Pharmacologic PPAR-γ Activation Reprograms Bone Marrow Macrophages and Partially Rescues HSPC Mobilization in Human and Murine Diabetes

Studies performed in experimental and human diabetes have consistently shown an impaired mobilization of hematopoietic stem/progenitor cells (HSPC) from the bone marrow (BM) to peripheral blood (PB) after stimulation with granulocyte-colony stimulation factor (G-CSF) (1). Such mobilization defect precedes the reduction of circulating HSPC (2,3), which, in turn, is a driver of micro- and macrovascular complications (4,5). On this basis, it has been hypothesized that restoring normal HSPC levels could exert protective effects against the future risk of complications (6,7). It is therefore of interest to identify therapeutic strategies to counter the so-called “diabetic stem cell mobilopathy” (8).

We have previously described a molecular pathway contributing to mobilopathy. Hyperglycemia stimulates myelopoiesis by BM progenitors (9), thereby expanding proinflammatory (M1) macrophages, which secrete excess amounts of oncostatin M (OSM) (10). In turn, OSM signals in BM mesenchymal stem/stromal cells (MSC) and induces the expression and release of CXCL12 (10). This effect requires cytosolic p66Shc, serving as an adaptor protein for OSM receptor (OSMR) and leading to Cxcl12 expression (11). CXCL12 is the major regulator of HSPC traffic in and out the BM: high levels of CXCL12 keep HSPC in their niches (12), whereas mobilizing stimuli act by reducing intramarrow CXCL12 concentrations (like G-CSF [13]) by desensitizing cells to CXCL12 (like plerixafor [14]) or by inverting the gradient of CXCL12 from the BM to PB (dipeptidyl peptidase 4 [DPP4] inhibitors [15]). The observation that DPP-4 inhibitors and plerixafor mobilize HSPC in diabetes (7) supports the concept that an ineffective CXCL12 switch causes diabetes-associated G-CSF unresponsiveness (1). In addition, the BM of diabetic rodents and patients with diabetes undergoes extensive remodeling, including adipocyte accumulation (8,16,17). BM adipocytes contribute to forming the HSPC niche and their expansion after BM injury favors hematopoietic reconstitution (18). Adipocytes are a source of CXCL12 and other niche gene products capable of retaining HSPC in the niche and preventing their mobilization (18–20).

Since OSM inhibition or deletion reduced BM Cxcl12 expression and rescued HSPC mobilization in diabetic mice (10,11), and given that proinflammatory M1 macrophages selectively produce OSM (10), we herein tested a pharmacologic strategy of macrophage reprogramming to counter mobilopathy. To this end, we took advantage of prior observations that PPAR-γ promotes M2 macrophage polarization (21). As PPAR-γ prominently regulates adipose tissue biology (22), we also explored the interplay among the OSM-p66Shc pathway, BM adipogenesis, and HSPC mobilization.

C57BL/6J wild-type (Wt) mice were purchased from The Jackson Laboratory and established as a colony since 2001. p66Shc^−/− mice were obtained from Pier Giuseppe Pelicci’s laboratory (European Institute of Oncology, Milan, Italy); a colony was established in 2010 and backcrossed on the C57BL/6J background for >10 generations. Osm^−/− mice on the C57BL/6J background were obtained from GSK (Stevenage, U.K.), and a colony was established since 2015. p66Shc^−/−Osm^−/− double-knockout (DKO) mice were generated by crossing p66Shc^−/− mice and Osm^−/−. For all the experiments, 2- to 4-month-old male mice were used; whenever female mice were included, the male:female ratio was 1:1 in all groups of each experiment. Diabetes was induced in 2-month-old animals by a single intraperitoneal injection of streptozotocin (STZ) at 175 mg/kg in 100 mmol/L Na-Citrate buffer, pH 4.5. Blood glucose was measured using a glucometer (FreeStyle; Abbott, Abbott Park, IL). HSPC mobilization was induced by injecting animals subcutaneously with 200 μg/kg/die human recombinant G-CSF daily for four consecutive days. Pioglitazone (DOC Generici, Milano, Italy) was given to animals by daily gavage at 20 mg/kg for 4 weeks. Diabetic mice were treated with pioglitazone since the onset of diabetes. For evaluation of in vivo OSM effects, 3-month-old male mice were injected subcutaneously with PBS or carrier-free recombinant mouse OSM (495-MO/CF; R&D Systems, Minneapolis, MN) at 0.5 mg per injection every 6 h for 48 h before analysis. Total WBC count was performed using the CELL-DYN Emerald hematology analyzer (Abbott) on fresh EDTA-treated mouse blood.

