Bone marrow adipose tissue-derived stem cell factor mediates metabolic regulation of hematopoiesis.

Hematopoiesis is dynamically regulated by metabolic cues in homeostatic and stressed conditions; however, the cellular and molecular mechanisms mediating the metabolic sensing and regulation remain largely obscure. Bone marrow adipose tissue remodels in various metabolic conditions and has been recently proposed as a niche for hematopoietic stem cells after irradiation. Here, we investigated the role of marrow adipose tissue-derived hematopoietic cytokine stem cell factor in unperturbed hematopoiesis by selectively ablating the Kitl gene from adipocytes and bone marrow stroma cells using Adipoq-Cre and Osx1-Cre, respectively. We found that both Adipoq-Kitl knockout (KO) and Osx1-Kitl KO mice diminished hematopoietic stem and progenitor cells and myeloid progenitors in the bone marrow and developed macrocytic anemia at the steady-state. The composition and differentiation of hematopoietic progenitor cells in the bone marrow dynamically responded to metabolic challenges including high fat diet, β3-adrenergic activation, thermoneutrality, and aging. However, such responses, particularly within the myeloid compartment, were largely impaired in Adipoq-Kitl KO mice. Our data demonstrate that marrow adipose tissue provides stem cell factor essentially for hematopoiesis both at the steady state and upon metabolic stresses. Copyright

Introduction

The metabolic and hematopoietic systems demonstrate dynamic and complex interplays in health and disease. On the one hand, a plethora of blood cells including granulocytes, monocytes and macrophages, mast cells, and lymphocytes contribute to the physiological and pathological regulation of energy intake and expenditure, glucose and lipid metabolism, bone remodeling, and the aging process.[1-4] On the other hand, hematopoietic stem cells (HSC), myeloid and lymphoid progenitors, and their mature progeny not only impose different bioenergetic demands during development,[5-8] but also show flexibility and plasticity in their maintenance, differentiation, and function in response to various metabolic disturbances, such as obesity, hyperglycemia, and aging.[9-16] A large body of evidence indicates that both intrinsic and extrinsic factors drive the hematopoietic process; however, the cellular and molecular mechanisms underlying the metabolic regulation of hematopoiesis remain for the most part elusive.

In adults, the bone marrow (BM) microenvironment provides niches that support the renewal, quiescence, and differentiation of hematopoietic stem and progenitor cells (HSPC).[17-19]

Recent studies have started to unveil the complexity and heterogeneity of niche cell types, niche factors, and their actions. BM mesenchymal stem cells [BMSC, also known as skeletal stem cells (SSC)] and their adipogenic, osteogenic, and chondrogenic progeny are major contributors of niche factors, such as stem cell factor (SCF) and CXC chemokine ligand 12 (CXCL12).[20-22] The sympathetic nervous system (SNS) extensively innervates the bone and BM to control hematopoietic homeostasis and regeneration via direct actions on HSPC and indirect actions on the niche.[23] In addition, signals from the vascular endothelial cells and the HSC progeny such as macrophages and megakaryocytes have also been shown to contribute to different aspects of HSPC regulation.[17] Nevertheless, whether these niche constituents mediate the sensing of metabolic cues and subsequent remodeling in hematopoiesis has not yet been determined.

White adipose tissue (WAT) that stores excess energy and brown adipose tissue (BAT) that dissipates energy as heat are key determinants of metabolic homeostasis. The role of BM adipose tissue (MAT), the third major adipose depot in the body, is just beginning to be revealed. Developmentally, BM adipocytes arise from the same Osterix+ skeletal lineage as osteoblasts and chondro-cytes.[24-26] Anatomically, constitutive MAT (cMAT) is found in the most distal portion of the tibia and tail vertebrae while regulated MAT (rMAT) is found in the proximal skeletal sites.[27-29] Although cMAT is relatively stable, rMAT expands in conditions like obesity, diabetes, caloric restriction, and aging.[27-29] Functionally, there are tripartite interactions between MAT, bone, and hematopoiesis, yet their mechanistic characteristics are still not fully understood.[30] An early study taking advantage of the genetic and pharmacological inhibition of adipogenesis suggested MAT to be a negative regulator of the hematopoietic microenvironment.[31] In contrast, recent work demonstrated that MAT supports HSC regeneration and myeloerythroid maturation following irradiation and reconstitution, partially by secreting SCF.[32,33]

The close relationship between hematopoiesis and metabolism is also represented by their regulation by common growth factors and cytokines. SCF and its receptor KIT play an essential role in the survival, migration, and differentiation of multiple stem and progenitor cells including HSPC.[34] In the hematopoietic system, loss-of-function mutations in SCF/KIT cause macrocytic anemia while gain-of-function mutations lead to systematic mastocytosis, acute myeloid leukemia, and lymphoma.[35,36] In the bone marrow niche, SCF is expressed in LEPR+ stroma cells, endothelial cells, and adipocytes, but not in osteoblasts or hematopoietic cells.[22,32] Deleting SCF selectively in these positive niche cells leads to defects in HSC maintenance.[22,32] In the metabolic system, SCF has been shown to promote the differentiation of brown adipocytes from human pluripotent stem cells and to be essential to mitochondrial function and energy expenditure in mice.[37,38] However, the cellular source of SCF in regulating systemic metabolism has not been determined. Here, we investigated the contribution of adipose-derived SCF in regulating energy and glucose metabolism and in mediating the effect of metabolic stresses on unperturbed hematopoiesis.

Methods

Mice

All mice used in this study were purchased from the Jackson Laboratory, including Kitl (stock n. 017861), Adipoq-Cre (stock n. 010803), Osx1-Cre (stock n. 006361), and Kitl (stock n. 017860). All animals were kept on a 14 hour (h):10 h light: dark cycle in the animal facility at the University of Minnesota, Minneapolis, MN, USA. Mice were group-housed, with free access to water and either a standard chow diet or 60% high fat diet (Research Diets, D12492). 1 mg/kg BW of CL-316, 243 (R&D Systems, #1499, diluted in saline) were intraperitoneally (i.p.) injected when indicated. For thermoneutral housing, mice born at 22°C were transferred to a room maintained continuously at 30°C at the age indicated. All procedures involving animals were conducted within Institutional Animal Care and Use Committee guidelines under approved protocols.

