Novel thiazolidinedione analog reduces a negative impact on bone and mesenchymal stem cell properties in obese mice compared to classical thiazolidinediones

Objective The use of thiazolidinediones (TZDs) as insulin sensitizers has been shown to have side effects including increased accumulation of bone marrow adipocytes (BMAds) associated with a higher fracture risk and bone loss. A novel TZD analog MSDC-0602K with low affinity to PPARγ has been developed to reduce adverse effects of TZD therapy. However, the effect of MSDC-0602K on bone phenotype and bone marrow mesenchymal stem cells (BM-MSCs) in relation to obesity has not been intensively studied yet. Methods Here, we investigated whether 8-week treatment with MSDC-0602K has a less detrimental effect on bone loss and BM-MSC properties in obese mice in comparison to first generation of TZDs, pioglitazone. Bone parameters (bone microstructure, bone marrow adiposity, bone strength) were examined by μCT and 3-point bending test. Primary BM-MSCs were isolated and measured for osteoblast and adipocyte differentiation. Cellular senescence, bioenergetic profiling, nutrient consumption and insulin signaling were also determined. Results The findings demonstrate that MSDC-0602K improved bone parameters along with increased proportion of smaller BMAds in tibia of obese mice when compared to pioglitazone. Further, primary BM-MSCs isolated from treated mice and human BM-MSCs revealed decreased adipocyte and higher osteoblast differentiation accompanied with less inflammatory and senescent phenotype induced by MSDC-0602K vs. pioglitazone. These changes were further reflected by increased glycolytic activity differently affecting glutamine and glucose cellular metabolism in MSDC-0602K-treated cells compared to pioglitazone, associated with higher osteogenesis. Conclusion Our study provides novel insights into the action of MSDC-0602K in obese mice, characterized by the absence of detrimental effects on bone quality and BM-MSC metabolism when compared to classical TZDs and thus suggesting a potential therapeutical use of MSDC-0602K in both metabolic and bone diseases.


INTRODUCTION
Obesity is accompanied with ectopic fat accumulation in non-adipose organs, including bones, leading to increased bone marrow (BM) adiposity, which is associated with an increased risk of bone fractures and osteoporosis in obese and diabetic patients [1]. Several studies including our recent observations, show that obesity changes the BM microenvironment and impacts BM mesenchymal stem cell (BM-MSC) properties by increased adipogenesis, which contributes to bone fragility induced by metabolic complications [2e6].
In obesity and type 2 diabetes several approaches are used to treat or prevent the detrimental effects of metabolic diseases, including physical activity, dietary or pharmacological treatment. The use of thiazolidinediones (TZDs) has been for a long time considered as an appropriate treatment for metabolic complications by improving insulin sensitivity in patients with type 2 diabetes. However, the adverse effects induced by first generation of TZDs in terms of weight gain, cardiovascular complications and bone loss have reduced their use in the clinical practice [7]. The negative effect of TZDs on bone physiology is associated with increased BM adiposity leading to bone loss [7]. The use of these insulin-sensitizing drugs usually has a secondary detrimental effect on bone physiology because of the activation of peroxisome proliferator-activated receptor-g (PPARg) regulating adipocyte (AD) differentiation both in the periphery and in the BM, where they also induce osteoclast activation [5,8]. Therefore, further studies are needed to dissect the molecular effects of this type of antidiabetic treatments on bone and fat physiology. A previous study by Stechschulte et al. [9] demonstrated that changes in post-translational modification of PPARg can reduce the negative effect on bone metabolism while maintaining the positive effect on energy metabolism. Along these lines, a novel "PPARg-sparing" TZD analog MSDC-0602K ("second generation of TZDs") was developed to minimize the adverse side effects of TZDs on bone and fat metabolism [10e14]. MSDC-0602K was purposely designed to decrease direct binding to PPARg [11,15], but it maintains its inhibitory effect on mitochondrial pyruvate carrier (MPC), which likely contributes to the beneficial effect on energy metabolism and glucose uptake [11,12,16]. Testing of MSDC-0602K drug in a clinical trial (phase 2b) showed promising results in terms of lowering glucose and insulin levels and improving liver steatosis without side effects in obese subjects [17]. Fukunaga et al. [13] recently reported that MSDC-0602K showed no side effects in lean mice with respect to bone parameters compared to rosiglitazone. However, the impact of this novel TZD analog on bone metabolism in obese conditions has not been studied yet. Thus, the main objective of this study was to determine whether MSDC-0602K has a less detrimental effect on bone structure and molecular properties of BM-MSCs compared to a typical TZD pioglitazone, using a model of diet-induced obesity in male C57BL/6N mice.

MATERIALS AND METHODS
Additional methods are described in the Supplementary Material.

Animals and dietary interventions
Male C57BL/6N mice (Charles River Laboratories, Sulzfeld, Germany) were maintained at 22 C with 12-hour lightedark cycle (light from 6:00 a.m.). Upon arrival and before the start of the experiment all mice had free access to water and standard chow diet (ND, Ssniff Spezialdieten GmbH, Soest, Germany). At the age of 12 weeks, mice were randomly divided into 5 groups (n ¼ 8e10, repeated in 3 independent experiments) and fed for 8 weeks with: i) normal diet (ND) (3.4% wt/wt as lipids; Rat/Mouse-Maintenance extrudate; Ssniff Spezialdieten GmbH, Soest, Germany); ii) high-fat diet (HFD, lipid content, w35% wt/wt, mainly corn oil [18]; iii) HFD þ P, HFD supplemented with 50 mg pioglitazone/kg diet (Actos, Takeda, Japan); iv) HFD þ M, HFD supplemented with 330 mg MSDC-0602K/kg diet (Cirius Therapeutics, USA) [10]. Diets were stored at À20 C, in sealed plastic bags filled with nitrogen. The dose of pioglitazone and MSDC-0602K was the same as in the previous studies using these compounds in mice [10,18]. Body weight and 24-hour food consumption were measured every week. Mice were sacrificed after 8 weeks of dietary interventions in the fedstate by cervical dislocation under diethyl ether anesthesia. Tissue samples and primary isolated mBM-MSCs were used for subsequent molecular analyses. All experiments were performed according to the guideline of the Institute of Physiology of the Czech Academy of Sciences and were approved under protocol number 81/2016.

