Learn more: PMC Disclaimer | PMC Copyright Notice
De novo generation of adipocytes from circulating progenitor cells in mouse and human adipose tissue
Abstract
White adipocytes in adults are typically derived from tissue resident mesenchymal progenitors. The recent identification of de novo production of adipocytes from bone marrow progenitor-derived cells in mice challenges this paradigm and indicates an alternative lineage specification that adipocytes exist. We hypothesized that alternative lineage specification of white adipocytes is also present in human adipose tissue. Bone marrow from transgenic mice in which luciferase expression is governed by the adipocyte-restricted adiponectin gene promoter was adoptively transferred to wild-type recipient mice. Light emission was quantitated in recipients by in vivo imaging and direct enzyme assay. Adipocytes were also obtained from human recipients of hematopoietic stem cell transplantation. DNA was isolated, and microsatellite polymorphisms were exploited to quantify donor/recipient chimerism. Luciferase emission was detected from major fat depots of transplanted mice. No light emission was observed from intestines, liver, or lungs. Up to 35% of adipocytes in humans were generated from donor marrow cells in the absence of cell fusion. Nontransplanted mice and stromal-vascular fraction samples were used as negative and positive controls for the mouse and human experiments, respectively. This study provides evidence for a nontissue resident origin of an adipocyte subpopulation in both mice and humans.—Gavin, K. M., Gutman, J. A., Kohrt, W. M., Wei, Q., Shea, K. L., Miller, H. L., Sullivan, T. M., Erickson, P. F., Helm, K. M., Acosta, A. S., Childs, C. R., Musselwhite, E., Varella-Garcia, M., Kelly, K., Majka, S. M., Klemm, D. J. De novo generation of adipocytes from circulating progenitor cells in mouse and human adipose tissue.
Adipocytes, the primary fat-storing cells of the body, are initially generated during fetal development in humans (1). New adipocytes are produced throughout life to replace extant fat cells as they die or to increase the storage capacity of adipose tissue (2, 3). In addition to serving as a primary energy reservoir, adipocytes play a central role in metabolic homeostasis by releasing adipokines including leptin, which regulates satiety and energy expenditure, and adiponectin (AdipoQ), which promotes insulin sensitivity and suppresses inflammation (4, 5).
Adipocytes in different body locations display distinct features (6–8). This concept is exemplified by the link between increased visceral adiposity and the appearance of chronic metabolic disturbances, whereas fat accumulation in subcutaneous or peripheral depots is relatively benign (7, 9, 10). This connection is also evident in the redistribution of fat to visceral depots that occurs with aging and/or loss of gonadal hormone production and the associated elevated risk for metabolic comorbidities (11–13). These depot-specific associations with disease risk may be related to differences in intrinsic gene expression signatures that are specified early in development. In fact, differences in the expression of genes related to embryonic development and pattern specification have been reported both between subcutaneous and intra-abdominal fat depots (14) as well as between different subcutaneous (flank vs. intrascapular) and intra-abdominal (epididymal vs. perirenal vs. mesenteric) depots (15).
Functionally discrete adipocyte populations are also present within individual adipose tissue depots (16–19). Varlamov et al. (20) reported striking heterogeneity in free fatty acid and glucose uptake by white adipocytes in adipose tissue explants from Rhesus macaques and fatty acid uptake by in-vitro–differentiated human adipocytes. Variability in fatty acid uptake was also observed in mouse white adipose tissue adipocytes by Katz et al. (21) who sorted fat cells into high- and low-uptake populations by flow cytometry. Each of these populations retained a similar high- or low-uptake phenotype even after dedifferentiation and subsequent redifferentiation. Heritability of variable phenotypic features highlights the existence of distinct adipocytes within individual depots. Thus, phenotypic and metabolic differences between and within adipose tissue depots likely play a central role in shaping overall metabolic health.
Distinct adipocyte populations are generated via a variety of mechanisms, but production from distinct progenitors and developmental pathways is noteworthy because the cell origin and microenvironment during differentiation may dictate function. Studies by Tchkonia et al. (8) and Hoffstedt et al. (22) suggest that visceral adipocytes are generated from preadipocytes with low adipogenic potential, whereas subcutaneous preadipocytes differentiate readily. Thus, visceral fat expands primarily through storage of lipid in existing large, insulin-resistant and inflammatory fat cells rather than by production of new, smaller, insulin-sensitive adipocytes.
Fate-mapping studies further underscore distinct lineage origins for adipocyte subpopulations, opening a Pandora’s box of debate (23). Most white adipocytes may be derived from endothelial/mesenchymal progenitors (24), whereas production of conventional thermogenic brown adipocytes requires endothelial (24) and myogenic stages (25). Additionally, a small population of cephalic fat cells is derived from neuroectoderm rather than the previously assumed mesoderm origin of all adipocytes (26).
Because developmental lineages can be recapitulated during progenitor differentiation, remodeling, repair, and disease, it is important to define the origin of adipocytes in the adult. The previous work of our laboratory (27–29) and others (30, 31) demonstrated the de novo production of bona fide adipocytes from bone marrow (BM)-derived progenitor cells in the major fat depots of mice. This finding was surprising because although BM-derived hematopoietic cells exhibited more plasticity than originally imagined, adipose origins were typically thought to be tissue resident mesenchyme. Bone marrow progenitor (BMP)-derived adipocytes preferentially accumulated over time in visceral fat depots, and differentiation was enhanced in females, suggesting a potential estrogen-mediated effect (28). High-fat feeding or thiazolidinedione treatment increased their accumulation (27), which illustrated that the global body and tissue microenvironment affects specification of the progenitors. Global gene expression profiling indicated that this BMP-derived subpopulation possesses a potentially detrimental phenotype, including minimal expression of genes related to mitochondrial fuel oxidation, and elevated production of inflammatory cytokines (28). We concluded that the sex- and depot-specific accumulation of marrow-derived adipocytes may explain, in part, adipose tissue heterogeneity, changes in adipose tissue composition with aging, and the detrimental impact of increased visceral adiposity. However, the translation to clinical studies of human adipose tissue remained uncertain.
Here, we report evidence that a subpopulation of adipocytes is produced from nontissue resident progenitor cells using an adoptive transfer model of BM from mice in which luciferase expression is governed by the adipocyte-restricted AdipoQ gene promoter. Excitingly, we also detected donor-derived adipocytes in flow cytometry-purified adipocytes from human hematopoietic stem cell (HSC) transplant recipients by exploiting microsatellite polymorphisms that distinguish donor and recipient DNA. These results suggest that the study of these adipocytes may be clinically relevant.