Assignment of mice to treatments or experimental groups was always determined randomly, based on a computer-generated sequence. All animal studies were approved by the Animal Care and Use Committee of Venetian Institute of Molecular Medicine and by the Italian Health Ministry.

Patients with diabetes and individuals without diabetes were recruited at the University Hospital of Padova. The study for G-CSF–induced mobilization was approved by the local ethics committee, was conducted in accordance with the Declaration of Helsinki as revised in 2000, and is registered at ClinicalTrials.gov (NCT01102699). This was a prospective, parallel-group study of direct BM stimulation with human recombinant G-CSF in subjects with and subjects without diabetes. Specific methods for quantifying blood cells and HSPC have previously been published (2). Informed consent was obtained from all human participants.

Flow cytometry was performed on murine BM or EDTA-treated PB. BM cells were isolated by flushing femurs and tibias with ice-cold MACS Separation Buffer (Miltenyi Biotec, Gladbach, Germany) through a 40 μmol/L cell strainer. PB or BM cells (100 μL) were incubated with antibodies for 15 min in the dark at room temperature. After red blood cell lysis, samples were resuspended in 200 μL PBS, and data were acquired with a FACSCanto (BD Biosciences) cytometer followed by analysis using FlowJo software (Tree Star, Inc.). Antibodies for Lin⁻c-kit⁺Sca-1⁺ (LKS) cells were anti–Lineage Cocktail, Pacific Blue (clone, B255716; BioLegend); anti–CD117 (c-Kit)-FITC (clone, 2B8; Thermo Fisher Scientific, Waltham, MA); and anti–Ly-6A/E (Sca-1)-PE (clone, D7; Thermo Fisher Scientific). BM macrophages were stained with the following antibodies: anti–Ly-6G (Gr-1)-PE (clone, RB6-8C5; eBioscience), anti-CD115 (c-fms)–Alexa Fluor 488 (clone, AFS98; Thermo Fisher Scientific), and F4/80 APC-eFluor 780 (clone, BM8; Thermo Fisher Scientific).

Human peripheral blood mononuclear cells (PBMCs) were isolated from buffy coat from healthy donors by Histopaque-1077 (Sigma-Aldrich) density gradient centrifugation followed by a high-density hyperosmotic Percoll (GE Healthcare) gradient. PBMCs were seeded at 5 × 10⁴ cells/cm² in RPMI 1640 medium (Corning) containing 10% FBS, 2 mmol/L glutamine, penicillin-streptomycin, and 20 nmol/L recombinant human macrophage–colony stimulating factor (M-CSF; ImmunoTools) for 7 days to obtain resting (M0) macrophages. Medium was changed every 3 days. Polarization stimuli were added for 48 h as follows: for M1 macrophages with lipopolysaccharide (1 μg/mL) (Sigma-Aldrich) and rh interferon-γ (rh IFN-γ [10 ng/mL]) (ImmunoTools) and for M2 macrophages with rh interleukin-4 (rh IL-4 [20 ng/mL]) (ImmunoTools) and rh IL-13 (5 ng/mL) (ImmunoTools). Pioglitazone (10 μmol/L) (Sigma-Aldrich) was added 16 h before polarization and maintained throughout it. Conditioned medium was obtained by adding serum-free medium without stimuli for 48 h.

