Discovery of a stem-like multipotent cell fate

Abstract

Adipose derived stem cells (ASCs) can be obtained from lipoaspirates and induced in vitro to differentiate into bone, cartilage, and fat. Using this powerful model system we show that after in vitro adipose differentiation a population of cells retain stem-like qualities including multipotency. They are lipid (-), retain the ability to propagate, express two known stem cell markers, and maintain the capacity for trilineage differentiation into chondrocytes, adipocytes, and osteoblasts. However, these cells are not traditional stem cells because gene expression analysis showed an overall expression profile similar to that of adipocytes. In addition to broadening our understanding of cellular multipotency, our work may be particularly relevant to obesity-associated metabolic disorders. The adipose expandability hypothesis proposes that inability to differentiate new adipocytes is a primary cause of metabolic syndrome in obesity, including diabetes and cardiovascular disease. Here we have defined a differentiation-resistant stem-like multipotent cell population that may be involved in regulation of adipose expandability in vivo and may therefore play key roles in the comorbidities of obesity.

Introduction

Currently 35% of adults in the United States are obese [1], a condition associated with comorbidities including cardiovascular disease [2], diabetes [3], and even cancer [4]. Obesity comes in two forms, distinguished by the ability of subcutaneous fat depots to grow by the differentiation of new fat cells (hyperplasia) or alternatively, by the enlargement of existing adipocytes to accommodate greater lipid storage (hypertrophy) [5]. In hypertrophic obesity, lipids are also stored as ectopic visceral fat causing lipotoxicity, which contributes to adverse health outcomes and comorbidities including diabetes and cardiovascular disease [6].

The adipose tissue expandability hypothesis states that when excess energy cannot be stored in subcutaneous fat depots (i.e., more adipocytes cannot be formed), the existing adipocytes compensate by becoming larger (storing more lipid per cell), and lipids are also stored in other body regions (ectopically). The ability to differentiate new adipocytes from preadipocytes determines the limit on subcutaneous adipose tissue expandability [7]. Here we use adipose-derived stem cells (ASCs) [8] as a model system to investigate adipose differentiation in vitro.

ASCs are adult stem cells of the mesodermal lineage that can easily be purified from subcutaneous fat obtained from liposuction aspirates and can be differentiated into chondrocytes, adipocytes, and osteoblasts by adding the appropriate cocktail of hormones to the culture media [8,9]. However, we found that adipose differentiation always appears incomplete-a substantial subpopulation of lipid (-) cells is always present. Here we investigate these cells to determine whether they represent bona fide stem cells. The International Society for Cellular Therapy (ISCT) defines ASCs based on the following criteria: cells must adhere to plastic; cells must express three surface markers CD73, CD90, and CD105; and cells must have trilineage differentiation potential to produce bone, cartilage and fat [10]. Here we show that the lipid (-) cells satisfy most of these criteria: they adhere to plastic, they express two of the three markers, and they retain trilineage differentiation capacity. Surprisingly, however, their gene expression profile strongly resembles that of adipocytes, suggesting that multipotency can be retained in previously unknown cell populations that may exist on a continuum between classical stem cells and differentiated fates.

Material and methods

Cell culture

Cell lines were grown in plastic 6-well dishes with growth media maintained in a humidified 5% CO₂ incubator at 37°C. The growth media consist of Dulbecco’s Modified Eagle Medium (DMEM, ThermoFisher #11965118) supplemented with 10% fetal bovine serum (ThermoFisher #10082147), 1X Penicillin/Streptomycin (ThermoFisher #15140122), and 1X Glutamax (ThermoFisher #35050061). Cell washes were performed with DPBS (ThermoFisher #14190144) and passages with Trypsin-EDTA 0.25% (ThermoFisher #25200056). Media was always changed every 72-96 hours under sterile hood for both growth and differentiation.

