Spatial and Biochemical Interactions between Bone Marrow Adipose Tissue and Hematopoietic Stem and Progenitor Cells in Rhesus Macaques

Abstract

Recent developments in in situ microscopy have enabled unparalleled resolution of the architecture of the bone marrow (BM) niche for murine hematopoietic stem and progenitor cells (HSPCs). However, the extent to which these observations can be extrapolated to human BM remains unknown. In humans, adipose tissue occupies a significant portion of the BM medullary cavity, making quantitative immunofluorescent analysis difficult due to lipid-mediated light scattering. In this study, we employed optical clearing, confocal microscopy and the nearest neighbor analysis to determine the spatial distribution of CD34+ HSPCs in the BM in a translationally relevant rhesus macaque model. Immunofluorescent analysis revealed that femoral BM adipocytes are associated with the branches of vascular sinusoids, with half of HSPCs localizing in close proximity of the nearest BM adipocyte. Immunofluorescent microscopy and flow cytometry analysis demonstrate that BM adipose tissue exists as a multicellular niche consisted of adipocytes, endothelial cells, granulocytes, and macrophages. Analysis of the BM adipose tissue conditioned media using liquid chromatography-tandem mass spectrometry revealed the presence of multiple bioactive proteins involved in regulation of hematopoiesis, inflammation and bone development with many predicted to reside inside microvesicles. Pretreatment of purified HSPCs with BM adipose tissue-conditioned media, comprising soluble and exosomal/microvesicle-derived factors, led to enhanced proliferation and an increase in granulocyte-monocyte differentiation potential ex vivo. Our work translationally extends extensive studies in murine models, indicating that BM adipose tissue is a central paracrine regulator of hematopoiesis and in nonhuman primates and possibly in humans.

Introduction

The bone marrow (BM) microenvironment is organized into distinct cellular niches that regulate various homeostatic and regenerative aspects of hematopoietic stem and progenitor cell (HSPC) function, including self-renewal, proliferation, and differentiation [1, 2]. Under homeostatic conditions, HSPCs exist in a quiescent state, but can proliferate and differentiate into mature blood cells in response to infection [3–5], irradiation [6, 7], obesity [8], and with age [9]. Irradiation, obesity, aging, and other conditions, including caloric restriction, physical inactivity, and estrogen depletion, are also associated with the expansion of BM adipose tissue derived from adipo-osteogenic mesenchymal progenitor cells [10]. In humans, BM adipose tissue occupies a significant portion of the bone medullary cavity that exhibits age-dependent changes in extent and potentially function [10]. In contrast, mice do not accumulate a significant amount of BM adipose tissue until advanced age [8, 10–12]. These species differences limit the translatability of mouse data to humans.

Although adipocytes represent one of the most abundant cell types present in the human BM, their functional role in regulation of hematopoiesis remains controversial. While earlier studies demonstrated that BM adipocytes inhibit hematopoiesis in mice [13], later studies showed that BM adipocytes are positive regulators of HSPC repopulation following irradiation [14] or metabolic challenge [15]. Since a majority of studies of BM adipose tissue have employed rodent models, a greater understanding of the primate BM environment and the function of BM adipose tissue is necessary to determine the relationship between hematopoiesis and BM adipogenesis in humans. Although humans and rodents share many aspects of hematopoietic function, there are important species differences in the hematopoietic differentiation cascade [16] as well as the length of the HSPC lifecycle [17–22]. Furthermore, compared to mice, humans have evolved to accommodate a significantly higher number of blood cells, which imposes a proportionally greater proliferative demand on their hematopoietic cells.

In the present study, we explored the spatial and functional relationship between BM adipocytes and HSPCs using rhesus macaques, whose hematopoietic and skeletal systems share significant similarity with that of humans [23–25]. To reduce light scattering imposed by an adipocyte-rich environment, we employed optical tissue clearing and performed confocal microscopy and nearest neighbor analyses to determine the distribution of HSPCs within the BM. Furthermore, we characterized the cellular secretome of isolated BM adipose tissue and tested its effects on HSPC proliferation and differentiation. Our studies demonstrate that BM adipose tissue plays an active paracrine role in regulation of the BM microenvironment, with implications for both hematopoiesis and bone development.

Materials and Methods

Animals

BM samples used in this study were obtained from 4 juvenile male and 10 adult male rhesus macaques studied under protocols previously approved by the Institutional Animal Care and Use Committee of the Oregon National Primate Research Center. Animals were maintained on a regular chow diet consisting of two daily meals of Purina LabDiet fiber-balanced monkey chow (15% calories from fat, 27% from protein, and 59% from carbohydrates; no. 5000; Purina Mills, St. Louis, MO), supplemented with fruits and vegetables. Animals were euthanized according to procedures recommended by the Panel on Euthanasia of the American Veterinary Association. The night prior to necropsy, food was withheld. Before necropsy, animals were sedated with ketamine in the home cage, transported to the necropsy suite, treated with pentobarbital (25 mg/kg), and exsanguinated by severance of the descending aorta.

Preparation of bone sections, immunohistochemistry, and optical clearing

Femur bones were cut in half cross-sectionally and fixed for 48 hours at 4°C in 4% paraformaldehyde/PBS. Following fixation, the femurs were sectioned into 2–3 mm adjacent slices using a diamond-coated saw (Mar-med, model 80 mini-band bone saw). Bone sections were stored for up to 2 months at 4°C in PBS supplemented with 5 mM sodium azide. For the experiments shown in Figure 1C (fluorescent dye staining), bone sections were placed on a rocking platform and incubated at room temperature for 4 hours with HCS LipidTox Deep Red (Invitrogen, Carlsbad, CA, 1:200) and isolectin GS-IB4, Alexa Fluor^™ 488 conjugate (Invitrogen, 20 μg/ml) in the blocking solution containing 0.5 ml 10% donkey serum, 0.25% Triton-X100/PBS. Longer incubation with fluorescent dyes and the detergent may result in higher background and fragmentation of lipid droplets. Following incubation, bone sections were washed 6 times x 15 minutes in 0.25% Triton-X100/PBS and once with PBS.