BM cells were obtained by flushing with sterile ice-cold PBS both femurs and tibia of 3-month-old mice. Red blood cells were lysed with ammonium-chloride-potassium lysing buffer. To obtain resting (M0) macrophages, we plated 150,000 cells/cm² on 6-well plates (BD Falcon, New York, NY) with RPMI 1640 medium (Corning) supplemented with glutamine 2 mmol/L, penicillin-streptomycin, and 10% FBS + 10 nmol/L murine M-CSF for 7 days without medium change. For M1 macrophages, cells were incubated 48 h with lipopolysaccharide (1 μg/mL) and mouse recombinant IFN-γ (10 ng/mL). For M2 polarization, resting macrophages were incubated with IL-4 (20 ng/mL) and IL-13 (5 ng/mL). Pioglitazone (10 μmol/L) (Sigma-Aldrich) was added 16 h before polarization and maintained throughout it. Conditioned medium was obtained by adding serum-free medium without stimuli for 48 h.

Murine BM-Derived MSC (BM-MSC) were isolated by flushing the BM of 3 month-old-mice and cultivating adherent cells with minimum essential medium (MEM)-α medium containing 10% FBS, glutamine 2 mmol/L, and penicillin-streptomycin. Passage 3–6 was used in all experiments. For gene expression analysis, cells were treated with murine recombinant oncostatin-M (R&D Systems) for 48 h in serum-free media.

Human BM-MSC were obtain from the BM of patients undergoing orthopedic surgery at the University Hospital of Padova (ethics committee protocol no. 2868P). Aspirates were washed twice with ice-cold sterile PBS, and the cell pellets were plated on tissue culture petri dishes (BD Falcon) with complete mesenchymal medium (MesenCult Mesenchymal Stem Cell Stimulatory Supplement (human); STEMCELL Technologies Inc.). The medium was changed when fibroblast-like cells began to appear and then every other day. Experiments were performed with MCS up to passage 5.

Femurs were fixed in 4% paraformaldehyde, decalcified, and cut in longitudinal sections. Sections were either stained with hematoxylin-eosin (H-E) or processed for immunofluorescence. Hoechst was used to visualize cellularity, whereas an anti-perilipin antibody identified adipocytes. In separate experiments, adipocytes were isolated as previously described (23). For further detail, see Supplementary Material.

Murine BM-MSC were seeded into 12-well plates (BD Falcon) and grown with MEM-α medium containing 10% FBS, glutamine 2 mmol/L, and penicillin-streptomycin until a confluent layer was reached. Adipogenic differentiation was induced by incubating the cells for 8 days with DMEM/F12 (1:1) medium (Sigma-Aldrich) containing 10% FBS, glutamine 2 mmol/L, and penicillin-streptomycin; 170 μmol/L pantothenic acid (MP Biomedicals); 33 μmol/L biotin (MP Biomedicals); 66 nmol/L insulin (Sigma-Aldrich); 0.25 mmol/L 3-isobutyl-1-methylxanthine (Sigma-Aldrich); 100 nmol/L dexamethasone (Sigma-Aldrich); 1 nmol/L 3,3′,5-triiodo-l-thyronine; Sigma-Aldrich); and 10 μmol/L rosiglitazone. Medium was changed every 3 days. For the oil red O (ORO) staining, cells were fixed with 4% paraformaldehyde for 10 min and after a rinse with PBS were stained for 15 min at room temperature with a 0.5% solution (w/v) of ORO in isopropanol. ORO was eluted by adding 100% isopropanol, and absorbance was read at 540 nm. Alternatively, after fixation cells were stained with 10 μmol/L BODIPY 493/503 (Invitrogen, Milano, Italy) for 30 min and counterstained with H33342 (Sigma-Aldrich) for 5 min.

RNA was isolated from flushed BM or cells by use of QIAzol or with a Total RNA Purification Micro Kit (Norgen Biotek) and quantified with a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific). cDNA was synthesized from 500 ng RNA using a SensiFAST cDNA Synthesis Kit (Bioline, London, U.K.). Quantitive PCR was performed using SensiFAST SYBR Lo-ROX Kit (Bioline) via a QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific). A list of primers can be found in Supplementary Table 1.

GSE27017 data sets (24) were browsed using the open-source GEO2R software. After defining two groups that contained BM adipocyte sample data and epididymal adipocyte sample data, respectively, we extracted and analyzed expression values of selected genes.

