High-Fat Diet–Induced Obesity Promotes Expansion of Bone Marrow Adipose Tissue and Impairs Skeletal Stem Cell Functions in Mice

Obesity represents a risk factor for development of insulin resistance and type 2 diabetes. In addition, it has been associated with increased adipocyte formation in the bone marrow (BM) along with increased risk for bone fragility fractures. However, little is known on the cellular mechanisms that link obesity, BM adiposity, and bone fragility. Thus, in an obesity intervention study in C57BL/6J mice fed with a high-fat diet (HFD) for 12 weeks, we investigated the molecular and cellular phenotype of bone marrow adipose tissue (BMAT), BM progenitor cells, and BM microenvironment in comparison to peripheral adipose tissue (AT). HFD decreased trabecular bone mass by 29%, cortical thickness by 5%, and increased BM adiposity by 184%. In contrast to peripheral AT, BMAT did not exhibit pro-in ﬂ ammatory phenotype. BM progenitor cells isolated from HFD mice exhibited decreased mRNA levels of in ﬂ ammatory genes ( Tnf a , IL1 b , Lcn2 ) and did not manifest an insulin resistant phenotype evidenced by normal levels of pAKT after insulin stimulation as well as normal levels of insulin signaling genes. In addition, BM progenitor cells manifested enhanced adipocyte differentiation in HFD condition. Thus, our data demonstrate that BMAT expansion in response to HFD exerts a deleterious effect on the skeleton. Continuous recruitment of progenitor cells to adipogenesis leads to progenitor cell exhaustion, decreased recruitment to osteoblastic cells, and decreased bone formation. In addition, the absence of insulin resistance and in ﬂ ammation in the BM suggest that BMAT buffers extra energy in the form of triglycerides and thus plays a role in whole-body energy homeostasis. © 2018 The Authors. Journal of Bone and Mineral Research Published by Wiley Periodicals, Inc. HFD. This results in osteoblast recruitment and decreased bone formation.


Introduction
O besity is a major health problem worldwide and represents a significant risk factor for insulin resistance, type 2 diabetes (T2D), and cardiovascular diseases. (1) Recent epidemiological studies have reported an association between obesity and increased risk of fragility fractures. (2) However, the biological mechanisms underlying this phenomenon are not known.
In addition to the expansion of peripheral adipose tissue (AT) in obesity, a similar expansion in bone marrow adipose tissue (BMAT) takes place. (3,4) BMAT is a newly recognized endocrineactive fat depot that is present in the non-hematopoietic marrow and in close contact with hematopoietic and bone tissues. (5,6) BMAT responds to metabolic needs of the organism as evidenced by significant expansion in a number of pathophysiological conditions, e.g., aging, diabetes, obesity, and anorexia nervosa or during pharmacological treatment with thiazolidinediones, glucocorticoids, or chemotherapy. (3,4,(7)(8)(9) BMAT originates from the bone marrow skeletal (also known as stromal or mesenchymal) stem cells (BM-MSC) present in the bone marrow (BM) stroma. BM-MSC give rise, in addition to adipocytes, to osteoblasts and hematopoietic-supporting stroma. (9) The differentiation capacity of BM-MSC is regulated via several secreted factors (e.g., sFRP-1, Dlk1/Pref-1), immune molecules (e.g., interleukins, TNFa, (10) microRNAs (11) ), and systemic hormones (e.g., parathyroid hormone (12) and insulin-like growth factors (13) ) present within the BM-MSC niche, which dictate lineage fate. (5,14,15) Moreover, BM-MSC lie in the vicinity of hematopoietic stem cells (HSC), whose microenvironment is affected in obesity and thus may participate in the regulation of BM-MSC differentiation and bone homeostasis. (16)(17)(18)(19) However, the possible relationship between increased BM adiposity, bone mass, and bone fragility are not completely determined.
Despite increased attention on the regulation of bone versus fat formation in the BM as a regulatory factor for bone formation, little is known regarding the metabolic function of BMAT and its contribution to whole-body energy metabolism. Although there is extensive literature on the role of visceral, subcutaneous, and brown adipose depots (VAT, SAT, BAT, respectively) in the development of insulin resista n c ea n dT 2 D ,f e ws t u d i e sh a v e investigated the function of BMAT as an adipose storage organ. (6,8,20,21) In obesity and T2D, peripheral AT develops insulin resistance and exhibits impaired insulin signaling. (22) However, there is no clear information about insulin sensitivity in BMAT. (23,24) In the current study, we hypothesized that obesity leads to metabolic changes in the BM microenvironment affecting BM-MSC biology, BMAT phenotype, and skeletal homeostasis. Using a mouse model of high-fat diet (HFD)-induced obesity, we investigated changes in bone mass, bone marrow fat mass, as well as the molecular phenotype and insulin sensitivity of BMAT and BM-MSC as compared with peripheral AT. We demonstrated that in response to HFD, there is an expansion in BMAT volume that is associated with decreased bone mass and impaired differentiation of BM-MSC.