Culture volumes for 6-well dishes (VWR Cat #10062-892) was 2 ml/well. For 16-chamber slides (VWR Cat #62407-350), 200 ul was used, but to avoid contamination and evaporation during 2-3 week differentiations slides were housed inside a sterile plastic petri dish.

Adipocyte differentiation

For adipocyte differentiation, cells were cultured to confluence. Then, the growth media was removed and replaced with adipogenic differentiation media consisting of 10% FBS, 1% Penicillin/Streptomycin, 1X Glutamax, 1.0 μM Dexamethasone, 0.5 mM IBMX (3-isobutyl-1-methylxanthine), 0.2 mM Indomethacin, and 10.0 μM Insulin.

Chondrocyte differentiation

For chondrocyte differentiation, cells were cultured to confluence in 6-well dishes. Then, the growth media was removed, cells trypsinized and concentrated at 200 g for 5 minutes into 2-3 micromass 10 μL droplets in a new dish (i.e., one well was concentrated to 3 drops). They were allowed to adhere at 37°C in an atmosphere of 5% CO₂ for 45 minutes, followed by gentle flooding of the new well with chondrocyte differentiation media (2 mL). Chondrocyte media is composed of 1% FBS, 1% PenStrep, 6.25 μg/mL Insulin, 10 ng/mL TGF-β1, and 50 nM Ascorbate-2-Phosphate.

Osteoblast differentiation

For osteoblast differentiation, cells were cultured to confluence. Then, the growth media was removed and replaced with osteoblast differentiation media consisting of 10% FBS, 1% PenStrep, 50 μM ascorbate-2-phosphate, 0.1 μM dexamethasone, and 10 mM β-glycerophosphate.

Trilineage differentiation

Confluent cells grown in plastic 6-well dishes were treated with adipogenic differentiation media (day 0). Media changes were performed under a sterile hood every 72-96 hours for 27 days. These adipocyte cells were then trypsinized and centrifuged at 200 g for 5 minutes. The floating lipid (+) cells were discarded, and the lipid (-) cells were re-plated and placed into standard growth media. Once again, these cells were grown to confluency with regular media changes every 72-96 hours. On day 45, the trilineage differentiation was initiated by placing an individual population of cells (2 wells) into either chondrogenic differentiation media, adipogenic differentiation media, or osteogenic differentiation media. These cell populations were permitted to differentiate with regular media changes every 72-96 hours, for 26 days. On day 66, the cells were fixed and stained using Alcian Blue, Oil Red O, and Von Kossa and imaged. In brief, cells were washed twice with DPBS, fixed in 4% paraformaldehyde (diluted from 20%, Electron Microscopy Sciences #15713-S) in PBS for 1 hour (for adipocytes, supplemented with 1% calcium chloride).

For adipocytes, cells were stained with Oil Red O (ORO; Santa Cruz Biotechnology #sc-203749) working solution for 5 minutes, then washed in 60% ethanol for 5 minutes and rinsed 3 times with diH₂O and imaged immediately under DIC microscopy on an inverted microscope. The ORO working solution was prepared by dissolving 0.1% w/v OrO powder directly into 99% Isopropanol; this was then diluted to 60% isopropanol in diH₂O and filtered through a 0.2 μm filter prior to use.

For osteoblasts, fixed cells were rinsed twice in DPBS, and then incubated in 2% Silver Nitrate Solution (5 ml diH₂O with 0.1 g silver nitrate powder, Sigma-Aldrich #209139) for 30 min in the dark. After a DPBS wash, cells were air dried and imaged on an inverted DIC microscope; sufficient darkening occurred spontaneously without use of UV light to further develop the signal.

For chondrocytes, fixed cells were washed in diH₂O twice and then incubated with 1% wt/vol Alcian Blue (Sigma-Aldrich #A3157) in 0.1N HCl (pH 1.0) for 30 min; they were then washed in 0.1N HCl for 5 minutes to remove excess stain, and imaged under DIC microscopy.

