MCP-1 Feedback Loop Between Adipocytes and Mesenchymal Stromal Cells Causes Fat Accumulation and Contributes to Hematopoietic Stem Cell Rarefaction in the Bone Marrow of Patients With Diabetes

Type 2 diabetes (T2D) is a major risk factor for cardiovascular disease because of its damaging impact on the micro- and macrovasculature. Patients with T2D have twice the chance of dying from ischemic events as subjects without diabetes (1), and those who survive suffer a markedly slower and more arduous recovery (2). This worse outcome has been attributed to the deleterious effect of T2D on angiogenesis and vasculogenesis, two intertwined reparative processes brought about by local and circulating cells (3). A defect in the mobilization of progenitor cells, such as CD34⁺ cells, from the bone marrow (BM) represents a common feature (4) and a predictor for cardiovascular mortality in patients with diabetes (5,6). Moreover, the reduced availability and functionality of circulating vasculogenic cells may be the consequence of pathogenic processes occurring within the BM microenvironment (7).

Excessive adiposity, typically associated with T2D, does not spare the BM. However, the marrow is not an ectopic fat deposit, like organs that normally contain only small amounts of fat, such as the liver, skeletal muscle, and heart. Instead, fat is a physiological component of the BM, representing ∼10% of the total fat mass in a lean individual (8). In patients with T2D, BM adipose tissue (BMAT) is remarkably increased at the expense of the hematopoietic tissue. Other pathological conditions, such as calorie restriction and radiation exposure, as well as aging and obesity have been associated with BM adiposity (9–12).

BMAT shares some phenotypic similarities with but constitutes a distinct category from both white and brown adipose tissue. Moreover, its function and pathogenic implications remain largely unknown (13). BM adipocytes (ADs) originate from resident mesenchymal stromal cells (MSCs), renamed recently as skeletal stem cells (14), which can also differentiate into osteoblastic and chondrogenic lineages (11). The relationship between BM-ADs and other marrow cell populations is increasingly complex. Several studies suggest that BMAT accumulation is deleterious to the integrity of bone and the hematopoietic niche. However, this association is not universal and depends on the experimental and clinical context. For example, in mice, both the expansion of endogenous BMAT, as in obesity or aging, and exogenous AD transplantation cause a decrease in Lin^negSca1^poscKit^pos (LSK), long-term CD34^neg (LT), and short-term CD34^pos (ST) hematopoietic stem cells (HSCs), which also contain multipotent progenitor cells (11). In contrast, a recent study showed that BM-ADs support regenerative hematopoiesis after irradiation or chemotherapy (15). BMAT and bone are also intimately linked, as their levels are often found to be inversely correlated, with a reduced bone-to-BMAT ratio being considered a risk factor for bone fractures (16). Mature osteoblasts inhibit the differentiation of MSCs into the adipose tissue through paracrine mechanisms. Conversely, mature BM-ADs influence bone mass by inhibiting the progression of BM-MSCs to osteoblasts and enhancing osteoclastogenesis (17). In addition, BMAT functions as an endocrine organ, surpassing the more widespread white fat as a source of adiponectin (ADIPOQ) (12) and secreting an extensive number of cytokines and adipokines, such as leptin (LEP), resistin (RETN), and monocyte chemoattractant protein-1 (MCP-1/CCL2) (18,19).

MCP-1 is expressed by various cell types, either constitutively or after induction by oxidative stress, cytokines, or growth factors (20). Circulating levels of MCP-1 are elevated in obese patients and patients with T2D (21). Furthermore, gene deletion (22) and pharmacological inhibition of MCP-1 reduce visceral white fat volume in obese mice (23,24), suggesting direct participation in adipogenesis. Activation of MCP-1 receptor CCR2 leads to the induction of the transcription of MCP-1–induced protein (MCPIP), which promotes AD differentiation in a peroxisome proliferator–activated receptor γ (PPARγ)–independent manner (25,26). MCP-1 is also pivotally involved in the modulation and recruitment of macrophages and T cells, which accounts for its participation in the induction of fat tissue inflammation in T2D and obesity (27).

The current study aims to unveil the molecular mechanisms and consequences of AD accumulation in the BM of patients with T2D. Results highlight a feedback-loop mechanism by which mature BM-ADs from subjects with T2D (T2D BM-ADs) release high levels of MCP-1, which in turn stimulates the differentiation of BM-MSCs into new ADs. Importantly, in vivo antagonism of the MCP-1/CCR2 signaling in T2D mice not only decreased BMAT accumulation but also rescued LT-HSC depletion.