Figure 1. Characterization of the adipocyte niche in the rhesus BM.

(A) Spatial organization of vascular niches and BM adipose tissue in the rhesus macaque femur. Central sinus, centrally located vein; sinusoids, branches of central sinus; artery, central arterial blood vessel (smaller arterioles are not shown). Distal and central parts of the femur are enriched in BM adipose tissue; proximal femur is enriched in hematopoietic cells. (B) Bone sections representing the proximal (p), medial (m), and distal (d) parts of the femur from a male rhesus macaque 2 years and 229 days of age. (C) Representative image of the proximal femur stained with Hoechst (DNA), isolectin GS-IB4 Alexa 488 (blood vessels), and HCS LipidTox Deep Red (adipocytes); arrows, adipocytes associated with blood vessels; CS, central sinus. (D) Low- and (E) high-magnification images of BM stained with antibodies to CD31 and the lipid dye HCS LipidTox Deep Red; E, right panel, BM adipocytes (a) stained with antibodies to perilipin-1 without HCS LipidTox Deep Red. To retain HCS LipidTox Deep Red in the lipid droplets of BM adipocytes, immunohistochemistry experiments presented in this figure were performed without optical clearing.

For the experiments shown in Figures 1D–E and 2C, bone sections were incubated for 4 hours at 4°C in the blocking solution and then with HCS LipidTox Deep Red (where indicated), sheep anti-human CD31 (R&D Systems, Minneapolis, MN, clone AF806), rabbit anti-Perilipin A (Fitzgerald Industries International, Acton, MA), and mouse anti-human CD34 (BD Pharmingen, San Jose, CA, clone 581) antibodies (1:100 dilution) in the blocking solution 4°C for 24 hours. Sections were washed 6 times x 15 minutes with 0.25% Triton X-100/PBS, incubated with fluorescent secondary antibodies (donkey) (Invitrogen, 1:1000 dilution) and Hoechst 33342 (Sigma-Aldrich, St. Louis, MO, 10 μg/ml) for an additional 24 hours, and then washed extensively with 0.25% Triton-X 100/PBS.

For tissue clearing (Figures 2A and B and 3B), each bone section was washed sequentially (4 ml, polypropylene tubes, 20 minutes each at room temperature) with the following solutions: PBS, 50% methanol/PBS, 80% methanol/DI water, 100% methanol, 20% DMSO/methanol twice, 80% methanol/DI water, 50% methanol/PBS, PBS twice, 1% Triton X-100/PBS twice before further staining procedures. Immunohistochemistry procedures were performed at 4° C on a rocking platform as described above. Stained bone sections were washed with following solutions (20 min per wash): 10 ml 50% methanol /PBS, 10 ml 70% methanol/PBS; 10 ml 100% methanol. Excess methanol was absorbed with a paper towel and the samples were incubated in 2 ml Visikol HISTO-1 at room temperature for 2 hours followed by 2 ml Visikol HISTO-2 (Visikol, Inc., Whitehouse Station, NJ) for additional 2–4 hours. Samples retained fluorescence for 3–4 days.

Figure 2. Primate CD34+ HSPCs localize in close proximity to BM adipocytes.

(A) Representative confocal image of optically cleared BM from the proximal metaphysis of a 7-year-old male rhesus macaque stained with antibodies to CD34 and CD31; s, sinusoids; arrows, CD31+CD34+ endothelium; a, adipocytes; a representative example of the adipocyte niche is outlined by the dotted line. (B and C) High-magnification confocal images depicting the localization of CD34+ HSPCs (asterisks) next to the BM adipocytes (a). (D) Nearest-neighbor analysis of distances between CD34+ HSPCs vs random dots and BM adipocytes; representative traces from a single animal; “NND,” nearest neighbor distance. Inset depicts the theoretical position of HSPCs (red) or random dots (blue) relative to adipocytes; pm, plasma membrane; ldm, lipid droplet membrane. (E) NNDs between HSPCs and BM adipocytes. Bars are means ± SD; n=3 males of 7–10 years of age; each point represents one animal and was calculated as the average NND derived from 4 nonadjacent confocal slices. (F and G) Digital analysis of anatomical structures in the BM of a 7-year old male rhesus macaque; CD34+ HSPCs (red), BM adipocytes (grey shapes), CD31+CD34+ endothelium (yellow lines), random dots (blues). Anatomical boundaries from the same BM sample: central sinus (F); the bone (G).

Figure 3. BM occupancy by adipocytes regulate their spatial proximity to HSPCs.

(A) Representative images of bone sections from a 7-year-old adult (ADT) and a 2-year-old juvenile (JUV) male rhesus macaques. (B) Representative confocal image depicting the localization of CD34+ HSPCs (asterisks) next to BM adipocytes (a) in the femoral BM of a juvenile rhesus macaque; arrows, CD34+ capillaries. (C) Nearest neighbor distances (NNDs) between HSPCs and BM adipocytes. Bars are means ± SD; n=3 nonadjacent confocal slices derived from one juvenile animal (JUV1). (D) The principle of Euclidian NND analysis; algorithm computes the distances between the centroid of each object classified as “HSPC” (3 cells are shown) and the nearest coordinate associated with the outer boundary of the BM adipocyte (a); solid red lines, NNDs; black dotted lines, distances between HSPC#2 and distant adipocytes. (E) Scatter plot showing the distribution of NNDs between HSPCs and BM adipocytes across 4 experimental animals (3 adults and 1 juvenile). Each data column represents one nonadjacent confocal slice; bar-median NND.