Flow cytometry

Bone marrow cells were isolated by flushing the femur in Ca2+ and Mg2+ free PBS with 1% heat-inactivated bovine serum. Cells were dissociated to a single cell suspension by gently passing through a 25-gauge needle and then filtering through a 70-mm nylon mesh. Red blood cells from BM were removed by ammonium-chloride-potassium lysing buffer. For flow analyses, BM cells were stained with a cocktail of biotin-conjugated lineage antibodies CD3e, B220, Ter119, Mac-1 and Gr-1 (Biolegend, 133307), CD4 (Biolegend, 100403), CD5 (Biolegend, 100603), CD8 (Biolegend, 100703), followed by Streptavidin-AF488 (Biolegend, 405235). Cells were then stained with CD127-APC (eBioscience, 17-1271-82), c-Kit-APC-eFluor780 (eBioscience, 47-1171-82), Sca-1-Super Bright 436 (eBioscience, 62-5981-82), CD34-PE (Biolegend, 152204), and FcγR-PerCP-eFluor710 (eBioscience, 46-0161-80), CD150-BV605 (Biolegend, 115927), and CD48-BUV395 (BDBioscience, 740236). Multicolor analysis for progenitor and stem-cell quantification was performed on a 3-laser-LSRII flow cytometer (BD).[12,39] HSPC was defined as Lin−Sca-1+c-Kit+ (LSK); phenotypic LT-HSC was defined as CD150+CD48−Lin−Sca−1+c-Kit+; myeloid progenitor (MP) was defined as Lin−CD127−Sca−1−c-Kit+; common lymphoid progenitor (CLP) was defined as Lin−CD127+Sca−1+c-Kit+; common myeloid progenitor (CMP) was defined as Lin−CD127-Sca−1−c-Kit+CD34+ FcγR−; megakaryocyte-erythrocyte progenitor (MEP) was defined as Lin−CD127−Sca−1−c-Kit+CD34−FcγR−; granulocyte-monocyte progenitor (GMP) was defined as Lin−CD127−Sca−1−c-Kit+CD34+FcγR+. SYTOX™ Green Dead Cell Stain (Thermo Fisher Scientific, S34860) was used to exclude dead cells. Absolute number was obtained by using counting beads (Thermo Fisher Scientific, C36950) as instructed by the manufacturer.

Metabolic assays

Body composition was assessed using an EchoMRI system by which fat and lean mass measured by magnetic resonance scanning were normalized to body weight for fat percentage and lean percentage. Adipose tissue weight was determined by dissecting and weighing indicated adipose depots. For glucose tolerance tests, 16 h-fasted mice were injected i.p. with glucose (1.5 g/kg body weight). Blood glucose from tail-vein blood collected at the designated times was measured using a Bayer Contour Glucometer (9545C).

Cell culture

The stromal vascular faction (SVF) cells derived from iWAT were obtained as previously described.[38] SVF cells were cultured in DMEM/F12 (Corning, 10-090-CV) containing 10% FBS (GenClone, 25-514), 1% penicillin/streptomycin (Gibco, 10378016), 20 nM insulin (Sigma), and 1 nM triiodothyronine (T3, Sigma, T6397). Two days after becoming confluent (defined as Day 0), SVF cells were induced with DMEM/F12 containing 10% FBS, 1% penicillin/streptomycin, 0.5 mM isobutylmethylxanthine (IBMX, Sigma, I7018), 125 μM indomethacin (Sigma, I7378), 1 μM dexamethasone (Sigma, D4902), 20 nM insulin, and 1 nM T3 for 48 h. Cells were maintained in DMEM/F12 containing 10% FBS, 1% penicillin/streptomycin, 20 nM insulin, and 1 nM T3 until lipid drop appeared. This medium was replenished every two days.

Oil Red O staining

Cells were washed with PBS and then fixed with 10% formaldehyde (Sigma) for 1 h. After washing with 60% isopropanol, fixed cells were stained with Oil Red O solution (2 mg/mL in 60% isopropyl alcohol, Sigma, O-0625) for ten minutes (min), and rinsed five times with pure H2O before photographic images were taken.

Genotyping and quantitative real-time polymerase chain reaction

Primers used for DNA amplification were: floxed-Kitl-F, CGAG-GTAGGGGAAAAGAACC; floxed-Kitl-R, GGATCTTCCCA-GAGGTTGGA; excised-Kitl-F, GGAAAAGAACCAAGTGAAGTC; excised-Kitl-R, ACGGGGAAAGACCTCCGGTCC; Adipoq-Cre-F, GGAAAAGAACCAAGTGAAGTC; Adipoq-Cre-R, ACGGGGAAAGACCTCCGGTCC. DNA for verifying knockout was isolated using the Quick-DNA Plus Kit (Zymo Research, #D4074). Genotyping was performed according to instructions in the manual provided by the Jackson Laboratory.

Total RNA from tissues was isolated using TRIzol (Invitrogen, 15596018). RNA was reverse-transcribed using iScript cDNA Synthesis Kits (Bio-Rad, 170-8891). Quantitative real-time poly merase chain reaction (qRT-PCR) was performed using SYBR Green Supermix (Bio-Rad, 1725124) with a C1000 Thermal Cycler (Bio-Rad) following the manufacturers’ instructions. For thermal cycling: 95°C, 3 min; then 40 cycles of 95°C, 10 seconds (s) and 60°C, 30 s. The primers used for qRT-PCR were long Kitl (flanking exon 5 and 6): 5’-GCCAGAAACTAGATCCTTTACTCCTGA-3’ and 5’-ACATAAATGGTTTTGTGACACTGACTCTG-3’; short Kitl (flanking junctions between exon 5/7 and exon 8/9): 5’-CCC-GAGAAAGGGAAAGCCG-3’ and 5’-ATTCTCTCTCTTTCT-GTTGCAACATACTT-3’; total Kitl (flanking exon 2 and 3): 5’-TCTGCGGGAATCCTGTGACT-3’ and 5’-CGGCGA-CATAGTTGAGGGTTAT-3’; excised Kitl (flanking exon 1 and 2, exon1 was floxed): 5’-CAGCGCTGCCTTTCCTTATGA-3’ and 5’-ATCAGTCACAGGATTCCCGC-3’, and the housekeeping gene 36b4: 5’-AGATGCAGCAGATCCGCAT-3’ and 5’-GTTCTTGCCCATCAGCACC-3’. Delta-delta Ct analysis was used to calculate relative gene expression.