MATERIALS AND METHODS
AdipoQ-luciferase mouse model
Mice harboring an AdipoQ-cre transgene [B6;FVB-Tg(Adipoq-cre)1Evdr/J] were purchased from The Jackson Laboratory (Bar Harbor, ME, USA), and LSLLuciferase mice were obtained from the National Cancer Institute Mouse Repository (Frederick, MD, USA). Mice with adipocyte-targeted luciferase were generated by mating AdipoQ-cre mice to LSLLuciferase mice. Male offspring hemizygotic at both alleles from 8 to 16 wk of age were used as BM donors in transplant studies. Recipient mice were irradiated with a split dose of 6 Gy using an X-ray source. Immediately following irradiation, mice were injected in the retro-orbital venous plexus with 5 × 106 BM cells from opposite sex AdipoQcreLSLLuciferase mice. At 8 wk after transplantation, mice were assessed for chimerism by fluorescence-activated cell sorting analysis of peripheral blood. Only transplanted mice with >95% of peripheral blood cells of donor origin were used in subsequent studies. Transplanted mice were maintained in a climate-controlled room with alternating 12 h periods of light and dark. The mice were fed ad libitum.
All animal studies and procedures were performed under a protocol approved by the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus.
In vivo imaging and luciferase assays
Luciferase activity was measured in transplant-naïve wild-type mice and AdipoQcreLSLLuciferase marrow recipient mice transplanted with luciferase-expressing BM in an IVIS Imaging System 50 (Caliper Life Sciences, Hopkinton, MA, USA). Animals were lightly anesthetized with pentobarbital (65 mg/kg, i.p.) and injected with d-luciferin (120 mg/kg, 100 μl retro-orbital). Measurements were initiated 3 min after luciferin injection, and luminescence was integrated over 5 min. At the end of the study, mice were euthanized immediately after whole-body imaging. Gonadal and dorsal fat pads were quickly harvested and returned to the imager for analysis of isolated fat pad luminescence. Data were processed with Living Image 3.0 software (PerkinElmer, Waltham, MA, USA).
In collateral experiments, mouse tissues including gonadal, subcutaneous, perirenal, mesenteric, and epicardial adipose tissue, and skeletal muscle, lung, and liver were harvested from transplant recipients. One-hundred milligram portions of each tissue were homogenized in Dual-Luciferase Assay Kit Passive Lysis Buffer (Promega Corp., Madison, WI, USA) according to the manufacturer’s directions. Homogenates were centrifuged for 10 min at 13,000 g, and the clear supernatant was reserved for determination of luciferase activity. To further demonstrate that luciferase expression was restricted to adipocytes, adipose tissue was digested with collagenase and separated into adipocytes and stromal fractions by flotation/centrifugation. Adipocytes were transferred to a clean tube and stored at −80°C. Stromal cells were resuspended in HBSS and incubated with magnetic beads linked to anti-CD11b antibodies, washed, and separated into CD11b-positive (CD11bPOS) and -negative (CD11bNEG) subfractions as directed by the manufacturer of the bead/magnet system (Miltenyi Biotec, Inc., Auburn, CA, USA) and stored at −80°C until use.
Luciferase activity in the tissue fractions was assayed with the luciferase assay system with cell culture lysis reagent (Promega Corp.). Thawed adipocytes were mixed in a volume of lysis reagent equal to the adipocyte suspension and incubated on ice for 10 min. The mixture was then centrifuged at 16,000 g for 10 min and the subnatant transferred to a clean tube. Thawed stromal cells were mixed with 50 µl lysis reagent and lysates prepared as for adipocytes. Luciferase activity in 20 µl lysate was measured on a Turner BioSystems 20/20 luminometer (Promega Corp.).
Human research subjects
Research subjects were recruited from a population of patients who had received HSC transplants as part of their clinical treatment for various hematologic malignancies [acute lymphoid leukemia (n = 2), aplastic anemia (n = 1), non-Hodgkin’s lymphoma (n = 2), Hodgkin’s lymphoma (n = 1), myelodysplastic syndrome (n = 1), and chronic myeloid leukemia (n = 1)] at the University of Colorado Hospital (Aurora, CO, USA). Inclusion criteria for the study were 1) age 18–75 yr; 2) history of allogeneic HCS transplantation (>180 d after transplant); 3) minimal calcineurin inhibitors (cyclosporine <100 mg 2×/d or tacrolimus <1 mg 2 times per day); and 4) ≤10 mg prednisone daily. Volunteers were excluded from participation if they 1) did not meet inclusion criteria; 2) were taking immunosuppression therapy in excess of the doses listed in the inclusion criteria; 3) had a history of bleeding disorders or were on blood thinners (such as warfarin or Plavix) that could not be withheld for the biopsy; 4) were unable to withhold aspirin or nonsteroidal anti-inflammatory drugs for 1 wk prior to the biopsy; 5) had evidence of infection or active graft vs. host disease; or 6) had a prior history of allergies to local anesthetics. A total of 6 men and 2 women with subcutaneous fat in the lower abdominal area suitable for small-volume liposuction participated. A medical history and physical examination were completed by the study physician before the biopsy was completed, and a single dose of oral antibiotic was taken by all participants 1 h before the procedure to reduce the risk for infection. All patients provided their written informed consent before participating in this study, which was approved by the Colorado Multiple Institutional Review Board.
Adipose tissue biopsy
Abdominal subcutaneous adipose tissue biopsies were obtained by needle aspiration using the Coleman’s manual vacuum technique (32) at the Clinical and Translational Research Center Outpatient Clinic at the University of Colorado Hospital. The biopsy site adjacent to the umbilicus was anesthetized with 1% lidocaine, and 1–2 g adipose tissue was aspirated with a 10 ml syringe and a 14-gauge needle.
Separation of adipocytes from stromal cells
Collected fat tissue was rinsed with sterile PBS. The tissue fragments were incubated in a Krebs-Ringer buffer containing 1 mg/ml collagenase [catalog number (cat. no.) C2139; Sigma-Aldrich, St. Louis, MO, USA] for 1 h at 37°C on an orbital shaker. Following tissue digestion, the suspension was allowed to flow through a 150 μm filter to remove any remaining tissue fragments. The cell suspension was centrifuged at 300 g for 5–8 min to sediment stromal cells. The floating adipocytes were gently transferred to a clean tube and washed with several milliliters of wash buffer (PBS containing 2% calf serum). After a second spin, the adipocytes were transferred to a clean 12 × 75 mm polypropylene tube, and the volume was corrected with wash buffer to 0.5 ml.
Staining of adipocytes for flow sorting
Human TruStain FcX (Fc Receptor Blocking Solution, cat. no. 422301; BioLegend, San Diego, CA, USA) was added at 5 µl/100 µl cell suspension, and cells were incubated for 10 min. LipidTOX Deep Red Neutral Lipid Stain (cat. no. H34477; Life Technologies, Thermo Fisher Scientific, Inc., Waltham, MA, USA) was then added to the adipocyte suspension at a 1:200 dilution and mixed gently. Lineage antibodies [anti-CD3 for T lymphocytes (clone HIT3a, cat. no. 300319), anti-CD19 for B lymphocytes (clone HIB19, cat. no. 302205), anti-CD14 for myelomonocytic cells (clone M5E2, cat. no. 301803), anti-CD15 for granulocytes (clone HI98, cat. no. 301903), anti-CD56 for NK cells (clone HCD56, cat. no. 318303), and anti-CD123 for megakaryocytes and basophilic granulocytes (clone 6H6, cat. no. 306013); all from BioLegend] were added at 0.25 µg/1–10 × 106 cells and mixed gently. Comparable results were obtained with cells stained only with antibodies to the pan-leukocyte marker, CD45 (cat. no. 304005, clone HI30; BioLegend). Samples were incubated at room temperature in the dark for 25 min. Following incubation, samples were spun at 150 g for 2 min to float the adipocytes, and the subnatant was removed carefully. The cells were washed with 0.5 ml wash buffer, and the centrifugation was repeated. After removal of the second subnatant, the volume was adjusted back to 0.5 ml. DAPI (cat. no. D9564; Sigma-Aldrich) was added to a final concentration of 1 µg/ml, and the cells were sorted within 15 min.