Patients undergoing hip replacement surgery were recruited under informed consent at the Avon Orthopaedic Centre, Southmead Hospital. The study protocol complied with the Declaration of Helsinki, was covered by institutional ethical approval (REC14/SW/1083 and REC14/WA/1005), and was registered as an observational clinical study in the National Institute for Health Research Clinical Research Network Portfolio, UK Clinical Trials Gateway, and ClinicalTrials.gov.

T2D was diagnosed according to the American Diabetes Association guidelines. Specifically, it was defined as 1) patient/referring doctor reports a previous diagnosis of diabetes, 2) HbA_1c >48 mmol/mol, and 3) not requiring insulin treatment for at least 12 months after the initial diagnosis. Exclusion criteria comprised acute disease/infection, immune diseases, current or past hematological disorders or malignancy, unstable angina, recent (within 6 months) myocardial infarction or stroke, critical limb ischemia, liver failure, dialysis due to renal failure, pregnancy, and lack of consent to participate. Available data on patients’ characteristics are provided in Supplementary Table 1.

BM samples were obtained from scooped femur heads remaining from hip replacement surgery. During the replacement procedure, the femoral head was removed with a saw, and the proximal femoral canal was opened with reamers and rasps. The BM displaced into the wound was scooped into a sterile pot with a curette. The superfluous cancellous bone was removed from the femoral neck and proximal metaphysis and placed with the marrow. The sample was decanted into a collection tube with 0.5 mol/L EDTA, pH 8. Only material that would otherwise be discarded was collected for study.

Bone samples were fixed in 1% paraformaldehyde (Thermo Fisher Scientific, Loughborough, U.K.), decalcified for 16 h with 10% formic acid, and embedded in paraffin. Blocks were sectioned on a rotary microtome at 2 μm, and the samples were stained with hematoxylin and eosin (H&E; Sigma-Aldrich, Gillingham, U.K.).

BM samples were stratified on Ficoll Histopaque 1077 (Thermo Fisher Scientific) and centrifuged without acceleration or brake at 300g for 30 min at 25°C. A total of 1 × 10⁷ mononuclear cells from the Ficoll separation was seeded in plastic flasks in α–minimum essential medium Eagle (αMEM) basal media (Thermo Fisher Scientific) supplemented with 20% FBS (Thermo Fisher Scientific) for 48 h at 37°C, 5% CO₂. The adherent cells were considered BM-MSCs and expanded in αMEM supplemented with 20% FBS.

Mature ADs from the floating phase after stratification were collected, strained through a 100-μm filter, and washed with Hanks’ balanced salt solution (Sigma-Aldrich). The cells were then cultured using a published method (28). Briefly, 1 × 10⁷ cells were seeded in a well of a 12-well culture plate (CELLSTAR; Greiner Bio-One, Cardiff, U.K.), which was filled to the brim with αMEM containing 20% FBS. A 22-mm borosilicate glass coverslip (VWR International, Leighton, U.K.) was added on top of the well. Mature ADs were incubated for 4 days until adhesion to the glass plate and then flipped and put inside a well of a six-well plate containing αMEM. ADs were then incubated 48 h in FBS-free αMEM. The conditioned media (CM) was then harvested and filtered through a 22-μm filter (Sigma-Aldrich). Unilocularity of ADs was evaluated by microscopy, and absence of foam-cell contamination was confirmed by negative expression of the macrophage markers CD11b and CD68 mRNA by quantitative PCR (qPCR) (Supplementary Fig. 1).

Chemokines were quantified using ELISA kits on 100 μL of CM, following the manufacturer’s protocol (Bio-Techne, Minneapolis, MN). Glycated and total hemoglobin was measured using ELISA kits (Generon, Slough, U.K.) using serum diluted 1:500.

RNA was purified using Tri-Reagent (Sigma-Aldrich) per the manufacturer’s protocol and repurified using acid-phenol phase separation. RNA was quantified using a NanoDrop (Thermo Fisher Scientific).

cDNA was generated using the High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific) and 100 ng total RNA following the manufacturer’s protocol. A T100 Thermal Cycler (Bio-Rad, Watford, U.K.) was used.

mRNA expression was assayed using real-time qPCR. The reaction was made with Power SYBR Green Master Mix (Thermo Fisher Scientific) and 200 nmol/L gene-specific primers (Supplementary Table 2) in a QuantStudio 6 Flex RT-PCR machine (Thermo Fisher Scientific).

