ITGAV and ITGA5 diversely regulate proliferation and adipogenic differentiation of human adipose derived stem cells

The fate of human adipose tissue stem cells (ASCs) is largely determined by biochemical and mechanical cues from the extracellular matrix (ECM), which are sensed and transmitted by integrins. It is well known that specific ECM constituents influence ASC proliferation and differentiation. Nevertheless, knowledge on how individual integrins regulate distinct processes is still limited. We performed gene profiling of 18 alpha integrins in sorted ASCs and adipocytes, identifying downregulations of RGD-motif binding integrins integrin-alpha-V (ITGAV) and integrin-alpha-5 (ITGA5), upregulation of laminin binding and leukocyte-specific integrins and individual regulations of collagen and LDV-receptors in differentiated adipocytes in-vivo. Gene function analyses in in-vitro cultured ASCs unraveled differential functions of ITGA5 and ITGAV. Knockdown of ITGAV, but not ITGA5 reduced proliferation, caused p21Cip1 induction, repression of survivin and specific regulation of Hippo pathway mediator TAZ. Gene knockdown of both integrins promoted adipogenic differentiation, while transgenic expression impaired adipogenesis. Inhibition of ITGAV using cilengitide resulted in a similar phenotype, mimicking loss of pan-ITGAV expression using RNAi. Herein we show ASC specific integrin expression patterns and demonstrate distinct regulating roles of both integrins in human ASCs and adipocyte physiology suggesting a negative impact of RDG-motif signaling on adipogenic differentiation of ASCs via ITGA5 and ITGAV.

Scientific RepoRts | 6:28889 | DOI: 10.1038/srep28889 differentiation and cell death has been connected to integrin-dependent adhesion [10][11][12][13] . Controlled by extracellular mechanical cues such as ECM rigidity or cell-cell contacts, the Hippo pathway mediates its signaling by modulating the expression and activity of the two major downstream effectors Yes-associated protein (YAP) and transcriptional co-activator with PDZ-binding motif (TAZ). Both proteins act as transcriptional co-factors controlling the expression of Hippo pathway target genes such as connective tissue growth factor (CTGF) 14 or survivin 15 . All of the mentioned signaling pathways are involved in the regulation of proliferation, migration and differentiation and thus plainly able to influence cell destiny 16,17 and tissue development. Therefore, a fundamental understanding of matrix-integrin interactions is important in order to elucidate basic ECM requirements of ASCs.
Previous studies have shown that cell proliferation of primary ASCs is favored by the presence of RGD-motif containing substrates such as fibronectin or vitronectin 18 . Signals from those ECM constituents are mainly recognized by integrin-alpha-5 (ITGA5) and integrin-alpha-V (ITGAV). While ITGAV serves as a subunit for integrin receptors binding RGD-motif containing substrates such as vitronectin, fibronectin and fibrinogen 19 , ITGA5 is mainly part of fibronectin-and osteopontin binding receptors 20 . Thus far, ITGAV has been shown to play an important role in the regulation of cancer growth and metastasis 21 . Especially the ITGAV/ITGB3 heterodimer has been connected with tumor neoangiogenesis via high levels of bFGF and tumor-necrosis factor α (TNFA) whereas the ITGAV/ITGB1 receptor is implicated in tumor cell proliferation via BCL2 and p53 activity 22 . However, the role of ITGAV as a fibronectin receptor and its impact on ASC cell physiology and adipogenic differentiation remains poorly defined.
To continue the research into the role of integrins in tissue remodeling, we herein analyzed the integrin expression profile of sorted primary ASCs and adipocytes in-vivo. We identified RGD-binding integrins as being consistently repressed during adipogenesis and outlined differential roles for ITGAV and ITGA5 in ASC proliferation and differentiation. We defined a molecular mechanism how these integrins diversely regulate ASC proliferation and showed that the specific phenotype can be induced by pharmacological inhibition of ITGAV signaling using cilengitide.