Glucose tolerance test
Intraperitoneal glucose tolerance test (GTT) was performed using 1 mg of glucose/g body weight in overnight fasted mice as previously published [10].

Biochemical analysis of plasma
Blood glucose levels were measured by OneTouch Ultra glucometers (LifeScan, Milpitas, CA, USA), and plasma insulin levels were determined by the Sensitive Rat Insulin RIA Kit (Millipore, Billerica, MA, USA).

Isolation of mBM-MSCs and mHSCs
mBM-MSCs were isolated from the bones of front and hind limbs of C57BL/6N male mice (after 8 weeks of dietary treatments) following previous protocols with the small modifications [3]. After bone crushing, collagenase digestion (StemCell, Vancouver, BC, Canada) and negative selection of CD45, CD31 and Ter119 cells (Miltenyi, Bergisch Gladbach, Germany) mBM-MSCs were obtained. mBM-MSCs were subcultured in growth medium (MEM alpha (Thermo Fisher Scientific, Waltham, MA, USA) þ 20% FBS (Thermo Fisher Scientific, Waltham, MA, USA) þ 1% penicillin/streptomycin (P/S) (Thermo Fisher Scientific, Waltham, MA, USA) þ 0.5% Amphotericin B (Merck, Darmstadt, Germany) þ 1% Glutamax (Thermo Fisher Scientific, Waltham, MA, USA) þ 1% MEM NEAA (Thermo Fisher Scientific, Waltham, MA, USA) þ 1% sodium pyruvate (Thermo Fisher Scientific, Waltham, MA, USA)) and cultivated for further analysis. The positive fraction with mHSCs was divided into three samples. One was  [25]. However, if a researcher is only interested in adipocytes and bone, a more efficient CESA could be used. In this study, Hexabrix, a non-toxic staining agent, was used to visualize BMAT in bones. Since bone is strongly attenuating in nature, we opted to use a CESA that was unable to interact with adipocytes, but able to generally interact with other BM constituents. This would increase the contrast differences between bone, BM and adipocytes and consequently facilitate the desired BMAT analyses. Hexabrix is well-known in the field of CECT (in vivo and ex vivo), and has extensively, but not exclusively, been used in studies concerning the visualization and quantification of sulfated glycosaminoglycans (sGAGs) in cartilage tissue. Both sGAGs and Hexabrix are negatively charged at physiological pH (7.4) and will consequently repel each other [26e28]. Since adipocytes contain many types of lipid molecules (free fatty acids, TG etc.) that create a hydrophobic environment containing biomolecules bearing negatively charged functional groups (phosphate and/or carboxylate groups), we expected a similar repelling effect of this CESA towards adipocytes, rendering adipocytes darker compared to the surrounding BM in the CECT images. This combined with a computed LogD (a metric for the distribution of a molecule between a hydrophobic and a hydrophilic phase at a certain pH) at pH 7.4 of À1.52 indicates that this molecule will unlikely accumulate in hydrophobic areas (e.g. BMAds). These Hexabrix-stained CECT images of BMAT were confirmed by H&Estained sections from the same bone area and thus validating this compound to visualize BMAT.
The staining solution was prepared by mixing commercial HexabrixÒ solution (Guerbet, 10 mL, 320 mgI/mL) with a 1x PBS (phosphatebuffered saline, 40 mL, 10 mM). Formalin-fixed proximal tibias (right leg) of the mice were transferred to a 1.5 mL Eppendorf tube containing 1 mL of staining solution. These Eppendorf tubes were placed on a shaker plate (gentle shaking) at ambient temperature. The samples were stained for at least 3 days, after which they were scanned.

mCT image acquisition and reconstruction
For image acquisition, the samples were removed from the Eppendorf tube and wrapped in parafilm to prevent dehydration while exposed to X-rays. Samples were imaged using a Phoenix NanoTom M (GE Measurement and Control Solutions, Boston, MA, USA). A diamondcoated tungsten target was used. The system was operated with the following acquisition parameters: voltage ¼ 60e70 kV, MicroCT, Billerica, MA, USA), we initiated the analysis of the mCT data using CTAn (Bruker MicroCT, Billerica, MA, USA). First, we selected the volume of interest (VOI) in the proximal metaphysis of the tibia starting 250 slices below the growth plate and covering 500 slices in the distal direction. In this VOI, binarization of the dataset was performed using a global threshold (130e255). This threshold masked both cortical and trabecular bone. Based on this selection, a denoised mask for the bone marrow combined with the trabecular bone was segmented by performing a sequence of VOI shrink-wrap, closing (2e10; increments of 2) and opening (2e10; increments of 2) operations (i.e. everything inside the cortical bone was selected). The newly generated mask was projected on the grey value image, generating a new VOI. In this new VOI, the segmentation of adipocytes was performed using a global threshold (1e59). For the analysis of individual adipocytes, we used the Avizo 3D (version 2021.1, Thermo Fisher Scientific, Waltham, MA, USA) software. First, the adipocytes were binarized and leftover noise was removed using an interactive thresholding (1e255) and despeckle (speckle size ¼ 7 mm Â 7 mm x 7 mm) module. This was followed by the segmentation of individual adipocytes using a combination of thickness map computation and the H-extrema watershed module. Next, a border kill module was applied, which removed adipocytes that were cut by the bounding box. Then, a 3D label analysis was performed that allowed a final filtering of the data based on shape (sphericity >0.5) and volume (>4000 mm 3 ). Finally, sphericity, volume (mm 3 ), area (mm 2 ), thickness (mm) and number of adipocytes were computed.