Freshly isolated BM cells or cultured BM-MSCs were labeled with primary antibodies in staining buffer (Hanks’ balanced salt solution supplemented with 1% BSA) and acquired with an LSR Fortessa ×20 (BD Biosciences, Oxford, U.K.), and quantification of the antigenic profile was performed using FlowJo v10 software (FlowJo, LLC). Antibodies used are listed in Supplementary Table 3.

Adipose differentiation was induced by treatment with 1 μmol/L dexamethasone, 0.5 mmol/L isobutylmethylxanthine, 10 μg/mL insulin, and 1 μmol/L indomethacin (all from Sigma-Aldrich) in DMEM supplemented with 10% FBS for 21 days, changing the media every 3 days. Cells were fixed with 10% formaldehyde, treated with 60% isopropanol, and exposed to Oil Red O (ORO) 0.3% solution (all from Sigma-Aldrich) for 30 min and then washed with H₂O. Osteoblast differentiation was induced by treatment with 0.1 mmol/L dexamethasone, 50 μg/mL l-ascorbic acid 2-phosphate, and 10 mmol/L β-glycerophosphate (all from Sigma-Aldrich) in DMEM supplemented with 10% FBS for 14 days, changing the media every 3 days. Cells were fixed with ice-cold methanol (Sigma-Aldrich) and treated with a solution of 0.4 mg/mL nitro-blue tetrazolium chloride and 0.19 mg/mL 5-bromo-4-chloro-3’-indolyphosphate p-toluidine (both from Thermo Fisher Scientific) to assess alkaline phosphatase activity.

Experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals (The Institute of Laboratory Animal Resources, 1996) and with the approval of the University of Bristol and the British Home Office (license 30/2811). As a model of T2D and obesity, we used 4-week-old male obese leptin-receptor homozygous mutant C57BLKS/J-Lepr^db/Lepr^db/Dock7+ (Db/Db) mice (Charles River Laboratories, Harlow, U.K.). Age- and sex-matched lean heterozygous C57BLKS/J-Lepr^db/Lepr^WT Dock7+ (Wt/Db) mice served as controls. Animals were fed standard chow (Charles River Laboratories) and provided water ad libitum.

Mice were randomly allocated to receive 4 g/kg/day of the MCP-1 receptor antagonist RS504393 (Tocris Bioscience, Bristol, U.K.) diluted in DMSO in drinking water (1% of water volume) or the equivalent volume of DMSO for 8 weeks. Glucose was measured in urine using Diastix colorimetric reagent strips (Bayer, Reading, U.K.).

Mouse femurs were cut longitudinally. Pictures were taken by a Fast1384 Qicam camera (QImaging) using Infinity Analyze capture software. Cells were counted and analyzed using ImageJ software (National Institutes of Health).

Whole BM cells were obtained from the mouse tibia. The marrow was flushed from the bone using PBS and then subjected to red blood cell lysis. Cells were then washed and used for the appropriate experiment.

Values are presented as mean ± SEM. Two-tailed independent-samples t test was used to compare groups with T2D and without diabetes assuming unequal variances. One- and two-way ANOVA were used to compare four group experiments, followed by multivariate analysis to compare each group individually. Nonparametric sample groups were evaluated using the Kruskal-Wallis test. P values <0.05 were considered statistically significant.

We confirm patients with T2D show a remarkable remodeling of the BM at a cellular level, as described in another cohort evaluated previously (10). H&E staining shows a large increase in the area covered by ADs as compared with control subjects without diabetes (Fig. 1A and B). This effect is due to an increase in both AD number and size (Fig. 1C and D). The average AD size was 4,487 ± 675 and 7,812 ± 666 μm² in patients without diabetes and with T2D, respectively, corresponding to a 1.74-fold increase in the former group (P < 0.05).

We next investigated if adipogenic predetermination of BM-MSCs was the cause and source of AD accumulation in the BM of subjects with diabetes. We first measured the effect of T2D on the expression of a spectrum of genes implicated in adipogenesis (Fig. 2A). Results show that T2D induces an increase in the mRNA levels of proadipogenic genes CEBPa, PDGFRb, and IGFR1. In addition, T2D caused the downregulation of the antiadipogenic genes SIRT1 and GLI1. STAT5a, which can be pro- and antiadipogenic, was also downregulated in T2D BM-MSCs. FACS analysis of cultured BM-MSCs showed a twofold increase in PDGFRβ protein levels in the group with T2D, whereas the preadipocyte factor PREF-1 remained unchanged (Fig. 2B).