Flow sorting of adipocytes
Adipocytes were sorted using a MoFlo XDP cell sorter with Summit 4.3 software (Beckman Coulter, Fullerton, CA, USA). A 70 μm nozzle tip was used with a sheath pressure of 60 ψ and a drop drive frequency of 97,000 Hz and amplitude of 15 V. The sheath fluid was IsoFlow (Beckman Coulter). The sample and collection tubes were maintained at 5°C using an attached Haake recirculating water bath (Thermo Fisher Scientific, Inc.). To keep cells in suspension, the MoFlo was equipped with a SmartSampler sample station with the sample agitation set to maintain an agitation cycle of 4 s on and 5 s off. The sample flow rate was set to a pressure differential of <0.4 ψ. Sort mode was set to Purify 1. Forward angle light scatter and side light scatter were collected using log scales. Appropriate signal compensation was set using single-color control samples.
RT-PCR analysis of adipocyte purity
PCR primers for human targets included CD45 forward 5′-TTG GCT TTG CCT TTC TGG ACA CAG-3′ and reverse 5′-AAG TGG AAC ACT GGG CAT CTT TGC-3′; AdipoQ forward 5′-TGG TGA GAA GGG TGA GAA-3′ and reverse 5′-AGA TCT TGG TAA AGC GAA TG-3′; perilipin (Plin)-1 forward 5′-CTG CCG GTG GTG AGT GGC AC-3′ and reverse 5′-CAC AGA GGC CAC CAG GGG GT-3′; CD11b forward 5′-GTG GCA AGG AAT GTA TTT GAG TG-3′ and reverse 5′-CAG AGC CAG GTC ATA AGT CAC-3′; platelet-derived growth factor receptor α (PDGFRα) forward 5′-TTC CTC TGC CTG ACA TTG AC-3′ and reverse 5′-ATG TCT TCA ATG GTC TCG TCC-3′; Flk1 forward 5′-ATA GAA GGT GCC CAG GAA AAG-3′ and reverse 5′-GTC TTC AGT TCC CCT CCA TTG-3′; and glyceraldehyde 3-phosphate dehydrogenase forward 5′-GGA GCG AGA TCC CTC CAA AAT-3′ and reverse 5′-GGC TGT TGT CAT ACT TCT CAT GG-3′.
RNA was prepared using the RNeasy Micro Kit and Qiashredders (Qiagen, Inc., Valencia, CA, USA). Adipocytes were sorted directly into Buffer RLT (Qiagen, Inc.), and RNA extraction was performed immediately after sample collection. RNA was eluted in RNase free water (Qiagen, Inc.) and analyzed on a NanoDrop ND-1000 (Thermo Fisher Scientific, Inc.) for concentration. cDNA was prepared using the Invitrogen SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific, Inc.). cDNA was diluted 20-fold, and 5 µl was loaded per reaction. Primers were diluted from a stock concentration of 10 µM to 0.5 µM final concentration in the reactions. PCR master mix was the DyNAmo Flash SYBR Green qPCR reagent (Thermo Fisher Scientific, Inc.) at a 1× concentration. Reactions were cycled on a PTC-200 thermocycler (Bio-Rad, Hercules, CA, USA) with an initial denaturation at 95°C for 7 min, followed by 39 cycles consisting of denaturation at 95°C for 45 s and annealing and extension at 60°C for 10 s. After a final extension at 72°C for 5 min, the reactions were held at 4°C. Products were visualized on a 2% agarose gel with ethidium bromide.
Chimerism analysis
Human short tandem repeat (STR) analysis was completed by the Molecular Diagnostic Laboratory at the Children’s Hospital Colorado, the same facility at which all clinical chimerism measurements were conducted for the participants. The validated commercial assay used is sufficiently sensitive to detect a 1% minor cell population with 10 ng DNA, the amount contained in ∼1700 human cells (based on 6 pg chromosomal DNA per diploid cell). We assumed that ∼10% of the flow-purified adipocytes would be of donor origin based on our previous studies with mouse subcutaneous adipose tissue. To achieve a 95% confidence level of detection (confidence interval 1.5%), the minimum sample size was determined to be 4250 cells. We collected DNA from >106 adipocytes from all subjects.
DNA was isolated from stromal cells and flow-purified adipocytes using QiAmp DNA Micro Purification reagents (Qiagen, Inc.). PCR chimerism analysis was performed with the Beckman Coulter GeXP GenomeLab Human STR Primer Set (cat. no. A20100; AB Sciex, Framingham, MA, USA) followed by fragment separation by capillary electrophoresis. Primer concentrations were optimized to provide balanced amplification from each locus. Loci known to be problematic were excluded from the GenomeLab primer set.
Distinct STR profiles obtained by chimerism analysis were compared to analyses from BM samples obtained prior to transplantation (for clinical purposes) to determine the percentage of cells that are donor vs. recipient derived. At the Molecular Diagnostic Laboratory, posttransplant chimerism is routinely calculated using an average of at least 2 informative loci. The standard chimerism calculation is the following: peak height of recipient-specific allele/sum of peak heights of recipient-specific and donor-specific alleles × 100%.
Ploidy analysis
Krishan’s staining protocol was used to quantify cellular DNA content (33). Cells were analyzed on a Beckman Coulter FC500 flow cytometer. Krishan’s staining and flow analysis is a standard protocol accepted for determination of ploidy and cell cycle (34). Aggregates were excluded from the analysis using the peak vs. integral gating method. To achieve a 95% confidence level for detection of donor-derived adipocytes produced by cell fusion (assuming 10% of total adipocytes would be of donor origin) required analysis of at least 2100 events per subject (confidence interval 1.9%). We analyzed at least 9000 events per subject using ModFit LT software (Verity Software House, Topsham, ME, USA).