Histology and immunostaining

Long bones were fixed in 10% neutral buffered formalin followed by three days of decalcification in 14% EDTA, followed by paraffin or OCT embedding. Bones embedded in paraffin were sectioned at 5 μm thickness using a microtome (Olympus Cut 4060) and stained with Hematoxylin and Eosin (H&E). For quantification of the BM adipocytes, the BM areas 4.5 mm from growth plate were selected. Bones embedded in OCT were sectioned at 7 μm thickness using a cryostat (Leica). Sections were blocked with 3% bovine serum albumin, 0.2% TWEEN 20 in Tris-buffered saline, incubated with chicken-anti-GFP (Aves, GFP-1020, 1:1,000), rabbit-anti-perilipin (Cell Signaling Technology, #9349, 1:200) overnight, and secondary antibodies (Alexa Fluor 488 anti-Chicken IgG and Alexa Fluor 674 anti-Rabbit IgG, Life Technologies, 1:400) for 1 h. A Nikon system was used for fluorescence detection.

Western blot

Tissue proteins were extracted using RIPA buffer with freshly added proteinase inhibitors. Protein concentrations were determined using BCA Protein Assay Kit (Pierce). Equal amounts of protein samples were subjected to western blot. The following antibodies were used: anti-UCP1 (Abcam, ab209483, 1:5000 dilution), anti-PGC-1a (Bioworld, BS72263, 1:500 dilution), anti-PER-ILIPIN (Cell Signaling Technology, 9349T, 1:1000 dilution), anti-PPARγ (CusAb, CSB-PA018424LA01HU, 1:500 dilution), anti-COX4 (Proteintech, 11242-1-AP, 1:1000 dilution), and anti-ACTIN (Sigma, A5441, 1:5000 dilution). Densitometry was performed using Image J. Relative band density was calculated by dividing the densitometry of target protein with loading control from the same membrane.

Isolation of bone marrow adipocytes and supernatant

A 0.6 mL microcentrifuge tube was cut open at the bottom and placed into a 1.5 mL microcentrifuge tube. Long bones were snipped both ends and placed in the prepared 0.6 mL microcentrifuge tube. BM was flowed out by quick centrifuge (from 0 to 10,000 rpm, RT). Red blood cells from BM were lysed by ammonium-chloride-potassium lysing buffer. After 3,000 rpm centrifugation for 5 min at RT, floating adipocytes were collected as BM adipocytes from the top layer.[40] The middle layer was collected and remaining cells were spun down by centrifugation (12,000 rpm, 30 s, RT). The supernatant was collected for SCF measurement using an ELISA (Thermo Fisher Scientific, EMKITL).

Colony-forming unit assay

The colony-forming unit (CFU) assay was performed using MethoCult™ GF M3434 (Stem cell) according to the manufacturer’s instructions. Briefly, BM cells were flushed from one femur, and filtered through a 40 um cell strainer. Spleen was minced and pressed through a 40 um cell strainer to obtain single cell suspension. 2×104 BM cells or 2×105 spleen cells were plated in methylcellulose, and the CFU were counted after 12 days.

Statistical analysis

All data are presented as mean±Standard Deviation (SD) or mean±Standard Error of Mean (SEM) as indicated in the figure legends. The statistical significance between two groups was determined by unpaired two-tailed Student t-test (Microsoft Excel or GraphPad Prism 7). Datasets involving more than two groups were assessed by one-way ANOVA with Tukey’s correction for multiple comparisons using GraphPad Prism 7. Two-way ANOVA followed by Tukey’s correction or Sidak’s correction for multiple comparisons was performed using GraphPad Prism 7 to examine data with two independent variables. *P<0.05; **P<0.01; ***P<0.001.

Results

Adipocyte-derived stem cell factor is not essential for brown fat function in vivo

To determine whether adipocyte-derived SCF cell-autonomously regulates BAT function in vivo, we generated fat-specific SCF KO mice by crossing Adiponectin (Adipoq)-Cre mice with Kitl mice. The adipose stromal vascular faction (SVF) cells derived from Kitl controls and Adipoq-Cre KO mice were isolated and differentiated into adipocytes in culture. The Kitl gene was specifically deleted (Online Supplementary Figure S1A) and the Kitl mRNA reduced its level (Online Supplementary Figure S1B) in KO cells when Adipoq-Cre started to be expressed after adipogenic induction (Online Supplementary Figure S1B). Control and KO SVF cells showed similar capacity in adipogenic differentiation, determined by Oil Red O staining (Figure 1A). Consistent with the previous finding that SCF is required for UCP1 expression,[37] we found that differentiated brown adipocytes from Adipoq-Cre mice had much lower levels of UCP1 protein compared to Kitl cells (Figure 1B and Online Supplementary Figure S1C). Expression of PGC-1α and PERILIPIN was unaffected (Figure 1B).

Figure 1.

Role of adipose stem cell factor (SCF) in brown fat function. (A and B) Stromal vascular faction (SVF) cells from iWAT of Kitl and Adipoq-Cre mice were differentiated into adipocytes in vitro and analyzed by Oil O Red staining (A) and western blotting (B). Densitometry of UCP1 shown in Online Supplementary Figure S1C. (C and D) Expression of proteins involved in adipogenesis and thermogenesis in BAT (C) and iWAT (D) from 7-week old Kitl and Adipoq-Cre male mice. Densitometry of UCP1 shown in Online Supplementary Figure S1D and E. (E and F) Kitl and Adipoq-Cre mice (n=6) were treated with CL 316,243 for seven days. Ucp1 mRNA levels (E) and UCP1 protein levels (F) in BAT and iWAT were determined. Densitometry of UCP1 shown in Online Supplementary Figure S1J and K. Data are presented as mean±Standard Deviation.