Next, we examined the effect of T2D on BM-MSC differentiation into ADs or osteoblasts. Quantification of ORO-positive cells shows an increase in the adipogenic activity of T2D BM-MSCs (Fig. 2Ci and ii). Real-time qPCR analysis also shows that T2D BM-MSCs express higher transcript levels of the AD differentiation markers PPARγ, ADIPOQ, and fatty acid–binding protein 4 (FABP4) (Fig. 2Ciii–v). These results indicate that T2D programs BM-MSCs with an enhanced propensity into adipogenesis. In contrast, T2D did not influence the osteogenic differentiation of BM-MSCs, as assessed by the alkaline phosphatase activity assay (Fig. 2Di and ii).

We next investigated the possibility that T2D BM-ADs directly interfere with BM-MSC differentiation, creating an imbalance in the typical lineage specification. BM-MSCs from subjects without diabetes were exposed to AD or osteoblast differentiation media, which were supplemented with the CM from cultured BM-ADs from subjects without diabetes (ND-CM) or BM-ADs from subjects with T2D (T2D-CM) or no CM as a control group (No CM). We observed that the T2D-CM increases the differentiation of BM-MSCs into ORO-positive ADs over the levels found in groups treated with ND-CM or No CM (Fig. 3Ai and ii). The enhanced activation of BM-MSC adipogenesis by the T2D-CM was associated with a greater induction of the ADIPOQ and PPARγ expression (Fig. 3Aiii and iv), whereas FABP4 was similarly upregulated among all groups (Fig. 3Av). These data suggest T2D causes an incremental feedback loop by which BM-ADs will increase MSC fate toward adipogenesis, which, in turn, will stimulate more BM-MSCs to do the same.

We next examined the effect of T2D on osteoblast differentiation. Osteoblastogenesis assays indicate that the ND-CM increases BM-MSC differentiation into alkaline phosphatase–positive osteoblasts. In contrast, this stimulatory effect was negated to the T2D-CM (Fig. 3Bi and ii). The same pattern was confirmed when assessing the mRNA expression of the osteoblast marker RUNX2 (Fig. 3Biii). Altogether, these results indicate that BM-ADs of patients with T2D promote a paracrine induction of BM-MSC adipogenesis at the expense of osteoblastogenesis.

To identify which factor secreted by BM-ADs is responsible for the regulation of BM-MSC adipogenic determination, we investigated by qPCR the mRNA expression of multiple adipokines and cytokines in freshly harvested human BM-ADs. We found that the adipokines LEP and RETN were upregulated in T2D as compared with BM-ADs from subjects without diabetes, whereas IGF1, PEDF, and ADIPOQ were downregulated (Fig. 4Ai). Cytokine expression analysis revealed a large increase in TNFa and MCP-1 levels in T2D BM-ADs as well as a modest increase in matrix metalloproteinase 2 (MMP2). We also observed that T2D reduced the levels of IL18, MIF, and angiopoietin 1 (ANGPT1) (Fig. 4Aii). We next verified if the observed transcriptional changes were associated with similar modifications of secreted proteins. ELISA of the AD CM confirmed that the levels of LEP and RETN were increased in T2D (Fig. 4Bi and ii), and ADIPOQ was decreased (Fig. 4Biii), whereas PEDF protein levels were comparable between the two groups (Fig. 4Biv). We observed a striking increase in the secretion of MCP-1 protein by T2D BM-ADs as well as a milder increase in MMP2 and interferon-γ (Fig. 4Bv–vii). Tumor necrosis factor-α was similar between the two groups (Fig. 4Bviii), and ANGPT1 was increased in the T2D BM-AD CM (Fig. 4Bix). Altogether, these data indicate T2D has a profound effect on the BM-AD secretome, with MCP-1 being the most prominently affected factor. This latter phenomenon is well conserved among species, as we have confirmed the elevated MCP-1 levels in the CM of BM-ADs from obese, diabetic Db/Db mice as compared with lean nondiabetic (ND) Wt/Db controls (Supplementary Fig. 2).