Micro-computed tomography scanning (mCT)
Proximal tibias of mice fed for 12 weeks with HFD or with normal diet (ND) as control were scanned with a high-resolution mCT system (vivaCT40; Scanco Medical, Bassersdorf, Switzerland), resulting in three-dimensional (3D) reconstruction of cubic voxel sizes 10.5 Â 10.5 Â 10.5 mm 3 . A detailed description for the quantification of 3D microarchitecture of trabecular and cortical bone has been presented previously. (25) Isolation of bone marrow skeletal stem cells (BM-MSC) and hematopoietic stem cells (HSC) BM-MSC were isolated from the bones of front and hind limbs of C57BL/6J male mice (after 12 weeks of HFD) following the protocol of Zhu and colleagues and Houlihan and colleagues (26,27) with a slight modification. After bone crushing, a collagenase digestion (StemCell, Vancouver, Canada; 5% collagenase) and negative selection of CD45, CD31, and Ter119 cells was done to obtain BM-MSC. The positive fraction with hematopoietic stem cells (HSC) was also kept for further analysis. Cells were cultivated and expanded in vitro up to passage 3. The BM-MSC were pooled from 2 mice in each group and used for subsequent analyses.

Isolation of bone marrow adipose tissue (BMAT)
BMAT were isolated from long bones based on the protocol of Liu and colleagues and Lanske and colleagues by flushing the BM, quick high-speed spinning, and removing HSC. (28,29) Then the pellets containing the BMAT and HSC were resuspended in phosphate buffered saline (PBS) to let BMAT float on the top of the liquid suspension. HSC pellet was lysed with erythrocyte lysis buffer to remove red blood cells.

Isolation of adipose-derived stromal (mesenchymal) stem cells (AT-MSC)
Epididymal fat pads were mechanically dissociated using the gentleMACS Dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany) and collagenase digested at 37°C for 45 minutes in HBSS (Gibco, Life Technologies, Carlsbad, CA, USA) containing 2% BSA (American Bioanalytical, Natick, MA, USA) and 1% collagenase (Sigma-Aldrich, St. Louis, MO, USA). Samples were filtered through 300-mm-diameter pore nylon mesh and centrifuged. The adipocyte layer and the supernatant were aspirated and the pelleted cells were collected as the stromalvascular fraction (SVF) containing AT-MSC. The cells were then treated with red blood cell lysis buffer and washed with PBS and plated or directly harvested for further analysis. The cells were then plated at a density of 200,000 cells/well in 6-well plates.