Image acquisition and analysis

Bone sections were transferred into a 35-mm glass-bottom culture dishes (MatTek, Ashland, MA) containing 300 μl of Visikol HISTO-2. Confocal microscopy was performed using a SP5 AOBS spectral confocal system (Leica, Wetzlar, Germany). Images were collected in sequential mode, at 2-μm x 300–400 intervals, using HC PL FLUOTAR 10.0 × 0.30 or ×20 PL APO NA 0.70 dry objectives. For each bone section, 3 random fields of 600 to 800-μm z-stacks were analyzed, including the central (adjacent to central sinus), middle, and outer (adjacent to the bone surface) areas. Confocal images were opened with the LOCI plug-in Fiji data browser and analyzed as follows. For detection of CD34+ cells, each confocal slice was smoothened using the median filter (rolling ball value, 2.0) and subjected to background subtraction and thresholding. CD34+ cells were detected using the “Analyze Particles” function and the XY coordinates (centroids) of each cell were recorded. Adipocyte boundaries were detected manually, or using the adipocyte detection algorithm (Supplementary Materials). In the absence of lipid staining, adipocytes appear as dark circular object, representing the interior of the lipid droplet. Thus, the adipocyte detection algorithm underestimates the physical boundaries of the cell by a small fraction proportional to the combined thickness of the cytoplasm and the plasma membrane.

Nearest-neighbor distance analysis was performed using two alternative methods. For calculating probability densities of distance distribution (Figure 2D and Supplementary Table S1), we used the Mosaic Interaction analysis plug-in for Fiji [26]. Random dot analysis was performed after masking adipocyte areas in the image. The number of random dots was normalized to the number of CD34+ cells in each field of view. For each animal, the probability densities were averaged (n=3–4 confocal slices) and used for the binning procedures. For each animal, the average probability densities corresponding to each 10-micron bin were combined to produce the probability density of the sum. The probability density distribution corresponding to the nearest distances between HSPCs and adipocytes were compared with that of the random dots, using the Kolmogorov-Smirnov test (Supplementary Table S1). For calculating Euclidian nearest neighbor distances (Figure 2E and 3C, E and F), we used the brute-force method that computes the distances between the centroid of each object being classified as HSPC and the nearest pixel associated with the BM adipocytes (Figure 3D).

Isolation of BM adipose tissue and conditioned media

BM adipose tissue was isolated from the femur of 6 adult male rhesus macaques. BM was collected at necropsy, placed in 25 ml X-Vivo™ 10 media (Lonza, Basel, Switzerland) and processed within 1 hour. BM was gently disrupted using a 25-ml syringe loaded with a blunt needle, and the cell suspension was filtered through a 100-μm cell strainer (Life Sciences, Durham, NC). The cell suspension was layered on a Ficoll density gradient followed by centrifugation for 30 min at 1500 rpm at room temperature. The BM adipocyte-enriched fraction was collected from the top of the gradient, washed at room temperature twice with 30 ml X-Vivo^™ 10 media by gravity, and incubated for 0 or 48 hours at 37°C in 4 Millicell insets (EMD Millipore, pore size, 4-μm; diameter, 12-mm) in a total volume of 1–1.5 ml. Conditioned media from the bottom of the wells was collected, pooled and centrifuged at 300xg for 10 min to remove cell debris, followed by centrifugation for 2 hours at 50,000xg. We also tested conditioned media prepared by centrifugation for 1 hour at 20,000xg, which contained similar biological activity towards the CD34+ cells as the 50,000xg supernatant. Concentrated conditioned media was aliquoted and snap-frozen in liquid nitrogen. For secretome analysis, BM adipocyte fraction conditioned media was dialyzed against PBS and processed as described below.

Secretome analysis

Conditioned media was diluted with 1 ml of ice-cold homogenization buffer (150 mM NaCl, 50 mM HEPES, pH 8.5), mixed with SDS to a final concentration of 2.5%, and sonicated three times at 2 watts for 5 seconds each with 30-second rests between the sonication pulses. Final protein concentration was determined using the Pierce bicinchoninic acid (BCA) protein assay with a BSA standard (Thermo Fisher Scientific, Rockford IL). Sample extraction and analysis was conducted as described in Supplementary Materials.

ex vivo staining of isolated BM adipose tissue

200 μl of the BM adipocyte fraction from the top of the gradient were transferred to the bottom of a 15-ml conical tube and incubated for 15 minutes at 37°C in the presence of Hoechst 33342 (10 μg/ml), isolectin GS-IB4, Alexa Fluor^™ 488 conjugate (20 μg/ml), HCS LipidTox Deep Red and the cell viability dye calcein red-orange AM (Invitrogen, 6 μg/ml). Cells were diluted in 1 ml 4% paraformaldehyde/PBS and fixed at room temperature for 15 minutes. Following fixation, cells were diluted with 10 ml PBS and centrifuged for 7 minutes at 200xg. 50–100 μl of floating cell suspension were transferred onto the glass slide and overlaid with the round coverslip without sealing. To prevent adipocyte squishing between glass surfaces, a thin glass spacer was introduced at the edge of the coverslip. The slides were carefully inverted and the images were recorded using a Leica SP5 AOBS spectral confocal system. Approximately 30–50% of BM adipocytes appeared viable after 48 hours of ex vivo incubation. For histological evaluation, a small aliquot of isolated BM adipose tissue was transferred onto a glass slide followed by standard Hematoxylin Eosin or Wright Giemsa staining. Images were collected using a VS120 Virtual Slide Microscope (Olympus Corporation, Tokyo, Japan) using a 20x objective.