Continuous data are expressed as mean ± SE unless otherwise specified, whereas categorical data are presented as percentage or as fold of basal. Normality was checked using the Kolmogorov-Smirnov test, and nonnormal data were log transformed before analysis. Comparison between two or more groups was performed using the Student t test and ANOVA for normal variables or the Mann-Whitney U test and Kruskal-Wallis test for nonnormal variables that could not be log transformed (e.g., because of frequent zero values). All tests were two tailed. Bonferroni adjustment was used to account for multiple testing. Biological replicates are shown for every experiment. Whenever there were <7 of 10 replicates, individual data points are shown. Significance was conventionally accepted at P < 0.05. Number of biological replicates and significance are reported in the figure legends.

The data sets generated during or analyzed during the current study are available from the corresponding author upon reasonable request.

We first tested whether PPAR-γ activation modified BM macrophage phenotype and reduced OSM signaling. We found that, in vitro, the antidiabetes PPAR-γ agonist pioglitazone reduced expression of the M1 gene Tnfa by ∼50% and increased that of the M2 gene Mrc ∼4× in murine BM-derived macrophages, concomitantly reducing Osm expression in M1-polarized macrophages (Fig. 1A). In contrast to the conditioned medium of control M1 macrophages, the condition medium of pioglitazone-treated M1 macrophages failed to upregulate Cxcl12 in murine BM-MSC (Fig. 1B and C). The surge in OSM protein concentration found in the conditioned medium of M1 macrophages was significantly blunted by pioglitazone (Fig. 1D), suggesting that reduction of OSM in the medium was responsible for blocking Cxcl12 induction. Indeed, the conditioned medium from Osm^−/− BM macrophages, treated with either vehicle or pioglitazone, failed to upregulate Cxcl12 in BM-MSC (Fig. 1E).

To verify the relevance of this effect in vivo, we used mice with type 1–like diabetes induced by STZ. This model was chosen for two reasons: 1) mobilopathy is complete and severe in STZ-induced diabetes, while it is milder in mice with high-fat diet–induced type 2–like diabetes (Supplementary Fig. 1), and 2) pioglitazone is expected to induce less confounding metabolic effects in STZ than in high-fat diet mice. We treated diabetic and nondiabetic mice with pioglitazone for 4 weeks. In diabetic mice, PPAR-γ activation nonsignificantly reduced glucose levels (361 ± 21 vs. 412 ± 25 mg/dL; P = 0.44) and did not affect body weight (Fig. 2A and B), indicating that pioglitazone exerted no substantial systemic metabolic effects in this model. Successful PPAR-γ stimulation in the BM was confirmed by transactivation of typical target genes (Plin1, 2×, and Abca1, 3×), transrepression of the M1 macrophage gene Il1b (0.4×), and induction of M2 macrophage genes Cd36 (>40×) and Mertk (5×) (Fig. 2C). Treatment of diabetic mice with pioglitazone did not improve features of enhanced myelopoiesis associated with diabetes, including the raise in granulocyte-to-lymphocyte ratio in PB and the increased number of macrophages in the BM (Fig. 2D and E). However, pioglitazone restored normal suppression of BM macrophages (Fig. 2F) and rescued the Cxcl12 switch (70% suppression; P = 0.015) (Fig. 2G) after G-CSF in diabetic mice. Suppression of OSM and CXCL12 levels in the BM plasma by pioglitazone was confirmed at the protein level (Fig. 2H). While nondiabetic mice promptly mobilized HSPC approximately eightfold after G-CSF, diabetic mice treated with vehicle were fully unresponsive to G-CSF (Fig. 2I). On the contrary, diabetic mice treated with pioglitazone showed a partial rescue of HSPC mobilization (4.1×).

