Commensal Gut Microbiota Immunomodulatory Actions in Bone Marrow and Liver have Catabolic Effects on Skeletal Homeostasis in Health

Despite knowledge the gut microbiota regulates bone mass, mechanisms governing the normal gut microbiota’s osteoimmunomodulatory effects on skeletal remodeling and homeostasis are unclear in the healthy adult skeleton. Young adult specific-pathogen-free and germ-free mice were used to delineate the commensal microbiota’s immunoregulatory effects on osteoblastogenesis, osteoclastogenesis, marrow T-cell hematopoiesis, and extra-skeletal endocrine organ function. We report the commensal microbiota has anti-anabolic effects suppressing osteoblastogenesis and pro-catabolic effects enhancing osteoclastogenesis, which drive bone loss in health. Suppression of Sp7(Osterix) and Igf1 in bone, and serum IGF1, in specific-pathogen-free mice suggest the commensal microbiota’s anti-osteoblastic actions are mediated via local disruption of IGF1-signaling. Differences in the RANKL/OPG Axis in vivo, and RANKL-induced maturation of osteoclast-precursors in vitro, indicate the commensal microbiota induces sustained changes in RANKL-mediated osteoclastogenesis. Candidate mechanisms mediating commensal microbiota’s pro-osteoclastic actions include altered marrow effector CD4+T-cells and a novel Gut-Liver-Bone Axis. The previously unidentified Gut-Liver-Bone Axis intriguingly implies the normal gut microbiota’s osteoimmunomodulatory actions are partly mediated via immunostimulatory effects in the liver. The molecular underpinnings defining commensal gut microbiota immunomodulatory actions on physiologic bone remodeling are highly relevant in advancing the understanding of normal osteoimmunological processes, having implications for the prevention of skeletal deterioration in health and disease.


Results
Commensal microbiota has anti-anabolic effects on trabecular bone remodeling. Histomorphometric and micro-CT studies were carried out in the trabecular bone compartment, based on the anatomical region being the most metabolically active site of bone remodeling in the young adult bone organ 31,32 . Histomorphometric analysis revealed decreased trabecular bone area (B.Ar/T.Ar) in the distal femur of SPF vs. GF mice (Fig. 1b,c). Micro-CT analysis of the proximal tibia demonstrated an osteopenic trabecular bone phenotype in SPF vs. GF mice (Fig. 1e). Consistent with the reduced trabecular bone quantity found in the distal femur via histomorphometry (Fig. 1b), micro-CT analysis revealed decreased trabecular bone volume (BV/TV) in the proximal tibia of SPF mice (Fig. 1f). Inferior trabecular bone micro-architecture properties in the proximal tibia of SPF mice were characterized by reduced trabecular number (Tb.N) (Fig. 1g), and a trend towards increased trabecular separation (Tb.Sp) (Fig. 1i). Micro-CT analysis of cortical bone parameters revealed no differences in femur length, and marginally reduced cortical bone (Ct.Ar/Tt.Ar) in the femur mid-diaphysis of SPF mice ( Supplementary Fig. S1).
Recognizing that the normal gut microbiota blunts the accrual of bone mass during bone modeling (growth) processes in the developing C57BL/6 skeleton [19][20][21] , dynamic bone formation indices were analyzed  to validate that alterations in bone remodeling are contributing to the osteopenic trabecular skeletal phenotype found in young adult SPF vs. GF mice. Mineral apposition rate (MAR) (Fig. 1k) and bone formation rate (BFR) (Fig. 1l) were suppressed in the distal femur trabecular bone of SPF mice, which demonstrates that the commensal gut microbiota has anti-anabolic effects on trabecular bone remodeling in the adult skeleton.
Commensal microbiota suppresses osteoblast differentiation and function. Considering the blunted dynamic bone formation indices found in the remodeling trabecular bone of SPF mice (Fig. 1j-l), bone marrow stromal cell (BMSC) osteoblast-progenitors were isolated to elucidate commensal microbiota induced alterations in osteoblast potential ( Fig. 2a-g). BMSC expansion over time was blunted in SPF vs. GF BMSC cultures (Fig. 2a), and gene expression studies in untreated day-4 BMSC cultures revealed alterations in intrinsic mesenchymal-stromal cell differentiation potential (Fig. 2b-e). Col2a1 (Fig. 2c), an early marker for chondrogenesis, and Runx2 (Fig. 2d) and Sp7 (Fig. 2e), transcription factors critical for commitment to and maturation in the osteoblastic lineage 33 , were significantly decreased in day-4 BMSC cultures from SPF vs. GF mice. Consistent with the reduced osteoblastic differentiation potential observed in untreated day-4 BMSC cultures from SPF mice (Fig. 2d,e), 21-days mineralization treatment (von Kossa assay) induced less mineralization in SPF vs. GF BMSC cultures (Fig. 2f,g).