Surface marker immunofluorescence and oil red O (ORO) staining

ASCs were cultured in chamber slides to ~95% confluency and analyzed. For differentiated cells, ASCs were cultured in growth media in chamber slides until confluent, and differentiated in adipocyte media for 18-21 days. Protocol A is described in Koopman et al. 2001 [11]. Briefly, media was aspirated and cells were fixed in 3.7% formaldehyde (VWR #97064-888) in PBS. After 1 hr the cells were rinsed 3x in diH₂O before permeabilization in 0.5% Triton X-100 in PBS for 5 min, followed by rinsing 3x in PBS and incubation with primary antibodies in PBS overnight at 4°C. The cells were then rinsed 3x in PBS, then incubated with secondary antibodies in PBS for 1 hr, and rinsed again 3x in PBS. Cells were then incubated in Oil Red O working solution (see below) for 30 min at room temperature, then rinsed 3x in diH₂O and then in gently flowing tap water for 10 minutes before mounting in Prolong Gold Antifade Reagent with DAPI (Cell Signaling #8961S). Negative controls lacking primary antibodies were included in all experiments. Protocol B is published on the Cell Signaling website (https://www.cellsignal.com/contents/resources-protocols/immunofluorescence-general-protocol/if). Briefly, cells are fixed in 4% methanol-free formaldehyde (Polysciences, Inc #18814-10) in PBS for 15 min at room temperature, followed by 3 rinses in PBS, blocking for 1 hour (1X PBS/5% Goat serum albumin/0.3% Triton^TM X-100), then incubated with primary antibodies overnight. Samples were then incubated with secondary antibodies and washed as in Protocol A, followed by mounting in Prolong Gold Antifade Reagent with DAPI.

Primary antibodies: CD90, BD Pharmingen 550402 used at 1:100 dilution; CD105, ThermoFisher PA5-16895 used at 1:50 dilution. Secondary (FITC) antibody was Santa Cruz sc-2010, used at 1:1000 dilution.

For the Oil Red O working solution, a stock was prepared as follows: 10 mg of ORO powder (Santa Cruz Biotechnology #sc-203749) was dissolved in 2 ml of 60% Triethyl phosphate (Santa Cruz #sc-251322). This was then further diluted to a 36% Triethyl-phosphate solution by combining 1.2 ml of this stock with 0.8 ml diH₂O, and filtered through a 0.2 μm vacuum filter (VWR #10040-460) to generate the working solution used in cell staining.

Laser confocal microscopy

Imaging and z-stack construction were performed using an Olympus FV1200 Laser Scanning Microscope equipped with FV10-ASW Viewer software (Olympus).

FACS sorting & microarray

ASC080414AF2 clonal cells (passage 12) were grown to confluency and differentiated in adipocyte media (AM) for 23 days. After trypsin treatment cells were FACS separated based on side scatter (SSC) using gates from ‘high SSC’ (lipid +) to ‘low SSC’ (lipid -). ASC080414AF2 stem cells were grown in parallel in growth media to approx. 70% confluency and then FACS sorted using the ‘low SSC’ gate to control for effects of sorting and gating. All cells were directly sorted into DNA/RNA Shield Buffer from the Zymo Duet DNA/RNA MiniPrep Plus kit (Zymo #D7003), followed by RNA extraction as per provided protocol. RNA was submitted to the Johns Hopkins University Microarray Core for Human PrimeView Gene Expression Array (Affymetrix) analysis.

Principal component analysis & hierarchical clustering analysis

Microarray data was processed using the Expression Console 1.4.1.46 for PCA and Transcriptome Array Console v3.1 for Hierarchical Clustering Analysis plus differential gene expression analysis.

Go analysis

The goatools v. 0.7.11 python package was used, with the ‘find_enrichment.py’ command using defaults (propagation of counts to parent terms). The PrimeView probe set and associated GO terms were used as the population data while Tables S1 and S2 were used as study data.

Cells used in the study (Table 1)

Table 1.

