Introduction

The bone, long considered a merely structural organ, is now seen as a highly dynamic tissue in which various processes, including hematopoiesis, bone remodeling, immunity and metabolism are tightly regulated^1,2. These processes rely on specialized microenvironments, termed niches, that play a critical role in cell fate regulation³, with the bone marrow adipose tissue (BMAT) being an emerging niche component^4–8. The bone marrow (BM) niche regulates hematopoietic stem and progenitor cell (HSPC) quiescence, self-renewal and commitment towards differentiated cells thanks to a combination of soluble factors and cell-cell interactions^1,3,9,10. Osteoblasts, adipocytes and endothelial cells form the BM niche along with a wide range of BM stromal cells (BMSCs) ^1,11. While the involvement of these specific cell types in the regulation of hematopoiesis is generally accepted, the extent of their action is controversial, especially for cells that form part of the adipocytic lineage. Mature BM adipocytes were first described to have a predominantly negative effect on hematopoietic recovery post-myeloablation^4,12,13. Subsequent studies showing a positive effect of the BM adipocyte lineage on hematopoietic recovery called these results into question^5,14,15. Recent work has uncovered a more nuanced role of non-lipidated BM adipogenic precursors in the control of bone homeostasis^6,7 and favoring hematopoietic recovery upon irradiation^8,16. As for other adipose depots, the BMAT is plastic and can be remodeled upon injury and dietary or pharmacological stimulation^13,17–21.

Bone and BM homeostasis is influenced by extramedullary mechanisms, including the incompletely understood gut-BM cross-talk^22–25. Communication between the gut and distant organs can be mediated by different signaling molecules, including bile acids (BAs), metabolites derived from the liver and modified by the gut microbiome²⁶. BAs are signaling molecules that modulate whole-body metabolism^27–29 through activation of specific receptors that include the G protein-coupled receptor Takeda-G-protein-receptor-5 (TGR5)³⁰. BA-TGR5 signaling has been studied in a multitude of organs^28,29, and profoundly affects adipose tissue physiology through several mechanisms, ranging from beiging^31,32 and suppression of chronic inflammation in resident macrophages of the white adipose tissue (WAT)³³, to stimulation of fat oxidation in brown adipose tissue (BAT)^34–36. In addition, TGR5 expression was observed in human BMSC-derived adipocytes³², but the effect of TGR5 on BMAT and thus on BM function is unknown. Previous reports have shown that BAs impact fetal and adult hematopoiesis acting as chemical chaperones^37–39 and, more recently, through direct interaction with myeloid blood cell precursors via the vitamin D receptor⁴⁰. However, the effects at the level of TGR5, a BA-specific and potentially druggable receptor, remain to be unraveled.

Here, we report that TGR5 regulates the BM adipocytic lineage and hematopoietic recovery upon transplantation. We show that the loss of TGR5 markedly decreases BMAT in chow- and high-fat-diet (CD and HFD, respectively) conditions. Concomitantly, we observe an increase of immunophenotypically-defined adipocyte-progenitor cells (APCs)¹³ and a hastened hematopoietic recovery upon BM transplantation that is not dependent on the hematopoietic cells transplanted but on the genotype of the recipient.

Results

We first aimed to define the presence of TGR5 in primitive and differentiated hematopoietic populations. For this, we took advantage of a GFP reporter mouse model driven by the Tgr5 promoter (TGR5:GFP)⁴¹. Flow cytometry analysis of hematopoietic populations showed consistent GFP signal throughout the main hematopoietic stem and progenitor (HSPC) populations in the adult mouse BM (Fig. 1A). We observed significant levels of GFP positivity in the lineage negative (Lin⁻), c-Kit⁺, Sca1⁺ cells (LKS), a population highly enriched in murine hematopoietic stem and multipotent progenitor cells (Fig. 1B). We detected similar frequencies of GFP⁺ cells in restricted myeloid progenitors with no marked differences in the percentage of GFP⁺ cells between the different immunophenotypically defined subpopulations (granulocyte-monocyte progenitors (GMP), common myeloid progenitors (CMP) and megakaryocyte-erythrocyte progenitors (MEP) (Fig. 1C). GFP expression was further detected by immunohistochemistry in the BM (Fig. 1D), spleen (Fig. S1A) and thymus (Fig. S1B), confirming the expression of TGR5 in different immune organs. Flow cytometry analysis of fully differentiated hematopoietic populations in peripheral blood (Fig. S1C) showed that TGR5 is mostly expressed in cells of the myeloid lineage with lower expression in lymphocytes (Fig. S1D). Taken together, these findings indicate that TGR5 is present in primitive and differentiated hematopoietic cells.