Dynamic bone histomorphometry
Mice were injected with calcein 30 mg/kg (Fluka Chemie, Buchs, Switzerland) at 9 and 2 days, respectively, before euthanasia. OsteoMeasure analysis software (OsteoMetrics, Decatur, GA, USA) was used to evaluate mineral apposition rate (MAR) (in mm/d) in the frozen sections of tibia and vertebrae according to standard protocols. (30,31)

HFD-induced obesity decreased trabecular and cortical bone mass
To determine the effect of obesity on BMAT and bone metabolism, C57BL/6J male mice were fed with a high-fat diet (HFD) (60% kcal of fat) for 12 weeks. After 12 weeks, HFD mice exhibited increased body weight (29%), body fat mass (212%), and impaired glucose tolerance compared with normal chow diet mice (ND) (Supplemental Fig. S1A-D).
mCT of proximal tibia showed decreased trabecular bone volume per total volume (BV/TV) and trabecular number (Tb.N), and increased trabecular separation (Tb.Sep), whereas trabecular thickness (Tb.Th) was unchanged in HFD mice compared with ND mice (Fig. 1A). Similarly, cortical bone parameters BV/TV and thickness were reduced in HFD mice at diaphysis (  Dynamic bone histomorphometry performed on tibia and vertebrae showed a trend toward decreased mineral apposition rate (MAR) and a significant decrease in bone formation rate (BFR) in tibia, whereas both parameters were significantly reduced in the vertebrae of HFD mice ( Fig. 1D-F). Moreover, serum levels of bone formation marker procollagen type 1 Nterminal propeptide (P1NP) decreased, but bone resorption marker C-telopeptide of type I collagen (CTX-1) did not change in HFD mice (Fig. 1G). Similarly, histological analyses of proximal tibia sections using tartrate-resistant acid phosphatase (TRAP) staining (Fig. 1H) showed no changes in osteoclasts surface per total bone surface (Fig. 1I) or number of osteoclasts per total bone surface (data not shown), suggesting that the osteoclast activity did not change with HFD.
Bone marrow adipose tissue (BMAT) volume increased in HFD-induced obesity To evaluate BMAT volume, BMAT was visualized by osmiumtetraoxide staining in long bones (tibias) ex vivo after 12 weeks of HFD. Histological sections of osmium-stained bones demonstrated the strong affinity of the osmium for BMAT ( Fig. 2A). mCT imaging of osmium tetroxide-stained BMAT revealed increased BMAT volume in tibia, especially in the proximal metaphysis and in the distal diaphysis ( Fig. 2B-D). In ND mice, most of the BMAT was present in distal tibia (Fig. 2D). Further, histomorphometric analysis of hematoxylin-eosin-stained sections of the bones confirmed a significant increase of BM adiposity in proximal tibia in HFD mice (Fig. 2E, F). These changes in the BM were accompanied by increased BM adipocyte number and size (adipocyte area) (Fig. 2G, H). Perilipin staining, a specific marker of adipocytes, localized the area of adipocytes (Fig. 2I). In addition, in the whole group of mice, tibia BMAT volume was negatively correlated with BV/TV and Tb.N and positively correlated with Tb.Sep (Supplemental Fig. S3A-C).

HFD-induced obesity altered BM-MSC stemness and differentiation
To determine the cellular mechanisms of enhanced adipogenesis and decreased bone mass in obese mice, we first examined the cellular and molecular phenotype of BM-MSC. HFD reduced the percentage of CD73þ and Sca1/CD140aþ cells in BM-MSC ( Fig. 3A) assuming that the proportion of progenitor populations changed. The short-term proliferation rate and colony-forming units-fibroblast (CFU-f) of primary cultures were decreased in HFD mice (Fig. 3B, C). We did not find evidence for increased number of senescent cells stained positive for senescence-associated bgalactosidase, nor for increased levels of senescent gene markers (p16, p21) in BM-MSC cultures of HFD mice (Fig. 3D, E). Further, BM-MSC obtained from HFD mice exhibited increased adipocytic (AD) differentiation capacity compared with ND mice measured by mRNA gene expression of adipogenic genes, ie, Pparg2, Lep, Adipoq, Fsp27 (Fig. 3F). Osteoblast (OB) differentiation of BM-MSC obtained from HFD mice revealed increased alkaline phosphatase (ALP) activity (Fig. 3G). mRNA levels of specific osteoblastic genes (ie, Alp, Runx2, Ocn) did not show differences of BM-MSC in HFD mice compared with ND (Fig. 3H).