Cells used in the study

Cell Line	Source	BMI	Depot	Sex	Donor Age	Passage at Experiment	Figure
ASC100610B	Zen-Bio, Inc.	23.3	Abdomen	F	40	p6	Figure 5B, 5C
ASC021606	Zen-Bio, Inc.	32.1	Abdomen	F	46	p4	Figures 4, 5A
ASC080414A-derived clonal line	Zen-Bio, Inc.	25.1	Abdomen	F	39	p12	Figures 2, 3, 6
0912	DeCicco-Skinner lab, AU	31.6	Breast	M	27	p5	Figure 1

Results

We observed that adipose-differentiated human ASCs always result in two apparently distinct populations, lipid (+) adipocytes and lipid (-) cells (Figure 1). We hypothesized these lipid (-) cells might represent a previously unknown population of quiescent stem cells maintained even during culture in differentiation-inducing conditions, but to study them further we needed to isolate them.

Figure 1.

Fixed differentiated cells stained with Oil Red O (ORO) and nuclei counterstained with DAPI to demonstrate the existence of two populations: lipid-positive adipocytes and lipid-minus cells.

We found that by FACS sorting on side scatter alone we were able to cleanly separate lipid (-) from lipid (+) cells as measured both by lipidTOX dye and by visual inspection of sorted cells (Figure 2). Consistent with their quiescent stem-cell status, the lipid (-) population demonstrated good replicative ability while the lipid (+) cells do not appreciably replicate (Figure 2C). We extracted RNA from differentiated FACS-sorted lipid (+), lipid (-), and matching stem cells (~70% confluent, collected through the same gate as lipid (-) on FACS), in triplicate. Because the FACS-sorted lipid (-) population yielded very low RNA levels, we pooled all three replicates and obtained seven total PrimeView microarray gene expression profiles covering over 49,000 individual loci and all annotated human genes.

Figure 2.

Demonstration that FACS can effectively separate lipid (-) from lipid (+) cells. (A) FACS-sorted stem cells showing Gate 1 captures stem cells. (B) FACS of adipose differentiated cells on side scatter (SSC) and forward scatter (FSC). Gates 1 and 4 cleanly separate two populations, lipid (+) and lipid (-), as shown by lipidTOX green dye readout, and by visual inspection in (C). (C) FACS sorted differentiated cells were visually inspected (i) and (iii) and allowed to propagate 10 days in growth media, (ii) vs. (iv). After 10 days only the lipid (-) cells had propagated, (ii) vs. (iv).

We used Principal Component Analysis (PCA) to evaluate relative gene expression profile differences. PCA captures the variation between extremely large complex datasets into new variables (components) that capture the between-dataset variation [12]. Plotting the top three components as axes on a 3D graph and locating individual datasets within that space graphically represents the relative similarity of datasets, with more similar samples clustering together [12]. In the case of our FACS-sorted cells, the top three PCA components capture 97.4% of the variation between the datasets, with PC1 alone capturing 78.8%. This component represents a clear differentiation axis with stem cells at one end and adipocytes at the other; the lipid (-) cells lie between them but closer to the adipocyte samples (Figure 3A).

Figure 3.

Analysis of gene expression data from FACS-sorted stem cells and lipid-minus cells. A. Principal Component Analysis (PCA) used to characterize gene expression profiles of FACS-separated adipocytes (red), actively growing stem cells (blue) and lipid-minus cells (yellow). B. Hierarchical clustering of 39 genes statistically up-regulated (P < 0.05 by ANOVA and over 2-fold change) in lipid-minus relative to adipocytes. Arrows indicate probes for Gremlin-1 and -2, and PTGS2/COX-2. C. Individual gene expression data from microarray. Microarray data for each gene was normalized to β-actin and displayed on a linear scale. Error bars represent standard deviation of three replicate microarrays; due to limited RNA yield the lipid (-) replicates were pooled into a single microarray. Student’s 2-tailed T-test was used to compare adipocytes to stem cells, with * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