By staining BM sections for perilipin, we observed that, in both diabetic and nondiabetic mice, treatment with pioglitazone increased BM adipocytes (Fig. 3A and B). This effect is known to occur also in humans (25,26) and is consistent with the ability of PPAR-γ to promote adipogenic differentiation of MSC (27). Importantly, STZ-induced diabetes per se resulted in accumulation of BM adipocytes (Fig. 3C and D). Average adipocyte area was not affected, ruling out that increased perilipin staining was due to hypertrophy (Fig. 3E). Excess adipocytes are believed to contribute to BM remodeling and HSPC mobilopathy in diabetes, owing to the expression and release of CXCL12 (19) and other paracrine factors (e.g., MCP-1) (28). In fact, by immunofluorescence staining of the BM, we found an intact CXCL12 signal close to adipocytes (Fig. 3F). Also, Cxcl12 gene expression in freshly isolated BM adipocytes (BMAds) was markedly higher than in the whole BM (Fig. 3G). Since Cxcl12 expression in BM-MSC is controlled by OSM through p66Shc (11), we verified whether the same was in adipocytes. Using an ex vivo organoid culture system (Fig. 3H), we found that stimulation of visceral adipose tissue (VAT) with OSM induced Cxcl12 expression in Wt (approximately threefold; P = 0.03) but not in p66Shc^−/− cells (Fig. 3I). As origin and properties of BM adipose tissue differ from those of VAT (24), we measured Cxcl12 expression after treating mice with recombinant OSM or vehicle: while OSM induced Cxcl12 in the unfractioned BM, the very high Cxcl12 expression in freshly isolated BMAds could not be further increased by OSM (Fig. 3J). To explain this site-specific effect of OSM, we mined publicly available gene expression profiles comparing BMAds with peripheral adipose tissue: this analysis confirmed higher expression of Cxcl12 in BMAds and showed that BMAds had significantly lower gene expression for OSMR and its cognate gp130 and LIF receptors than VAT (Supplementary Fig. 2). This finding can explain why BMAd Cxcl12 expression could not be stimulated by OSM. However, consistent with the notion that OSM drives osteogenic at the expense of adipogenic differentiation of MSC (29), we found that Osm^−/− mice displayed strikingly increased amounts of BM adipocytes compared with Wt mice either in the diabetic or nondiabetic condition (Fig. 3K and L). Immunostaining signal for intact CXCL12 around BM adipocytes was diminished by ∼40% in Osm^−/− versus Wt mice (P = 0.013) (Fig. 3M and N), and Cxcl12 gene expression was lower in freshly isolated BMAds of Osm^−/− mice compared with Wt mice (Fig. 3O). Thus, despite the fact that Cxcl12 expression was very high in BMAds and could not be further stimulated by OSM, these data still support that OSM can control Cxcl12 expression in BMAds as in MSC.

We reasoned that suppression of Cxcl12 in BMAds could contribute to the rescued HSPC mobilization that we have previously shown in diabetic Osm^−/− mice (11), and thus we checked whether the same was true in pioglitazone-treated mice. While the expression of Osm and Cxcl12 was reduced by pioglitazone in the unfractioned diabetic BM, expression of Cxcl12 specifically in BMAds was unaffected and remained elevated (Fig. 4A and B). Therefore, contrary to what was observed in Osm^−/− mice (Fig. 3M–O), sustained Cxcl12 expression by the expanded BM adipose tissue in pioglitazone-treated mice could limit HSPC mobilization.

As an additional pathway, we explored DPP4, which is released by VAT (30) and degrades CXCL12 (31). We found that gene expression of Dpp4 was not significantly different between BMAds and the rest of BM (Fig. 4D) or VAT (Supplementary Fig. 2); neither Dpp4 expression nor DPP4 activity was affected by pioglitazone treatment (Fig. 4D and E).