Flow analysis of adipocyte-bound and adipocyte-free fractions

BM adipose tissue was isolated from both femurs of adult male rhesus macaques as follows. BM tissue was gently disrupted using a 25-ml syringe loaded with a blunt needle, and the cell suspension was filtered through a 70-μm cell strainer (Life Sciences). This suspension was centrifuged for 10 minutes at 300xg at room temperature, and the top adipocyte-containing layer was transferred to a tube containing 20 ml X-Vivo™ 10 (room temperature). The tissue was mixed by gentle inversion and left at room temperature for 15 minutes. For isolating the adipocyte-free fraction, the cell pellet was resuspended in 10 ml erythrocyte lysis buffer and incubated at 37°C for 5 minutes. The cell suspension was diluted with 40 ml PBS and centrifuged for 10 minutes at room temperature. The resulting cell pellet was washed with PBS, resuspended in 1 ml PBS, and kept on ice. For isolating the adipocyte-bound fraction, the top adipocyte layer was transferred to a new tube containing 10 ml of collagenase solution (30 mg collagenase type II (GIBCO, Waltham, MA) dissolved in 3.5% BSA/PBS/2 mM CaCl₂) and incubated at 37°C for 40 minutes in a water bath. During incubation, the tube was swirled by hand every five minutes and tissue lysis was monitored to ensure complete digestion. The cell suspension was diluted with PBS, washed by centrifugation twice, and resuspended in 1 ml PBS. Adipocyte-bound and adipocyte-free fractions were stained with anti-CD45 PE-Cy7 (BD Biosciences, clone D058–1283) using the eBioscience FoxP3/Transcription Factor Staining Buffer Set according to the manufacturer’s instructions (ThermoFisher Scientific). Cells were collected on a BD LSRII and analyzed as described below.

Isolation of CD34+ cells and flow cytometry

Mononuclear cells were collected from the Ficoll gradient interface and HSPCs were isolated using rhesus-specific antibodies to CD34 (BD Pharmingen, clone 563) and the Miltenyi Biotec (Bergisch Gladbach, Germany) anti-PE MACS magnetic bead system using the supplied buffers and reagents. Specifically, 3–4×10⁸ BM mononuclear cells derived from a single femur were resuspended in 1 ml wash buffer and mixed with 200 μl FcR blocking reagent (Miltenyi Biotec) and 100 μl PE-conjugated antibodies to CD34 (BD Pharmingen, clone 563). Cells were incubated on a rocking platform for 20 min at 12–15°C, washed by centrifugation at 1000 rpm twice with 20 ml wash buffer, resuspended in 0.5–1 ml wash buffer supplied with 200 μl anti-PE MicroBeads (Miltenyi Biotec) and incubated for an additional 15 min. Cells were washed with 20 ml of wash buffer, resuspended in 1 ml wash buffer and isolated on the magnetic MS columns (Miltenyi Biotec) according to the manufacturer’s instructions. To increase cell viability, CD34+ cells were eluted in 1 ml StemSpan SFEMII media (Stem Cells Technologies, Vancouver BC, Canada). Typical cell yields were ~2–4×10⁶ CD34+ cells from 4×10⁸ BM mononuclear cells derived from one femur. CD34+ cells were cryopreserved using CryoStor CS10 (Stem Cells Technologies). For flow cytometry analysis, CD34+ cells were stained with the viability dye Zombie Aqua (BioLegend, San Diego, CA). Samples were then blocked for 45 min in PBS containing 5% each human and mouse serum, sodium azide, and EDTA (FACS blocking buffer) at 4°C. Samples were stained in FACS blocking buffer using anti-CD34-PE (BD Pharmingen, clone 563), anti-CD38-APC (Caprico Biotechnologies, clone OKT10), and anti-Ki67-FITC (BD Pharmingen) for an additional 45 min at 4°C. Finally, samples were washed in FACS buffer and analyzed using an LSRII flow cytometer (BD Biosciences, San Jose, CA). Data were acquired using BD FACS Diva (BD Biosciences) and post-analysis performed using FlowJo v10 (Tree Star, Ashland, OR) and Prism v6 (GraphPad Software, San Diego, CA).

Proliferation assay

The rate of CD34+ cell division in liquid culture was quantified using a CFSE-Cell labeling kit (Abcam, Cambridge, MA). Briefly, 1–2×10⁶ CD34+ cells were washed with 15 ml PBS, resuspended in 1 ml PBS at 37°C and labelled with 1 μM CFSE for 10 min under gentle agitation. The reaction was quenched at room temperature for 5 min by adding 15 ml X-Vivo™ 10 media (Lonza), centrifuged for 5 min, washed once with X-Vivo™ 10 media and resuspended in 1–2 ml StemSpan SFEMII media (Stem Cell Technologies). CFSE-labelled CD34+ cells were cultured ex vivo in non-adhesive tissue culture untreated plates for indicated periods of time supplied with human cytokines (100 ng/ml stem cell factor (SCF), thrombopoietin (TPO), and Fms-related tyrosine kinase 3 (Flt3); (Stem Cell Technologies) [24] in the continuous presence or absence of BM adipose tissue conditioned media (25% of total volume). At each time point, an aliquot of cell suspension was washed, transferred into an 8-well imaging plate (Lab-Tek II chambered #1.5 German Coverglass System; Nunc, Roskilde, Denmark) precoated with RetroNectin (Takara Shuzo Co., LTD, Kyoto, Japan), and cells were allowed to adhere to the bottom of the plate for 30 min. Adhered cells were fixed with 2% paraformaldehyde/PBS for 10 min, washed 3 times with PBS, stained with Hoechst (10 μg/ml) and analyzed by confocal microscopy as described below. Parallel cultures of CD34+ cells maintained without CFSE were stained with mouse anti-Ki67-FITC (BD Pharmingen) and mouse anti-human CD38 (Caprico Biotechnologies, clone OKT10). For CFSE and antibody staining experiments, confocal images were collected using 5 random fields per each well (200–400 cells per each time point) and the HC PL FLUOTAR 10.0 × 0.30 .30 or ×20 PL APO NA 0.70 objectives. Images were processed in Fiji. The Hoechst channel was used to generate the binary mask and the outline of each cell (region of interest) was detected using the “Analyze Particles” algorithm. Regions of interest were stored in the ROI manager and then applied sequentially onto channel 1 (Hoechst) and channel 2 (CFSE). Mean fluorescent intensities of individual cells were calculated at days 1, 2, 3, and 5.