SPF mice have decreased Bglap and Igf1 in bone, and lower serum IGF1. In light of the blunted mineralization in SPF trabecular bone (Fig. 1k) and BMSC cultures (Fig. 2f,g), Bglap(Osteocalcin) was assessed as a marker for mature osteoblast function in bone marrow and calvaria (Fig. 2h). While bone marrow and calvaria are both heterogeneous in their cellular composition, the rationale for including calvaria is based on its more homogeneous stromal-osteoblastic cellular composition better reflecting osteoblast specific gene expression. Corroborating the suppressed osteoblast function detected in vivo (Fig. 1k) and in vitro (Fig. 2f,g) in SPF vs.
GF mice, Bglap was marginally down-regulated in SPF bone marrow, and significantly decreased in SPF calvaria (Fig. 2h).
Considering Sp7 (Osterix) was decreased in BMSC cultures from SPF mice (Fig. 2e), and that IGF1 signaling upregulates Sp7 mediated osteoblast maturation and function 33,34 , Igf1 expression and serum IGF1 were evaluated (Fig. 2i,j). Igf1 was decreased in SPF vs. GF bone marrow and calvaria (Fig. 2i), which suggest the commensal gut microbiota's inhibitory effects on osteoblastogenesis are potentially mediated through blunted IGF1 signaling in osteoblastic cells 35 . Indirect evidence from a transgenic mouse model locally deficient in IGF1 signaling within osteoblastic cells, which has a skeletal phenotype (reduced BV/TV, decreased Tb.N, increased Tb.Sp) 35 strikingly similar to SPF vs. GF mice (Fig. 1), indicates the commensal gut microbiota impact on trabecular bone morphology may occur in part by impaired IGF1 signaling within skeletal tissue. Consistent with the reduced Igf1 expression in SPF vs. GF bone tissues (Fig. 2i), serum IGF1 was 18.3% lower in SPF mice (Fig. 2j). Appreciating that liver-derived IGF1 constitutes 70% of circulating IGF1 36 , differences were ruled out in liver Igf1 expression (Fig. 2i). While the suppressed osteoblastogenesis phenotype in SPF vs GF mice (Fig. 1j-l; Fig. 2) delineates the mRNA assessed as markers of osteoblastogenic potential. Relative quantification of mRNA was performed via the comparative C T method (ΔΔCT); Gapdh was utilized as an internal control gene; data expressed as fold difference relative to SPF. (f,g) von Kossa mineralization assay (21 day mineralization treatment) (n = 4/gp). (f) Representative von Kossa stained culture images. (g) Mineralization area per well area. (a-g) BMSC assays carried out in duplicate (two technical replicate) cultures; n-values represent biological replicates per group. (h-j) Commensal microbiota in vivo regulation of osteoblastogenesis. 11 to 12 week-old male SPF & GF mice were euthanized; tissues were harvested. (h,i) RNA was isolated from marrow (n = 4/gp), calvaria (n = 4/gp), liver (n = 6/gp) for qRT-PCR analysis of candidate osteogenic genes. (h) Bglap(Osteocalcin) mRNA assessed as a marker of mature osteoblast function, and (i) Igf1 mRNA assessed as a critical osteoblastic signaling factor. Relative quantification of mRNA was performed via the comparative C T method (ΔΔCT); Gapdh was utilized as an internal control gene; data expressed as fold difference relative to SPF. (j) Serum was isolated from whole blood (n = 10/gp); ELISA analysis of IGF1 levels. Data reported as mean ± SEM. *p < 0.05 vs. SPF; **p < 0.01 vs. SPF. commensal microbiota's catabolic effects on skeletal remodeling are mediated in part by blunted osteoblast bone formation, recognizing that bone remodeling occurs through dual osteoclast-osteoblast actions, investigations were carried out to elucidate the commensal microbiota impact on osteoclastogenesis.
Commensal microbiota enhances osteoclast size and eroded bone perimeter. Histomorphometric analysis of tartrate-resistant acid phosphatase (TRAP) stained distal femur sections was performed to investigate the commensal microbiota's effects on in vivo osteoclastogenesis ( Fig. 3a-g). SPF vs. GF mice had similar numbers of osteoclasts lining the trabecular bone perimeter (N.Oc/B.Pm) (Fig. 3b), which suggest the commensal microbiota does not alter the commitment of monocyte-myeloid cells to the osteoclast lineage. The average osteoclast cell size (Oc.Ar/Oc) was 2.5× larger in SPF mice (Fig. 3c), which resulted in a 2× greater osteoclast perimeter per bone perimeter (Oc.Pm/B.Pm) in SPF vs. GF mice (Fig. 3d). The substantially increased Oc.Ar/Oc (Fig. 3c) in SPF mice, implies the commensal microbiota enhances osteoclast maturation.
Eroded bone perimeter analysis ( Fig. 3e-g) was performed to assess alterations in osteoclast function. SPF vs. GF mice had an increased eroded perimeter per bone perimeter (E.Pm/B.Pm) (Fig. 3e), demonstrating that Relative quantification of mRNA was performed via the comparative C T method (ΔΔCT); Gapdh was utilized as an internal control gene; data expressed as fold difference relative to SPF. Data reported as mean ± SEM. *p < 0.05 vs. SPF; **p < 0.01 vs SPF; ***p < 0.001 vs SPF. the commensal microbiota upregulates bone resorption in health. Findings that the increased osteoclast-positive eroded perimeter per bone perimeter (Oc + E.Pm/B.Pm) (Fig. 3f), paralleled the 2× greater Oc.Pm/B.Pm (Fig. 3d) in SPF vs. GF mice, indicates the commensal microbiota's pro-resorptive actions are attributed to mechanisms enhancing osteoclast size/maturation (Fig. 3c).