Chromosome aneusomy
After separation of the adipocytes from the stromal cells by centrifugation, a portion of the sample was taken for investigation of chromosome aneusomy by FISH completed by the Molecular Pathology/Cytogenetics Shared Resource at the University of Colorado Cancer Center, according to their validated techniques. FITC-conjugated anti-CD45 antibody (cat. no. 304005, clone HI30) was added at 0.25 µg/1–10 × 106 cells and mixed gently. Samples were incubated in the dark on ice for 25 min. Following incubation, samples were spun at 150 g for 2 min to float the adipocytes, and the subnatant was removed carefully. The cells were washed with 0.5 ml wash buffer, and the centrifugation was repeated. After removal of the second subnatant, cells were fixed in 4% paraformaldehyde (4× cell volume, cat. no. 18505; Ted Pella, Inc., Redding, CA, USA) for 30 min at 4°C, rinsed with PBS 2 times, and permeabilized with 70% methanol (4× cell volume, cat. no. 179337; Sigma-Aldrich) for 1 h at 4°C. Cells were washed by centrifugation and brought to volume with flow buffer. DAPI was added to a final concentration of 1 µg/ml, and the cells were sorted within 15 min.
Free-floating CD45NEG/DAPIPOS adipocytes were sorted directly onto microscope slides, which were air-dried. Each slide was incubated in 0.008% pepsin/0.01 M HCl at 37°C for 5 min, fixed in 1% formaldehyde solution for 10 min, and dehydrated in graded ethanol series. Probe mix was applied to the selected 12-mm2–diameter hybridization area, which was covered with a glass coverslip and sealed with rubber cement. DNA codenaturation was performed for 10 min in an 86°C dry oven, and hybridization was allowed to occur in a moist chamber at 37°C for ∼40 h. Posthybridization washes were performed with saline-sodium citrate/0.3% Nonidet P-40 at 72°C for 2 min and saline-sodium citrate 2 min at room temperature, and chromatin was counterstained with DAPI (0.3 µg/ml in Vectashield Mounting Medium; Vector Laboratories, Inc., Burlingame, CA, USA). Analysis was performed on an epifluorescence microscope using single-interference filter sets specific for wavelengths of blue (DAPI), aqua, green (FITC), yellow, and red (Texas red). Tentatively, 50 cells were analyzed per specimen for copy number of each of the genomic targets (95% confidence level; 8.3% confidence interval; and minimum required number of nuclei, 37).
Sample size and statistics
Because measurements of donor-derived adipocytes have not previously been conducted in human adipose tissue, minimum sample sizes were based on percentages of donor-derived adipocytes in the subcutaneous adipose tissue of BM-transplanted mice (27, 28). To achieve a 95% confidence level for detection, minimum sample sizes were calculated assuming that ∼10% of the total adipocyte population would be derived from donor cells. The online calculator available at MaCorr Research (Toronto, ON, Canada; www.macorr.com) was used for all sample size calculations. Repeated biopsies in a subset of participants were used to confirm results and assess accumulation of BMP-derived adipocytes over time. Linear regression analysis was performed with Prism 6.0e software (GraphPad Software, Inc., La Jolla, CA, USA).
RESULTS
Detection of BM-derived adipocytes using a lineage-specific luciferase reporter
To demonstrate a BM origin for certain adipocytes in adult mice, we performed adoptive transfer of undifferentiated BM in combination with the AdipoQ gene promoter, which provides stringent adipocyte-specific expression in transgenic systems (23, 35). BM was isolated from donor mice in which luciferase expression was governed by the AdipoQ promoter (AdipoQcreLSLLuciferase mice). Whole-body light emission increased over time in wild-type mice transplanted with whole BM from the AdipoQcreLSLLuciferase donors (Fig. 1A). No light emission was observed in transplant-naïve control mice (Fig. 1B). Light emission from internal organs in the thoracic and abdominal cavities was also measured. Intense signals were detected from gonadal adipose tissue, and less-intense emission was recorded from subcutaneous and mesenteric adipose tissue (Fig. 1C). Relative light unit emission was also high in cardiac-associated adipose tissue. No luciferase activity was detected in nonadipose tissue including intestines, liver, or lungs, demonstrating adipose-restricted luciferase expression. Luciferase enzyme activity was also measured in homogenates of various adipose and nonadipose tissues (Fig. 1D). In concordance with the in vivo imaging results, luciferase activity was highest in epi/pericardial and gonadal adipose tissue, with lower levels measured in subcutaneous, mesenteric, and perirenal adipose tissue. Light emission detected in liver, lung, and skeletal muscle was not significantly increased above baseline. To further confirm that luciferase expression was restricted to adipocytes within the adipose depot, we assayed free-floating adipocytes or CD11bPOS or CD11bNEG stromal-vascular cells isolated from collagenase-digested gonadal fat. CD11bPOS cells are myeloid lineage cells with a recognized HSC BM origin. Of the total luciferase activity detected in the gonadal depot, >95% was present in the adipocyte lysate, ∼3% in CD11bPOS stromal cells, and <1% in CD11bNEG stromal cells (Fig. 1E).
Adipocytes are produced from BMP cells in mice. Eight-week-old female C57Bl/6 wild-type mice were transplanted with BM from mice in which luciferase expression was targeted to adipocytes (AdipoQcreLSLLuciferase). A–C) Luciferase activity (light emission) was measured in the recipients in an IVIS Imaging System 50, and data were acquired and analyzed with Living Image 3.0 software. A) Representative luminescence images of BM-transplant recipients at 2, 8, 14, and 18 wk after transplant. All images are at the same scale. B) Luminescence measured in a transplant-naïve mouse 16 wk after transplant. Scale is the same as in (A). C) At 16 wk, some animals received a second injection of luciferin and were immediately sacrificed so that the abdominal and thoracic cavities could be opened to expose internal organs. Representative image of light emission showing adipose tissue-restricted luciferase activity is shown. SubQ, subcutaneous. D) Adipose and nonadipose tissues were harvested from transplant recipients 8 wk after transplant and digested with Passive Lysis Buffer from the Dual-Luciferase Assay Kit. Luciferase activity was measured in a Turner BioSystems 20/20 luminometer. Luciferase activity in each fraction was corrected for lysate protein concentration (n = 8; data are presented as means ± sd). RLU, relative light unit. E) Gonadal adipose tissue was digested with collagenase, and adipocytes were separated from stromal cells by flotation. CD11bPOS and CD11bNEG stromal populations were prepared by magnetic bead separation. Luciferase activity was measured in lysates of the 3 fractions as described in (D) (n = 2; data are presented as means ± sd).
Flow cytometric analysis and isolation of intact human subcutaneous adipocytes
To assess whether the de novo generation of adipocytes from BM occurs in humans, adult males (n = 6) and females (n = 2) who received allogeneic BM HSCs, mobilized peripheral blood stem cells (PBSCs), or cord blood HSC transplants as part of standard clinical treatment for hematologic malignancies were enrolled in the study (Table 1). The patients ranged in age from 24 to 75 yr. Hematopoietic engraftment was assessed at the clinical appointment closest to adipose tissue biopsy. Abdominal subcutaneous adipose tissue biopsies were collected between 12 and 53 mo after transplant (Table 2).
TABLE 1.