Intriguingly, pioglitazone significantly reduced p66Shc expression in the whole BM and in BMAds (Fig. 4C). Since p66Shc regulates HSPC mobilization in different ways (11,32), we explored the interplay among Osm and p66Shc. Consistent with prior studies showing that p66Shc-derived H₂O₂ sustains adipogenesis (33), we found less adipocyte accumulation in the BM of diabetic p66Shc^−/− versus Wt mice (Fig. 5A–C). We then evaluated whether and how p66Shc affected the adipogenic differentiation of BM-derived MSC. In vitro, MSC isolated from the BM of p66Shc^−/− had a 60% lower adipogenic differentiation capacity compared with that of Wt MSC (Fig. 5D and E). Adipogenesis by p66Shc^−/− BM-MSC could be partially rescued by transfection with a vector encoding for Wt but not with that encoding for 36^Ser→Ala p66Shc mutant (Fig. 5F and G), which does not translocate to mitochondria (34). Since OSMR recruits cytosolic p66Shc to activate the cascade culminating in Cxcl12 expression (11,35), it is possible that Osm deletion increased the pool of p66Shc available for mitochondrial translocation, thereby stimulating adipogenesis. To verify this hypothesis, we generated Osm^−/−p66Shc^−/− DKO mice. DKO mice were viable and fertile, with a normal reproductive fitness and no overt pathologic phenotype (data not shown). The extent of BM adipocyte accumulation in DKO mice was even greater than that observed in Osm^−/− mice, ruling out that adipogenesis induced by Osm deficiency depended on p66Shc (Fig. 5H and I). As expected, BMAds from DKO expressed markedly lower levels of Cxcl12 compared with Wt mice (Supplementary Fig. 4). DKO mice showed high levels of phenotypically (LKS) and functionally (colonies) defined HSPC in the unstimulated PB (Fig. 5J and K). In addition, despite massively increased BM adipocytes, HSPC mobilization after G-CSF in diabetic DKO mice was restored to normal (Fig. 5L).

To establish the clinical impact of pioglitazone therapy on HSPC mobilization, we performed in vitro experiments with human macrophages. Pioglitazone reduced OSM expression in M1-polarized macrophages by ∼50% (Fig. 6A). Consequently, the ability of pioglitazone-treated M1 macrophage–conditioned medium to induce CXCL12 expression in human BM-MSC was abolished (Fig. 6B), which was consistent with what was observed with murine cells. We then analyzed data from a study wherein patients with and patients without diabetes received low-dose G-CSF to test HSPC mobilization (2) and divided them in pioglitazone users and nonusers (Table 1). While control subjects without diabetes had doubled PB-HSPC levels 24 h after G-CSF, this was not observed in patients with type 1 diabetes (T1D) (1.0×; P = 0.96) or patients with type 2 diabetes (T2D) not taking pioglitazone (1.2×; P = 0.50). On the contrary, patients with T2D patients on pioglitazone treatment displayed a significant 1.7× mobilization of CD34⁺ HSPC (Fig. 6C and D). This beneficial effect occurred despite the fact that T2D patients on pioglitazone were older, more obese, and had a higher prevalence of hypertension and atherosclerotic disease (Table 1)—factors that are expected to negatively impact HSPC availability (17). The fold change in CD34⁺ HSPC levels in pioglitazone-treated patients with diabetes was higher than in T1D patients and T2D patients not taking pioglitazone combined but still lower than that observed in individuals without diabetes.

Countering the mechanisms driving chronic shortage of circulating HSPC is expected to improve micro- and macrovascular outcomes (7). Enhancing HSPC mobilization in patients with diabetes could also improve the outcome of autologous transplantation for the treatment of blood cancer (1). We show that PPAR-γ is a new candidate target to rescue the impaired HSPC mobilization consistently observed in human and experimental diabetes. PPAR-γ has previously been identified as a regulator of HSPC homeostasis (36), mainly by driving an expansion of BMAds. Indeed, PPAR-γ activation switches MSC differentiation from osteoblasts to adipocytes (37). Adipocytes contribute to forming the HSPC niche in the BM by providing soluble mediators that are critically important in regulating cellular quiescence, maintenance, and traffic (18). Under stress conditions, such as after irradiation or chemotherapy, BMAds expand and accelerate recovery of the BM niche and of hematopoietic function (18,38). By producing the HSPC-retaining chemokine CXCL12, BMAds exert antimobilizing activities (19,39). Indeed, adipocyte accumulation is a feature of the diabetic BM and associated with HSPC mobilopathy (16,17). Yet, we found that PPAR-γ activation partially rescued HSPC mobilization, despite increasing BMAds. In vitro, pioglitazone reprogrammed macrophages toward M2 and inhibited OSM signaling and production of CXCL12 by BM stromal cells. Suppression of OSM and CXCL12 was confirmed in vivo, supporting this mechanistic link. Many effects of pioglitazone were recapitulated by Osm deletion. According to our previous findings (11), Cxcl12 expression is controlled by OSM in BM-MSC via p66Shc. While the same appears to be true in VAT, BMAds expressed much higher levels of Cxcl12 that could not be stimulated further, possibly because of lower expression of OSMRs. Nonetheless, in the absence of OSM, BMAds expressed less Cxcl12, being arguably devoid of HSPC-retaining activity. Unexpectedly, despite pioglitazone-suppressed OSM gene and protein expression in the BM, pioglitazone failed to suppress Cxcl12 expression by BMAds, which remained high. This reflects that factors other than OSM concomitantly regulate Cxcl12 and can explain why pioglitazone only partially rescued HSPC mobilization.