Based on this, we conclude that the lipid (-) population does not represent traditional stem cells at least by gene expression analysis. However, the lack of lipid accumulation suggests lipid (-) cells may be at least partly undifferentiated, so we searched for genes that might contribute to this shift of the lipid (-) cells toward stem cells and away from adipocytes in PCA space (Figure 3A). In total 39 genes are statistically upregulated in lipid (-) relative to adipocytes by ANOVA (p < 0.05) and at least twofold expression change (Figure 3B; Table S1). Using unsupervised hierarchical clustering [13] to group the samples based on the gene expression profiles of these 39 genes produced a consistent grouping of lipid (-) cells with adipocytes, rather than stem cells (Figure 3B) similar to the result seen with PCA analysis (Figure 3A). These genes give insight into the mechanisms by which these quiescent stem cells are maintained in the population: they overexpress PTGS2/COX-2, which is a direct target of dexamethasone [14], a differentiation-inducing glucocorticoid hormone added to the adipocyte-differentiation media [8,9]. Furthermore, lipid (-) cells upregulate Gremlin-1 and Gremlin-2, which are antagonists of BMP signaling [15]. Gremlin-1 has been reported to be highly expressed in undifferentiated adipose stem cells or preadipocytes [16] consistent with its expression here in a quiescent stem cell population. Therefore, this lipid (-) cell population has specific mechanisms for counteracting the differentiation-inducing effects of adipose-differentiation media. Additionally, lipid (-) upregulated genes include CD1d, an MHC class I molecule that presents lipid antigens; tumor necrosis factor member 11 (TNFRSF11B, or Osteoprotegerin), involved in bone development; FAM20A, involved in hematopoiesis and tooth development; and Homeobox C13 (HOXC13), involved in hair development (Table S1). Interestingly, HOXC13 has recently emerged in meta-analysis of GWAS studies as a gene affecting fat distribution [17], suggesting it plays unappreciated adipose-related functions, and highlighting the value of our microarray data. The BMP antagonist Noggin was higher in lipid (-) than in either stem cells or lipid (+) adipocytes (Figure 3C) but due to its low overall expression this was not scored as statistically significant and is not included in Table S1.

These data suggest that lipid (-) cells may be primed toward non-adipose lineages including bone. Gene Ontology (GO) analysis of Table S1 confirms this, as GO:0031214, ‘biomineral tissue development’ was statistically over-represented after Bonferroni correction (Table S3). In addition, two smooth-muscle-related terms appear in this list: GO:0048660, ‘regulation of smooth muscle cell proliferation’ and GO:2000097, ‘regulation of smooth muscle cell-matrix adhesion’ suggesting that the lipid (-) cells may be poised between mesodermal fates known to be within the differentiating range of these cells [8,9]. These data support the idea that lipid (-) cells occupy a different fate space than adipocytes, as suggested by our PCA analysis (Figure 3A).

Genes upregulated in adipocytes relative to lipid (-) cells are generally adipocyte-specific lipid or glucose metabolism genes (Table S2). For example, adiponectin (ADIPOQ) and perilipin-1 (PLIN1) were significantly up, as was adipogenesis regulatory factor (ADRIF) (Table S2; Figure 3C). The most strongly upregulated genes include mitochondrial glycerol-3-phosphate dehydrogenase 1 (GPD1), a metabolic link between glucose and lipid metabolism [18] and a known adipocyte-specific gene [19], and glycerol-3-phosphate acyltransferase, (GAPM), a gene linked to a congenital lipidistrophy disorder [20]. The top adipocyte upregulated gene was phosphoenolpyruvate carboxykinase 1 (PCK1), which regulates the critical step in adipocyte glyceroneogenesis [21]. In total 141 genes were statistically upregulated in adipocytes relative to lipid (-) cells (Table S2). Gene Ontology (GO) analysis of Table S2 confirmed this: statistically enriched are GO:0019432, ‘triglyceride biosynthetic process’, GO:0046460, ‘neutral lipid biosynthetic process’, GO:0032868, ‘response to insulin’, and strikingly, GO:0005811, ‘lipid droplet’ (Table S4).