MCP-1 and its receptor CCR2 have been implicated in obesity through the promotion of AD differentiation and consequent increase in adipose tissue mass (26,29). However, it remains unknown if MCP-1 plays a role in fat accumulation within the BM of patients with T2D. To verify this possibility, we performed adipogenesis assays on human BM-MSCs stimulated with No CM, ND-CM, or T2D-CM as in the previous experiment, but with the addition of the MCP-1 receptor antagonist RS504393 or its vehicle. Inhibition of MCP-1 signaling moderately decreased BM-MSC adipogenesis in the No-CM and ND-CM groups (Fig. 5A and B). In the T2D-CM group, in which adipogenesis was increased, MCP-1 antagonism resulted in a striking inhibition to levels even lower than the No-CM group (Fig. 5A and B). These phenomena were associated with consensual changes in the expression of AD markers. The antagonist reduced the ADIPOQ mRNA expression in BM-MSCs exposed to No CM or ND-CM and completely abrogated the inductive action exerted by the T2D-CM (Fig. 5C). Second, RS504393 selectively inhibited the T2D-CM–induced upregulation of PPARγ (Fig. 5D). Third, the antagonist downregulated FABP4 mRNA levels in all experimental groups (Fig. 5E). Fourth, the antagonist inhibited MCPIP, which was selectively upregulated in BM-MSCs following exposure to T2D-CM (Fig. 5F). This latter result confirms the effective inhibition of molecular signaling downstream to the MCP-1/CCR2 duo.

We posited that in vivo administration of the same antagonist would result in a decrease in BM-AD accumulation. To validate this hypothesis, we exposed T2D Db/Db mice to either 4 mg/kg/day of RS504393 (DR group) or DMSO (DD group) in drinking water for 8 weeks. We established a third group of ND Wt/Db (wild-type [WT]) to receive DMSO and be used as a reference group. During the 8-week experiment, no significant differences in weight were observed between the DD and DR groups (Supplementary Fig. 3). In contrast, we observed that RS504393 treatment had an inhibitory effect on body weight gain (Fig. 6Ai) without altering food intake (Supplementary Fig. 3). We found this lowered growth rate associated with selective changes in specific fat deposits. Although the weight of either pericardial or inguinal fat was similar in DD and DR groups, RS504393 treatment resulted in a reduction of the epididymal fat weight (Fig. 6Aii). Histology of the epididymal fat shows that the DR-treated mice exhibited smaller ADs compared with the DD-treated mice (Fig. 6Aiii and iv). This indicates that the antagonist selectively reduces epididymal fat accumulation by acting on AD hypertrophy. Interestingly, we also observed RS504393 treatment ameliorated the metabolic control in T2D mice as indicated by mitigation of glycosuria (Fig. 6Bi) and reduction of blood glycated hemoglobin levels (Fig. 6Bii).

After 8 weeks of treatment, when mice were 12 weeks of age, AD coverage at the metaphysis of the femur averaged 4% of the BM space in the ND WT group and 85% in the T2D DD group (Fig. 7Ai and ii). RS504393 treatment blunted the effect of T2D on fat encroachment, lowering the total area covered by ADs from 85 to 73% (Fig. 7Ai and ii). The decreased BMAT coverage was not associated with differences in the average AD size (Fig. 7Aiii), although the DR group contained more small-size ADs (1–10 μm²), which, in white adipose tissue, are considered metabolically healthy compared with large ADs (Fig. 7Aiv). We also observed that the antagonist caused a decrease in the number of ADs (Fig. 7Av). This decrease in number in the absence of changes in average size suggests that MCP-1 most likely affects the differentiation of BM cells into ADs in situ.

To assess if the observed changes in BM-AD accumulation correspond to consensual modifications in adipogenesis markers in BM, we quantified the mRNA expression of FABP4, ADIPOQ, and PLIN1 (Fig. 7Bi–iii). We observed an increase in the mRNA levels of the three markers in the untreated T2D DD group, which correlated with the increased AD presence in the BM of these mice. In contrast, antagonism of MCP-1 signaling decreased the levels of these markers, in line with the observed AD decrease. Analysis of murine Mcp-1 mRNA levels in whole BM did not show any difference among the three groups (Fig. 7Ci). In contrast, Mcpip mRNA levels were increased in the DD group as compared with the WT group, with this increase being inhibited by RS504393 in the DR group (Fig. 7Cii). Serum and cell-free BM lavage concentrations of MCP-1 were elevated in T2D DD mice compared with WT controls, with no differences observed between the DD and DR groups (Fig. 7Ciii and iv).

The cortical bone thickness of the tibia was reduced by 50% in T2D mice regardless of treatment (Supplementary Fig. 4Ai and ii). The density of osteocytes embedded inside the bone was also reduced in the DD group compared with the WT group, with this reduction being blunted by RS504393 in the DR group (Supplementary Fig. 4Aiii). We also observed a twofold increase in osteoclast density in DD and DR mice with a corresponding decrease in osteoblast as compared with WT (Supplementary Fig. 4B and C).