Commensal microbiota modulates the RANKL/OPG Axis. Based on the in vivo histomorphometry findings revealing enhanced osteoclast size/maturation in SPF mice (Fig. 3c), the Tnfsf11(Rankl)/Tnfrsf11b(Opg) Axis was assessed ( Fig. 3h-j) to evaluate alterations in critical osteoclastic signaling factors 13,15,18 . RANKL, which signals at the RANK receptor on pre-osteoclast/osteoclast cells, is required for osteoclast differentiation/maturation. Due to OPG functioning as the RANK decoy receptor, it is imperative to assess the RANKL/OPG ratio when evaluating RANKL levels. Gene expression analysis revealed a trend towards a higher Tnfsf11:Tnfrsf11b ratio in marrow, and a significantly upregulated Tnfsf11:Tnfrsf11b ratio in calvaria of SPF vs. GF mice (Fig. 3j). The increased Tnfsf11:Tnfrsf11b ratio findings in SPF mice (Fig. 3j) were interestingly attributed to enhanced Tnfrsf11b expression (Fig. 3i), not alterations in Tnfsf11 expression (Fig. 3h). Recognizing that Tnfrsf11b is primarily expressed by stromal-osteoblastic cells in the bone environment, the findings that Tnfrsf11b upregulation is a trend in bone marrow and significant in calvaria (Fig. 3i) are in-line with calvaria vs. marrow having a more homogeneous stromal-osteoblastic cellular composition. Appreciating that an increased Tnfsf11:Tnfrsf11b ratio indicates higher levels of unbound RANKL available in the bone environment to activate RANK signaling, this notably implies that differences in in vivo RANKL signaling contribute to the superior osteoclast size/maturation phenotype in SPF vs. GF mice (Fig. 3c).

Commensal microbiota dynamically regulates osteoclast-precursor cell maturation.
Osteoclast-precursor (OCP) in vitro differentiation assays ( Fig. 4a-p) were utilized to further define the commensal microbiota's regulatory effects on osteoclast maturation, including the role of RANKL-signaling. Isolated marrow hematopoietic progenitor cells (HPCs) were labeled with CD11b MicroBeads, magnetic cell sorting was applied to separate CD11b neg HPCs, and cells were stimulated in culture (primed with CSF1) to enrich for CD11b neg osteoclast-precursor (OCP) cells having high osteoclastic potential [37][38][39] . CD11b neg OCP cultures were then subjected to stimulation with control (CSF1 alone) or treatment (CSF1 & RANKL) media for 3, 5 and 7 days 37-39 . Cultures stimulated for 3, 5 and 7 days were TRAP stained for cytomorphometric analysis of cellular differentiation endpoints ( Fig. 4; Supplementary Fig. S2), to evaluate cell level alterations in RANKL-induced osteoclast differentiation. Appreciating that day-3 of the culture system is when OCPs begin fusing to form multi-nucleated osteoclasts [37][38][39] , gene expression studies were carried out in day-3 OCP cultures ( Fig. 4e-h) to detect early transcription level alterations in RANKL-stimulated osteoclast differentiation.
Cytomorphometric analysis of day-3 OCP cultures showed no differences in SPF vs. GF OCP cellular endpoints ( Fig. 4a-d), which implies commensal microbiota immunomodulatory effects do not alter RANKL-induced early commitment of pre-osteoclastic cells to the osteoclast lineage. Day-3 OCP culture gene expression analysis revealed alterations in RANKL-stimulated osteoclastic genes, which unexpectedly indicated enhanced cell level differentiation potential outcomes in GF vs. SPF OCPs at later time points in the culture system. Nfatc1 (Fig. 4e), the master transcription factor for osteoclastogenesis, and Tnfrsf11a (Fig. 4f), the RANKL receptor and a surrogate marker for in vitro osteoclast maturation, were decreased in SPF vs. GF CD11b neg OCPs. Appreciating that signaling at both the CSF1R and RANK receptor are critical and necessary for osteoclast differentiation, differences in Csf1r (Fig. 4g) were ruled out to delineate that alterations in SPF vs. GF OCP outcomes are mediated by RANKL-stimulation.
Day-5 OCP culture cytomorphometric analysis elucidated cell level alterations in OCP differentiation potential ( Fig , which reflects terminal osteoclast differentiation potential in the OCP culture system, validated the observed superior in vivo osteoclast maturation phenotype noted in distal femur trabecular bone of SPF mice (Fig. 3c). The 2.5× larger osteoclast size (Oc.Ar/Oc) lining the trabecular bone in SPF mice (Fig. 3c), was remarkably paralleled by 2.5× larger Oc.Ar/Oc in SPF vs. GF OCP cultures (Fig. 4o). Number of nuclei per osteoclast (N.Nc/Oc), another indicator of osteoclast maturation, was also greater in day-7 SPF OCP cultures (Fig. 4p).