Figure 1.

We next aimed at defining the impact of a loss-of-function of TGR5 in hematopoiesis. Flow cytometry analysis of the BM of young germline TGR5 knock-out (Tgr5^−/−) mice showed no marked alterations in the percentage (Fig. 2A) or absolute number (Fig. 2B) of HSPC populations compared to their controls (Tgr5^+/+). Functionally, we could not detect differences in hematopoietic colony-forming unit (CFU) potential, and thus functional blood cell progenitors, between Tgr5^−/− mice and their Tgr5^+/+ counterparts (Fig. S2A). However, we observed an increase in the total number of CD45⁺ cells per hindlimb collected from Tgr5^−/− animals (Fig. 2C) that did not lead to an increase in CFUs per hindlimb (Fig. S2B). Complete blood counts (CBC) of peripheral blood showed a modest increase in the number of red blood cells and a decrease in eosinophils, but no change in other white blood cells or platelets in Tgr5^−/− mice (Fig. 2, D-H). Taken together, these results indicate that TGR5 is non-essential for steady-state hematopoiesis in young adult male mice.

Figure 2.

In the next step, we aimed at defining the hematopoietic cell-autonomous effects of a germline deletion of TGR5 upon hematopoietic stress. For this, we examined the repopulating capacity of Tgr5^−/− and Tgr5^+/+ BM cells when co-transplanted with competitor CD45.1.2 BM into a lethally irradiated wild-type recipient (CD45.1) (Fig. 2I). We found that, upon primary transplantation, Tgr5^−/− donor cells blunt short-term repopulating capacity, but do not affect long-term repopulation (Fig. 2J) nor short-term repopulating capacity in the secondary or tertiary transplant (Fig. S2C). Taken together, these findings indicate that Tgr5^−/− BM cells exhibit a transient deficit in short-term repopulating capacity that is corrected over serial transplantation in wild-type recipients, with no detectable long-term impairment of hematopoietic stem cell function.

Based on our findings that functional hematopoietic progenitors are quantitatively conserved at homeostasis and that the cell-autonomous effects disappear on serial transplantation (i.e., exposure to a wild-type BM niche), we hypothesized that an altered BM microenvironment in Tgr5^−/− animals may account for this phenotype. In agreement with the literature⁴², we found that young adult Tgr5^−/− male mice display a decrease in bone volume fraction (BV/TV) and a decrease in trabecular number (Tb.N) with no changes in trabecular thickness (Tb.Th) or separation (Tb.Sp) at the distal femur (Fig. 3, A and B). In addition, no alterations in cortical parameters were found at the level of the femoral diaphysis (Fig. S3, A and B), as evaluated by micro-computed tomography (µCT). We could not detect any changes in bone microarchitecture in the spine at the level of L4 in young male Tgr5^−/− mice (Fig. S3, C and D). These data suggest that the role of TGR5 on bone microarchitecture differs based on skeletal location, which is in line with the recently described differential regulation of the axial and the appendicular skeleton^43,44.

Figure 3.

To further study the overall composition of the BM microenvironment, we used osmium tetroxide (OsO₄) contrast-enhanced µCT⁴⁵ to visualize the tibial BMAT. The BMAT in the proximal tibia, also known as regulated BMAT (rBMAT), is regarded as the most dynamic and stimuli-responsive BMAT in opposition to that found in the distal tibia, which is enriched in the less responsive constitutive BMAT (cBMAT)⁴⁶. This analysis revealed a profound reduction in proximal tibia BMAT in young adult Tgr5^−/− animals as well as a less marked decrease in distal tibia BMAT compared to age-matched controls (Tgr5^+/+) (Fig. 3, C and D). As BMAT accumulates with age⁴⁷, we evaluated the tibias of 1-year-old Tgr5^−/− mice, which showed a similar decrease in BMAT compared to their age-matched controls, indicating that the expansion of BMAT with aging was blunted in the absence of TGR5 (Fig. 3, E and F). Dietary intervention has also been shown to modulate BMAT expansion and high-fat diet (HFD)-feeding robustly increases marrow adiposity, in general, and rBMAT, in particular^13,48,49. To investigate if TGR5 could control BMAT plasticity upon this intervention, we fed Tgr5^−/− and Tgr5^+/+ mice with HFD for 8 weeks and evaluated the tibial BMAT. In agreement with our data in young and middle-aged chow-diet (CD)-fed mice (Fig. 3, C-F), contrast-enhanced µCT analysis of OsO₄-stained tibias revealed a marked decrease in rBMAT along with a more moderate decrease in cBMAT in Tgr5^−/− mice compared to their wild-type controls (Fig. 3, G and H).