We observed that expression of two of the three classical stem-cell markers [10] were not apparently decreased upon adipose differentiation: CD90 expression appears to increase upon differentiation while CD105 is relatively unaffected (Figure 3C). To confirm this we performed immunofluorescence against these three markers, enabling us to examine both their differentiation behavior and their expression in lipid (-) cells. We found that both CD90 and CD105 are detected in lipid (-) cells after differentiation (Figure 4A), while CD73 was not reliably detected by immunofluorescence and thus is not shown. Consistent with our observation that lipid (-) cells are relatively undifferentiated (Figure 3A), both CD90 and CD105 are depleted in (differentiated) lipid (+) adipocytes, being predominantly detected in lipid (-) cells (Figure 4A-C). The staining pattern of CD90 was highly punctate, raising the concern that it might be marking cytoplasmic structures, so we performed DIC imaging along with immunofluorescence to confirm that signal is indeed co-planar with cell membranes (Figure 4B). Quantitation showed that cells generally express either a stem cell surface marker (green) or Oil Red O, but rarely both, and that similar proportions of cells are positive for CD90 and CD105 (Figure 4C).

Figure 4.

CD90 and CD105 surface markers are detected by immunofluorescence in differentiated cell line ASC021606 (BMI = 32.1). A. Cells were fixed and stained using Protocol A (Koopman, et al., 2001 [11]), with FITC signal (green) for surface markers and counterstained with Oil Red O for lipid content. No primary controls: identically processed samples but only stained with secondary (FITC) antibodies to control for background. All panels represent compressed Z-stacks taken and displayed under identical conditions. Scale bars = 20 μm. B. DIC images (planar single images, not stacks) showing that surface marker signal is co-located with membrane. Scale bars = 20 μm. C. Quantitation of green and red cells observed after CD90 or CD105 immunofluorescence with Oil Red O staining. Data represent the average and standard deviation of three fields.

We confirmed both stem cell markers are detectable in stem cells and that the signal is also co-planar with the membrane as revealed by DIC imaging (Figure 5A). We observed a consistent punctate CD90 pattern and tested whether it was potentially an artifact of fixation method or use of Oil Red O in our staining protocol. While omitting Oil Red O increased CD90 signal intensity, the signal was still punctate (Figure 5B); similarly using a different (Oil Red O-free) protocol (see Methods) also revealed the punctate pattern (Figure 5C). We therefore conclude that the CD90 pattern observed here is not caused by treatment of samples with Oil Red O, nor is it limited to a single cell type or BMI since samples from both obese (Figure 5A) and normal (Figure 5B, 5C) donors produce a consistent pattern. Both Protocols A and B use formaldehyde (one is methanol-free and the other is not): methanol fixation did not yield good immunofluorescence signal so we cannot rule out the possibility that formaldehyde fixation induces punctate CD90 signals. However, we consider this unlikely because the CD105 signal was relatively more diffuse in both stem and differentiated cells imaged on the same slides and treated the same as CD90 (Figures 4A, 4B, 5A).

Figure 5.

CD90 and CD105 surface markers are detectable by immunofluorescence in stem cells. A. Stem cells (line ASC021606, BMI = 32.1) processed using Protocol A, imaged from a single plane to show membranes by DIC. B, C. Controls to verify punctate CD90 signal is still evident without Oil Red O treatment and in a different cell line (ASC100610B, BMI = 23.3); two fixing and staining protocols used (see Methods). B. Protocol A, (Koopman et al. 2001 [11]) omitting ORO gives much stronger signal than with ORO (compare to A). C. Protocol B gives fainter signal but has lower background. All panels in B and C represent compressed Z-stacks taken and displayed under identical conditions. Scale bars = 20 μm.