Results

Adipocytes and vascular sinusoids form a distinct niche in rhesus BM

To study the spatial organization of the BM niche in rhesus macaques, freshly isolated femurs were fixed, sectioned into circular segments, prepared for staining, and analyzed by confocal immunofluorescence microscopy (Supplementary Figure S1A and B). Morphologically, BM adipose tissue exhibited a gradient of distribution along the longitudinal axis of the femur, being enriched in the distal metaphysis and in the diaphysis. The proximal metaphysis contained a high fraction of hematopoietic cells, while BM adipose tissue was enriched in the area surrounding the central sinus (Figure 1A, B and D, and Supplementary Figure S1B). The lateral gradient of BM adipose tissue segregation was more apparent in the femur of 1 to 3-year-old juvenile rhesus macaque (equivalent of 2.5 to 7.5-year-old humans; Figure 1B). In 7 to 10-year-old adult rhesus macaques (equivalent of 18 to 28-year-old humans), the femur was progressively more abundant in BM adipose tissue, effacing the proximal part of the metaphysis (Supplementary Figure S1B). In older animals, the entire femur was composed of BM adipose tissue (data not shown), consistent with the relocation of hematopoiesis from within long bones to the axial skeleton as observed in adult humans [10].

Immunofluorescent analysis of BM from juvenile animals showed that lipid droplet-laden adipocytes were associated with isolectin-positive blood vessels (Figure 1C, and Supplementary Figure S2A). Furthermore, BM adipocytes identified by perilipin staining formed frequent contacts with CD31+ branches of the central vascular sinus (Figure 1E, arrows). Consistent with recent ultrastructural electron microscopy studies of rodent BM adipose tissue [27], a subpopulation of rhesus BM adipocytes associated with sinusoids contained smaller lipid droplets surrounding the large central lipid droplet (Figure 1E, middle panel and Figure S2A). Using negative-contrast staining, BM adipocytes appeared to form contiguous contacts with the vascular endothelium outlining the bifurcations (branching points) of sinusoidal vessels (Figure 1E, right panel and Supplementary Figure S2B and C). Thus, adipocytes and vascular sinusoids form a distinct BM niche in rhesus macaques.

HSPCs are spatially associated with BM adipocytes in adult BM

To delineate the spatial distribution of HSPCs in the adult proximal metaphysis that contains a mixed population of hematopoietic cells and adipocytes (Supplementary Figure S1B), we performed whole-mount microscopy of BM using validated antibodies to the CD34 epitope expressed on HSPCs. Our initial attempts to visualize hematopoietic and endothelial cells in rhesus macaque BM failed due to significant light scattering produced by the highly refractile lipid droplets of BM adipocytes. To reduce light scattering, immunolabelled bone sections were optically cleared using Visikol [28], resulting in significantly improved optical transparency of the lipid-rich BM (Supplementary Figure S1C). Confocal microscopy analysis of the adult femur revealed that CD34+ HSPCs were frequently detected in close proximity with, or directly attached to, the BM adipocyte plasma membrane (Figure 2A–C, F and G and Supplementary Figure S3). Nearest-neighbor analysis [26] showed that approximately 50% of CD34+ HSPCs were detected at high probability within 10 microns (approximately one cell diameter) of the nearest adipocyte (Figure 2D, E and Supplementary Table S1). The spatial localization of HSPCs in proximity to BM adipocytes was significantly different from that of randomly placed dots (Figure 2D and Supplementary Table S1). A minor fraction of HSPCs localized near the CD34+/CD31+ endothelium (Figures 2A, arrows, F, G and Supplementary Figure S4, arrows and Supplementary Table S1), presumably representing BM arterioles [29, 30], although previous studies have shown that both CD34 and CD31 are also expressed on human [31–34] and rhesus [35–37] endothelial progenitor cells. Furthermore, CD34+ HSPC distribution was not biased toward the central sinus or the bone surface (Figure 2F and G). In summary, our studies demonstrate that approximately half of adult HSPCs were associated with BM adipocytes, while the rest of HSPCs localized within 30 microns of the nearest adipocyte.

BM adipocyte crowding controls their spatial separation from HSPCs in juvenile BM

Compared to the adult BM, the juvenile BM contains fewer adipocytes that occupy a proportionally smaller area of the BM (Figure 2A and B and Figure 3A and B). To determine the spatial distribution of HSPCs in the BM, we computed Euclidian distances between the centroid of each HSPC and the nearest BM adipocyte (“Materials and Methods” and Figure 3D). As observed with adult BM, almost half of HSPCs were detected within 10 microns of the nearest adipocyte in the juvenile BM (Figure 3C). Importantly, spatial data analysis from several experimental animals suggests that the median and maximal distances from HSPCs to the nearest adipocyte correlated inversely with the total area occupied by BM adipocytes (Figure 3E and F; Supplementary Figure S5). Collectively, the nearest neighbor distance analysis demonstrated that a significant fraction of HSPCs in adult and juvenile animals may form nonrandom contacts with BM adipocytes, and that cell density inside BM may control the spatial separation between BM adipocytes and HSPCs.