HFD-induced obesity did not lead to a pro-inflammatory BM microenvironment
To investigate the mechanism underlying the BM-MSC changes in lineage commitment, we examined changes in BM microenvironment. Because obesity is associated with increased inflammation in peripheral AT, we examined whether obesogenic condition in mice has a similar effect on BM microenvironment. Immunophenotyping showed an altered composition of immune cells in BM with decreased number of CD19þ B cells and increased number of CD4þ T lymphocytes as well as CD45þ hematopoietic cells in HFD mice (Fig. 4A). However, mRNA levels of inflammatory gene markers in HSC revealed decreased inflammatory signature in BM under HFD condition evidenced by decreased mRNA expression of Il1b, which was more manifested after 20 weeks of HFD by decreased mRNA levels of Il6 and Lipocalin 2 (Lcn2) in addition to Il1b (Supplemental Fig. S4A, B). To confirm the low pro-inflammatory phenotype, isolated CD45þ hematopoietic cells exposed to lipopolysaccharide (LPS) exhibited decreased inflammatory response measured by mRNA levels of Il1b and Mcp1 in HFD mice compared with ND mice (Fig. 4C). In addition, the response to LPS stimulation was similar in HFD compared with ND mice with a respect to activation of NF-kB pathway, one of the key inflammatory pathways, as measured by phospho IkBa and p65 activity (Fig. 4D).
HFD-induced obesity did not exhibit a pro-inflammatory phenotype of BMAT compared with peripheral adipose tissue (AT) Changes in BM microenvironment in response to HFD can be mediated by the accumulated BMAT. Thus, we characterized the molecular phenotype of BMAT isolated from long bones in response to HFD-induced obesity and compared to visceral adipose tissue (VAT) with known pro-inflammatory phenotype as well as brown adipose tissue (BAT). Under basal conditions of ND, the gene expression profile of BMAT showed higher mRNA levels of pro-inflammatory genes (Tnfa, Il1b, and Lcn2) compared with peripheral adipose depot: VAT and BAT (Fig. 5A). In contrast, genes related to insulin signaling (Insr, Irs1, Igf1) and adipogenesis (Pparg2, Ap2, Adipoq, Lep) were highly expressed in VAT and BAT compared with BMAT (Fig. 5B, C). Molecular signature of BMAT was different from BAT except for Dio2, which showed similar mRNA levels in BMAT and BAT (Fig. 5D).

HFD-induced obesity did not impair insulin sensitivity in BM-MSC compared with AT-MSC
Insulin resistance accompanies HFD-induced obesity and may play a role in mediating changes in BM-MSC niche. To test this hypothesis, we examined insulin signaling and responsiveness in BM-MSC obtained from HFD mice and compared with AT-MSC derived from VAT previously shown to exhibit insulin resistance and inflammation in HFD-induced obesity. (32) In HFD mice, VAT AT-MSC exhibited impaired adipocytic (AD) differentiation (Supplemental Fig. S5A, B) and under basal conditions decreased gene expression of Pparg2 and increased mRNA levels of Tnfa, a marker of inflammatory status in AT. On the other hand, BM-MSC exhibited increased gene expression of Pparg2 and decreased gene expression of Tnfa (Fig. 6A, B), suggesting the opposite metabolic phenotype. In addition, under HFD insulin-stimulated activation of pAKT(S473)/ totalAKT was decreased in undifferentiated and AD differentiated progeny in VAT AT-MSC, whereas BM-MSC maintained insulin responsiveness under basal and following AD differentiation (Fig. 6C-F).