The study of an enriched OCP population (CD11b neg ) facilitated elucidating subtle alterations in the osteoclastogenic capacity of SPF vs. GF mice, which appears to be secondary to commensal microbiota immunoregulatory effects modulating RANKL-induced early osteoclast fusion genes. Dcstamp, a transmembrane protein critical for osteoclast fusion which is upregulated by RANKL-signaling 40 , was profoundly increased in day-3 GF OCP cultures (Fig. 4h). Appreciating that Dcstamp is essential for RANKL-induced osteoclastogenesis 40 , and considering Dcstamp was substantially increased in CD11b neg OCPs from GF vs. SPF mice (Fig. 4h), it appears the absence of the commensal microbiota in vivo immunostimulation sensitizes CD11b neg OCPs to RANKL-stimulated osteoclast differentiation/maturation. The realization that the enhanced osteoclast differentiation endpoints found in the GF vs. SPF OCP cultures at day-5 (Fig. 4k,l) were lost at day-7 (Fig. 4o,p), suggests that the sensitized RANKL-signaling in GF CD11b neg OCPs was transient.
The altered susceptibility of SPF vs. GF OCPs to RANKL stimulation in the OCP culture system provides novel insight into the commensal microbiota immunomodulatory impact on physiologic osteoclastogenesis. Osteoclasts are critical for normal skeletal turnover and repair processes, which is evidenced by anti-resorptive medications' deleterious effects (osteonecrosis of the jaw, atypical femoral fractures). Considering that RANKL is locally upregulated at osseous micro-damage 41 and early fracture sites 42 to transiently stimulate osteoclastic bone resorption necessary for normal osseous repair processes, based on the finding that RANKL-stimulation  11 week-old male SPF & GF mice were euthanized; bone marrow harvested; hematopoietic progenitor cells (HPCs) isolated. Magnetic cell sorting was applied to separate CD11b neg HPCs, which were then stimulated in culture (primed with CSF1) to enrich for CD11b neg osteoclast-precursor (OCP) cells having high osteoclastic potential. CD11b neg OCP cultures were then stimulated with control (CSF1 alone) or treatment (CSF1 & RANKL) media for 3, 5 and 7 days. Cytomorphometric cellular differentiation endpoints were analyzed in TRAP stained CD11b neg OCP cultures at day-3, day-5, and day-7 to evaluate cell level alterations in RANKL-induced osteoclast differentiation; TRAP + cell with > 3 nuclei considered an osteoclast. Gene expression studies were carried out in CD11b neg OCP cultures at day-3 to detect early transcription level alterations in RANKL-stimulated osteoclast differentiation. (a-d) Day-3 TRAP stain assay (n = 4/gp). (a) Representative images (200X) of CD11b neg OCP induced a more rapid onset and shorter duration pro-osteoclastogenic effects in GF vs. SPF OCPs (Fig. 4), the absence of the commensal microbiota may be advantageous for acute skeletal repair. Recognizing that increased chronic systemic inflammation has catabolic effects on systemic skeletal remodeling 12 and disrupts local osseous tissue repair 43 , the enhanced terminal osteoclast maturation phenotype shown in SPF mice ( Fig. 3c; Fig. 4m-p) implies that commensal microbiota pro-osteoclastic immunostimulatory effects are detrimental to skeletal tissue homeostasis in health.
Due to the proximity of the resident gut microbiota, upregulated cytokine expression was anticipated in the ileum of SPF mice. There was no difference in Tnf, Csf1, Ccl2 or Cxcl1 and surprisingly Il6 was elevated in the GF ileum (Fig. 5b). The increased Il6 expression in the ileum of GF vs. SPF mice (Fig. 5b) is consistent with a prior report in the colon of GF vs. Conv mice 21 , which implies that in absence of exogenous gut microbiota immuno-stimulation, Il6 is endogenously upregulated to support gut homeostasis/function.
Unexpectedly, Tnf, Il6, Csf1, Ccl2 and Cxcl1 were all elevated in the liver (Fig. 5c) of SPF vs. GF mice. These findings are in line with recent seminal reports demonstrating that commensal gut microbiota derived ligands translocate into the circulation under physiologic conditions to directly stimulate the liver innate-immune response in health 45,46 . Corroborating the increased innate-immune cytokine expression findings in liver, but not ileum of SPF mice (Fig. 5b,c), the frequency of CD11b + LY6G − F4/80 + LY6C hi (inflammatory monocyte) cells 47,48 was enhanced in the draining liver (celiac, portal) lymph nodes (LLNs), but not the mesenteric lymph nodes (MLNs) of SPF mice (Fig. 5e). The upregulated innate-immune cytokine expression in liver and inflammatory monocyte cell frequency in draining LLNs, but no differences in ileum and MLNs of SPF mice, led to the postulation that the commensal gut microbiota has osteoimmunomodulatory actions mediated through a previously unidentified Gut-Liver-Bone Axis (Fig. 5f).