Subject characteristics
| Patient number | Transplant type | Donor information | HLA matching | Hematopoietic engraftment |
|---|---|---|---|---|
| 1 | Allogeneic PBSC | Matched unrelated donor | 10/10 | >95% donor |
| 2 | Allogeneic PBSC | Matched sibling | 10/10 | 100% donor |
| 3 | Allogeneic BM | Matched sibling | 10/10 | >95% donor |
| 4 | Double cord blood | Unrelated cord blood | 2/10, 6/10 | 100% 6/10 cord |
| 5 | Allogeneic PBSC | Matched sibling | 10/10 | 100% donor |
| 6 | Allogeneic PBSC | Matched sibling | 10/10 | 100% donor |
| 7 | Allogeneic PBSC | Matched sibling | 10/10 | 100% donor |
| 8 | Double cord blood | Unrelated cord blood | 8/10, 6/10 | 100% 8/10 cord |
HLA, human leukocyte antigen.
TABLE 2.
Adipose tissue chimerism results
| Patient number | Biopsy number | Age at biopsy (yr) | BMI at biopsy (kg/m2) | Posttransplant duration (mo) | Adipocyte chimerism | Stromal chimerism | ||
|---|---|---|---|---|---|---|---|---|
| % Donor | % Recipient | % Donor | % Recipient | |||||
| Patients with single biopsy | ||||||||
| 1 | 1 | 63 | 28.4 | 12 | <5 | >95 | 35 | 65 |
| 2 | 1 | 75 | 26.3 | 12 | 11 | 89 | 42 | 58 |
| 3 | 1 | 45 | 29.8 | 14 | 11 | 89 | 22 | 78 |
| 4 | 1 | 24 | 24.1 | 18 | 0 | 100 | ND | ND |
| 5 | 1 | 33 | 30.1 | 48 | 35 | 65 | 61 | 39 |
| Patients with repeat biopsy | ||||||||
| 6 | 1 | 59 | 25.4 | 28 | 5 | 95 | 31 | 69 |
| 2 | 60 | 24.4 | 35 | 13a | 87 | 27 | 73 | |
| 7 | 1 | 54 | 25.1 | 31 | <5 | >95 | 29 | 71 |
| 2 | 55 | 24.8 | 43 | 12 | 88 | 25 | 75 | |
| 8 | 1 | 41 | 31.8 | 43 | 9 | 91 | 11 | 89 |
| 2 | 42 | 32.2 | 53 | 14 | 86 | 30 | 70 | |
ND, not done. aDue to paucity of the sample, this value was calculated from the percent donor chimerism in unsorted adipocytes (percent DAPIPOS events plus percent LinPOS events).
Our strategy to detect human BMP-derived adipocytes relied on quantifying microsatellite polymorphisms in flow cytometry-purified adipocytes by STR analysis of donor- and recipient-specific informative loci. Human adipocytes were selected and isolated using a multistep flow-sorting strategy based on methods we devised for purifying mouse fat cells (29, 36). We previously demonstrated that these strategies provide purified murine adipocytes free from stromal hematopoietic and mesenchymal cell contamination (36). Free-floating, buoyant adipocytes were collected from collagenase-digested human abdominal subcutaneous adipose tissue single-cell suspensions, following centrifugation to remove the bulk of stromal-vascular cells.
Adipocytes were separated from potential stromal cells based on their size [forward scatter (FSC)] and high refractility [side scatter (SSC)] (Fig. 2A). Cell clusters were eliminated based on SSC signal height vs. signal width. Lipid-containing cells were collected based on high LipidTOX fluorescence, and dead cells and free nuclei were removed based on their staining with DAPI. Finally, false-positive circulating cells arising from donor marrow were excluded based on the presence of hematopoietic lineage markers, either CD45 alone, or a mixture of hematopoietic lineage-specific antibodies (CD3, CD19, CD14, CD15, CD56, and CD123).
Multistep flow cytometry/sorting strategy to purify human adipocytes. A) Gating strategy for adipocyte isolation is diagramed from top to bottom. In the first step, adipocytes are identified by their large size and refractile properties in a plot of FSC vs. SSC. Singlets were identified optically by comparing SSC peak height to peak width. In the third step, LipidTOX fluorescence of the singlets identified lipid droplet-containing cells and allowed exclusion of dead cells and free nuclei based on DAPI fluorescence. Only lipidPOS/DAPINEG cells were collected. Finally, any remaining stromal contaminants are excluded based on their labeling with FITC-conjugated antibodies to CD45. PE, phycoerythrin channel. B) Immediately following flow sorting of the adipocytes, their purity was checked by flow cytometry analysis of a small portion of the sorted cells. All sorted cells were intact and alive (DAPINEG), and CD45NEG. C) RT-PCR was used to verify the purification strategy. The results show the presence of adipocyte markers (AdipoQ and Plin-1) in purified adipocytes and whole-adipose tissue homogenate. However, hematopoietic (CD45 and CD11b) and mesenchymal (PDGFRα and Flk1) markers were only detected in the tissue homogenate and essentially undetectable in flow-purified adipocytes. GAPDH, glyceraldehyde 3-phosphate dehydrogenase.
Collectively, percentages of dead cells/free nuclei and lineage-positive cells measured during flow sorting were <0.02% each in all samples, indicating a high initial purity for our adipocyte fractions. Immediately after sorting, a portion of the purified adipocytes was reanalyzed to assess their purity (Fig. 2B). All adipocytes were DAPINEG, indicating that the cells were intact and alive because dead cells and free nuclei exhibit bright DAPI fluorescence. Furthermore, all cells were CD45NEG (or negative for lineage markers), indicating the absence of stromal contamination.
Isolated adipocyte samples were assessed by RT-PCR to quantitate adipocyte, hematopoietic, and mesenchymal stromal markers (Fig. 2C). As expected, the adipocyte markers AdipoQ and Plin-1 were detected in the flow cytometry-purified adipocytes as well as whole-adipose tissue homogenates. However, hematopoietic (CD45 and CD11b) and mesenchymal (PDGFRα and Flk1) markers were only detected in the whole-adipose tissue homogenate, not in purified adipocyte fractions. Thus, we have demonstrated the successful isolation and purity of human subcutaneous adipocytes using multistep flow cytometric enrichment.
De novo generation of adipocytes from BM is recapitulated in humans
Donor/recipient chimerism of the subcutaneous depot biopsies was assessed in purified flow-sorted adipocyte populations by PCR amplification of STRs at 16 highly polymorphic loci. Degree of chimerism was also quantified in the adipose stromal population. Representative electropherograms for flow cytometry-purified adipocytes (Fig. 3A) and adipose stromal cells (Fig. 3B) for one participant are shown. All patients exhibited ≥95% donor-derived hematopoietic engraftment in the BM and circulation prior to the time of biopsy (Table 1). Donor DNA was detected in flow cytometry-purified adipocytes in 7 of 8 patients, indicating production of adipocytes from transplanted hematopoietic progenitor cells. Donor engraftment in the adipocyte population ranged from <5 to 35% in the initial biopsy (Table 2). A second adipose tissue sample was obtained from 3 subjects (n = 2 female, n = 1 male) 7–10 mo following the initial biopsy. Donor chimerism was higher in all 3 secondary samples than in the first and ranged from 12 to 14% (Fig. 4 and Table 2). Regression analysis indicated a positive correlation between donor-derived adipocyte accumulation and time posttransplant. Regression analysis did not show a positive correlation between adipocyte chimerism and subject age or body mass index (BMI).