BM adipose tissue comprises several cell types

The close association of HSPCs with BM adipocytes demonstrated by confocal microscopy implies a potential interaction between these cells. To begin delineating paracrine signaling between BM adipocytes and HSPCs, we first developed the protocol for isolating and ex vivo culturing of BM adipose tissue. Our preliminary studies demonstrated that collagenase digestion of the rhesus BM resulted in adipocyte lysis. To overcome this technical problem, we applied a gentle mechanical force for disrupting BM tissue. Femoral BM was mechanically dissociated, passed through a filter to get remove large cellular aggregates, and BM adipocytes were isolated by floatation, as described in Materials and Methods. The morphology and viability of BM cells was examined by confocal microscopy following fluorescent dye labeling, which was performed either immediately after cell isolation or following a 48-hour ex vivo incubation period. BM adipocytes labelled with a neutral lipid-specific dye appeared in small groups (n=1–5 adipocytes) and were frequently associated with isolectin-positive cells and tubular structures, presumably representing endothelial cells and fragments of vascular sinusoids, (Figure 4A). BM adipocytes appeared to form contacts with groups of smaller cells (Figure 4A and B). These cells associated with BM adipocytes remained viable even after a 48-hour ex vivo incubation period (Figure 4A, lower panel). Surprisingly, BM adipocytes appeared to be inviable immediately after isolation, possibly due to the mechanical stress, but recovered following ex vivo incubation, as evidenced by the accumulation of a cell viability dye in the adipocyte cytoplasm (Figure 4A, lower panel, arrows). Histological assessment of freshly isolated BM adipose tissue revealed the presence of cells with both uni- and multilobulated nuclear morphology (characteristic features of monocytes/marcrophages and granulocytes, respectively) associated with the plasma membrane of BM adipocytes (Figure 4B and C). To further assess the composition of leukocytes bound to BM adipocytes, we separated the adipocyte-bound and adipocyte-free leukocyte fractions and assessed them by flow cytometry (Materials and Methods). This analysis consisted of a basic anti-CD45 stain (pan-leukocyte) compared against each cell’s side-scatter profile (measure of cell complexity and granularity). We found that granulocytes were highly enriched in the adipocyte-bound fraction, while mononuclear cells (lymphocytes and monocytes) were enriched in the adipocyte-free fraction (Figure 4D and E).

Figure 4. BM adipose tissue is a multicellular niche.

(A) Morphological assessment of isolated BM adipose tissue before (0 h) and after 48-hour ex vivo incubation. Isolated BM adipose tissue was labeled with the cell viability dye calcein red-orange AM, endothelial-specific isolectin GS-IB4, Alexa Fluor^™ 488 conjugate, lipid droplet-specific HCS LipidTox Deep Red, and nuclear Hoechst 33342; BF, fluorescent images overlaid with bright field; max, maximal z-projections; z1-z3, confocal planes; arrows, calcein-positive BM adipocytes. (B and C) Histological evaluation of isolated BM adipose tissue stained with Wright Giemsa (WG) and Hematoxylin-Eosin (HE); arrows, multilobar cells (granulocytes); arrowhead, putative oil released by adipocytes. (D) Isolation of adipocyte-bound (AB) and adipocyte free (AF) fractions. BM was disrupted mechanically and separated by centrifugation. AB fraction was treated with collagenase to release bound leukocytes and AF fraction was cleared of red blood cells (RBC) before flow cytometry analysis. (E) AB and AF fractions were stained with anti-CD45 antibodies (pan-leukocyte) and compared against each cell’s side scatter profile (measure of cell complexity and granularity); % of live cell populations are indicated.

BM adipose tissue stimulates the proliferation and differentiation of HSPCs

Previous studies have shown that vascular sinusoids provide a supportive niche and exit sites for mature hematopoietic cells [38–42]. Because our microscopy analysis demonstrated that a significant fraction of CD34+ cells localized next to adipocytes, we hypothesized that BM adipocytes may secrete paracrine factors that support HSPC function en route to the circulation. Thus, we tested the effect of rhesus BM adipose tissue conditioned media on the proliferation and differentiation rates of isolated HSPCs ex vivo. CD34+ HSPCs were labelled with the cell proliferation dye carboxyfluorescein succinimidyl ester (CFSE) and cultured ex vivo with the human hematopoietic cytokines SCF, TPO, and Flt3L [24] in the presence of control or BM adipose tissue conditioned media collected after a 48-hour incubation period. BM adipose tissue-treated CD34+ cells exhibited accelerated proliferation compared to control cells, as evidenced by the growing proportion of CFSE^dim cells after 2–3 days in culture. Untreated control cells, by contrast, remained relatively quiescent, and their average fluorescence reached half-maximum values only at day 5 (Figure 5A and B and Supplementary Figure S6). Liquid culture of CD34+ cells maintained in the presence of BM adipose tissue conditioned media produced a higher number of Ki67+/CD38+ colonies (Figure 5C and D). Consistent with this finding, microscopic and flow cytometry analyses of CD34+ cells isolated from rhesus femur showed that CD34+/CD38+ cells were more abundant and more proliferative (based on Ki67 expression) than CD34+/CD38- cells (Supplementary Figure S7). To test the effect of BM adipose tissue on HSPC differentiation, purified CD34+ HSPCs were incubated in the presence of control or BM adipose tissue conditioned media and tested in a colony-forming assay. CD34+ cells pre-treated with BM adipose tissue conditioned media formed a higher number of granulocyte-monocyte (GM) colony-forming units compared to control conditions (Figure 5E and F) when cultured on MethoCult supplied with 100 ng/ml SCF, TPO, and Flt3L. Collectively, these results implicate factors released by BM adipose tissue in enhancing proliferation and differentiation of CD34+ HSPCs ex vivo.

Figure 5. BM adipose tissue stimulates proliferation and differentiation of HSPCs ex vivo.