Bone strength analyses
The femora isolated from C57BL/6N male mice after 8-week-long treatment with HFD or HFD supplemented with pioglitazone and MSDC-0602K were tested in a three-point bending test using an ElectroForce testing system (TestBench LM1, EnduraTEC Systems Group, Bose Corp., Minnetonka, MN, USA). A standard protocol as Original Article described in previous work was used in this experiment [29,30]. Span length and radius of curvature of the supports were 7 mm and 2 mm, respectively. In the period between dissection and mechanical testing, the bones were fixed in 4% paraformaldehyde at 4 C for the first 48 h after which they were stored in PBS at 4 C. The bones were placed with the anterior surface pointing downward and were subjected to a small stabilizing preload (1 N) and two conditioning cycles before loading until failure at a rate of 0.1 mm/s. The following parameters were derived from the loadedisplacement curve: 1) bone strength (N), determined as the ultimate load during the three-point bending test; 2) work-to-failure (mJ), determined as the area under the loade displacement curve, representing the energy absorbed by the bone before breaking and 3) bone stiffness (N/mm), calculated as the slope of the linear proportion of the loaded-displacement curve, representing the elastic rigidity. 2.9. Statistical analysis All data are representative of at least two independent experiments of similar results performed in triplicates unless otherwise indicated. The statistical significance of the differences in the means of experimental groups was determined by unpaired t-test or ANOVA and Bonferroni or Tukey post hoc tests using GraphPad Prism 5.0a software. The data are presented as means AE SEM. p value < 0.05 was considered statistically significant.

MSDC-0602K is less detrimental than pioglitazone on bone parameters in obese mice
To determine the effect of novel TZD analog, MSDC-0602K on metabolic and bone parameters in high-fat diet (HFD)-induced obesity, C57BL/6N male mice were fed for 8 weeks with HFD or HFD supplemented with first generation of TZDs, pioglitazone (HFD þ P) or MSDC-0602K (HFD þ M). As shown by Bardova et al. [10] and the present study, MSDC-0602K administered to HFD mice had a similar positive impact as pioglitazone on metabolic parameters, including reduction in fasting glycemia and insulinemia and improvement of glucose tolerance (Figure. S1AeD), as well as decreased adipose tissue (AT) inflammation [10], but with a more pronounced effect with respect to reduction of weight gain (Figure. S1EeF). This may indicate less severe side effects on AT function when compared to the first generation of TZDs.
As an extension of the above study with a focus on MSDC-0602Kinduced impact on bone physiology, mCT analyses of proximal tibia and L5 vertebrae were performed in treated mice. While we did not observe significant changes in bone microstructure of proximal tibia (data not shown), there were more pronounced changes in L5 vertebrae ( Figure 1AeE). HFD induced increased cortical porosity (Ct.Po) in L5, which was decreased in mice treated with HFD þ M mice ( Figure 1A). On the other hand, cortical thickness (Ct.Th) in L5 was decreased in HFD mice with a trend to increase in HFD þ M group ( Figure 1B). Further, trabecular number (Tb.N) in L5 was decreased in HFD þ P compared to HFD mice, while HFD þ M group showed a less detrimental effect on this parameter ( Figure 1C). Other trabecular parameters were not changed (Figure. S1GeI). These changes are demonstrated in the representative 3D images of L5 trabecular and cortical bone in treated mice ( Figure 1DeE). Further, ratio of circulating levels of bone formation marker procollagen type 1 N-terminal propeptide (P1NP) and bone resorption marker tartrate-resistant acid phosphatase (TRAP) (P1NP/TRAP) showed higher bone formation rate in HFD þ M group compared to normal diet (ND) ( Figure 1F). Bone strength measured by three-point bending test revealed stronger femora in HFD þ M mice compared to HFD and HFD þ P group ( Figure 1G). Together these data showed that 8-week-long preventive treatment of HFD mice with MSDC-0602K reduced a negative impact on bone parameters with increased resistance to mechanical stress than typical TZD, pioglitazone.