Representative PCR analysis of polymorphic STR for detection of chimerism. DNA was extracted from flow cytometry-purified subcutaneous adipocytes and adipose stromal cells 43 mo after transplant. Figure shows electropherograms for locus D18S51. The figures show the position of a PCR product common to both donor and recipient, as well as donor-specific and recipient-specific products. Analysis of donor-specific and recipient-specific peak areas indicated that ∼12% of the purified adipocyte population (A) and 25% of the stromal cells (B) were derived from donor hematopoietic cells.
Donor-derived adipocytes accumulate over time in humans. Donor adipocyte chimerism values from Table 2 are plotted vs. months after transplant. Colored data points (blue, red, and green) represent values from participants for whom initial and secondary biopsies were obtained. Linear regression for all data points (solid black line, R2 = 0.3074) is shown. Linear regression for all points except the 35% value at 48 mo (hashed black line, R2 = 0.2332) is shown.
Human BMP-derived adipocytes are not a byproduct of cell fusion
BM-HSCs were characterized as having more differentiation potential than restricted to blood lineages (37). However, this finding was complicated by subsequent analyses using terminal lineage-specific and nuclear reporters in mice, which demonstrated the existence of HSC fusion with cardiac myocytes (38) and skeletal muscle (39). One difference between these studies and ours is the normal process of myocyte fusion that occurs in muscle tissue but is absent in adipose tissue. However, we utilized 2 methods of ploidy analysis to demonstrate normal DNA and chromosomal content per cell, to exclude the possibility of cell fusion to contribute to the degree of measured adipocyte chimerism. First, propidium iodide-stained adipocytes were analyzed by flow cytometry to quantitate DNA by fluorescence intensity. None of the free-floating adipocytes was polyploid (Fig. 5), indicating that all adipocytes evaluated, including those generated from donor marrow cells, are diploid and are not the product of cell fusion. Second, FISH was used to exclude aneuploidy. In each FISH probe set (Table 3), a mix of 1 locus specific and 3 centromere probes was used, and in combination, these 2 sets covered 8 genomic regions located in 8 different chromosomes (1/3 of total haploid genome). A normal disomic cell presents 2 copies of each target. Therefore, it is clear that all isolated adipocyte specimens have predominantly disomic cells (Fig. 6). Cytogenetic results for all patients are summarized in Table 4, showing the mean copy number per cell with sd, and percentage of cells with ≤2, or ≥3 , copies for each genomic target.
Ploidy analysis demonstrates the absence of cell fusion in the human adipocyte populations. Free-floating adipocytes from 6 subjects were permeabilized and stained with propidium iodide. Figure shows flow cytometric cell cycle analysis with singlet discrimination. ModFit software distinguished peaks for G1 and G2 cells, and events to the right of the G2 peak were considered polyploid. Avg., average.
TABLE 3.
FISH probe sets
| Probe set | Genomic target | Reagent name | Chromosomal location | Fluor |
|---|---|---|---|---|
| I | Chromosome 1 centromere | CEP 1 (D1Z5) | 1p11.1-q11.1 Alpha Satellite DNA | SpectrumOrange |
| Chromosome 6 centromere | CEP 6 (D6Z1) | 6p11.1-q11 Alpha Satellite DNA | SpectrumAqua | |
| Chromosome 11 centromere | CEP 11 (D11Z1) | 11p11.11-q11 Alpha Satellite DNA | SpectrumGreen | |
| KRAS gene | LSI KRAS | 12p12.1 | SpectrumGold | |
| II (UroVysion) | Chromosome 3 centromere | CEP 3 | 3p11.1-q11.1 Alpha Satellite DNA | SpectrumRed |
| Chromosome 7 centromere | CEP 7 | 7p11.1-q11.1 Alpha Satellite DNA | SpectrumGreen | |
| CDKN2A gene | LSI p16 | 9p21 | SpectrumGold | |
| Chromosome 17 centromere | CEP 17 | 17p11.1-q11.1 Alpha Satellite DNA | SpectrumAqua |
CDKN2A, cyclin-dependent kinase inhibitor 2A; KRAS, Kirsten rat sarcoma viral oncogene homolog.
No evidence for cell fusion in the purified adipocyte population, determined by molecular cytogenetics. Interphase nuclei from subjects 3 (A) and 5 (B), hybridized respectively with probe sets I and II, show 2 copies of each genomic target per cell (disomy).
TABLE 4.
Cytogenetic results
| Patient number | FISH probe | T1 in set I = CEP 1; T1 in set II = CEP 3 | T2 in set I = CEP 11; T2 in set II = CEP 7 | T3 in set I = KRAS; T2 in set II = CDKN2A | T4 in set I = CEP 6; T4 in set II = CEP 17 | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean T1 | sd T1 | Cells with ≤2 T1 signals (%) | Cells with ≥3 T1 signals (%) | Mean T2 | sd T2 | Cells with ≤2 T2 signals (%) | Cells with ≥ 3 T2 signals (%) | Mean T3 | sd T3 | Cells with ≤2 T3 signals (%) | Cells with ≥3 T3 signals (%) | Mean T4 | sd T4 | Cells with ≤2 T4 signals (%) | Cells with ≥3 T4 signals (%) | ||
| 2 | Set I | 2.2 | 0.7 | 82 | 18 | 1.7 | 0.6 | 98 | 2 | 1.8 | 0.5 | 98 | 2 | 1.8 | 0.5 | 96 | 4 |
| Set II | 2.1 | 0.4 | 92 | 8 | 1.9 | 0.4 | 100 | 0 | 1.9 | 0.4 | 98 | 2 | 1.9 | 0.3 | 100 | 0 | |
| 3 | Set I | 2.1 | 0.5 | 92 | 8 | 2 | 0.5 | 94 | 8 | 2 | 0.6 | 90 | 8 | 1.8 | 0.5 | 96 | 4 |
| Set II | 2.3 | 1 | 72 | 28 | 2.2 | 0.6 | 78 | 26 | 1.9 | 0.9 | 80 | 20 | 1.6 | 0.8 | 92 | 8 | |
| 5 | Set I | 2.2 | 0.7 | 85 | 15 | 2.2 | 0.6 | 90 | 20 | 2.3 | 0.8 | 75 | 35 | 2.2 | 0.7 | 85 | 15 |
| Set II | 2.2 | 1 | 81 | 19 | 1.8 | 0.5 | 100 | 0 | 1.8 | 0.6 | 95 | 5 | 1.9 | 0.5 | 95 | 5 | |
| 6 | Set I | 1.9 | 0.9 | 88 | 12 | 2 | 0.3 | 98 | 2 | 2.1 | 0.5 | 90 | 10 | 1.9 | 0.6 | 92 | 8 |
| Set II | 1.9 | 0.8 | 89 | 11 | 1.7 | 0.5 | 100 | 0 | 1.6 | 0.8 | 93 | 7 | 1.7 | 0.8 | 93 | 7 | |
| 7 | Set I | 2.3 | 0.7 | 76 | 24 | 2 | 0.2 | 96 | 4 | 2 | 0.5 | 90 | 10 | 2.1 | 0.5 | 90 | 10 |
| Set II | 2 | 0.5 | 94 | 6 | 2 | 0.2 | 100 | 0 | 1.9 | 0.3 | 100 | 0 | 1.9 | 0.2 | 100 | 0 | |
| 8 | Set I | 2.2 | 0.7 | 80 | 20 | 2 | 0.2 | 100 | 0 | 1.9 | 0.5 | 92 | 8 | 1.9 | 0.4 | 96 | 4 |
| Set II | 2.1 | 0.3 | 90 | 10 | 1.9 | 0.4 | 100 | 0 | 1.9 | 0.5 | 96 | 4 | 2.1 | 0.4 | 90 | 10 | |
CDKN2A, cyclin-dependent kinase inhibitor 2A; KRAS, Kirsten rat sarcoma viral oncogene homolog.