Finally, we sought to evaluate whether the robust decrease in rBMAT with a more preserved cBMAT was exclusive of the appendicular skeleton or was also present in axial structures. For this, we chose the caudal spine, as this segment of the vertebral column contains a gradient where rBMAT gives way to cBMAT in a proximal- to-distal manner⁴⁷. µCT scanning of OsO₄-stained caudal spines showed this adipocytic gradient, so-called the yellow-to-red transition. at the level of the second caudal vertebrae (CA2) in young male Tgr5^+/+ mice (Fig. S3E). Analogously to our findings for the rBMAT in tibia, Tgr5^−/− mice presented a marked decrease in BMAT at the level of CA2 (Fig. S3F, G). Histological preparations of the same spine segments from an independent cohort further confirmed our µCT data (Fig. S3H).

Taken together, our findings indicate a mutual loss of trabecular bone and marrow adiposity in Tgr5^−/− mice. Specifically, loss of TGR5 decreased trabecular bone in the appendicular skeleton without affecting bone structure at the level of the axial skeleton. Conversely, lack of TGR5 markedly reduced BMAT volume in both the appendicular and axial skeleton. The decrease in BMAT occurred fundamentally at the expense of rBMAT both at homeostasis and under HFD-induced expansion of BMAT, suggesting that TGR5 might be a key player in the regulation of this tissue.

Given the changes in both BMAT and bone architecture, we hypothesized that BM stroma cells (BMSCs) obtained from Tgr5^−/− mice would present a defect in both adipocytic and osteoblastic differentiation. Surprisingly, and contrary to our hypothesis, Tgr5^−/− BMSC cultures differentiated better into adipocytes, as evidenced by Oil Red O staining (ORO) (Fig. 4A, and Fig. S4A) and digital holographic microscopy (Fig. S4, B and C)⁵⁰. Moreover, Tgr5^+/+ and Tgr5^−/− BMSCs displayed comparable osteoblastic differentiation, evaluated by alkaline phosphatase staining (ALP) (Fig. S4D) and calcium deposition, as evidenced by alizarin red staining (Fig. S4E). Taken together, these in vitro data indicate that TGR5 deletion does not cause an intrinsic defect in in vitro BMSC differentiation.

Figure 4.

Recent reports have revealed that lipodystrophy or specific ablation of pre-adipocytic populations in the BM leads to massive trabecular bone invasion of the medullary cavity, indicating that these populations negatively regulate the mineralized compartment of the BM microenvironment^{4,6–8,42,51,52}. Considering this, we hypothesized that our findings in Tgr5^−/− mice, (i.e., decreased BMAT and trabecular bone together with an increased in vitro adipocyte differentiation), could be explained by an in vivo blockage of lipidation of adipocyte progenitor cells (APCs) that is rescued upon forced induction of in vitro adipocytic differentiation. In order to assess this, we analyzed the BM stroma based on the expression of Sca1 and CD24 for APC quantification¹³. In agreement with our hypothesis, we found an increase in immunophenotypic APCs (CD45⁻Ter119⁻CD31⁻Sca1⁺CD24⁻) in the stroma of Tgr5^−/− mice (Fig. 4B, and Fig. S4F). Finally, we observed an alteration in the clonogenic capacity of BMSCs, evidenced by an increase in fibroblast colony-forming units (CFU-F) in Tgr5^−/− BM and compared to wildtype controls (Fig. 4C, and Fig. S4G), which agrees with a shift towards less differentiated BM stromal populations in Tgr5^−/− animals. Taken together, our findings indicate that TGR5 loss-of-function perturbs the bone-BMAT-hematopoietic tissue equilibrium within the marrow cavity and suggests that TGR5 signaling is necessary for BM pre-adipocytes to fully mature.