Circulating levels of COOH-terminal telopeptide, a marker of bone resorption, did not change among all groups, whereas RANKL and OPG levels were decreased in both groups with diabetes regardless of treatment compared with those without diabetes, resulting in an unchanged RANKL to OPG ratio (Supplementary Fig. 4Di–iv). Sclerostin, a circulating marker of osteocyte activity, was reduced in the DD group compared with ND, with this reduction being attenuated by the MCP-1 antagonist (Supplementary Fig. 4Dv).

HSCs are significantly affected both regarding amount and functionality by T2D (10,30), a result that was confirmed in the present investigation. We further analyzed by flow cytometry whether RS504393 could influence HSC availability in the BM of T2D mice. We defined BM-HSCs as lineage-negative (Lin^neg), cKit^pos, and Sca-1^pos cells. LT-HSCs were characterized as Lin^negcKit^posSca-1^posCD34^neg, whereas ST-HSCs as Lin^negcKit^posSca-1^posCD34^pos (Fig. 8Ai). The percentage of HSCs in the BM of the T2D mice was lower than that of the ND WT group (Fig. 8Aii), with no difference between the DD and DR groups. When considering HSC subpopulations according to their maturity, we found that T2D reduced the relative abundance of LT-HSCs (Fig. 8Aiii) while increasing the abundance of ST-HSCs (Fig. 8Aiv). This unbalance was corrected by treatment with the MCP-1 antagonist in the T2D DR group (Fig. 8Aiii and iv), suggesting that T2D impinges upon the HSC fate determination and that MCP-1 inhibition blocks this effect. Stem cell factor (SCF) has been shown to be secreted by BM-ADs and able to support hematopoiesis (15). However, the levels of SCF in the cell-free BM lavage were similar among the studied groups (Fig. 8Av). To verify if MCP-1 acts directly on HSCs, we exposed BM mononuclear cells to the CM of BM-ADs from subjects without diabetes or T2D BM-ADs for 96 h and analyzed the antigenic phenotypes of the HSC population by flow cytometry (Fig. 8Bi). Compared with the No-CM group, both the ND-CM and T2D-CM groups showed a slight but significant decrease in total HSCs (Fig. 8Bii) and LT-HSCs (Fig. 8Biii), which was associated with a corresponding increase in ST-HSC (Fig. 8Biv). Cotreatment with RS504393 did not affect LT- and SC-HSC levels compared with the vehicle while still inhibiting MCPIP expression (Fig. 8Bii–v).

In this study, we provide novel extensive evidence about the cellular and molecular mechanisms underpinning fat accumulation in the BM of patients with T2D and also show BMAT could adversely influence the BM microenvironment by paracrine signaling to neighbor cells.

In animal models of obesity, AD hypertrophy precedes hyperplasia in response to the increase in calories requiring storage (31). These dynamic changes are hard to monitor in real time in patients. Our study captured an advanced stage of BM remodeling in subjects with T2D, in whom a twofold accumulation of BMAT was associated with increased AD size and density. Paracrine factors released by enlarged ADs may induce neighboring pre-ADs to grow and differentiate, strengthening the adipose tissue energy reservoir. In line with this, we observed BM-ADs paracrinally stimulate the intrinsic capacity of the BM-MSC pool to differentiate into new ADs. In theory, a reduced clearance may also contribute to expanding the AD pool. However, evidence indicates the canonical apoptotic molecular machinery is upregulated in the visceral adipose tissue of patients with T2D (32,33). Therefore, apoptosis may not be a good explanation for enhanced adiposity in our setting. More likely, hyperglycemia and oxidative stress, which we showed to be activated in BM from subjects with diabetes (10,30), could be responsible for a dysregulated differentiation of precursor cells into ADs. A recent report has attributed a pathogenic role to reactive oxygen species in peripheral obesity (34). Moreover, reactive oxygen species are acknowledged to facilitate differentiation of pre-ADs into mature ADs by accelerating mitotic clonal expansion (35,36). In line with this, we show that BM-MSCs of patients with T2D have an increased intrinsic adipogenic potential, without changing their osteogenic potential. Undifferentiated T2D BM-MSCs show upregulated expression of proadipogenic factors and downregulation of antiadipogenic factors at steady state. C/EBPα is central for AD differentiation and lipid accumulation (37). Hence, high levels of C/EBPα suppose a priming of BM-MSCs toward adipogenesis. PDGFRβ is pivotally implicated in the control of adipogenesis, as supported by lineage-tracing experiments in mice showing PDGFRβ-positive mural stem cells are the originators of adipose tissue hyperplasia under high-fat diet conditions (38). Our finding of a twofold increase in PDGFRβ-positive BM-MSCs of patients with T2D further supports a readiness toward fat accumulation.