(A-B) Dye dilution assay using CD34+ cells isolated from a 2-year old male rhesus macaque. (A) Representative fields of view showing CFSE-labelled CD34+ cells after incubation in suspension for 1, 2, 3, and 5 days in StemSpan media supplemented with 100 ng/ml SCF, TPO and Flt3L in the presence or absence BM adipose tissue conditioned media (CM). At each time point, a cell aliquot was collected and cells were adhered to fibronectin-coated plates. Cells were fixed and stained with Hoechst, and images were recorded by confocal microscopy. (B) Quantification of intracellular CFSE fluorescence of CD34+ cells isolated from a 2–year old male rhesus macaque. Each data point represents an individual CD34+ cell. Mean CFSE fluorescent values are indicated; T-test * p<0.05. CFSE-labeling experiment was repeated twice, using a different animal (Supplementary Figure S7). (C) Immunolabeling of HSPCs; unlabeled CD34+ cells were incubated in the presence or absence BM adipose tissue conditioned media for 5 days, adhered to the imaging plates and stained with Hoechst and antibodies to Ki67 and CD38. (D) Bars are means ± SEM, n=4394 cells; ^** p<0.001. This experiment was repeated twice with similar results. (E) CFU assay; representative examples of granulocyte-monocyte (GM), erythroid (E) and multilineage progenitor (GEMM) colonies recorded on day 10; insets, early hematopoietic colonies recorded on day 7. (F) 2000 cells were plated onto MethoCult™ (H4435 Enriched) supplied with 100 ng/ml SCF, TPO and Flt3L. Bars are means ± SEM, n=3 animals (7-year-old males). * T-test p<0.01. Each biological data point represents the mean of three replicate experiments.

BM adipose tissue is a rich source of biologically active proteins

To identify potential BM adipocyte-derived mediators of HSPC proliferation and differentiation, we carried out a quantitative proteomic analysis of the BM adipose tissue-conditioned media clarified by 50,000xg centrifugation (Materials and Methods). The resulting supernatant containing both secreted soluble and microvesicle/exosomal proteins was analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Overall, 994 proteins were found to be released from BM adipose tissue (Supplementary Table S2 and Supplementary Figure S8). Several of the most abundant proteins identified in BM adipose tissue-conditioned media, including TGFB1 [43], FBLN1 [44], IGFBP2 [45] LGALS1 [46, 47], TIMP1 [48–50], and C3 [51], have been implicated in the positive regulation of HSPC differentiation, motility, adhesion, and mobilization from the BM to the peripheral blood (Figure 6A and Supplementary Table S2). Importantly, BM adipose tissue-conditioned media was highly enriched in several S100 calcium-binding proteins, with S100A8/9 showing the highest levels (Figure 6A and Supplementary Table S2). String pathway analysis of BM adipose tissue secretome revealed an enrichment for complement, focal adhesion, and extracellular-receptor interactions/proteolytic remodeling pathways (Figure 6B). A significant fraction of proteins identified in the BM adipose tissue secretome contained serpine, EGF-like, and growth factor receptor cysteine-rich domains (Figure 6C). Using the DAVID Bioinformatics tool, we determined that 430 proteins (43%) of the total number identified in BM adipose tissue conditioned media were of microvesicular/exosomal origin (Supplementary Table S2). Collectively, this proteomic analysis of BM adipose tissue-conditioned media reveals its complex biochemical composition and enrichment for proteins with potential paracrine activity.

Figure 6. BM adipose tissue is a rich source of hematopoietic factors and adhesion molecules.

(A) String analysis of the top 200 proteins released by BM adipocytes. The graph represents the String interaction network depicting the KEGG pathway enrichments analysis. Additional analysis is included in Supplementary Table S2. BM adipose tissue was isolated from six 7 to 10-year old rhesus macaque males. BM adipose tissue was incubated at 37°C for 48 hours. Following incubation, media was collected, cleared by ultracentrifugation, dialyzed against PBS, and analyzed by mass spectrometry as described in “Materials and Methods.” (B) KEGG pathways analysis of BM adipose tissue secretome (20 top pathways are shown). (C) Top protein domains identified in BM adipose tissue secretome.

Discussion

In this study, we performed a detailed quantitative analysis of the spatial distribution of HSPCs in the BM of rhesus macaques. Immunofluorescent analysis of the BM niche in humans and nonhuman primates has been limited due to extensive fatty infiltration that results in lipid-mediated light scattering. While bone optical clearing has been previously used to study the organization of the BM niche in mice [38], this methodology had not been applied for clearing of the BM in larger mammals. In the present study, we demonstrated the utility of the clearing agent Visikol [28] for the analysis of the adipocyte-rich BM in rhesus macaques. Furthermore, we developed the automated image algorithm for nearest neighbor distance analysis to calculate the proximity between HSPCs and BM adipocytes. Co-localization and proximities between BM cells and cellular structures have been widely regarded as indicative of their functional interaction [13, 14, 38, 52–54]. Such a role for paracrine effects of BM adipose tissue in particular is supported by previous studies detailing the well-described ability of adipocytes to secrete regulatory factors as either soluble protein [55] or in microvesicles/exosomes [56]. Additional mechanisms involving BM adipocytes themselves include the interaction of BM adipocytes with metastasizing prostate cancer cells through the direct transfer of lipid droplet-derived fatty acids to provide energy for proliferation [57], and to other cell types such as macrophages through lipid-containing exosomes that modulate differentiation [58]. Thus, the spatial association of CD34+ HSPCs with BM adipocytes implies that these cell types may communicate bidirectionally via local paracrine signals. Importantly, our microscopic observations are consistent with earlier studies demonstrating that a fraction of human CD34+ HSPCs localize in close proximity to BM adipocytes [52]. To our knowledge, however, no rodent studies to date have addressed the spatial relationship between HSPCs and BM adipocytes.

The functional role of BM adipocytes in regulation of hematopoiesis has been controversial. Two studies have shown that BM adipocytes act as negative regulators of hematopoiesis in vivo. The initial study by Naveiras et al. demonstrated that fatless mice or wild-type mice treated with a peroxisome proliferator-activated receptor-gamma (PPARγ) inhibitor that attenuates adipogenesis exhibit enhanced HSPC engraftment after irradiation [13]. Furthermore, fatless mice lacking BM adipocytes also presented with a compensatory increase in osteogenesis following hematopoietic ablation. This study was confirmed by a recent report demonstrating that BM adipocytes suppress hematopoietic recovery and osteogenesis after irradiation [59]. Because the transplanted HSPCs tend to home near the endosteal bone surfaces after BM ablation by irradiation [60], it is possible that BM adipocytes inhibit HSPC engraftment by an indirect mechanism through the suppression of the osteoblast lineage. Importantly, osteoblasts and adipocytes originate from the common BM mesenchymal progenitor cells [10] and hence enhanced adipogenesis can directly inhibit osteogenesis, leading to a reduced engraftment of HSPCs.