Given the accumulation of APCs in Tgr5^−/− mice, we next investigated the impact of the associated BM microenvironment alterations in stress hematopoiesis. Based on the first report of the negative effect of lipidated BM adipocytes on hematopoietic recovery⁴ and recent work showing that non-lipidated adipocytic precursors improve hematopoietic recovery¹⁶, we reasoned that the displacement of the BM adipocyte differentiation axis towards non-lipidated progenitors observed in the BM microenvironment of Tgr5^−/− mice would correlate with an improved early hematopoietic recovery upon irradiation and transplantation. To assess this, we generated inverse chimeras (Fig. 4D) and evaluated early hematopoietic recovery using CBCs as a direct readout of the functionality of the system for the first 4 weeks post-irradiation. We observed lower mortality post-transplant in the Tgr5^−/− recipients (Fig. 4E) and faster hematopoietic recovery in Tgr5^−/− recipients upon lethal irradiation and transplantation with wild-type donor cells. Tgr5^−/− recipients recovered their platelet levels faster than the Tgr5^+/+ recipients, showing values consistently above 200.000/µL one week sooner; platelet levels below this threshold are considered a risk factor for bleeding events (Fig. 4F)^53,54. The recovery was also faster for white blood cells (WBCs) (Fig. 4G) without a clear predominance of any cell type as shown for neutrophils (Fig. 4H), monocytes (Fig. 4I) or lymphocytes (Fig. 4J). Of note, Tgr5^−/− recipients had their neutrophil levels over 500.000 cells/µL, the threshold value for infection risk^53,54, on day 15 post-irradiation (uncorrected p-value 0.028) but this difference was no longer statistically significant upon multiple-comparison correction. We observed no major differences in the recovery of red blood cells (Fig. 4K). Taken together, our findings suggest that the loss of TGR5 leads to a BM stroma that is beneficial for early hematopoietic recovery.

Discussion

In the current study, we aimed at investigating the role of TGR5 on BM homeostasis. We showed that TGR5 loss-of-function leads to the accumulation of adipocytic precursors and a simultaneous loss of BMAT, suggesting that TGR5 is necessary for the in vivo maturation of BM adipocytes in young male mice. We also found that Tgr5^−/− mice receiving BM transplantation after lethal irradiation recover faster than their Tgr5^+/+ counterparts, which could be a consequence of the accumulation of adipocyte precursors¹⁶, a decrease of lipidated adipocytes⁴, or a combination of both.

The BM adipocyte differentiation axis is gaining attention as a fundamental component of bone and blood physiology, with several recent studies focusing on the BM stroma population as a key player in the maintenance of the balance between bone and BM space^8,55. Ablation of the BM adipocyte precursor population leads to massive accumulation of bone in the medullary cavity^4,7,14,52,54 which in turn reduces the volume available for hematopoietic tissue. In agreement with this model, we observe that male Tgr5^−/− mice display an increase in adipocytic precursors, a decrease in bone, and an expansion of CD45⁺ cells in the BM.

BMAT has been described as a highly dynamic fat depot that responds to dietary changes, but it has distinct characteristics compared with extramedullary fat, both from a biological and regulatory perspective⁵⁶. For instance, and surprisingly, both nutrient deprivation (e.g. caloric restriction and anorexia) and excess (e.g. HFD feeding) result in BMAT expansion^49,57–59. Although counterintuitive, this nutrient-mediated control of BMAT ensures energy conservation in BM to support the energy-consuming process of hematopoiesis and bone remodeling^60,61 . Another difference between BMAT and peripheral adipose tissues is the absence of HFD-induced pro-inflammatory response in the BMAT compared with the latter⁴⁹. While the role of TGR5 in extramedullary adipose tissues has been extensively studied, and shown to be important for beige remodeling^31,32, lipolysis³², and protection against diet-induced immune infiltration of the white adipose tissue^29,33, the effects of an activated TGR5 signaling pathway in BMAT and its impact on hematopoiesis is unknown and will require future investigation. Of note, the BA ursodeoxycholic acid (UDCA) is routinely used in clinics in the context of BM transplantation, where supplementation with UDCA has been shown to reduce post-transplant complications and improve overall survival⁶². Available reports have attributed these benefits to the unique feature of tauro-conjugated form of UDCA (TUDCA) as chemical chaperones, improving protein folding in fetal liver HSCs³⁸ or after acute BM insult³⁹. In addition, UDCA and TUDCA are hydrophilic BAs, and part of their beneficial effects could be a consequence of changes in the hydrophobicity of the BA pool^63,64. However, BAs signal through dedicated BA receptors, but the receptor-mediated signaling of UDCA and TUDCA remains unclear. Although some studies proposed UDCA as a TGR5 agonist^65–68, effects occur mainly at supra-physiological doses of UDCA. In addition, the EC₅₀ of UDCA for TGR5 is more than 60 times higher than that of the strongest endogenous TGR5 agonist, lithocholic acid (LCA)⁶⁹, indicating that UDCA is a poor TGR5 agonist. Based on this notion, it will be interesting to investigate whether, in these context-dependent conditions, the benefits of UDCA upon stress hematopoiesis result from the inability to activate TGR5 signaling in the BMAT. Future work will be needed to confirm this possibility.