As a means to substantiate the authors' novel Gut-Liver-Bone axis theory, the expression of pattern-recognition receptors (toll-like receptors, NOD-like receptors) and critical downstream signal transduction factors were evaluated in SPF vs. GF livers ( Supplementary Fig. S3). Toll-like receptor (Tlr)2, which primarily recognizes extracellular microbial cell wall components, and Tlr3, which detects viral nucleic acids, were upregulated in the liver of SPF vs. GF mice (Supplementary Fig. S3a). Findings that critical regulators of TLR2 mediated signal transduction (Myd88, Tirap(Mal), Irak4) were increased in SPF livers, while there was no difference in indispensable regulators of TLR3 mediated signal transduction (Ticam1(Trif)) ( Supplementary Fig. S3b), indicate that the commensal gut microbiota's pro-inflammatory actions in the liver are primarily mediated through TLR2 signaling. Notably, this investigation supports findings from prior reports demonstrating that normal gut microbiota derived ligands translocate into the circulation under physiologic conditions to directly stimulate the liver innate-immune response in health 45,46 . While pattern-recognition receptor signaling in the liver has been extensively investigated under pathophysiological states, the current report delineates that resident gut microbes substantially upregulate TLR2 signaling in the liver in health.
Congruent with the enhanced Tnf expression in marrow and liver (Fig. 5a,c), serum TNF levels were significantly elevated in SPF mice (Fig. 5g). Serum CSF1 was similar in SPF vs. GF mice (Fig. 5h), which is consistent with knowledge that hepatic/splenic macrophages tightly regulate circulating CSF1 levels via receptor-mediated endocytosis 49 .
T H 17, CD4 + IL17 + and CD4 + IFNγ + cells are increased in SPF marrow. Despite knowledge that the commensal gut microbiota directs T-lymphocyte mediated immunity, the commensal gut microbiota immuno-regulatory effects on marrow CD4 + /CD8 + T-cell hematopoiesis in the healthy adult skeleton is unclear. Recognizing the field of osteoimmunology has shown that specific bone marrow T-lymphocytic cells regulate osteoclastogenesis and bone remodeling in the adult skeleton [12][13][14][15]  ( Supplementary Fig. S4), transcription factor expression analysis (Fig. 6a-c) and intracellular cytokine expression analysis (Fig. 6d-g) elucidated differences in the frequency of specific helper CD4 + T-cell subsets and their characteristic cytokines.
While a recent experimental postmenopausal osteoporosis study in C57BL/6 SPF vs. GF mice demonstrated the commensal gut microbiota induces the upregulation of marrow TNF, IL17a, and IFNγ in sex steroid deprivation states 28 , the current study reveals the commensal gut microbiota enhances marrow T-lymphocyte expression of osteoclastogenic cytokines in health. The flow cytometry findings demonstrate the complex nature of lymphocytic cell interactions with bone cells in the marrow environment, providing novel insight into commensal microbiota in vivo immuno-stimulatory effects which could have altered the osteoclast maturation phenotype found in the in vitro OCP system (Fig. 4). Considering that IFNγ suppresses osteoclastogenesis through inhibition of RANKL-signaling via direct targeting of maturing osteoclast precursors 50, 51 , the suppressed sensitivity to RANKL-stimulation observed in day-5 SPF CD11b neg OCP cultures ( Fig. 4i-l) may be attributed to the enhanced frequency of marrow CD4 + IFNγ + cells in SPF vs. GF mice (Fig. 6g). Recognizing that IL17a is a pro-resorptive cytokine having potent synergistic effects on TNF pro-osteoclastic actions 14, 52 , the higher marrow Tnf (Fig. 5a)/ circulating TNF (Fig. 5g) levels and increased frequency of marrow CD4 + IL17a + cells (Fig. 6e,f) in SPF vs. GF mice may mediate the enhanced terminal osteoclast maturation potential exhibited in day-7 SPF OCP cultures (Fig. 4m-p).

Pro-inflammatory effector helper T-cells are increased in LLNs, but not MLNs of SPF mice.
Microbial-associated molecular patterns (MAMPs) derived from normal gut microbiota structural components (lipopolysaccharide, peptidoglycan, nucleic acids, etc.) are recognized by host cells via pattern-recognition receptors (toll-like receptors, NOD-like receptors, etc.). Central to the authors' postulated novel Gut-Bone-Liver Axis (Fig. 5f), normal gut microbiota derived (MAMP) byproducts circulating through liver sinusoidal vessels have been shown to induce innate-immune signaling at liver sinusoid endothelial cells and closely residing tissue resident Kupffer (liver tissue macrophage) and dendritic cells 45,46 . Considering tissue resident dendritic cells in gut  (Fig. 7) to further explore the commensal gut microbiota osteoimmunomodulatory effects mediated through hepato-gastrointestinal tissues.