In addition to exerting negative effects on osteogenesis, BM adipocytes may also directly regulate HSPC function though local paracrine signaling. For example, recent studies in mice demonstrated that BM adipocytes produce stem cell factor (SCF), which appears to be essential for hematopoietic regeneration after irradiation or chemotherapy [14] and for the recovery of myelopoiesis following metabolic challenge with a high-fat diet [15]. Consistent with these studies, human acute myeloid leukemia has been shown to disrupt the BM adipocyte niche, leading to impaired myelo-erythropoiesis, whereas in vivo administration of PPARγ agonists stimulated BM adipogenesis and also reversed leukemia-induced hematopoietic failure [61]. Importantly, both obesity [8] and aging [9] are also associated with an increase in BM adipogenesis, which coincides with the development of myeloid lineage-biased hematopoiesis. Collectively, these studies are consistent with our results showing that BM adipose tissue releases factors that accelerate the proliferation of mature CD34+/CD38+progenitors and myeloid differentiation of HSPCs ex vivo. This result is also consistent with a previous report that human cord blood-derived mature CD34+/CD38+ progenitors exhibit a higher differentiation potential than the primitive CD34+/CD38- progenitors [62]. Hence, it is possible that the systemic metabolic state induced by high-fat diet or aging changes the secretome of BM adipocytes, resulting in activation and differentiation of HSPCs to accommodate a greater proliferative demand on the hematopoietic system.

The close proximity between CD34+ HSPCs and BM adipocytes reported in the present and other [52] studies implies that HSPCs may directly interact with the BM adipocytes or with the extracellular matrix surrounding adipocytes. Consistent with the latter idea, our proteomic analysis reveals that BM adipose tissue releases a variety of focal adhesion and collagen molecules that may regulate the retention of HSPCs and other cell types in the cellular neighborhood of BM adipocytes and vascular sinusoids. Remarkably, BM adipocytes appear to form physical contacts with vascular sinusoids and isolectin-positive cells, as demonstrated in bone sections and isolated BM adipose tissue. Furthermore, we demonstrate that BM adipocytes can be isolated as a cellular ensemble with leukocytes, while our preliminary flow cytometry analysis suggests that granulocytes are particularly enriched in the BM adipocyte fraction. Neutrophils and other types of granulocytes, as well as other immune cells, likely contribute to the physiological in situ secretory activity of the niche. Indeed, our proteomic analysis shows that the BM adipose tissue secretome is highly enriched in the myeloid cell-derived proinflammatory factors S100A8/9 [63–68] and TGFB1, which, at lower doses, reportedly enhance myeloid-biased HSPC differentiation [69, 70] and stimulate the growth and differentiation of monocyte-granulocyte precursors [71–73]. Consistent with our flow cytometry results, our proteome analysis shows that BM adipose tissue-conditioned media is also enriched in several neutrophil markers, including matrix metallopeptidases 2 and 9 (MMP2 and 9), and myeloperoxidase. The release of MMP2 and MMP by neutrophils has been shown to cause rapid HSPC mobilization from the BM [74, 75]. Thus, BM adipocytes may participate in activation of BM neutrophils in response to pathophysiological conditions. Consistent with this model, a recent study demonstrates that peripheral adipose tissue lipolysis activates and attracts neutrophils, triggering high-fat diet-induced macrophage infiltration in white adipose tissue [76]. Thus, BM adipocytes may act as the central relay station regulating the biogenesis of different myeloid cells, ranging from early hematopoietic progenitors to granulocytes.

Future studies will be needed to understand the mechanisms mediating cross-talk between adipocytes, HSPCs, and other niche cells and to clarify the role of extracellular vesicles as potential regulators of hematopoietic fate [77–80]. Furthermore, the role of BM adipocytes in osteogenesis requires further studies. In fact, several factors identified in BM adipose tissue-conditioned media (IGFB2 [81], IGFBP4 [82], IGFBP7 [83] and TGFB1 [84]) have been implicated in a variety of osteogenic processes ranging from bone resorption to bone formation. Since hematopoiesis, as an organ system, relies on the integration of intrinsic programs with extrinsic crosstalk, faithful modeling of the BM niche is critical to translate and validate observations and studies of the nonhuman primate BM niche can fill the gap toward validation of observations made in mice. The present study represents an important advancement toward our understanding of the spatial organization of the BM niche in nonhuman primates and will allow for a more comprehensive phenotypic characterization of rhesus HSPCs. Given that humans and rhesus macaques share many of the same hematopoietic phenotypes [23–25], the present study provides fundamental metrics of spatial organization and function in primates for focused validation in human hematopoiesis.

Highlights.

• Bone marrow architecture in the femur of rhesus macaques was studies using optical clearing, confocal microscopy, and spatial analysis

• Hematopoietic stem and progenitor cells are associated with bone marrow adipocytes

• Isolated bone marrow adipose tissue exists as a niche containing vascular endothelium, granulocytes and monocytes

• Bone marrow adipose tissue stimulates proliferation and myeloid differentiation of hematopoietic stem and progenitor cells ex vivo

• Bone marrow adipose tissue releases bioactive soluble and exosomal/microvesicular proteins implicated in regulation of hematopoiesis and osteogenesis

Abbreviations:

BM

bone marrow

BM

adipose tissue

HSPC

hematopoietic stem and progenitor cell