Finally, novel therapeutic strategies could be envisioned in the future to pharmacologically mimic TGR5 inhibition in the BMAT. Although several selective TGR5 agonists have been developed, only a few TGR5 antagonists have been discovered, namely triamterene⁷⁰ and SBI-115⁷¹. However, these compounds have only been used in vitro, and additional in vivo research and optimization will be required to pharmacologically block TGR5 in the context of hematopoietic stem cell transplantation or other conditions linked to stress hematopoiesis, in which a fast production of mature blood cells is required. Our results indicate that deletion of TGR5 modulates the BM adipocyte lineage and suggest that strategies aimed at blocking the activity of this receptor might be therapeutically attractive.

Author contribution

AAC, AP and FS: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, writing – review and editing. SFL, VD, AJ, UK, DPP: Methodology. KS and ON Conceptualization, Resources, Supervision, Funding acquisition, Validation, Investigation, Writing – original draft, Project administration, Writing – review and editing.

Acknowledgements

We wish to thank the Histology Core Facility, Center for Phenogenomics, Bioscreening Facilities, Electron Microscopy facilities and BioImaging and Optics Platform at EPFL as well as the In Vivo Imaging Facilities at AGORA (UNIL/EPFL) and Flow Cytometry facilities at AGORA (UNIL/EPFL) for their continuous support and guidance for this project. This work was funded by the Ecole Polytechnique Fédérale de Lausanne (EPFL), the University of Lausanne (UNIL), the Swiss National Science Foundation (SNSF N° 310030_189178, Sinergia CRSII5_180317/1 to K.S. and SNSF Professorship grant PP00P3_144857 and 183725 to O.N.), and La Caixa Foundation (to K:S). Alessia Perino was supported by a fellowship from AXA Research Fund. Frédérica Schyrr was funded by the SNSF MD-PhD grant 183986.

Materials and Methods

Animal studies

Generation of mouse models and tissue collection

To evaluate the expression of TGR5 in peripheral blood and BM, we used a transgenic mouse reporter model (TGR5:GFP), that co-expresses human TGR5 and the enhanced green fluorescent protein (GFP) linked by the foot-and-mouth disease virus (F2a) cleaving peptide sequence under the control of the endogenous Tgr5 mouse promoter ^41,72.

Mice were housed with ad libitum access to water and food and kept under a 12-h dark/12-h light cycle with a temperature of 22 °C ± 1 °C and a humidity of 60% ± 20%. 8- to 12-week-old male mice fed chow diet (CD, SAFE 150) were used for all experiments. 1-year-old male mice fed (CD, SAFE 150) were used where indicated. For the high-fat diet experiment, 8-week-old male mice were fed HFD (Research Diets D12492) for 12 weeks. The whole-body TGR5 knock-out mouse model (Tgr5^−/−) has been previously described ⁷³. C57BL/6J (CD45.2) and C57BL/6J Ly5.1 (CD45.1) mice were bred in-house at the EPFL facility. Double congenic mice (CD45.1/.2) were bred in-house, by crossing CD45.2 and CD45.1 mice.

BM transplantation

For transplantation assays, recipient mice were irradiated (lethal x-ray irradiation 8.5 Gy, split in two 4.25 Gy doses 4 h apart using an RS-2000 X-ray irradiator (RAD SOURCE). Transplanted cells were administered the day following irradiation via tail-vein injection. For two weeks after lethal irradiation, mice were treated in drinking water with paracetamol (500 mg Dafalgan in 250 ml water). No antibiotics were provided in drinking water. For the primary competitive transplant setting, 300.000 cells of each the CD45.2 donor (Tgr5^−/− or Tgr5^+/+) and the CD45.1/.2 competitor were co-administered. For the secondary and tertiary transplant, 4 million cells were collected from the primary donor and transplanted into the recipient. For the inverse chimera experiments where the CD45.2 mice were the recipients (Tgr5^−/− or Tgr5^+/+), 250.000 CD45.1 total BM cells were administered.