Consistent with prior reports that lymphocytic cells are reduced in the gastrointestinal tissues of GF vs. SPF mice 1,53,54 , the frequency of overall CD3 + CD4 + CD8 − (helper) T-cells, CD3 + CD4 − CD8 + (cytotoxic) T-cells, and CD3 + CD4 − CD8 -TCRγδ + (gamma-delta) T-cells were decreased in the MLNs of GF mice (Fig. 7a-c). Transcription factor expression analysis of lymph node cells ( Fig. 7d-f) was supplemented with cell surface activation markers, due to experimental design not including intracellular cytokine expression analysis. In line with a previous investigation in SPF vs GF mice 54 , the frequency of CD4 + CD25 + FOXP3 + (T REG ) cells were upregulated in MLNs from GF mice (Fig. 7d), which may be attributed to the recognition of self-antigens or a hypersensitivity to food antigens in the absence of commensal gut microbiota immuno-stimulation. While there were no differences in the MLNs, the frequencies of CD4 + CD196 + RORγt + (T H 17) cells and CD4 + 183 + T-bet + (T H 1) cells were marginally (Fig. 7e) and significantly (Fig. 7f) increased in the LLNs of SPF vs. GF mice.
The lack of alterations in T H 1/T H 17 cells in SPF MLNs is consistent with current knowledge the MLNs function as the primary site for oral tolerance induction to commensal gut microbiota antigens 55,56 . In view of indirect evidence that supra-physiologic T H 1/T H 17 cell priming in MLNs drives pathophysiologic inflammation in gastrointestinal disease states 56,57 , the finding that SPF mice have similarly increased effector helper T-cell populations in LLNs suggests that commensal gut microbiota pro-inflammatory actions in health are mediated through the liver. The realization that resident gut microbes induce sustained alterations in the adaptive-immune response in draining LLNs, but not MLNs of SPF mice, supports the authors' postulate that the commensal gut microbiota has osteoimmunomodulatory actions mediated through a previously unidentified Gut-Liver-Bone Axis.
GF mice colonization with SPF mice gut microbiota normalizes osteoimmunomodulatory outcomes. A conventionalization study was executed to validate that the commensal gut microbiota mediates the osteoimmunomodulatory effects found in the SPF vs. GF mouse model (Figs 1-7). Conventionalized (ConvD) mice were generated by microbially associating 8 week-old GF C57BL/6 littermate mice with gut microbiota from age matched SPF C57BL/6 littermate mice (Fig. 8a). Four weeks following conventionalization there were no differences in trabecular bone (Fig. 8b-f), critical osteoblastic signaling factors (Fig. 8g,h), or critical osteoclastic signaling factors (Fig. 8i,j) in ConvD vs SPF mice. Pro-inflammatory innate-immune cytokines normalized in the marrow (Fig. 8k) and liver (Fig. 8m) of ConvD mice. Findings that Tnf and Cxcl1 were upregulated in the ileum of ConvD vs SPF mice (Fig. 8l) is in line with an acute innate immune response to the recently associated microbiota colonizing the proximal gut tissues, which is likely secondary to a transient increase in neutrophils expressing high levels of TNF and CXCL1.

Discussion
This osteoimmunology study in young adult mice introduces the commensal gut microbiota as a critical immunoregulator of osteoclast-osteoblast mediated bone remodeling in the healthy adult skeleton. As opposed to prior investigations of the normal gut microbiota's impact on physiological skeletal growth [19][20][21][22][23] and pathophysiological skeletal deterioration [28][29][30] , this report is the first to delineate the commensal gut microbiota's osteoimmunomodulatory effects on both osteoclastogenesis and osteoblastogenesis in the remodeling healthy adult skeleton.
Earlier investigations of the normal gut microbiota's influence on skeletal physiology in growing GF mouse models are essentially different than the current investigation of the normal gut microbiota's impact on bone remodeling in young adult GF mice. To properly appreciate our study outcomes, it is necessary to understand that bone modeling (growth) in the developing skeleton is a fundamentally different skeletal metabolic process than bone remodeling (turnover) in the adult skeleton. Sjogren et al. 21 reported the normal gut microbiota has inhibitory effects on bone mass accrual in the growing skeleton of C57BL/6 Conv vs. GF mice, which the authors attributed to increased osteoclast numbers secondary to immunostimulatory effects in the gut. In contrast, more recent work showing that the normal gut microbiota supports skeletal growth in BALB/c Conv vs. GF mice 22 , highlights the potential influence of mouse strain genetic background on gut microbiota osteoimmunoregulatory effects. Investigations in BALB/c Conv vs. GF mice 22 and CB6F1 (mixed BALB/c and C57BL/6 background) ConvD vs. GF mice 23 delineated that the normal gut microbiota enhances skeletal growth and bone formation through pro-anabolic actions upregulating liver and serum IGF1. While the current report and prior osteoimmunology investigations in the GF murine model focus on commensal gut microbiota induced host immune response effects, it should be emphasized that confounding study variables (e.g., maternal care effects, environmental factors, diet, dissimilar microbiota across SPF housing facilities, etc.) potentially impact outcomes reported within/ across studies.