BMAT composition is differently affected in obese mice treated with pioglitazone and MSDC-0602K
To further evaluate the impact of different diets on BM adiposity, the BMAT volume was analyzed using contrast-enhanced X-ray microfocus computed tomography (CECT). While mCT is unable to distinguish between distinct soft tissues due to their intrinsic low X-ray attenuation, CECT can do so by using X-ray opaque contrast-enhancing staining agents (CESAs). These CESAs enrich the X-ray attenuating atom content of the soft tissues in a specific manner. In this study, we selected Hexabrix, an anionic, iodinated CESA, which shows weaker interactions with non-adipocyte tissue constituents and is still excluded by the adipocytes. Moreover, it is a smaller molecule compared to the POM-based CESA, and thus it stains the whole tissue sample faster [26e28]. Image analysis of Hexabrix-stained BMAds confirmed an increased BM adiposity in HFD mice as in previous studies using OsO 4 [3,5]. A similar increase in BM adiposity was observed in the HFD þ P and HFD þ M  Original Article 8 groups, as depicted on the 3D representative images of BMAds in proximal tibia and BMAT volume evaluation (Figure 2AeB). The validation of the Hexabrix-stained BMAds was confirmed by the registering a cross-section of the CECT dataset with the corresponding hematoxylin-eosin (H&E)-stained section of the same tibia ( Figure S2A). Interestingly, further analyses of BMAT revealed higher number of BMAds in HFD þ P and HFD þ M groups compared to HFD, but with a larger proportion of smaller adipocytes in HFD þ M group compared to HFD þ P (Figure 2CeE). In addition, histomorphometric analysis of H&E-stained sections of proximal tibia confirmed the data from CECT, demonstrating a sustained increase in BMAds in the HFD, HFD þ P and HFD þ M groups compared to ND ( Figure S2B), while the size of BMAds was lower in HFD þ M mice (Figure 2FeG). These data highlight that MSDC-0602K treatment in HFD mice increased the proportion of smaller BMAds in tibia compared to HFD þ P.
3.3. MSDC-0602K improves lipidome and metabolome of circulating plasma in more extend than BM composition in comparison to pioglitazone in obese mice To further examine changes in systemic and BM microenvironment in treatment with different TZDs, a global lipidomic and metabolomic analysis was performed in circulating plasma and BM samples of treated mice using LC-MS. Overall, 904 polar metabolites and simple and complex lipids were annotated in these two matrices with an overlap of 85% ( Figure 3A). Partial least-squares discriminant analysis (PLS-DA) showed distinct separation of HFD þ P and HFD þ M groups from HFD in plasma, while less separation was observed for all three groups in case of BM ( Figure S3A-B). Only a few metabolites were unique or detected at a much higher intensity in plasma compared to BM (e.g., cholesteryl esters, bile acids), while BM contained more lipid classes and polar metabolites (e.g. lysophosphatidylglycerol (LPG), lysophosphatidylinositol (LPI), phosphatidylglycerol (PG), phosphatidylserine (PS), dipeptides, nucleotides), which usually maintain low concentrations in circulating plasma [32] ( Figure 3B, Table S4) suggesting that these uniquely present structural components and nutrients in BM may contribute to the fate and properties of BM-MSCs. These plasma/BM ratios were not affected by the treatment with TZD drugs apart from triacylglycerol (TG) which were reduced in HFD þ P and HFD þ M compared to HFD group. Next, we examined the lipidomic data from the perspective of the sum of abundances of all lipid species for each lipid class. In plasma, 13 out of 32 lipid classes were at higher concentration in HFD compared to HFD þ P and HFD þ M groups ( Figure 3C). Specifically, the sum of ether-linked TG was detected at 3-fold and 6.5-fold higher concentrations in HFD compared to HFD þ P and HFD þ M, respectively, followed by other lipid classes. On the other hand, only 2 lipid classes (TG, monoacylglycerol (MG)) were statistically altered in BM. In the case of TG, their intensity was 1.5-fold higher in HFD þ M compared to HFD and HFD þ P ( Figure 3D). Since the use of the sum of abundances may hide useful information for data interpretation, we also examined the detailed abundance patterns of all annotated lipid species along with polar metabolites. As Figure 3C shows, the top-50 metabolites in plasma belonged to lipids in line with previous data analysis with exception of lysophophatidylcholine (LPC). Among all annotated LPC species, 14 LPC species were statistically altered and formed two clusters ( Figure S3E). In the first cluster, some LPC species (e.g. LPC 19:0, LPC 20:0, LPC 18:0) were increased in HFD and decreased with pioglitazone and MSDC-0602K treatment, which are known to be associated with oxidative stress and inflammation [33,34]. On the other hand, the second cluster showed the opposite trend, i.e. low concentration in HFD compared to HFD þ P and HFD þ M groups (e.g. LPC 15:0, LPC 14: 0), which have been shown to regulate glucose uptake in cells [34]. Contrary to plasma profile, BM showed mostly TG species among the top-50 metabolites changed with diet with trend towards higher levels in MSDC-0602K treated group ( Figure S3D). Polar metabolites were not impacted with different interventions, besides three compounds including 3hydroxybutyric acid, dimethylglycine, stachydrine associated with diet content, which were significantly altered in plasma, but they were not observed in BM. We also detected the parent drugs (pioglitazone and MSDC-0602K) present in both plasma and BM with higher concentrations in circulating plasma vs. BM. Taken together, these data revealed that systemic treatment of obese mice with TZDs has a positive impact on decreasing circulating levels of lipids in HFD mice, while BM lipid content was changed differently, thus confirming data from mCT evaluation on different BMAT composition.