Blood collection for the experiments described in this work was performed at noon. Blood was obtained from the tail vein by means of a small incision with a scalpel blade and collected into EDTA-coated tubes. Complete blood counts were obtained with an Element HT5 hematology analyzer (Heska, USA)

µCT analysis

To evaluate bone microarchitecture and osmium tetroxide (OsO₄) stains, a SkyScanner 1276 (Bruker, Belgium) was used. For both applications, an 0.25 mm Al filter was used with a voltage of 200 kV and a current of 55 mA. Samples were wrapped in PBS-soaked paper towels and scanned inside a drinking straw sealed on both ends to avoid drying. Voxel size for both applications was set at 10×10×10 µm³.

Bone microarchitecture was evaluated according to the ASBMR guidelines ⁷⁴ using a custom CTan (Bruker, Belgium) script for automatic segmentation of trabecular bone in the distal femoral VOI, which was set 100 slices proximal to the distal growth plate and extended 200 slices towards the femoral diaphysis (slice thickness of 0.010 mm). The threshold used to binarize the calcified tissue was 40 on a 0-255 scale. Reconstruction of the scans was performed using NRecon (Bruker, Belgium) and further analysis were performed using CTan (Bruker, Belgium) with the minimum for CS to image conversion set at 0 and maximum set at 0.14. For the analysis of cortical parameters, the midpoint of the femur was determined, and the VOI was defined as the bone 50 slices (slice thickness of 0.010 mm) distal and proximal of the slice corresponding to the midpoint of the bone. All other parameters were kept the same as for the analysis of trabecular bone.

OsO₄ stained was performed on formalin-fixed, EDTA-decalcified tibias. For this, tibias were kept in 10% formalin for 24 hours at 4 °C. Following fixation, the bones were decalcified in PBS-0.5 M EDTA for 2 weeks refreshing of the decalcifying solution on day 7. OsO₄-was performed as previously described ⁴⁵. Reconstruction of the scans was performed as for calcified bone except for the minimum and maximum CS range, which was set to 0 and 0.30, respectively. OsO₄ was segmented using a threshold of 50 on a 0-255 scale. Images and quantification of BMAT were analyzed using Dragonfly software (ORS, Canada).

Flow cytometry

All antibodies used for this work as well as their dilutions are listed in Supplementary Table 1. All dilutions and cell suspensions were prepared in FACS buffer (2% serum + 1 mM EDTA in PBS).

Histology

Spleens and thymus were fixed in 10% formalin for 24 h at 4 °C. EDTA-decalcified bones (as described previously for OsO₄ stain), spleens and thymus were cut longitudinally in 3-4 µm sections for staining. For IHC, detection of rabbit anti-GFP (Abcam, ab6673,) was performed manually. After quenching with 3% H₂O₂ in PBS 1x for 10 min, a heat pretreatment using 0.1 M Tri-Na citrate pH 6 was applied at 60 °C in a water bath overnight. Primary antibodies were incubated overnight at 4 °C. After incubation of ImmPRESS HRP (Ready to use, Vector Laboratories), revelation was performed with DAB (3,3′-Diaminobenzidine, D5905, Sigma-Aldrich). Sections were counterstained with Harris hematoxylin and permanently mounted. All slides were imaged on an automated slide scanner (Olympus VS120-SL) at 20x or 40x.

Cell culture

Cell extraction

Culture of primary BMSCs was performed as reported in the literature ⁷⁵. 8-12-week-old male mice were used for these experiments. To isolate BMSCs from the BM, we collected femora and tibias, cut the epiphyses, and placed the bones in a 200 μL pipette tip with its end cut to fit in a 1.5 mL microcentrifuge tube. Bones were spun at 10.000 g for 10 seconds to flush out the BM from the medullary cavity. The BM plug was then transferred to a 15 mL conical tube containing 10 mL of ascorbic-acid free αMEM (containing 10% FBS and 1% penicillin-streptomycin) and this suspension filtered through a 70 μm mesh filter prior to plating in a 100 mm cell culture dish (Corning, USA). Cells were replated 72 hours later at a density of 62.500 cells/cm² in tissue-culture treated 12-well plates (i.e., 250.000 cells/well) and grown to confluency prior to differentiation. Culture medium was refreshed every other day until confluency⁷⁵.