We then sought to determine gene expression in thermogenic fat tissues in vivo. The levels of UCP1, PGC-1a, COX4, and PPARγ proteins in BAT and inguinal WAT (iWAT) were essentially the same between Kitl and Adipoq-Cre mice (Figure 1C and D and Online Supplementary Figure S1D and E). Kitl and Adipoq-Cre mice had similar body weight at 7, 14, and 28 weeks of age (Online Supplementary Figure S1F-H). No changes in the mass of BAT, iWAT, and gonadal WAT (gWAT) (Online Supplementary Figure S1F), composition of lean and fat mass (Online Supplementary Figure S1G), or fasting body weight (Online Supplementary Figure S1H) were observed. In addition, systemic glucose metabolism shown by the glucose tolerance test was also comparable between the two genotypes (Online Supplementary Figure S1I).

To assess sympathetic nerve-activated adaptive thermogenesis, we treated mice with a β3-adrenoceptor agonist, CL 316,243 for seven days. qRT-PCR and western blotting showed no difference in levels of Ucp1 mRNA (Figure 1E) or UCP1 protein (Figure 1F and Online Supplementary Figure S1J and K) in BAT or iWAT between Kitl and Adipoq-Cre mice. Together, these data demonstrate that SCF secreted by adipocytes is essential for UCP1 expression in vitro but is not essential for energy metabolism in vivo. It is possible that SCF from non-adipose cells or tissues may compensate for the loss of SCF in adipocytes, as no change in serum levels of SCF was observed in Adipoq-Cre mice (Online Supplementary Figure S1L).

Adipocyte-derived stem cell factor controls steady-state hematopoiesis

Marrow adipose tissue is an endocrine organ important for hematopoiesis and systemic metabolism.[27-30] MAT promotes the regeneration of hematopoietic stem cells after irradiation by secreting SCF.[32] We first confirmed that most BM adipocytes (approx. 77% in cMAT and approx. 80% in rMAT) expressed SCF by performing perilipin immuno-fluorescent staining on bone sections of Kitl knockin mice (Figure 2A and Online Supplementary Figure S2A). Compared to BAT and iWAT, the flushed BM expressed similar levels of the long isoform of Kitl that can be transcribed and cleaved into the soluble form of SCF, but much less of the short Kitl transcript that encodes the membrane-bound SCF (Online Supplementary Figure S2B).[34] Note that there are significant numbers of non-adipocytes in the adipose tissues and BM analyzed. In BM, Adipoq-Cre was recently shown to only label mature adipocytes, but not bone stroma, adipogenic progenitors, hematopoietic cells, bone lining cells, or osteoblast cells.[41] In Adipoq-Cre mice, the Kitl gene was specifically knocked down in marrow adipocytes (Online Supplementary Figure S2C). We could observe a significant loss of Kitl transcript in the flushed BM (Figure 2B) and SCF protein in the BM supernatant (Figure 2C), suggesting that MAT is a major source of SCF in the BM.

Figure 2.

Adipocyte-derived stem cell factor (SCF) is a niche factor for hematopoietic stem and progenitor cells. (A) Co-staining of Perilipin and EGFP in the cMAT and rMAT of Kitl mice. Scale bar=50 μm. (B and C) Levels of total Kitl mRNA (B) and SCF protein (C) in the flushed bone marrow (BM) from tibia of Kitl (n = 5) and Adipoq-Cre (n=6) 13-week old male mice. (D) BM cellularity in the femur of 13-week old Kitl (n = 7) and Adipoq-Cre (n=8) male mice. (E) Representative flow cytometry plots showing LSK and MP cells among the lineage− CD127− population in 13-week old Kitl and Adipoq-Cre male mice. Average frequencies are shown as inserts. (F) Quantification of absolute numbers of LSK and MP cells in the femur of 13-week old Kitl (n=7) and Adipoq-Cre (n=8) male mice and phenotypic LT-HSC in the femur of 8-month old Kitl and Adipoq-Cre (n=4) male mice. (G) Representative flow cytometry plots showing CMP, MEP, and GMP cells among the MP population in 13-week old Kitl and Adipoq-Cre male mice. Average frequencies are shown as inserts. (H) Absolute numbers of CMP, MEP, and GMP cells in the femur of 13-week old Kitl (n=7) and Adipoq-Cre (n=8) male mice. (I) The ratio of marrow MEP to GMP in 13-week old Kitl (n=7) and Adipoq-Cre (n=7) male mice. (J) The absolute number of CLP in 13-week old Kitl (n=7) and Adipoq-Cre (n=7) male mice. (K) The ratio of marrow CMP to CLP in 13-week old Kitl (n=7) and Adipoq-Cre (n=7) male mice. (L) Colony formation assay of 2×104 BM cells from Kitl and Adipoq-Cre (n=4) mice. (M) Spleen weight of Kitl and Adipoq-Cre (n=4) mice. (N) Colony formation assay of 2×105 splenic cells from Kitl and Adipoq-Cre (n=4) mice. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by unpaired student t-test (B-K) or one-way ANOVA (N).

We then quantified hematopoietic stem and progenitor cells (HSPC) in the BM of Adipoq-Cre mice by flow cytometry (Online Supplementary Figure S3).[12,39] We first characterized the Adipoq-Cre line and could not observe any potential Cre-specific defects in BM cellularity or HSPC numbers when compared to wild-type mice (Online Supplementary Figure S4). Thus, Kitl mice were used as controls for comparison in the following experiments. Loss of SCF specifically in adipocytes reduced marrow cellularity in male mice (Figure 2D). The frequency and also absolute number of lineage−Sca−1c-Kit (LSK) HSPC, phenotypic long-term (LT)-HSC (e.g. CD150CD48− LSK cells), and myeloid progenitors (MP) were all down-regulated in Adipoq-Cre male mice (Figure 2E and F). Within MP, common myeloid progenitors (CMP), megakaryocyte-erythrocyte progenitors (MEP), and granulocyte-monocyte progenitors (GMP) all showed decreased frequency and number (Figure 2G and H). Interestingly, there was a reduction in the ratio of MEP to GMP (Figure 2I), suggesting that the extent of dependence on SCF varies between different myeloid progenitors. On the other hand, common lymphoid progenitors (CLP) maintained their number (Figure 2J) and the ratio of CMP/CLP was reduced (Figure 2K) in Adipoq-Cre mice. Colony formation assay showed that BM HSPC, though reducing their numbers in the SCF-deficient environment, did not show functional decline when assessed in a complete medium in vitro (Figure 2L). It suggests that loss of adipose SCF results in defects in the niche environment but not intrinsically in HSPC. Whether BM adipocyte-derived SCF supports the long-term proliferation and self-renewal of HSPC requires future investigation. Hematopoietic stresses can mobilize HSPC outside the BM to sites like the spleen to expand hematopoiesis.[42] We found that spleen in Adipoq-Cre mice was slightly heavier (Figure 2M) and splenic cells from Adipoq-Cre mice formed significantly more colony-forming units (CFU) in vitro (Figure 2N), indicating a compensatory induction of splenic hematopoiesis when BM hematopoiesis was defective.