MSDC-0602K decreases adipocyte differentiation potential of mBM-MSCs compared to pioglitazone in obese mice
To determine cellular changes related to bone formation, primary murine BM-MSCs (mBM-MSCs) were isolated from treated mice to characterize their cellular and molecular phenotype. Stem cell properties of mBM-MSCs evaluated by colony-forming units-fibroblast (CFU-f) showed increased CFU-f in HFD þ P compared to HFD mice ( Figure S4AeB). The short-term proliferation rate of primary cultures did not show any differences between the groups ( Figure S4C). Further, as it was previously shown [3], mBM-MSCs from the HFD group manifested increased adipocyte (AD) differentiation potential compared to mBM-MSCs from the ND group, as measured by Oil Red O staining ( Figure 4A). This AD differentiation potential was further elevated in HFD þ P group. However, primary mBM-MSCs derived from HFD þ M did not show the same pattern as their AD differentiation potential was decreased compared to HFD þ P ( Figure 4A). These data were confirmed by gene expression analysis of adipogenic markers (Pparg2, Cebpa, Adipoq, Cd36, Fsp27, Insr), which were more pronounced with pioglitazone treatment compared to MSDC-0602K treatment, suggesting higher PPARg activity in HFD þ P than HFD þ M group ( Figure 4B). Osteoblast (OB) differentiation potential of primary mBM-MSCs measured by Alizarin staining (AZR) and alkaline phosphatase (ALP) activity ( Figure 4CeE) revealed impairment in the HFD and HFD þ P groups, while OB induction was improved in the HFD þ M group. These data were also confirmed by gene expression analysis of osteoblastic genes (Alpl, Oc, Col1a1, Bmp2, Dlx5, Msx2) and genes involved in Wnt signaling, (Ctnnb1, Lrp5) (Figure 4FeG). Taken together, these data document that MSDC-0602K treatment in HFDfed mice induces less AD potential and improves OB induction in primary mBM-MSCs compared to the first generation of TZD, pioglitazone, likely via lower activation of PPARg.
3.5. MSDC-0602K increases glycolytic activity along with decreased senescence of mBM-MSCs in comparison to pioglitazone in obese mice To further understand the impact of TZD treatment on cellular metabolism, we evaluated bioenergetic profile in primary mBM-MSCs obtained from treated mice in undifferentiated state using Seahorse bioanalyzer. We obtained simultaneous measurements of mitochondrial function via the oxygen consumption rate (OCR) and glycolysis via the extracellular acidification rate (ECAR) ( Figure 5). In glycolysis measurements, primary mBM-MSCs from the HFD þ M group revealed higher glycolytic capacity compared to the HFD group, while HFD þ P mBM-MSCs had lower glycolytic response ( Figure 5A, Figure S5A). On the other hand, measurements of cellular respiration showed higher maximal respiration rate induced by carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP) in HFD mBM-MSCs compared to ND mBM-MSCs ( Figure 5B, Figure S5B). This increase in HFD-induced OCR capacity was counteracted by treatment with pioglitazone and MSDC-0602K. Surprisingly, mBM-MSCs from HFD þ M group showed less reduction of maximal OCR than those from HFD þ P group ( Figure 5B, Figure S5B). It might indicate that glucose is in the case of HFD þ M metabolized by glycolysis and cannot fully support respiratory metabolism, which prefers other substrates. Further, measurements of reactive oxygen species (ROS) production in primary mBM-MSCs revealed increased senescent phenotype in mice from the HFD þ P group compared to HFD group ( Figure 5C), while mBM-MSCs obtained from HFD þ M were protected from this effect. This phenotype was also confirmed by increased expression of senescent markers (e.g. p53) and markers of senescence-associated secretory phenotype (SASP) (e.g. Fas, Fasgl, Vegfa, Vcam, Tnfa, Il10, Il1rn) [35] in primary mBM-MSCs in HFD and HFD þ P groups, while mBM-MSCs in HFD þ M group did not show such an increase of the selected markers (Figure 5DeE). The senescent phenotype was also manifested in primary hematopoietic stem cells (mHSCs) obtained from HFD þ P group, showing increased expression of p21, p16, Vegfa, Vcam compared to the ND and HFD groups ( Figure S5C), whereas HSCs from HFD þ M mice showed less senescent features. Indeed, these findings confirmed that both mBM-MSC and mHSC populations were affected by pioglitazone and MSDC-0602K treatment, which was associated with changes in BM microenvironment. In addition, gene expression of enzymes associated with mitochondrial metabolism (Pdk4, Pc, Mpc1 and Mpc2) ( Figure 5F) were upregulated in the HFD þ M group compared to HFD þ P, suggesting differential activation of mitochondria in these cells. Taken together, these data demonstrate that MSDC-0602K differently affects cellular metabolism and nutrient utilization in mBM-MSCs compared to pioglitazone with less side effects including senescence compromising their stem cell properties.
3.6. MSDC-0602K reduces insulin and inflammatory responsiveness in mHSCs in comparison to pioglitazone in obese mice As TZD treatment under diabetic conditions is known to affect immune cell function in AT by improving insulin signaling and reducing inflammatory responses [36e38], we investigated downstream signaling pathways related to AKT and inflammatory signaling in mHSCs (as progenitors of immune cells) isolated from HFD mice and mice fed HFD supplemented with TZDs. Insulin stimulation measured by AKT phosphorylation (pAKT S473/total AKT, and pAKT T308/total AKT) showed that insulin signaling in primary mHSCs of HFD mice was not impaired when compared to ND cells (Figure 6AeB), similarly as in primary mBM-MSCs isolated from C57BL/6J mice [3]. In addition, pioglitazone supplementation of HFD increased insulin responsiveness in the corresponding primary mHSCs compared to the HFD and HFD þ M groups mostly expressed by increased pAKT T308/total AKT ( Figure 6B), which was confirmed by gene expression of insulin responsive genes (Irs1, Irs2, Insr) ( Figure 6C). This was accompanied by increased gene expression of adipogenic genes (Pparg2, Cebpa, Cd36) ( Figure 6D). In addition, gene expression of inflammatory genes (Il1b, Tnfa, RelA) showed increased inflammatory response in mHSCs from the HFD þ P group compared to the HFD þ M group ( Figure S5D), suggesting that TZD has a different impact on mHSCs compared to cells in peripheral tissues. These data also correlate with other data on the PPARg upregulation in mHSCs related to osteoclast activation and bone resorption [39], which was not present in mice receiving HFD supplemented with MSDC-0602K ( Figure S5E).