A few specimens had a subset of cells with >2 copies for one of the targets. The prevalent reason for this was the occurrence of signal doublets, a phenomenon seen when cells are in mitotic division between the phases S and late metaphase, in which chromatin is duplicated, but sister chromatids are still attached, or when the genomic region tested has long stretches of repetitive sequences. The latter condition was specifically responsible for the higher percentage of cells carrying >3 signal copies in subjects 3, 5, 7, and 8 for chromosome enumeration probe (CEP) 1 (encompassing the 1q heterochromatin), one of the human genomic regions with highest variability in length (and copy number of repeats). Three specimens had between 20 and 28% of cells with >2 copies of CEP 3, another region well known for carrying long stretches of repetitive DNA sequences in some individuals. Based on these data, we conclude that the presence of polysomy for 8 different chromosomes was not detected in any of the 6 subjects.
DISCUSSION
In these studies, we describe the de novo production of adipocytes from BMPs in both adult mice and humans. We provide evidence that adipocytes derived from BMPs are indeed found in humans who have undergone previous HSC transplantation in the absence of cell fusion events. In addition, we support these data using a BM-transplantation model with a stringent adipocyte lineage-specific reporter, to provide evidence that BMPs contribute to de novo production of adipocytes. Although our current studies relied solely on adoptive transfer models, previous data showing production of adipocytes from circulating myeloid cells in a nonmyeloablative fate-mapping model (28) suggest that this process is not simply an artifact of HSC-transplant methods.
The production of adipocytes from transplanted marrow cells has also been reported by Sera et al. (30) and Tomiyama et al. (31), whereas other laboratories have failed to confirm our observations (23, 40). Such discrepancies are common in the study of adipocyte origins and fate determination. For example, Tran et al. (24) demonstrated production of white and brown adipocytes from endothelial cells using a VE-Cadherin–based fate-mapping strategy, whereas Berry and Rodeheffer (23) failed to detect production of adipocytes from endothelial cells using a receptor tyrosine kinase of the Tie family-based labeling model. Similarly, Rosenwald et al. (41, 42) reported production of beige adipocytes from white adipocytes via transdifferentiation using uncoupling protein 1-based fate-mapping models, whereas Wang et al. (43) failed to find evidence of transdifferentiation using an AdipoQ promoter-based labeling strategy. It is now widely recognized that different adipocyte-specific (adipocyte protein 2-CreBl, adipocyte protein 2-CreSalk, AdipoQ-cre, and AdipoQ-CreER) and preadipocyte-specific (PDGFRα-Cre, PDGFRα-CreERUCL, and PDGFRα-CreERJHU) models exhibit surprisingly different patterns of expression, with some models failing to label target populations (35, 44). Potential explanations for such discrepancies are numerous, and technical issues include mosaic (45–47) and variegated (48, 49) expression of reporter genes, silencing of reporter genes (50), sex of parent from which reporter is inherited (45, 51), failure of Cre recombinase to elicit recombination (52), and variability in reporter gene recombination (53). Moreover, different reporters (e.g., green fluorescent protein vs. Discosoma sp. red fluorescent protein) expressed from the same promoter and locus can exhibit wildly divergent spatiotemporal patterns of expression (54). Additionally, hardware and reagent limitations can inhibit the ability to detect, image, and sort such cells. It is clear that fate-mapping models and detection techniques must be selected carefully before concrete conclusions can be drawn. To our knowledge, no other laboratory has replicated our results using the same models or detection methods.
Because sophisticated labeling strategies traditionally utilized in basic fate-mapping studies are not possible in humans, we studied patients who had previously undergone BM, peripheral blood, or cord blood HSC transplantation for clinical indications as our initial proof-of-concept model. Measuring microsatellite polymorphisms allowed us to quantify donor/recipient chimerism in flow cytometry-purified adipocytes from human HSC-transplant recipients. Our data revealed a trend for greater donor adipocyte chimerism with increased time after transplant, suggesting that donor-derived adipocytes accumulate over time. This idea is further supported by an increase in adipocyte chimerism between initial and secondary biopsies in 3 of our participants and the increase in light emission in mice transplanted with BM from AdipoQ-luciferase donors. Total adipocyte number is fairly constant in adult humans (2) and mice (55), and the rate of adipocyte turnover in humans appears to be tightly regulated (2). Therefore, the time-related increase in donor-derived adipocytes observed in our human subjects suggests that the rate of their accumulation (production minus turnover) may be ultimately favored over conventional adipocytes.
Rydén et al. (56) recently reported similar engraftment of HSCs into the subcutaneous adipocyte population of 65 subjects who had received allogeneic BM or PBSC transplants using single nucleotide polymorphism and microsatellite analysis. Like our results, Rydén et al. (56) observed a correlation between the number of donor-derived adipocytes and time posttransplant but saw no correlation with subject age or sex. However, they did note a positive correlation between BMI and percentage of donor-derived adipocytes, which was not observed by us, perhaps due to our limited BMI range and small sample size. Importantly, donor-derived adipocytes comprised from 0.1 to 27% of the subcutaneous adipocyte population in their subjects, similar to our findings, and their mathematic modeling suggests that circulating progenitor cells contribute ∼10% of the subcutaneous adipocyte population over the entire life span. Determination of mixed donor and recipient genotypes from single-cell whole-genome sequencing was detected in 2 cells by Rydén et al. (56), evidence that some cell fusion events potentially contribute to what appear to be donor-derived adipocytes. Nevertheless, adipocytes of donor-specific genotype were also detected, evidence of direct differentiation events without fusion. Therefore, they suggest that it is possible that nontissue resident progenitors may influence adipocyte heterogeneity through both cell fusion events and adipocyte differentiation, although our results suggest that differentiation plays the predominant role. Rydén et al. (56) conclude that BM may be an important reservoir for adipocyte progenitors that could be exploited to treat metabolic disorders.