Interestingly, we found that, in addition to suppressing OSM, pioglitazone also reduced expression of p66Shc in the whole BM and in BMAds. Cxcl12 induction by OSM requires p66Shc also in adipocytes, but Osm deletion and p66Shc deletion exerted opposite effects on adipogenesis. As cytosolic p66Shc is recruited to allow OSMR signaling (35), in the absence of OSM, more p66Shc might translocate to mitochondria, where it potentiates adipogenesis (33), Using DKO mice, we ruled out that adipogenesis induced by Osm deletion could be prevented by deleting p66Shc. Thus, OSM and p66Shc cooperate to induce Cxcl12, but they regulate the adipogenesis independently. DKO mice mimicked pioglitazone-treated mice with regard to adipogenesis and HSPC mobilization. However, pioglitazone did not completely suppress Osm, did not affect BMAd-derived Cxcl12, and only partially rescued mobilization. In addition, we acknowledge that other effects of pioglitazone could contribute to the improved mobilization and are not necessarily limited to modulation of the OSM-p66Shc pathway.

In this study, we purportedly used a model of T1D to avoid the confounding metabolic effects exerted by pioglitazone in T2D models. Importantly, while diabetic patients were normally unresponsive to the HSPC-mobilizing effect of G-CSF (40), patients with diabetes on pioglitazone showed a partial recovery of HSPC release. This lends support to the clinical transferability of our findings. Notably, pioglitazone is known to increase BM adiposity also in humans (25), which is supposed to make long bones prone to fractures (41). Although we have no information on BM adipocyte content in our patients, the emerging scenario makes a striking parallel between mice and humans, with pioglitazone stimulating BM adipogenesis and being partially able to rescue HSPC mobilization. Interestingly, this occurred despite the fact that pioglitazone did not affect myelopoiesis. Thus, even in the absence of systemic metabolic and anti-inflammatory effects, pioglitazone disconnected HSPC mobilopathy from diabetes-associated myelopoiesis. Pioglitazone has the potential to improve hard outcomes in patients with diabetes (42), and, in view of the strong association between HSPC levels and adverse diabetes outcomes (4,5), we argue that part of this protection could be mediated by rescuing HSPC traffic. Strategies to limit BM adipogenesis might potentiate such effect and improve clinical benefits of PPAR-γ activation.

Funding. The study was supported by the following grants: European Foundation for the Study of Diabetes (EFSD)/Novartis 2013 grant to G.P.F., EFSD/Lilly 2016 grant to G.P.F., Ministry of University and Education PRIN grant 2015 to G.P.F., and Italian Diabetes Society/Lilly grant 2017 to G.P.F.

Duality of Interest. S.C., M.A., and G.P.F. are the inventors of a patent, held by the University of Padova, on the use of pharmacologic oncostatin M inhibition to allow stem cell mobilization.

Author Contributions. S.T., S.C., A.C., A.A., M.A., and G.P.F. designed the research. S.T., S.C., L.M., M.D., V.S., R.C., A.A., and M.A. performed the research. S.T., S.C., L.M., M.D., V.S., A.C., A.A., M.A., and G.P.F. analyzed the data. S.C., A.A., M.A., and G.P.F. wrote the manuscript. All authors reviewed and edited the manuscript. G.P.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.