MSDC-0602K manifests similar beneficial effects on hBM-MSC characteristics compared to classical TZDs
Following in vivo experiment in HFD mice, we further evaluated the effect of novel TZD analog MSDC-0602K on stem cell properties in human BM-MSCs (hBM-MSCs). We used immortalized hBM-MSCs cell line described in previous studies [2,21,22,40]. OB differentiation measured by ALPL expression in hBM-MSCs showed inhibitory effect of pioglitazone and rosiglitazone treatment, which was improved by MSDC-0602K treatment ( Figure 7A). Gene expression of MPC1 and MPC2 showed similar responses to different TZDs as in primary mBM-MSCs( Figure 7A). AD differentiation measured in hBM-MSCs by Nile Red staining confirmed a less stimulatory effect of MSDC-0602K when compared to first generation of TZDs (rosiglitazone and pioglitazone), similar to what was observed in mouse primary cells ( Figure 7B). Further, measurement of cellular respiration in undifferentiated hBM-MSCs in response to TZDs and MSDC-0602K showed inhibitory effect of classical TZDs on maximal respiration compared to vehicle, while MSDC-0602K did not manifest such inhibition ( Figure S6AeB), thus recapitulating the data from mBM-MSCs. These results confirmed the similar effects of MSDC-0602K treatment on hBM-MSCs as in mBM-MSCs, i.e. less detrimental effects on BM-MSC properties compared to typical TZDs.

BM-MSCs under MSDC-0602K treatment prefers glutamine over glucose utilization in comparison to AT-MSCs
Previous studies in myoblasts claimed that TZD's effect on cellular metabolism is mostly via MPC inhibition (i.e. via the inhibition of pyruvate transport) and changing substrate utilization in cells [41,42]. To further understand how TZDs and MSDC-0602K affect mitochondrial metabolism in hBM-MSCs, we analyzed how different types of TZDs affect mitochondrial function in intact cells. Our aim was to determine their inhibitory effect on the MPC pathway, the bioenergetic profile of hBM-MSCs, and the preferred energy source for mitochondrial metabolism.
Acute treatment with a lower concentration (10 mM) of TZDs and MSDC-0602K showed diminution of FCCP-induced maximal respiration compared to vehicle ( Figure S7AeB), which was not present after the administration of an MPC-specific inhibitor (UK5099). This effect was even more pronounced when a higher concentration (30 mM) of tested drugs was used (Figure 7CeD). These data suggest that TZDs do not act in hBM-MSCs via inhibition of the MPC pathway as it was demonstrated in myoblasts [41]. Further, measurement of glucose uptake in hBM-MSCs after acute treatment with first generation of TZDs (pioglitazone and rosiglitazone) and MSDC-0602K revealed lower glucose uptake in cells treated with MSDC-0602K and UK5099 compared to typical TZDs ( Figure 7E). On the other hand, intracellular levels of glutamate (product of glutamine metabolism after enzymatic reaction with glutaminase) were higher in MSDC-0602K-treated cells compared to pioglitazone ( Figure 7F), while glutamine intracellular levels were lower ( Figure 7G) suggesting  preferential utilization of glutamine in cellular metabolism after administration of novel TZD analog in comparison to the first generation of TZDs. To understand the effect of different TZDs on bone cells compared to peripheral tissues, we also analyzed their effect on mitochondrial metabolism in AT-MSCs (i.e. 3T3-L1). Interestingly, the lower concentration of TZDs and MSDC-0602K showed a slight increase in FCCP-induced maximal respiration compared to vehicle ( Figure S7Ce D), while at the higher concentration (30 mM) (Figure 8AeB), there was a significant inhibition of respiration induced by MSDC-0602K and UK5099 indicating an impact on MPC function. In addition, gene expression profiling of adipogenic and metabolic genes (Pparg2, Cebpa, Insr, Mpc1, Mpc2) showed differences between 3T3-L1 cells and mBM-MSCs ( Figure S7E), which suggests that these cells are set up differently to respond to metabolic stimuli. AD induction assessed by gene expression profiling of adipogenic genes (Pparg, Adipoq) and metabolic genes (Mpc1, Mpc2, Glut4) showed a stronger response to TZDs in 3T3-L1 cells than in BM-MSCs ( Figure 8CeD). Next, based on the findings of the inhibitory effect of TZDs on MPC [41], we tested the hypothesis that administration of MSDC-0602K and pioglitazone may lead to differential use of nutrients to affect cell differentiation potential under in vitro conditions. The gene expression of markers involved in glutamine metabolism (e.g. SLC1A5, GLS) was upregulated in response to MSDC-0602K treatment of hBM-MSCs compared to pioglitazone, suggesting a shift to increased glutamine utilization by cells ( Figure 8E), while no change or decrease of glutamine metabolism genes (e.g. Slc1a5) was observed in 3T3-L1 cells treated with MSDC-0602K compared to pioglitazone ( Figure 8F). These data were further confirmed by unchanged intracellular levels of glutamate ( Figure S7F) and higher levels of glutamine in MSDC-0602K-treated 3T3-L1 cells compared to pioglitazone ( Figure 8G), while glucose uptake was increased in these cells ( Figure 8H). These results indicate that the TZD analog MSDC-0602K works differently on the periphery compared to BM microenvironment. These findings are important regarding further evidence of the dual role of these drugs in different organ systems.