Using an in vitro approach, we showed that the secretome of T2D BM-ADs could bolster the fate commitment of naive BM-MSCs into ADs. This phenomenon creates a vicious cycle of fat accumulation in the BM from subjects with diabetes, as the appearance of mature ADs will send more adipogenic signals to the BM-MSCs, which will differentiate into more ADs. Furthermore, our findings suggest that it is not the BMAT volume that negatively modulates osteogenesis, but the pathophysiological status of the adipose tissue itself. For example, secretion of adiponectin by BM-ADs (39), which acts positively on osteoblast differentiation, is downregulated in patients with T2D. This decrease could explain why AD CM from subjects without diabetes promotes osteoblastogenesis, whereas T2D AD CM does not.

In line with our results, previous studies have shown that high glucose and hypoxia can induce MCP-1 expression in ADs and other cellular systems via reactive oxygen species (36,40). Analysis of AD transcripts and secretome showed the upregulation of LEP and RETN and downregulation of ADIPOQ in human BM-ADs, which follows the patterns observed in other white adipose deposits of patients with T2D (41). LEP reportedly induces adipogenesis and inhibits osteoblastogenesis in the BM (42). RETN is also relevant to bone remodeling, as it activates osteoclast differentiation (43). Moreover, BMAT is a major source of ADIPOQ (12). Hence, the decrease in ADIPOQ production may contribute to the decrease in circulating levels of ADIPOQ previously reported in patients with T2D (44). MMP2 could also participate in the positive-feedback loop of adipogenesis in the BM of individuals with T2D, possibly through a previously reported in vitro action on pre-AD differentiation (45).

We decided to conduct a more in-depth investigation on MCP-1 for several reasons. First, we know from the literature that MCP-1 can induce AD differentiation, osteoclastogenesis, and bone reabsorption via the transcription factor MCPIP (25,26,46) and that high levels of circulating MCP-1 are associated with insulin resistance through induction of an inflammatory response in adipose tissue (21). Second, we found that, within a spectrum of chemokines and growth factors differentially modulated in T2D BM-ADs, MCP-1 is upregulated by 10-fold as compared with counterparts without diabetes. Third, we demonstrated that this phenomenon is conserved among species, as MCP-1 levels are augmented in BM-ADs of T2D Db/Db mice. Nevertheless, no previous study has examined the possibility that BM-AD–produced MCP-1 drives the conversion of MSCs into new ADs in diabetes. We verified this possibility by performing an adipogenic induction assay on naive BM-MSCs stimulated with BM-AD CM in the presence of RS504393, an MCP-1 receptor antagonist, or its vehicle. Results indicate RS504393 inhibits the transcription of the adipogenic markers ADIPOQ, PPARγ, and FABP4. Importantly, in line with our hypothesis, this molecular response was associated with the complete abolition of the T2D BM-AD secretome’s ability to induce BM-MSC differentiation into mature ADs.

We next performed an in vivo validation study in T2D Db/Db mice, with the primary objective of demonstrating MCP-1 antagonism can halt the accumulation of BM-ADs. Systemic RS504393 for 8 weeks resulted in the inhibition of MCP-1 downstream effector MCPIP, downregulation of adipogenic factors FABP4, ADIPOQ, and PLIN1, and, last but not least, amelioration of several pathological features, including a significant effect on the designed primary end point. BMAT reduction by RS504393 was mainly due to a lowered number of ADs, which was associated with a shift toward smaller ADs. This suggests that MCP-1 antagonism not only limits adipogenesis but also promotes the formation of healthier and metabolically active ADs (47). Investigation of BM-AD content by osmium tetroxide and micro computed tomography imaging may allow a more direct approach to visualize global changes of adiposity throughout the bone (48).

Similar to the study published by Kang et al. (23), we observed a slower weight gain in treated Db/Db mice, which was reflected by a lowered epididymal fat pad weight through a reduction in AD size. In addition, RS504393 treatment reduced urinary glucose and plasma HbA_1c levels, thus suggesting MCP-1 receptor antagonism may benefit metabolic control. Further investigation is necessary to determine if reduced BM adiposity has contributed in improving insulin resistance in our experimental model.