Similar phenotypes were observed in female mice. There was a trending decline in BM cellularity in Adipoq-Cre+;Kitl female mice, compared to their control counterparts (Online Supplementary Figure S5A). Both the absolute number and frequency of LSK cells, MP, CMP, MEP, and GMP were declined when SCF was absent (Online Supplementary Figure S5B and C). CLP remained unchanged (Online Supplementary Figure S5B and C). Adipoq-Cre female mice also had reduced ratio of MEP to GMP (Online Supplementary Figure S5D). These data demon strate that adipocyte-derived niche factor SCF is essential for the maintenance of hematopoietic stem and myeloid progenitor cells.

We then performed the complete blood count test of the peripheral blood. Levels of red blood count (RBC), hemoglobin concentration (HGB), and hematocrit were all decreased in both male and female Adipoq-Cre mice when compared to Kitl controls (Table 1). The mean corpuscular volume (MCV) and the count of reticulocytes were both significantly increased in Adipoq-Cre mice (Table 1), showing that Adipoq-Cre mice developed the typical macrocytic anemia that is observed in animals and patients with loss-of-function mutations in the SCF/KIT pathway.[35,36,43] Meanwhile, Adipoq-Cre mice also had less platelets compared to control mice (Table 1). Within white blood cells, neutrophils, monocytes, and lymphocytes reduced their number in both sexes, while eosinophils and basophils were significantly declined in male Adipoq-Cre mice and showed a trending reduction in female Adipoq-Cre mice (Table 1). Taken together, these data from both male and female mice indicate that adipocytes compose a niche that produce SCF to maintain unperturbed hematopoiesis.

Table 1.

Decline in production of mature blood cells in the periphery of Adipoq-Cre mice.

Ablation of stem cell factor in Osterix+ cells impairs hematopoiesis

Adipoq-Cre-mediated ablation deleted SCF in all adipose tissues. To exclude the potential confounding effect of SCF from peripheral adipose tissues, we then conditionally deleted the Kitl gene in BM stroma cells using the Osterix (Osx1)-Cre, which marks progenitor cells that can be differentiated into MAT.[24-26] Loss of SCF in Osx1+ cells resulted in lower BM cellularity (Figure 3A) and drastically reduced populations of LSK, phenotypic LT-HSC, MP, CMP, GMP, MEP, and CLP in Osx1-Cre mice (Figure 3B). Such reductions were much larger than those observed in Adipoq-Cre mice (Figure 2), suggesting that SCF from non-MAT, Osx1-Cre-derived niche cells also contributed to HSPC defects in Osx1-Cre mice. Interestingly, heterozygous Osx1-Cre/+ mice also had decreased BM cellularity and frequencies of all HSPC (Figure 3A and B), suggesting the haploinsufficiency of SCF in hematopoietic regulation. Similar to findings in Adipoq-Cre mice, the MEP/GMP ratio was reduced (Figure 3C) and the CMP/CLP ratio showed a trending decrease (Figure 3D). As a result, Osx1-Cre mice suffered macrocytic anemia (Figure 3E). Furthermore, there were fewer platelets, neutrophils, and monocytes in Osx1-Cre mice (Figure 3E). Counts of eosinophils, basophils, and lymphocytes were similar between control and Osx1-Cre mice (Figure 3E). BM cells from Osx1-Cre KO mice could efficiently form CFU as wild-type controls (Figure 3F). suggesting defects in the niche environment not intrinsically in HSPC as a result of loss of adipose SCF. However, splenomegaly and profoundly increased CFU formation of splenic cells were observed in Osx1-Cre mice (Figure 3G and H), demonstrating a shift of hematopoiesis toward the spleen. Taking data from Adipoq-Cre-and Osx1-Cre-mediated knockout mice, we argue that, in the steady-state, SCF from MAT is essential for homeostasis of hematopoietic progenitors.

Figure 3.

Impaired hematopoiesis when stem cell factor (SCF) is ablated in Osx1 cells. (A) Bone marrow cellularity in the femur of 8-month old control (n=4), Osx1-Cre (n=5), and Osx1-Cre (n=4) male mice. (B) Frequencies of LSK, phenotypic LT-HSC, MP, CMP, GMP, MEP, and CLP populations in the bone marrow (BM) of 8-month old control (n=4), Osx1-Cre (n=5), and Osx1-Cre (n=4) male mice, determined by flow cytometry. (C and D) Ratios of MEP to GMP (C) and CMP to CLP (D) in 8-month old control and Osx1-Cre (n=4) male mice. (E) Complete blood count of 10-week old control (n=8, including 6 Kitl and 2 Osx1-Cre) and Osx1-Cre (n=7) male mice. (F) Colony formation assay of 2×104 BM cells from Kitl and Osx1-Cre (n=4) mice. (G) Spleen weight of Kitl and Osx1-Cre (n=4) mice. (H) Colony formation assay of 2×105 splenic cells from Kitl and Osx1-Cre (n=4) mice. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by one-way ANOVA followed with Tukey’s multiple comparison (A, B, and H) or unpaired Student t-test (C, E, and G).