DISCUSSION
Treatment of obesity and type 2 diabetes with insulin-sensitizers such as TZDs has become a major challenge in clinical practice because of their side effects that include weight gain, cardiovascular complications, and increased risk of fractures [43]. In addition, many drugs prescribed for metabolic diseases have a negative effect on bone metabolism because the target molecules play dual roles in different organ systems. Here, we report that a novel TZD analog MSDC-0602K ("TZD of second generation or PPARg-sparing TZD"), purposely designed to have a weak binding affinity to PPARg, manifested less detrimental effects on metabolic and molecular properties of bone and mBM-MSCs in HFD-induced obesity in male mice compared with the classical TZD pioglitazone. Interestingly, we also found differential effects of MSDC-0602K on cellular metabolism in hBM-MSCs when compared to peripheral AT cells, thus confirming our hypothesis about the impact of systemically administered drugs on different tissues. In our present study, we investigated the effects of 8-week-long administration of TZDs and a novel TZD analog MSDC-0602K on bone and energy metabolism in mice with HFD-induced obesity. The experimental design and doses of the tested drugs have been chosen according to previously published papers studying the metabolic flexibility of peripheral AT in response to dietary interventions in mice [10,18,44].
Even though previous studies in lean and HFD mice showed that of the classical TZDs, rosiglitazone has the most deleterious effect on bone loss in terms of inhibition of bone formation, increased osteoclastogenesis and induction of BMAT compared to pioglitazone [45e49], in our study we used pioglitazone in comparison to MSDC-0602K, as it is the only available PPARg agonist used to treat T2D patients [50e52]. Previous studies using pioglitazone in HFD-fed male rodents (4e6 weeks) demonstrated a negative impact on vertebral bone density, decreased chondrocyte differentiation along with increased adiposity compared to non-treated animals [49,53,54]. This correlates with our data on BMAT volume, bone mechanical properties, AD differentiation potential and senescent phenotype of mBM-MSCs in HFD þ P compared to HFD group. However, pioglitazone-induced detrimental effects on bone properties were not so profound compared to HFD, as in the present study we used C57BL/6N male mice and shorter dietary intervention (8 weeks vs 12e20 weeks) that did not respond to HFD administration by bone loss as is known to occur in the C57BL/6J substrain [3,5]. Nevertheless, MSDC-0602K treatment revealed a positive effect on mechanical properties in femur compared to pioglitazone treated mice, which might be explained by the changes in collagen content as OB differentiated mBM-MSCs isolated from HFD þ M mice showed higher expression of Col1a1 compared to primary cells of HFD þ P mice.
Interestingly, CECT analysis of bone microstructure and BMAT content showed improved bone parameters in L5 vertebra, increased proportion of smaller BMAds in tibia along with increased mechanical properties of bones in HFD þ M compared to HFD þ P mice. These findings support data from Fukunaga et al. [13] demonstrating a beneficial effect of MSDC-0602K compared to rosiglitazone on bone phenotype in lean mice after 6 months of treatment. Furthermore, our study investigated the cellular and molecular properties of primary mBM-MSCs isolated from treated mice, which has not been studied so intensively in previous reports with TZDs. Our data showed improved differentiation potential of mBM-MSCs towards OB and lower AD differentiation capacity in the HFD þ M compared to HFD þ P group, confirming reduced PPARg activation in MSDC-0602K-treated mice. These changes were accompanied by less pronounced senescent and inflammatory phenotype in primary mBM-MSCs derived from mice treated with the novel TZD analog compared to the classical TZD, pioglitazone, suggesting its less detrimental effect on BM-MSC properties. Even though we did not see a decrease in lipid content measured by metabolomics in BM of HFD þ M mice compared to the HFD þ P and HFD group, the detailed analyses of BMAd properties by CECT revealed the impact of MSDC-0602K treatment on the distribution of smaller BMAds in the tibia, which correlated with improved bone mechanical properties. These observations may be explained by increased lipolytic activity in BMAds present in BM, which might affect the lipid cycling and using this energy to be taken by other BM-MSCs to store free fatty acids in form of small lipid droplets. This statement is further supported by decreased gene expression of metabolic genes such as Insr, Pparg2, Cd36 in mBM-MSCs of HFD þ M mice compared to those from HFD þ P, which could correspond to a higher responsiveness to lipolytic than lipogenic signals. Based on recently published data on metabolic heterogeneity of murine and human mature adipocytes in peripheral AT using spatial and single-cell transcriptomics, one can hypothesize that MSDC-0602K treatment may affect the expansion of specific BM-MSC subpopulations with different responsiveness to insulin and thus further lead to different accumulation of lipid droplets [55,56]. Indeed, our recent study documented the presence of murine BM-MSC progenitors with a distinct metabolic program, which are differently responsive to exogenous metabolic stimuli that affect their differentiation fate [57]. However, further studies are needed to test this hypothesis using more high-throughput methods to analyze the functional heterogeneity of BM-MSCs in relation to different treatments. Regardless, our current data highlight the importance of BMAd metabolic and structural changes in response to antidiabetic drug treatment affecting bone quality. In addition, metabolomic analysis of BM revealed the presence of unique metabolites (e.g. dipeptides, nucleotides etc.) only in BM but not in the circulation, suggesting their involvement in cellular metabolism in BM niche and affecting stem cell fate and differentiation potential. In mammals, the uptake of small peptides