About the status of the bone, a secondary end point of our in vivo study, we found that the cortical bone thickness of Db/Db mice was half as thin as that of control ND Wt/Db mice, and RS504393 could not rescue this phenotype. However, we did observe changes in bone-regulating osteocytes. In untreated Db/Db mice, the density of osteocytes was greatly reduced as compared with the ND Wt/Db mice. Osteocytes are responsible for bone mechano-sensing, homeostasis, and remodeling, as they coordinate the activity of osteoclasts and osteoblasts (49). It has been shown that a reduction of osteocyte lacunae in older patients is associated with defects in bone remodeling and fracture healing (50,51). MCP-1 antagonist treatment of Db/Db mice increased osteocyte density in the femoral bone to an intermediate level between ND Wt/Db and untreated Db/Db, possibly suggesting this may be a viable strategy to improve bone health in T2D. Sclerostin is produced by the osteocytes, and its levels in the circulation followed the changes in osteocyte density, decreasing in T2D and being partially rescued by RS504393. However, these changes could be counterintuitive, as sclerostin has antianabolic effects on bone formation through inhibition of the Wnt signaling (52). Also, we could not observe any benefit on various cellular and molecular end points of bone formation and degradation, including osteoblast and osteoclast density, RANKL to OPG ratio, and circulating levels of COOH-terminal telopeptide, a specific marker for the degradation of mature type I collagen, which is elevated in patients with increased bone resorption.

We showed a decrease in total HSC levels in the two groups with T2D treated with the MCP-1 antagonist or its vehicle as compared with those without diabetes. These data confirm the depletive action of diabetes and also suggest that MCP-1 does not directly contribute to HSC rarefaction. However, when looking at the abundance of specific subpopulations, we found that MCP-1 antagonism rescued the T2D-associated primitive LT-HSC deficit, restoring the levels observed in ND mice. This was contrasted by the lack of effect of RS504393 in an in vitro assay, in which the antagonist failed to inhibit the effect AD CM on the relative abundance of CD34^pos and CD34^neg HSCs. Altogether, these findings suggest that secreted factors other than MCP-1 are responsible for the paracrine influence of T2D BM-ADs on HSC lineage progression and that treatment with RS504393 corrects the proportion of LT- and SC-HSCs indirectly through amelioration of general and local conditions, such as improved metabolic control and inhibition of BM adipogenesis. Reduced AD presence in the BM signifies a decrease in secretion of factors that negatively affect HSC maintenance, thus permitting the recovery of proper primitive LT-HSC levels. In line with this, BM reconstitution studies showed that inhibition of adipogenesis is beneficial for hematopoietic tissue recovery (53). In contrast, other studies suggest that BMAT is instrumental to HSC maintenance, mainly via the secretion of SCF (15). In this regard, we could not find any significant change in SCF levels in the cell-free murine BM lavage among studied groups.

In conclusion, this study highlights a key paracrine mechanism by which adipose tissue accumulates in the BM of subjects with diabetes and perpetuates BM-MSC adipogenic differentiation. This study demonstrates for the first time that MCP-1 inhibition provides a means for protection of the BM microenvironment in diabetes, reducing BMAT volume and AD size and supporting hematopoiesis. These results have important clinical and therapeutic implications for the maintenance of the BM integrity. We suggest that clinical MCP-1 receptor antagonism could pave the way to novel treatments of T2D and associated comorbidities.

Acknowledgments. The authors thank Dr. Andrew Herman, Lorena Sueiro Ballesteros, and the University of Bristol, Faculty of Biomedical Sciences Flow Cytometry Facility for assistance. The authors also thank the Italian Leukemia-Lymphomas Association, a section of Treviso, for the valid contribution to the project.

Funding. This study was supported by British Heart Foundation grant RG/13/17/30545, “Unravelling mechanisms of stem cell depletion for the preservation of regenerative fitness in patients with diabetes,” and Italian Ministry of Health, Istituto di Ricovero e Cura a Carattere Scientifico MultiMedica Ricerca Corrente and grant RF-2011-02346867.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.F.-M. contributed to the article by establishing the hypothesis and research protocol of this study, generating the data of Figs. 2–8, as well as drafting and correcting the manuscript. D.M. contributed to the article by generating the data of Fig. 1, correcting the manuscript, and participating in the scientific discussion on the structure and data of the paper. G.S. and M.S. contributed to the article by providing human BM histology samples, correcting the manuscript, and participating in the scientific discussion on the structure and data of the paper. N.S. and A.B. contributed to the article through surgical expertise and collection of human BM. P.M. is the principal investigator of the research project and contributed to the article by obtaining British Heart Foundation funding, generating ideas, participating in the scientific discussion on the structure and data of the paper, participating in the drafting and correction of the article, as well as providing invaluable mentoring to D.F.-M. D.F.-M. and P.M. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in poster form at the Keystone Symposia on Molecular and Cellular Biology: Bioenergetics and Metabolic Disease (J4), Keystone, CO, 21–25 January 2018.