with MRL-into-B6 transplantation whereas they were lower in MRL controls; the significance of these two changes is not clear.

Our study has several limitations. We did not examine the effects of sex on the gut microbiome and related differences in PTOA outcomes in mice; previous studies have shown DMM induces a stronger histologic phenotype in male mice than female mice [29]. The role of the microbiome in mediating these sex discrepancies is being actively explored by our laboratory. Additionally, we did not include behavioral or pain outcomes; a number of human studies have indicated the gut microbiome can affect pain, including in OA [8,57]. The possibility of microbiome transplantation as a pain-reducing modality for OA therapy should be explored in future studies. Furthermore, we only indirectly explored intestinal permeability; a future hypothesis-driven direct investigation would be beneficial.

We evaluated immunophenotypes in transplanted non-DMM”d animals, reasoning that differences in baseline immune cell populations would be correlated with OA outcome. However, almost certainly differences in cellular responses at timepoints following DMM (particularly within the acute inflammatory phase) would also be affected and may have a more direct influence on OA pathology. This should be evaluated in a future longitudinal study of immunophenotype changes at multiple timepoints post-DMM and will be a focus of future work. It is likely that the OA-preventative and therapeutic effects we see related to microbiome modification are not induced by microbes themselves, but rather immunomodulatory microbial proteomic/metabolomic products, which we imputed indirectly to gain early insight into pathways that should be the focus of future studies. Another limitation of the current study is our relatively early timepoint for animal euthanasia (8 weeks). Previous studies of MRL mice (and their parent strains) have used later timepoints of 12 to 16 weeks, and have found that some additional measures of cartilage regeneration, specifically in the ear, demonstrated further segregation at 16 weeks post-injury[28,58,59]. Future studies should extend our work to include later timepoints.

In summary, in the present study we report compelling evidence for a link between the gut microbiome and OA development in mice with a particular focus on microbiome modification as an OA preventative. Additionally, we present the first evidence that the OA-protected phenotype seen in MRL/MpJ mice is due, at least in part, to the gut microbiome. Our findings strengthen previous OA microbiome associations and offer insight into the underlying systemic immunophenotype changes that associate with improvements in OA outcomes. Future studies should extend our findings and evaluate the potential of individual microbiome clades we identify as OA preventatives and/or therapeutics.

Key Messages:

What is already known on this topic:

• DMM surgery performed on C57BL6/J (B6) mice is the standard OA mouse model

• Certain naturally-occurring mouse strains are protected from OA, including MRL/MpJ (MRL) mice

• Various gut microbiome alterations have been previously associated with the risk for human and mouse OA

What this study adds:

• The gut microbiome is significantly different in MRL mice compared to B6 mice

• Microbiome transplantation from OA-protected MRL mice to OA-susceptible B6 mice improves OA outcomes. Conversely, transplantation from B6 to MRL mice worsens OA outcomes

• Specific gut microbiome clades are strongly correlated with either OA risk or OA protection

• Baseline (pre-DMM) systemic immunophenotypes are different in MRL vs. B6 mice and are associated with both transplantation and OA protection

How this study might affect research, practice, or policy:

• This study adds to mounting evidence for the broad effects of the gut microbiome on chronic diseases like OA

• This study provides specific gut microbiome targets for future OA therapeutic research

• Individual immunological pathways, influenced by the microbiome, should be the focus of additional studies and targeted for the prevention and/or treatment of OA

Data Sharing Statement:

The data that support the findings of this study are available from the corresponding author, MAJ, upon reasonable request.