RGD-receptors ITGA5 and ITGAV are repressed during adipogenesis. Previous studies have
demonstrated, that culture of MSC on fibronectin or RGD-peptide coated plates promotes proliferation but not differentiation 18 . We found that all RGD-motif binding integrins were repressed during adipogenesis in-vivo, and therefore focused on ITGAV and ITGA5 for functional analyses. Pharmacological inhibitors 24,25 have been designed for both integrins and are currently being tested in clinical oncologic studies, which further prompted us to investigate their possible functional significance for tissue restoring applications.
First, we confirmed mRNA expression data by immunohistochemistry and immunoblot analysis of sorted ASCs and primary adipocyte lysates. ITGA5 was strongly expressed in undifferentiated cells, but downregulated in mature adipocytes ( Fig. 2A,D), which can be morphologically identified as "signet ring cells" in tissue sections and reveal a homogenous intracellular, peripheral expression pattern for adipocyte specific marker PLIN1 ( Supplementary Fig. 3). ITGAV levels were similarly high in undifferentiated cells (Fig. 2D). However, expression was not completely abolished in mature adipocytes (Fig. 2B,D). Instead, residual protein accumulations at specific sites between adipocyte membranes were observed. The dotted pattern might indicate either development of cell-cell contacts to enhance the mechanical properties of the tissue or existence of RGD-motif containing ECM constituents surrounding adipocytes. Testing the second possibility, fibronectin staining showed increased levels in the ECM of undifferentiated cells but did not stain protein accumulations at adipocyte-adipocyte contact sites (Fig. 2C).
To investigate whether the observed regulations were also detectable in in-vitro differentiated cells, we subjected ASCs to adipogenesis and determined ITGA5 and ITGAV expression by immunoblotting. ITGA5 expression in in-vitro differentiated adipocytes decreased, corresponding to our in-vivo observations. Surprisingly, ITGAV expression was slightly induced upon in-vitro differentiation, which we suspected to be due to the amount of non-differentiated cells (Fig. 2E, left panel). To enrich differentiated cells, lipid droplet containing cells were selectively isolated by means of an OPTIPREP TM based density centrifugation method 26 and subjected to immunoblotting (Fig. 2E, right panel). Similarly to in-vivo differentiated adipocytes, both integrins were strongly downregulated in differentiated cells. In contrast, cells subjected to adipogenesis that did not develop sufficient numbers of lipid droplets also showed reduced ITGA5 levels, but intriguingly showed an upregulation of ITGAV, which might explain increased ITGAV levels in unseparated samples. In summary, the results supported our mRNA data showing a decline in expression of both integrins during adipogenesis and therefore suggest an essential function in adipogenic differentiation.

Loss of ITGAV moderately induced cell death and reduced cell proliferation.
To further investigate the importance of ITGA5 and ITGAV in ASC physiology, we designed two independent shRNAs for each integrin and cloned them into lentiviral GATEWAY TM RNAi vectors 27 . For overexpression experiments, ITGA5 and ITGAV were expressed from corresponding lentiviral vectors (Fig. 3). In ASCs transduced with the RNAi vector, ITGA5 and ITGAV total protein levels were repressed to less than 10% of basal protein levels in control-shRNA (Ctr-shRNA) transduced cells (Fig. 3A). Analysis of ITGA5 and ITGAV cell surface levels by immunostaining and flow cytometry showed an average knockdown (KD) efficacy of 75-80% (Fig. 3B,C). Interestingly, we observed a compensatory upregulation of ITGA5 surface expression in ITGAV depleted cells. Vice versa, ITGAV was moderately upregulated in ITGA5-KD cells. As this regulation was not detected on mRNA level, increased synthesis, transport or protein stabilization of ITGA5 might occur in ASCs upon loss of ITGAV. As the cell morphology of ITGAV-KD cells differed from that of ITGA5-KD or Ctr-shRNA infected cells (Fig. 3G), we further analyzed cell proliferation and viability. Automated cell counting (Fig. 3E) and mitochondrial activity as a marker of cell proliferation (Fig. 3D) showed a strong decrease in cell proliferation in ITGAV-KD cells at day five post infection. In contrast, ITGA5 depletion did not significantly reduce cell numbers or proliferation. However, at later time points loss of ITGA5 resulted in similarly low cell numbers suggesting a cell proliferation regulating role of ITGA5 in ASCs by continuous depletion (Supplementary Fig. 2). Analysis of cell viability by annexinV/PI staining in transduced ASCs (Fig. 3F) showed a moderate, but not significant increase in apoptotic cell numbers (mean: 9.5% ± 6.8, p = 0.19) in ITGAV-KD but not ITGA5-KD cells, indicating that some cells undergo cell death upon loss of ITGAV within the investigated time.
Intracellular signaling pathways differentially regulated by ITGAV and ITGA5. Ligand binding to integrins activates many signaling pathways, including MAPK, AKT and Hippo pathways following SRC/