MAT-derived stem cell factor contributes to stressed hematopoiesis in obesity

Obesity is associated with increased MAT mass and altered hematopoietic and immune functions. We then asked whether MAT-derived SCF mediates the effect of high-fat diet (HFD) on hematopoiesis. Male Kitl and Adipoq-Cre mice gained similar body weight after HFD feeding for eight weeks (Online Supplementary Figure S6A). HFD significantly increased BM adiposity and cellularity, which could be slightly diminished by the loss of adipose SCF (Figure 4A-C). We did not observe any changes in Kitl gene expression or SCF secretion after HFD in the BM (Online Supplementary Figure S6B and C), indicating a potential compensation from non-adipocytes. Future experiments are required to determine SCF expression by different stroma cells during HFD. Flow cytometric assessment of HSPC showed that numbers of LSK, GMP, and CLP progenitors in wild-type male mice were increased after HFD (Figure 4D-F). Such induction was completely blunted in Adipoq-Cre males (Figure 4D-F). It is possible that other yet-to-be defined factors mediate the expansion of HSPC in response to HFD, but their complete function requires MAT-secreted SCF. Even though HFD did not affect numbers of CMP or MEP (Figure 4G and H), there was a reduction in the ratio of MEP to GMP (Figure 4I). The MEP/GMP ratio was lower in Adipoq-Cre males and could not be further reduced by HFD (Figure 4I). Consistent with the changes of HSPC in the BM, there was an increase in the number of granulocytes, monocytes, and lymphocytes in the peripheral blood of wild-type males after HFD (Figure 4J-L). The ratio of megakaryocyte (represented by platelet)-erythrocyte (MkE) to granulocyte-monocyte (GrMo) was down-regulated (Figure 4M), while the ratio of lymphocytes to all myeloid cells was up-regulated by HFD (Figure 4N). These data suggest that HFD preferentially promotes the hematopoietic differentiation toward the GrMo and lymphoid lineages, which may contribute to the development of inflammation and insulin resistance in obesity. In the peripheral blood of Adipoq-Cre male mice, however, HFD could not increase the number of granulocytes, monocytes or lymphocytes to the extent observed in Kitl males (Figure 4J-L). HFD-induced downregulation of MkE/GrMo ratio and upregulation of lymphoid/myeloid lineage ratio were both ablated in Adipoq-Cre male mice (Figure 4M and N). Other peripheral blood parameters including RBC count, HGB, MCV, and platelet count were either not affected by HFD or were similarly regulated between Kitl and Adipoq-Cre males (Online Supplementary Figure S6D). These results indicate that SCF from MAT is required for the skewed hematopoietic differentiation toward GrMo and lymphoid lineages during HFD.

Figure 4.

High-fat diet (HFD)-stressed hematopoiesis in control and Adipoq-Cre male mice. (A and B) Kitl and Adipoq-Cre male mice at the age of eight weeks were fed with normal chow (NC) or HFD (n=3-4) for another eight weeks. (A) Representative images of femur sections. (B) Quantification of adipocyte numbers in the BM from the growth plate to 4.5 mm away distally. (C-I) 8-week old Kitl (n=6 for each diet) and Adipoq-Cre (n=8 for NC and n=6 for HFD) male mice were fed with NC or HFD for eight weeks. BM cellularity (C), LSK number (D), GMP number (E), CLP number (F), CMP number (G), MEP number (H), and the MEP/GMP ratio (I) were determined by flow cytometry. (J-N) Complete blood count of NC-and HFD-fed Kitl (n=16 and 13, respectively) and Adipoq-Cre (n=12 and 6, respectively) male mice showing granulocyte number (J), monocyte number (K), lymphocyte number (L), megakaryocyte-erythrocyte (MkE) to granulo-cyte-monocyte (GrMo) ratio (M), and the ratio of lymphocyte to all myeloid cells including MkE and GrMo (N). Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by two-way ANOVA followed by multiple comparison using Tukey’s correction (B) or Sidak correction (C-N).

Sexual dimorphism is observed in obesity-associated inflammation and immune dysfunction, which is partially attributable to difference in hematopoiesis.[44,45] We then sought to determine the effect of MAT-derived SCF on obesity-stressed hematopoiesis in females. HFD-induced gain in body weight was similar between Kitl and Adipoq-Cre female mice (Online Supplementary Figure S7A). We found that, compared to males, HFD feeding in wild-type females did not affect BM cellularity (Online Supplementary Figure S7B) but LSK and all myeloid progenitors including CMP, MEP and GMP were all expanded (Online Supplementary Figure S7C-F), without a change in the relative ratio of MEP to GMP (Online Supplementary Figure S7G). The increase in the frequency of LSK, CMP, and MEP by HFD in females was abolished when SCF was deleted in adipocytes (Online Supplementary Figure S7C-E). However, the increase in GMP frequency after HFD in females seemed to be independent of adipocyte-derived SCF (Online Supplementary Figure S7F). In the peripheral blood, HFD feeding significantly augmented the red blood cell count and hemoglobin concentration in both Kitl and Adipoq-Cre female mice (Online Supplementary Figure S7H and I). Mean corpuscular volume (MCV), platelet count, and lymphocyte count were not affected by HFD in either Kitl or Adipoq-Cre female mice (Online Supplementary Figure S7J-L). Despite the increased frequency of GMP in the BM, there was a declining trend in peripheral granulocytes and significant downregulation of monocytes in HFD wild-type females, which was absent in Adipoq-Cre female mice (Online Supplementary Figure S7M and N). Reasons causing these discrepancies between the BM and the periphery are unclear, but may involve the production and turnover of mature cells, their release into the circulation, and recruitment to target tissues. Taken all these data from male and female mice, we conclude that MAT niche factor SCF is required for HFD-induced changes in HSPC maintenance and differentiation, despite the sex differences observed in such responses.

Adipocyte-derived stem cell factor partially mediates the β3-adrenergic regulation of hematopoietic stem and progenitor cells