by the Slc15A family of oligo/dipeptide transporters provides an effective and energysaving intracellular source of amino acids [58]. Recent study by Sharma et al. demonstrated the importance of amino acid metabolism on bone homeostasis and stem cell properties [59]. Besides MSDC-0602K, and among the novel TZDs [39,60] developed to avoid classical PPARg transcriptional activation, Choi et al. [61] demonstrated action of a non-agonistic PPARg ligand (SR1664) that blocks Cdk5-mediated phosphorylation at S273, thereby retaining its insulin-sensitizing effect, preventing weight gain in obese mice, and having no effect on bone formation in vitro. However, this compound has not been tested on bone parameters in mice with prolonged treatment. Another study by Stechschutle et al. reported that PPARg post-translational modifications at S273 and S112 decrease PPARg activation in resorptive osteoclasts and promote osteogenesis while maintaining its insulin-sensitizing effect [9]. Interestingly, this PPARg agonist (SR10171) had an anabolic effect on the long bones but not on the vertebrae, whereas rosiglitazone had a negative effect in both skeletal sites explained by different cellular composition in specific skeleton sites. Our data with MSDC-0602K showed the major effect in vertebrae, which could also correlate with different cellular composition and bone turnover in vertebra vs long bones. However, we did not evaluate BMAT in vertebra site, which might have explained the different impact on bone properties. As MSDC-0602K acts independently of PPARg, its insulin-sensitizing effect supposes to maintain its binding affinity to mitochondria and affect cellular metabolism via MPC [11,62]. The molecular mechanism of TZD action is mainly based on the inhibition of the MPC complex (MPC1 and MPC2) [16,41]. Pyruvate uptake across the mitochondrial inner membrane is a central branch point in cellular energy metabolism with the ability to balance glycolysis and oxidative phosphorylation and control catabolic and anabolic metabolism. Transport of pyruvate into the mitochondrial matrix by the MPC is an important and rate-limiting step in its metabolism [63]. The MPC inhibition can improve cellular glucose handling, as TZDs and UK5099 increase glucose uptake in cells [41,42,64e66]. In our study we tested the effect of MSDC-0602K on cellular metabolism in primary mBM-MSCs, which could explain the changes in stem cell differentiation potential. Surprisingly, we did not observe an inhibitory effect of MSDC-0602K and UK5099 on mitochondrial respiration in primary mBM-MSCs isolated from treated animals, which was also confirmed in hBM-MSCs compared to pioglitazone. However, we found increased glycolytic activity in MSDC-0602K-treated cells compared to pioglitazone treatment, suggesting a change in the utilization of various nutrients contributing to the TCA cycle. This is an interesting finding as MSDC-0602K appears to act differently in peripheral AT-MSCs, in which we were able to detect an inhibitory effect of both drugs, MSDC-0602K and UK5099, on maximal respiration. Previous studies have shown that MPC activity determines the fuel oxidized by mitochondria, which has direct implications in terms of cellular functions and cell fate [67,68]. Recent findings in brown adipocytes showed that MPC inhibition leads to increased mitochondrial respiration, which activates lipid cycling and energy expenditure by replenishing energy from fatty acid and glutamine uptake [42]. However, these experiments were performed on isolated mitochondria compared to our data in intact cells.
We hypothesized that MSDC-0602K might affect glutamine metabolism as MSDC-0602K treatment increased OB induction in primary mBM-MSCs, which is associated with increased glutamine metabolism [69]. Interestingly, glucose uptake was decreased, and glutamate content (a product of glutamine metabolism) was increased in MSDC-0602K-treated BM-MSCs compared to pioglitazone. These changes were accompanied by upregulation of genes associated with glutamine metabolism in MSDC-treated cells compared to pioglitazone. Based on our findings we hypothesize that the beneficial effect of MSDC-0602K on BM-MSCs compared to AT-MSCs, is due to different preferences of nutrient utilization (i.e. glutamine over glucose uptake) in these cells, which affect their metabolic and stem cell properties supporting osteoblast differentiation in BM-MSCs but not in AT-MSCs. However, more data are needed to elucidate the exact mechanism behind the inhibitory effect of TZD on MPC and mitochondrial content and oxidative capacity, as well as its consequences in terms of changes in BM-MSC properties in relation to bone metabolism. The present study brings several positive aspects. First, in contrast to previous studies investigating the impact of TZDs on bone, we performed a comprehensive analysis of bone and BMAd parameters along with cellular and molecular properties of primary mBM-MSCs from treated animals. Second, we applied a preventive strategy to investigate the effect of insulin-sensitizing drugs in the context of obesity and T2D. Third, we employed state-of-the-art methods to characterize cellular metabolism in BM-MSCs isolated from both mice and humans. On the other hand, our study has some limitations. Dietary interventions in HFD-fed male mice lasted only 8 weeks, suggesting that longer treatment could have a much greater effect on bone properties. Moreover, we did not test the novel TZD analog in female mice, which is important for future perspective use of this drug in women with metabolic complications associated with osteoporosis. Taken together, our study using HFD-induced obesity animal model shows that MSDC-0602K, as a novel insulin sensitizer, has reduced detrimental effects on bone parameters and BM-MSC phenotype by activation of glutamine metabolism as compared to the first generation of TZDs, i.e. pioglitazone. In addition, our findings provide novel insights on MSDC-0602K action in vivo and thus might further contribute to more specialized strategies in treatment of both bone and metabolic diseases.