The current investigation reveals that the normal gut microbiota has catabolic effects on skeletal homeostasis in the healthy adult skeleton, secondary to immunomodulatory actions suppressing osteoblastogenesis and enhancing osteoclastogenesis (Supplementary Fig. S6). SPF mice exhibited blunted bone formation in vivo, and reduced osteoblastic cell differentiation and mineralization in vitro, demonstrating the commensal microbiota has anti-anabolic effects on bone remodeling. Decreased Sp7(Osterix) in and Igf1 in bone tissues and IGF1 in serum from SPF mice, indicate the commensal microbiota's anti-osteoblastic actions are potentially mediated via suppression of local IGF1-signaling in skeletal tissues. Considering that prior investigations of mice on the BALB/c background have reported the gut microbiota enhances skeletal growth via upregulating liver derived IGF1 22,23 , current study findings demonstrating that resident gut microbes suppress skeletal tissue derived IGF1 in C57BL/6 mice highlights IGF1 as a potential genetic determinant modulating the gut microbiota's osteoimmunomodulatory effects. These findings are in line with prior investigations delineating that mouse strain background significantly impacts skeletal physiology 58,59 . Knowledge that BALB/c and C57BL/6 mice differ in polyreactive IgA abundance which impacts the generation of antigen-specific IgA and microbiota diversity 60 , provides indirect evidence that genetic influence on host -microbiota interactions may have implications for the gut microbiota impact on osteoimmunology.
While Sjogren et al. 21 reported the normal gut microbiota increases osteoclast numbers in growing Conv vs. GF mice, the current study has elucidated the commensal gut microbiota pro-catabolic effects on bone remodeling in the adult skeleton are secondary to immunomodulatory actions enhancing osteoclast size/maturation. Upregulated Tnfsf11:Tnfrsf11b (Rankl:Opg) ratio in vivo, and superior RANKL-stimulated terminal OCP maturation outcomes in vitro, in SPF vs. GF mice delineated that commensal gut microbiota pro-osteoclastic actions are related to sustained alterations in RANKL-signaling. Marrow Tnf, T H 17 cells, and CD4 + IL17a + T-cells were increased in SPF mice, implying the commensal microbiota's pro-osteoclastic actions are secondary to immunomodulatory effects directing marrow effector CD4 + T-cell hematopoiesis. Unexpected findings that SPF mice had upregulated pro-inflammatory innate-immune cytokines in liver, and increased pro-inflammatory innate-and adaptive-immune cells in LLNs, suggests the commensal gut microbiota's osteoimmunomodulatory actions are partly mediated by a previously unidentified Gut-Liver-Bone Axis (Supplementary Fig. S6).
Appreciating that all of the toll-like receptors (TLRs) are expressed in the liver, and that TLR2, TLR4, TLR5 and TLR9 recognize the majority of gut microbiota derived ligands in the portal blood 45,61 , future research is necessary to determine how signaling at specific TLRs in the liver regulates the normal gut microbiota's osteoimmunomodulatory effects. Recognizing that distinct commensal gut bacteria disproportionately modulate the host immune response 1-5 , ongoing research is indicated to discern whether specific gut commensals drive the upregulated T H 1/T H 17 phenotype found in the LLNs and marrow of SPF mice. Considering indirect evidence osteoblastic signaling factor. Relative quantification of mRNA was performed via the comparative C T method (ΔΔCT); Gapdh was utilized as an internal control gene; data expressed as fold difference relative to SPF. (h) Serum was isolated from whole blood (n = 4/gp); ELISA analysis of IGF1 levels. (i-j) In vivo regulation of osteoclastogenesis: qRT-PCR analysis in calvaria to assess alterations in the RANKL/OPG Axis. (i) Tnfsf11(Rankl) and Tnfrsf11b(Opg) mRNA levels. (j) Tnfsf11(Rankl):Tnfrsf11b(Opg) ratio. Relative quantification of mRNA was performed via the comparative C T method (ΔΔCT); Gapdh was utilized as an internal control gene; data expressed as fold difference relative to SPF. (k-n) In vivo pro-inflammatory cytokine expression: RNA was isolated from tissues and qRT-PCR analysis was performed in (k) bone marrow (n = 4/gp), (l) ileum (n = 4/ gp), (m) liver (n = 4/gp), (n) spleen (n = 4/gp) to assess Tnf, Il6, Csf1, Ccl2, Cxcl1 mRNA. Relative quantification of mRNA was performed via the comparative C T method (ΔΔCT); Gapdh was utilized as an internal control gene; data expressed as fold difference relative to SPF. Data are reported as mean ± SEM. *p < 0.05 vs. SPF; ***p < 0.001 vs. SPF. that probiotic administration protects against estrogen depletion induced bone loss in experimental murine osteoporosis models 28,62,63 , non-invasive interventions (dietary modulation, pre/probiotics administration) in the gut microbiome need to be researched as a means to optimize skeletal tissue remodeling and homeostasis in health.
Findings from this investigation of the impact of the commensal gut microbiota vs. studies assessing the impact of pathophysiologic gastrointestinal conditions on skeletal metabolism underscore the necessity of understanding health when discerning mechanisms causing disease. Mechanisms that have been extensively reported to drive pathophysiologic bone loss in inflammatory hepato-gastrointestinal disease states, including increased liver CXCL1, CCL2, CSF1, IL6 and TNF, upregulated circulating TNF, decreased circulating IGF1, increased marrow T H 17 cells/IL17a, unbalanced RANKL:OPG ratio, enhanced osteoclastogenesis, and blunted osteoblastogenesis 44,[64][65][66][67][68] , are astonishingly the same immunomodulatory mechanisms which appear to mediate the commensal gut microbiota's catabolic effects on skeletal homeostasis in health.