For CFU-F assays, BM was obtained in a similar manner and plated at a density of 1 million cells per well of a tissue-culture treated 6-well plate (i.e., 100.000 cells/cm²). CFU-F cultures were kept in αMEM containing 10% FBS and 1% penicillin-streptomycin and stained on day 7 post-plating with 0.1% toluidine blue in 10% formalin.

Media composition and differentiation cocktail

Differentiation into osteoblasts was started once confluency was reached. For osteoblastic differentiation, the base culture medium of ascorbic-acid free αMEM containing 10% FBS and 1% penicillin-streptomycin was supplemented with L-ascorbic acid (final concentration 8 mM) and β-glycerophosphate (final concentration 1 mM) and refreshed every other day until the endpoint of the assay.

Differentiation into adipocytes was started once confluency was reached. For adipocytic differentiation, base culture medium was high-glucose (4.5 g/L) DMEM supplemented with 10% FBS and 1% penicillin-streptomycin was used. Adipocytic induction medium containing dexamethasone (1 μM), isobutylmethylxanthine (IBMX, 0.5 mM), insulin (2 μM) and rosiglitazone (20 μM) was prepared. Cells were cultured in adipocytic differentiation medium for 4 days with refreshing on day 2. On day 4, culture medium was changed to maintenance medium, prepared by supplementing the base medium with only insulin (2 μM) and rosiglitazone (20 μM). Cells were kept in maintenance medium until the endpoint (Day 7 post-differentiation) with no medium refreshing.

Differentiation stains

For alkaline phosphatase (ALP) stain of osteoblast cultures, one tablet of SigmaFast (Sigma-Aldrich) was dissolved in 10 mL distilled water. Osteoblasts were fixed in 10% formalin for 1 minute, washed once with PBS-0.05% Tween 20 and incubated in the staining solution for 6 minutes. Cells were then washed with PBS-0.05% Tween 20 and imaged.

For alizarin red (AR) stain of calcium deposits in osteoblastic cultures after ALP staining, cells were washed once with distilled water and incubated for 45 minutes with AR stain (prepared fresh with 2 g of alizarin red in 100 mL distilled water and pH adjusted to 4.2). Cells were then washed with distilled water 4 times and imaged.

For Oil Red O (ORO) staining of adipocyte cultures, we fixed the cells in 10% formalin for 1 hour at room temperature. In the meantime, we filtered the ORO stock solution through filter paper as described in the literature ⁷⁵. Cells were stained with ORO for 15 minutes, washed 6 times with distilled water, imaged, and incubated for 10 minutes with 100% isopropanol to solubilize the stain. ORO was quantified using an iD3 plate reader (Agilent, USA) set to measure absorbance at 490 nm ⁷⁵.

Digital holographic microscopy

Digital holographic imaging was performed in black wall 96-well imaging plates (Corning, 353962) using Transmission DHM® T1000 (Lyncée Tec, Switzerland). Plates were pre-coated with Poly-ornithine (Sigma,100 mg/L) to prevent cell detachment. The cells were imaged live and without prior liquid manipulation using a 20x/0.4 NA microscope objectives. The best-focus phase images were reconstructed automatically in MATLAB (MathWorks, USA) from the acquired holograms and the average quantitative phase signal or optical path difference (OPD) was automatically measured using a fixed threshold value.

Ethical approval

Experiments were carried out in accordance with the Swiss law and with approval of the cantonal authorities (Service Vétérinaire de l’Etat de Vaud, authorization nos. 2990, 2990.1, 2990.2 and 3740). No statistical methods were used to predetermine sample size. Mice were randomly assigned to control and intervention groups. The investigators were not blinded to allocation during experiments and outcome assessment unless otherwise stated.

Statistics

Differences in groups were assessed as stated in the legend for each figure. GraphPad Prism 9 was used for all statistical analyses. The studies were either replicated and/or the results were confirmed by using different approaches (flow cytometry and immunohistochemistry, for example) yielding similar results. All P values ≤ 0.05 were considered significant. When the exact value is not provided, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 and ****P ≤ 0.0001.