Discussion
In the present study, we report that HFD-induced obesity in mice leads to increased BMAT formation, enhanced AD differentiation of BM-MSC, and reduced bone mass. Interestingly, BMAT, in contrast to extramedullary adipocytes, did not exhibit a pro-inflammatory phenotype and maintained its insulin responsiveness. Our study suggests that BMAT responds to metabolic stimuli and acts as a depot for storing extra energy during HFD when peripheral AT is incapable of mediating this function.

HFD-induced obesity led to decreased bone mass and BMAT expansion
Similar to other investigators studying BMAT in mice and humans, we observed that HFD increased BMAT formation as evidenced by increased BMAT volume, adipocyte number, and size. (3,4,(33)(34)(35)(36)(37) In addition, we observed significant and consistent decrease in trabecular and cortical bone mass after 12 weeks of HFD. Previous studies have reported inconsistent results regarding the impact of HFD on bone mass and bone quality. Some studies reported no change or increased or decreased bone mass with lower bone strength. (33)(34)(35)(36)38,39) These discrepancies can be explained by variations in experimental models, e.g., mouse strain, sex, length/type/ composition of diet (60% versus 45% fat or 10% corn oil), and possibly evaluation methods. However, the consistency observed in our study and some of the previous studies, (35,37,40,41) as well as the recent study of Bornstein and colleagues reported HFD induced reduction in the cortical and trabecular parameters in the long bones along with increased BMAT volume, (37) suggesting that bone mass reduction is a biological consequence of coherent mechanisms of cellular and molecular changes relevant for studies of bone fragility in human obesity. (42)(43)(44) HFD led to a progenitor cell "exhaustion" and enhanced adipocyte differentiation BMAT originate from progenitor cell population within the BM stroma. (45,46) We observed that HFD-induced obesity increased BMAT volume and is associated with decreased proliferative potential of progenitor cells and enhanced differentiation of BM-MSC into adipocytes. Based on a number of pathophysiological studies, an inverse relationship between BMAT and bone formation has been reported, (3,4,47,48) suggesting changes in BM-MSC commitment to adipocyte versus osteoblast lineage. (49) Our data support this hypothesis under HFD-induced obesity, as we observed enhanced adipocyte differentiation in BM-MSC obtained from HFD mice. Moreover, decreased number of Sca1þ cells in HFD suggests that the expansion of adipocytic progenitors (Sca1-) might occur as previously reported. (50) The decrease in BM-MSC and increased number of committed adipocytes may be caused by several mechanisms including changes in the BM-MSC niche due to local and systemic changes (51) or high-triglyceride load in the circulation increasing lipid uptake by skeleton. (52) HFD did not cause an "inflamed" BMAT We examined for a possible mechanism that links obesity, BMAT expansion, and decreased bone mass. Obesity is associated with increased inflammation in peripheral AT. (1,22) Our results demonstrated that after HFD, BMAT does not develop a "proinflammatory" phenotype similar to that observed in peripheral AT with impaired adipogenesis and insulin signaling along with increased inflammation. (22) In contrast to previous studies (20) that examined BMAT as part of the heterogeneous bone tissue, we isolated BMAT directly from long bones, which enhanced the internal validity of our study and point out that decreased bone mass is not the result of "inflamed" BMAT. HFD did not lead to a "pro-inflammatory" BM microenvironment BMAT and its progenitors exist within a close proximity to hematopoietic marrow. Previous studies reported obesityinduced impairment of hematopoiesis, including depletion of B lymphocytes and changes in HSC niche. (16,53,54) We observed decreased B lymphopoiesis and increased CD4þ T lymphocyte and CD45þ/CD34-HSC populations in HFD mice. These changes may be attributed to changes in adipokine secretion, having inhibitory effects on HSC proliferation and differentiation. (17,19,50) In vitro studies showed that adipocyte-derived factors such as adiponectin could impair B lymphopoiesis and increase myelopoiesis. (55,56) Indeed, in our study, we observed increased levels of adiponectin in BMAT along with decreased Blymphocyte population. In addition, LPS responsiveness in HSC exposed to the HFD environment was impaired, suggesting the presence of a "protective" mechanism against BM inflammation during the initial phases of BMAT expansion in obesity. Future studies need to determine the anti-inflammatory mechanisms and its relevance to the BMAT biology.