The BM is extensively innervated by the SNS.[23,46] β-adrenoceptors are expressed in Nestin+ SSC and their activation by the SNS mediates the circadian mobilization of HSC,[47,48] while the neuropathy of the BM niche contributes to the pathogenesis of myeloproliferative neoplasms.[49,50] In addition, MAT, particularly the rMAT, expresses all three β-adrenoceptors and undergoes remodeling when the sympathetic tone is elevated by cold or β-adrenergic agonists.[51,52] We then sought to determine whether MAT-secreted SCF mediates the effect of β3-adrenergic signaling on hematopoiesis. Treatment with CL 316,243 to activate the β3-adrenoceptor for seven days did not change body weight, body fat percentage, MAT mass, or BM cellularity in either Kitl or Adipoq-Cre mice, when compared to saline controls (Figure 5A-C and Online Supplementary Figure S8A and B). Scheller et al. recently showed that MAT, compared to peripheral WAT, relatively resists lipolysis and remodeling in response to CL 316,243.[51] We measured Kitl gene expression and found that CL 316,243 strongly induced Kitl mRNA in the BM but not iWAT (Figure 5D and E). Adipoq mRNA in MAT could be also induced by CL 316,243 (Online Supplementary Figure S8C), suggesting that β3-adrenergic activation is able to remodel MAT. CL 316,243 treatment significantly increased the numbers of LSK, MP, MEP, GMP, and CLP in the wild-type bone marrow (Figure 5F-K). Interestingly, the induction of LSK and MP, in particular MEP, by CL 316,243 was diminished in Adipoq-Cre mice (Figure 5F, G, and I). The number of CMP was not affected by CL 316,243 in either control or Adipoq-Cre mice (Figure 5H). CL 316,243-induced increase in GMP and CLP was not dependent on adipocyte-derived SCF (Figure 5J and K). As a result of the loss of adipose SCF, the reduction in MEP/GMP ratio was further decreased, while the increase in CLP/CMP ratio was further augmented in Adipoq-Cre mice upon CL 316,243 treatment (Figure 5L-O). These data indicate that SCF from MAT mediates some aspects of homeostatic responses in the BM upon the β3-adrenergic activation. Nonetheless, we cannot exclude the possibility that the role of CL 316,243 on hematopoiesis is affected by lipolysis in peripheral WAT.

Figure 5.

Sympathetic nervous system (SNS)-activated hematopoiesis in control and Adipoq-Cre mice. (A and B) 12-week old Kitl and Adipoq-Cre male mice were treated with the saline vehicle or CL 316,243 for one week. (A) Representative images of femur sections. (B) Quantification of adipocyte numbers in the bone marrow (BM) from the growth plate to 4.5 mm away distally. (C) BM cellularity of 12-week old Kitl (n=7 for vehicle and n=6 for CL) and Adipoq-Cre (n=8 for vehicle and n=6 for CL) male mice were treated with the saline vehicle or CL 316,243 for one week. (D and E) Relative Kitl mRNA levels in the BM (D) or iWAT (E) of wild-type mice treated with vehicle or CL 316,243 (n=6) for one week. (F-M) 12-week old Kitl (n=7 for vehicle and n=6 for CL) and Adipoq-Cre (n=8 for vehicle and n=6 for CL) male mice were treated with the saline vehicle or CL 316,243 for one week. LSK number (F), MP number (G), CMP number (H), MEP number (I), GMP number (J), CLP number (K), the MEP/GMP ratio (L), and the CLP/CMP ratio (M) were determined by flow cytometry. (N and O) CL 316,243 treatment-induced changes in MEP/GMP ratio (N) and CLP/CMP ratio (O) in Kitl (n=6) and Adipoq-Cre (n=6) male mice. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by two-way ANOVA followed by multiple comparison using Sidak correction (F-M) or unpaired, two-tailed Student t-test (D, N, and O).

Compared to the standard housing temperature (22°C), thermoneutrality (30°C) suppresses the SNS, promotes HSPC apoptosis, and increases the radiosensitivity of mice.[53] Consistent to the previous finding,[53] we did not observe any difference in BM cellularity (Online Supplementary Figure S9A) or the frequency of LSK, MP, and CLP (Online Supplementary Figure S9B) between mice housed at 22°C and 30°C. However, thermoneutrality increased the ratio of MEP to GMP and the ratio of CLP to CMP (Online Supplementary Figure S9C and D). After being housed at 30°C for one month, Adipoq-Cre mice had similar body weight, body composition, and MAT mass to Kitl mice (Online Supplementary Figure S9E and F). Loss of adipose SCF eliminated the rise in MEP/GMP ratio (Online Supplementary Figure S9G) but further augmented the increase in CLP/CMP ratio (Online Supplementary Figure S9H). Collectively, these data show that adipocyte-derived SCF mediates part of the environmental effects, particularly those via the β3-adrenergic signaling, on HSPC function.

MAT-provided stem cell factor in the aged hematopoietic stem and progenitor cell compartment

Aging-related changes in the hematopoietic system can be attributed to cell-intrinsic and microenvironmental alterations.[54] MAT expands as a function of age in both rodents and humans,[27] we then sought to determine whether SCF from MAT contributes to altered HSPC function during aging. Young (3 months old) and middle-aged (10 months old) male mice were analyzed; we did not observe any difference in body weight between Kitl and Adipoq-Cre mice at either age (Figure 6A). Aging slightly decreased BM cellularity in control mice, which was further down-regulated by the loss of adipose SCF (Figure 6B). Strikingly, the expansion of LSK and various myeloid progenitors including CMP, MEP, and GMP observed in middle-aged mice was totally abolished in Adipoq-Cre mice (Figure 6C-G). There were more CLP in 10-month old mice than in 3-month old mice, but there was no difference in these changes between genotypes (Figure 6H). Aging increased the ratios of both MEP to GMP and CLP to CMP; however, Kitl and Adipoq-Cre mice showed similar extent of increase as the interaction effect was not statistically significant (Figure 6I and J). These data demonstrate that SCF, as a MAT niche factor, is essential for the phenotypic expansion of HSC and myeloid progenitors during aging.

Figure 6.

The effect of aging on hematopoietic stem and progenitor cells (HSPC) in Kitl and Adipoq-Cre mice. (A) Body weight of Kitl and Adipoq-Cre (n=6) male mice at three [n=9 and 10 for control and knock-out (KO), respectively] and ten (n=6 and 5 for control and KO, respectively) months of age. (B-J) Flow cytometric analyses of bone marrow from Kitl and Adipoq-Cre (n=6) male mice at three (n=9 and 10 for control and KO, respectively) and ten (n=6 and 5 for control and KO, respectively) months of age, showing total cellularity (B), LSK number (C), MP number (D), CMP number (E), MEP number (F), GMP number (G), CLP number (H), MEP/GMP ratio (I), and CLP/CMP ratio (J). Data are presented as mean±Standard Deviation. Two-way ANOVA followed by multiple comparison using Tukey’s correction was performed. (Top) P-values for interaction and between groups. *P<0.05; **P<0.01; ***P<0.001 between 3M and 10M mice in indicated genotypes.