Given that the lipid (-) cells were not apparently ‘true’ stem cells based on their gene expression profile, we queried their multi-lineage stem potential. ASCs are considered stem cells if they retain ability to differentiate into bone, cartilage, and fat [10] so we devised an experiment to first differentiate, then isolate lipid (-) cells, re-grow them, and perform multi-lineage differentiation. We found that lipid (-) cells retain full multi-lineage potential and thus are functional stem cells (Figure 6). This finding fundamentally broadens our understanding of stem potential and quiescence, suggesting that a broader array of cells may retain multipotency than previously appreciated.

Figure 6.

Demonstration that lipid (-) cells are multipotent. After one round of adipogenic differentiation cells were float-separated to isolate the lipid (-) population, which were re-propagated for subsequent trilineage differentiation. Cell staining was used to verify (i) chondrocyte, (ii), adipocyte, and (iii) osteoblast lineages.

Discussion

In this study we show that a previously overlooked lipid (-) population continues to display two of three stem-cell markers even after adipose differentiation (Figure 4), and although its gene expression profile is relatively adipocyte-like (Figure 3) these cells are multipotent (Figure 6). Since they adhere to plastic, these cells satisfy the ISCT criteria for multipotent stem cells [10], with the exception of CD73 expression (Figure 3), suggesting that they are not ‘canonical’ stem cells. Therefore we have elucidated a novel stem-like multipotent fate that does not require other cell types for specification, instead being established by adipose-intrinsic cell fate determination in vitro. While quiescent stem cells have been reported in skin [22,23], gut [24], blood [25], and neurons [26], they have yet to be defined in adipose tissue [27], although active adipose stem cell depots have been reported in the adipose perivasculature [28-31]. Here we describe a population of quiescent stem cells that may constitute a source of ASCs from lipoaspiration [8,9] and under physiological regulation in vivo may activate into more classical stem-cell states (bearing the three surface markers [10]) specifically at times when new terminally differentiated adipocytes are needed by the organism.

This quiescent stem-like fate may be one mechanism used by the body to preserve adipose expandability. According to the adipose expandability hypothesis, when excess energy cannot be stored in subcutaneous fat depots through differentiation of new adipocytes, the existing adipocytes compensate by becoming larger (storing more lipid per cell), and lipids are also stored in other body regions (ectopically) [6,7]. Ectopic lipid accumulation in turn causes lipotoxic effects, leading to insulin resistance, apoptosis and inflammation. The ability to differentiate new adipocytes is limited by an unknown mechanism [7]; here we have defined an apparent cell-intrinsic mechanism determining efficiency of adipose cell differentiation in vitro. We hypothesize that when the lipid (-) stem-like multipotent cells are depleted or not adequately maintained, adipose expandability is lost and hypertrophic obesity results. Alternatively, if a stem cell population over-commits to lipid (-) cells, it may fail to differentiate new adipocytes when they are needed, thereby limiting adipose expandability at the other extreme: retaining too many differentiation-resistant quiescent stem cells.

In vivo data appears to support this general model. One study found an inverse relationship between Body Mass Index (BMI) and lipoaspirate-derived stem cells yields, consistent with a depletion in the stem-cell population under obese conditions [32]. In diabetic patients (generally also more obese) fewer pre-adipocytes are found in adipose tissues compared to the control (non-diabetic) population [33]. Even those stem cells retained in obese individuals may be compromised: the ability of preadipocytes to differentiate into mature adipocytes was impaired in hypertrophic obesity [34,35] and in both diabetes-predisposed [36] and diabetic [33,37] individuals. Taken together, these mechanisms would severely limit the ability of hypertrophically obese individuals to generate new adipocytes, which may drive ectopic lipid deposition.

Our discovery of a quiescent stem-like cell population in adult adipose tissue fundamentally expands our understanding of cellular multipotency. In light of the link between BMI and stem cell abundance and behavior, future studies are necessary to investigate the relative role of a stem-like fate in comorbidities of obesity.

Disclosure of conflict of interest

None.