No circulating progenitor-derived adipocytes were detected in one of our subjects despite complete hematopoietic engraftment. This was our first subject, and the absence of donor-derived adipocytes may simply reflect suboptimal DNA recovery from flow-sorted adipocytes at an early stage of the study. Alternately, based on previous murine studies (28), the sex (male), age (24 yr), low BMI (24.1 kg/m2), and short interval since transplant (18 mo) indicate that we would indeed expect this patient to have a low percentage of donor-derived adipocytes. Furthermore, a specific feature of this individual’s physiology, disease, or medical treatment may have blocked production of adipocytes from donor cells, although nothing remarkable was noted in this participant’s medical records.
The levels of human adipocyte chimerism measured in our experiments are consistent with the number of BMP-derived adipocytes we previously reported for BM-transplanted mice (28). In mice, subcutaneous adipose tissue consisted of 2–4% donor adipocytes within 8–12 wk of transplant. However, 10% marrow-derived adipocytes were detected in visceral depots, and exposure to either high-fat diet or the proadipogenic thiazolidinedione, rosiglitazone, further increased marrow-derived adipocyte numbers to 10–20% (27). A caveat to the present studies is the inability to detect the fraction of marrow-derived adipocytes arising from recipient cells prior to transplant. Therefore, our previous and current data likely underestimate the contribution of adipocytes generated from nontissue resident progenitor cells to the pool of total adipocytes. This conclusion is supported by our prior fate-mapping analysis in mice, in which all myeloid cells and BMP-derived adipocytes were genetically labeled throughout life, revealing higher percentages of chimerism (>20% in visceral fat) as compared to adoptive transfer models (28). Thus, marrow-derived adipocytes are likely more numerous than our current or previous BM-transplant studies indicate.
Small changes in cellular composition can elicit significant functional changes in adipose tissue. For example, increases in the number of small adipocytes of ˂1% show strong positive correlation with the expression of monocyte markers and inflammatory cytokines in subcutaneous fat of insulin-resistant men (57). Another example is the high correlation between the percent area of beige adipocytes (0.2–14% in murine retroperitoneal fat and 4–45% in inguinal fat) and β-agonist–induced weight loss (58). Thus, although marrow-derived adipocytes typically constituted <15% of human subcutaneous fat cells, even modest production of donor-derived adipocytes may have a significant impact on adipose tissue function. This contention is also supported by studies demonstrating that low-level engraftment (<1–5%) of progenitor cells into damaged tissues (e.g., lung, liver, heart, kidney, and bone) contributes to tissue regeneration, reduces inflammation, suppresses cell death, and mitigates cell growth arrest (59–66). These beneficial results are primarily mediated by paracrine factors released by the nonresident progenitor cells (59, 67, 68). Future studies investigating the consequences of marrow-derived adipocyte accumulation and the mechanisms by which it influences the local adipose tissue environment are necessary.
Previous experiments with nonablative BM-transplant mouse fate-mapping models demonstrated that adipocytes were produced from hematopoietic cells, supporting the contention that they were generated via normal differentiation processes (28) vs. an injury response to irradiation. We recognize the possibility that the circulating progenitor-derived adipocytes detected in our patients may arise as the result of disease or “injury” following BM transplant. We are preparing to address this issue in humans by detecting HSC-restricted chromosomal translocations (e.g., BCR-Abl) in flow cytometry-purified adipocytes from transplant-naive leukemia patients. Finally, even if future studies reveal that the development of BMP-derived adipocytes is not a part of normal human physiology but rather an artifact of disease or treatment, this does not diminish the importance of these findings. Understanding how the generation of nontissue resident-derived adipocytes plays a role in the pathophysiology of immunocompromised conditions and impacts the long-term health of HSC-transplant recipients is of great importance.
Another concern is the possible skewing of results due to selective enrichment or loss of subsets of adipocytes during flow cytometry sorting. We believe that this is unlikely because the lumen diameter of the MoFlo XDP fluidics system is ∼250 µm, which should accommodate virtually all adipocytes. Moreover, the entire suspension of free-floating adipocytes can be sorted without clogging of the instrument or changes in fluidic stability. In addition, FSC vs. SSC plots of adipocytes during sorting are smooth and lack abrupt changes that might indicate loss of adipocytes of specific sizes. Finally, cell counts performed by flow cytometry or by microscope/hemocytometer match closely, further indicating minimal loss of adipocytes during flow sorting.
In summary, we have demonstrated a nontissue resident origin of subcutaneous adipocytes in humans who have undergone previous HSC transplantation, challenging the existing paradigm that all adipocytes in the adult are generated from tissue resident mesenchymal progenitors (Fig. 7). Ploidy analysis complemented by cytogenetic studies indicated that de novo HSC-derived adipocytes are likely produced by direct differentiation from donor cells rather than through cell fusion, which is not characteristic of adipose cells. Donor-derived adipocytes appear to accumulate over time in HSC-transplant patients, suggesting that they may have an increasing impact on adipose tissue biology with age, although confirmation of this idea awaits further experimentation. Furthermore, murine studies in which luciferase expression was governed by the adipocyte-restricted AdipoQ gene promoter provide evidence that BMPs contribute to de novo production of adipocytes in murine models. Our results highlight the existence of a novel lineage of adipocytes in humans and mice. Determination of the distinct developmental pathway of these novel adipocytes and their clinical relevance requires further investigation in both human and murine models.
Schematic of adipocyte production from donor HSCs. Donor HSCs engraft recipient BM. Donor-derived adipocyte progenitor cells then travel from BM, via the circulation, to the adipose tissue where they undergo transition into bona fide adipocytes. Adipocytes produced from donor cells continue to accumulate over time.
Acknowledgments
The authors thank Drs. J. Lee Nelson and Vijayakrishna K. Gadi (Fred Hutchinson Cancer Research Center and University of Washington Medical Center, Seattle, WA, USA) for discussions on chimerism analysis in humans. They also thank Dr. Kristine Erlandson (University of Colorado Anschutz Medical Campus) for medical safety oversight of the project. This research was funded by U.S. National Institutes of Health (NIH) National Institute of Diabetes and Digestive and Kidney Diseases Grants R01 DK078966 (to D.J.K.) and R21 DK092718 (to W.M.K.); and NIH Eunice Kennedy Shriver National Institute of Child Health and Human Development Grant P50 HD073063 (to W.M.K.); and NIH National Institute on Aging Grants T32 AG000279 and F32 AG046957 (research fellowships to K.M.G.). This research was also supported in part by NIH National Center for Advancing Translational Sciences Colorado Clinical and Translational Science Awards Grant UL1 TR001082 and NIH National Cancer Institute Centers for Common Disease Genomics Grant P30 CA046934. The authors declare no conflicts of interest.
Glossary
| AdipoQ | adiponectin |
| BM | bone marrow |
| BMI | body mass index |
| BMP | bone marrow progenitor |
| cat. no. | catalog number |
| CEP | chromosome enumeration probe |
| FSC | forward scatter |
| HSC | hematopoietic stem cell |
| PBSC | peripheral blood stem cell |
| PDGFRα | platelet-derived growth factor receptor α |
| Plin | perilipin |
| SSC | side scatter |
| STR | short tandem repeat |