Discussion

Hematopoietic cytokines support the developmental processes of blood cells, at least partially through rewiring the cellular metabolism. Meanwhile, many of these cytokines also act directly on diverse metabolic tissues and cells. For example, erythropoietin improves glycemic control and insulin sensitivity, prevents obesity by acting on the hypothalamus, and attenuates adipose tissue inflammation.[55] Granulocyte-macrophage colony-stimulating factor regulates lipid metabolism in the liver.[56] Interleukin 4 inhibits adipogenesis, promotes lipolysis, and also disposes glucose by enhancing insulin action.[57,58] Interestingly, a specific hemopoietin cocktail composed of SCF, FLT3 ligand, IL-6, and vascular endothelial growth factor together with bone morphogenic protein 7 induces efficient differentiation of human pluripotent stem cells into functional brown adipocytes.[37] Our previous work showed that the expression of SCF in BAT is sensitive to food availability and environmental temperature.[38] SCF overexpression activates thermogenesis in BAT and reduces weight gain, while Kit mutant mice become obese as a result of reduced energy expenditure.[38] Consistently with these previous findings, in this study we found that adipose-derived SCF was required for thermogenic gene expression in cultured adipocytes, demonstrating a cell-autonomous effect of SCF. However, no changes in UCP1 expression, body weight, or glucose metabolism were observed in Adipoq-Cre mice, suggesting a possible compensation from SCF or other factors secreted by non-adipose cells. Determining additional sources of SCF in regulating systemic metabolism is warranted in future studies.

In the BM, SCF is secreted by endothelial cells,[22] stromal cells that can be labeled by LEPR,[6,59] CXCL12,[60] Nestin-GFPlow,[60] PDGFRa,[62] N-Cadherin,[63] and Prx1-Cre,[21] and adipocytes.[32] Loss of SCF from these niche cells all leads to reduced HSPC numbers.[64,65] It is still highly debatable whether osteoblasts express SCF, but HSC frequency and function were not affected by deleting SCF or CXCL12 from Col2.3+ osteoblasts or ablating Ocn+ osteoblasts cells.[21,22,66,67] A recent study by Zhou et al. elegantly showed that BM adipocytes proliferate after irradiation or chemotherapy, and deleting SCF using Adipoq-Cre/ER inhibits hematopoietic regeneration.[32] In this study, using Adipoq-Cre-and Osx1-Cre-mediated knockout of SCF, we determined that MAT-secreted SCF is essential for HSC maintenance and hematopoiesis also in the steady-state. Zhou et al. did not observe any deficiency in HSC frequency in non-irradiated Adipoq-Cre/ER mice, probably because they used heterozygous Kitl mice as controls, in which the GFP insert disrupted one allele of the Kitl gene. Haploinsufficiency of the SCF/KIT pathway has been well documented,[22,36] and we also observed substantial defects in HSPC in heterozygous Osx1-Cre+;Kitl mice. The discrepancies between the results of our studies and those of Zhou et al. could also lie in the different Cre lines used. Adipoq-Cre only labels mature adipocytes, but not bone stroma, adipogenic progenitors, hematopoietic cells, bone lining cells, or osteoblast cells.[41] However, Adipoq-CreER also recombines in a subset of LEPR+ stromal cells in the BM.[32] On the other hand, the Adipoq-Cre line has its limitations as the Cre expression is not restricted to adulthood. However, a minimal amount of MAT is present at the age of one week in mice and MAT rapidly expands afterwards.[52] We argue that the hematopoietic defects observed in adult Adipoq-Cre mice were largely due to SCF excision during adulthood. In the BM of young mice, adipocytes are relatively rare, compared to other cell types that also express SCF. Even though more than half of Kitl mRNA and SCF protein were lost in the BM of Adipoq-Cre mice, it is still unclear why such a profound effect could be observed when adipose SCF was absent. Note that we cannot rule out the contributions of peripheral adipocytes and other Osx1+ progeny to the hematopoietic defects observed in Adipoq-Cre mice and Osx1-Cre mice, respectively. Tools that can selectively target MAT are needed to solve this enigma.

There have been a handful of studies that investigated the role of diet and obesity on the composition of the HSPC compartment and blood cell production.[9-13,15,16] Despite the inconsistencies seen in these studies, such as the different mouse models, diets, and length of treatment employed, it is generally accepted that diet-induced obesity promotes the immediate expansion of LSK stem cells but push their differentiation skewed toward the myeloid and lymphoid lineages, which will result in long-term defects in hematopoiesis upon stress and infection. Of note, only male mice were assessed in these studies. Here, we consistently found that HFD activates myelopoiesis and lymphopoiesis in male mice, which may facilitate tissue inflammation in obesity. However, HFD in females did not change the ratio of MEP to GMP in the BM or the MkE/GrMo ratio in the peripheral blood. These discrepancies may help explain the sexual dimorphism in metabolic dysfunction associated with obesity in animals and humans. It has recently been reported that estrogen signaling could control the sexual dimorphism of HSPC development.[44,45] Shown in this study, HFD-induced alterations in the frequency of HSPC populations and the composition of blood cells, despite being divergent between males and females, were both dependent on adipose SCF, underscoring the fundamental niche function of MAT.

Similar to obesity, aging is also associated with increased HSC numbers but decreased regenerative potential and skewed differentiation toward myeloid cells.[54,68,69] Both cell-intrinsic and cell-extrinsic mechanisms are accountable to these aged-related changes. MAT substantially expand in aged humans and rodents, and has been proposed as a suppressor of hematopoiesis in aging.[70] However, in this study, by deleting SCF from adipose tissues, we were able to show that adipose-derived SCF is essential for the expansion of LSK stem cells and all myeloid precursor populations in middle-aged mice. It again supports the notion that MAT provides niche factors for the HSPC, particularly the myeloid compartment. Nevertheless, it is unclear whether such requirement of adipose SCF prevents hematopoietic aging or accelerates the exhaustion of HSPC in the bone marrow. Functional characterization of these HSPC in aged Adipoq-Cre mice is required in future experiments.

In summary, we demonstrate that MAT is a functionally important source of SCF in steady-state hematopoiesis and required for HSPC to cope with metabolic stresses in obesity and aging.