BMAT maintains an insulin-responsive phenotype in HFD-induced obesity The physiology of BMAT and its contribution to whole-body metabolism is still not well understood. In the present study, the molecular and cellular data on BMAT phenotype with increased expression of insulin-responsive genes raised a question whether BMAT insulin signaling is affected by obesity similar to peripheral AT. Recent animal studies revealed that insulinregulated glucose uptake in the skeleton measured by PET scanning was decreased under HFD, suggesting the presence of insulin resistance. (23,24,57) However, in these studies, they did not distinguish BMAT from the rest of the BM. Our findings in BMSC and BMAT point out that BMAT is a "metabolically responsive" tissue in HFD, and contrary to peripheral extramedullary AT, it may respond to increased metabolic demand coming from the BM microenvironment and serves as a buffering organ for excessive energy. However, longer exposure to HFD might switch the BMAT metabolic phenotype with a worse impact on bone. This needs to be addressed in future studies.
There are some limitations in our study. First, there may be a risk that the isolated BMAT preparations were contaminated with hematopoietic cells, which affects the molecular analysis of BMAT phenotype. However, we observed no difference in the number of CD45þ cells in BMAT preparations obtained from ND and HFD mice, suggesting the absence of significant hematopoietic cell contamination. The presence of hematopoietic cells would have increased the levels of pro-inflammatory cytokines, whereas we observed a decrease in their levels. Second, we employed collagenase digestion in cell isolation procedures, which may activate several inflammatory pathways. However, we employed collagenase isolation in both BM-MSC and AT-MSC, and the cells were cultured under identical conditions for several days, minimizing the confounding effects of collagenase on inflammation-associated gene expression. Third, in our study, we employed male mice because they are more prone to obesity-induced metabolic complications. (58,59) However, studies of female mice need to be performed to generalize our findings.
HFD-induced obesity in mice led to decreased bone mass, which is different from the observed increase in bone mass in obese humans, limiting the translational relevance of our study. However, BMAT expansion is a consistent finding in HFD mice Fig. 7. A working model for the effects of HFD-induced obesity on BMAT compared with peripheral AT. In HFD condition and compared with peripheral adipose tissue (AT), bone marrow microenvironment exhibits decreased inflammatory response of hematopoietic cells (HC) and adipocytic progenitors as well as maintenance of insulin signaling leading to enhanced adipogenesis. Peripheral AT exhibits higher levels of inflammation and impaired insulin signaling. The BMAT expansion can thus serve to buffer excess energy from HFD. This results in decreased osteoblast recruitment and consequently decreased bone formation. and in obese humans. (4,60) Thus, our study is relevant to the understanding the cellular, molecular, and metabolic consequences of increased bone marrow adipogenesis and provides insight into the possible mechanisms of bone fragility observed in obese humans.
In summary (Fig. 7), the present study proposes a model for the relationship between increased BMAT and decreased bone mass after HFD feeding. The absence of inflammation and continuous recruitment of pre-adipocytic cells from BM-MSC to buffer the excess calories leads to stem cell exhaustion and impaired osteoblast recruitment and bone formation. Importantly, our findings demonstrate the presence of a metabolically healthy BMAT capable of responding to metabolic demands in a low-inflammatory BM microenvironment, which is opposite to what is observed in peripheral AT, suggesting an unexpected role of BMAT in obesity and providing possible strategies for a treatment of obesity-induced bone fragility.

Disclosures
All authors state that they have no conflicts of interest.