This research defining the commensal gut microbiota immunomodulatory effects on osteoclast-osteoblast mediated skeletal remodeling in health calls into question what is "normal" bone remodeling, a phenomena that is currently poorly understood. Based on the generally accepted empirical theory governing bone remodeling, which dictates that osteoclastic-osteoblastic mediated actions are "coupled" (balanced) in health and "uncoupled" (unbalanced) in disease [12][13][14]18 , findings reported here alarmingly imply the commensal gut microbiota induces a pathologic skeletal dys-homeostasis in health. The realization that the commensal gut microbiota has catabolic effects on endogenously programmed bone remodeling infers that bone remodeling is a not a simple "coupled" vs. "uncoupled" process, but rather a physiological gradient. The introduction of the commensal gut microbiota as a critical immunoregulator of normal osteoclast-osteoblast mediated bone remodeling processes in the healthy adult skeleton advances our understanding of skeletal physiology, having significant implications for the prevention of skeletal deterioration in health and disease.

Methods
Animals. Germ-free (GF) C57BL/6 mice were obtained from GF & Gnotobiotic Mouse Facilities at University of Michigan, and bred and maintained in sterile isolators at Medical University of South Carolina (MUSC) Gnotobiotic Animal Core. Specific-pathogen-free (SPF) C57BL/6 mice were purchased from Taconic, and maintained in ventilated cages in a SPF vivarium at MUSC. Mice were aged to 11-12 weeks for experiments. Animal procedures were approved by the MUSC Institutional Animal Care and Use Committee, and carried out in accordance with the approved guidelines.
Conventionalization Study. Conventionalized (ConvD) mice were generated by transferring four 8 week-old GF C57BL/6 littermate mice from the MUSC Gnotobiotic Animal Core to the MUSC SPF vivarium facility. ConvD mice microbial association was performed via fecal inoculum derived from pooled fresh feces of four age matched SPF C57BL/6 littermate mice. Fresh feces from SPF mice was homogenized in PBS, and a fecal inoculum suspension was prepared at a concentration of 0.1 g/ml. At age 8 weeks, and weekly intervals thereafter until age 11 weeks, ConvD mice were administered fresh fecal inoculum (derived from age matched SPF mice) via oral gavage at an approximate dose of 200 uL per mouse; residual fecal inoculum suspension was deposited on the ConvD mice fur. Dirty bedding was transferred from SPF mice caging to ConvD mice caging during weekly inoculation treatments. ConvD and SPF mice were housed in the same SPF vivarium; cages were maintained side by side on the same rack, and on the same shelf during the entire procedure. Further precautions were taken to minimize cage effects, including that animal handling and cage changing/maintenance was carried out by designated personnel trained in gnotobiotic animal husbandry. One week following the final inoculation treatment, ConvD and SPF littermate mice were sacrificed at age 12 weeks.
Histomorphometry. Femurs were fixed in 10% phosphate-buffered-formalin, dehydrated in graded EtOH and xylene, and embedded undecalcified in modified-methylmethacrylate 69 . Serial para-sagittal sections were cut through distal femur for trabecular bone analyses. 4 um tartrate-resistant acid phosphatase (TRAP) stained sections were used for quantifying osteoclast cellular endpoints and assessing eroded bone perimeter. 8 um sections were stained with toluidine blue for bone area analysis. 8 um unstained sections were used for dynamic indices of bone formation. 20 mg/kg calcein was administered via intraperitoneal injection 5 and 2 days prior to euthanasia 69 . Analyses were limited to the secondary spongiosa, beginning 250 µm proximal to the growth plate and extended 1000 µm proximally (50 µm from endocortical surfaces). Data were collected semi-automatically via an Olympus BX61 microscope and Visiopharm software. Data are reported in accordance with standardized nomenclature 70 .
Specimens were scanned with Scanco Medical µCT 40 Scanner, using the following acquisition parameters: X-ray tube potential = 55 kVp; X-ray intensity = 145 µA; Integration time = 200 ms; Isotropic voxel size = 6 µm 3 . Calibrated three-dimensional images were reconstructed. Tibia trabecular bone morphology was analyzed using Analyze 12.0 Bone Microarchitecture Analysis software (Analyze Direct). For trabecular analysis, transverse CT slices were analyzed beginning 250 µm distal to the proximal growth plate and extending 1200 µm distally; fixed threshold of 1750 Hounsfield Units was used to discriminate mineralized tissue. Femur length and cortical bone morphology were analyzed using Analyze 12.0 Bone Microarchitecture Analysis software (Analyze Direct). For cortical analysis, transverse CT slices were analyzed in a 1000 µm segment of the mid-diaphysis; fixed threshold of 2500 Hounsfield Units was used to discriminate mineralized tissue. Data are reported in accordance with standardized nomenclature 71 .
Statistical Analysis. Unpaired t tests were performed using GraphPad Prism 6.0. Data are presented as mean ± SEM; significance is p < 0.05.