Figure 2. Expression of fibronectin receptors ITGAV and ITGA5 in-vivo and in-vitro.
Immunohistochemistry analysis of paraffin fixed human subcutaneous fat tissue sections using specific antibodies against (A) ITGA5, (B) ITGAV and (C) fibronectin. Adipocytes can be morphologically identified as "signet ring cells" containing large fat vacuoles in the images. Protein expression was visualized using horseradish peroxidase (brown) and hematoxylin counterstaining (blue). Note the small ITGAV protein accumulations indicated by arrows at connecting adipocyte membranes (B). (D) Sorted ASCs and primary adipocytes of two representative donors (Donor#1 and Donor#2) were subjected to immunoblotting analyzing the expression of ITGA5, ITGAV and adipocyte specific marker FAPB4. (E) Cell lysates of in-vitro differentiated ASCs (day 14, left panel; proliferating cells are labeled with "prolif ", differentiated cells are labeled with "diff ") and the same cells separated according to their lipid droplet content (LD, right panel; lipid droplet containing cells are labeled with "+LD", cells containing no lipid droplets are labeled with "−LD") were subjected to immunoblotting detecting ITGAV, ITGA5, GAPDH and the adipocyte specific markers FABP4 and PLIN1. As loading controls of the depicted immunoblots cropped images of total protein stains acquired before blotting are shown. Full length gels are presented in Supplementary Fig. 4. FAK activation. One of the most cited pathways affected by integrin signaling is mediated by extracellular signal-regulated kinases 1/2 (ERK 1/2) 28 , which is controlled by MAP kinase-kinase (MAP2K) mediated phosphorylation at Thr202/Tyr204 29 . Analysis of phospho-ERK1/2 in proliferating ITGAV-and ITGA5-KD cells did not show integrin specific differences in ERK1/2 phosphorylation at Thr202/Tyr204, indicating that this signaling pathway was not responsible for the observed cell cycle effects (Fig. 4A). Next we analyzed S473 phosphorylation of AKT, as previous findings suggested a PI3-kinase dependent regulation of AKT upon integrin signaling at this phosphorylation site 30 . However, AKT phosphorylation was almost undetectable in transduced ASCs and no integrin-specific differences could be observed (Fig. 4A). Next, the expression of the two major downstream effectors of the Hippo signaling cascade, YAP and TAZ, was analyzed. Immunoblotting showed that loss of neither ITGAV nor ITGA5 exerted an influence on YAP levels ( Fig. 4B), while TAZ levels were about 2-fold lower in ITGAV-KD cells than in ITGA5-KD or control infected cells. To evaluate whether this decrease affected the expression of known TAZ target genes we determined levels of CTGF and survivin in these cells by quantitative RT-PCR. As expected, CTGF levels showed a moderate decrease in ITGAV-KD cells. More impressive, survivin was strongly repressed in ITGAV but not in ITGA5-KD cells (Fig. 4B,D). The data suggested that ITGAV signaling is linked to the Hippo pathway by regulation of TAZ levels and repression of survivin, a cell cycle regulator and caspase inhibitor 31 , might be responsible for impaired proliferation in ITGAV-KD cells. Therefore, we performed rescue experiments by introducing transgenic survivin-cDNA or GFP-cDNA as control in ITGAV-KD cells and analyzed cell proliferation (Fig. 4E). Although survivin expression strongly overwhelmed levels in GFP-transduced cells, transgenic survivin was insufficient to rescue the proliferation phenotype induced by ITGAV-KD (Fig. 4F). This suggested that repression of survivin was not causative for the decrease in cell proliferation.
Loss of ITGAV mediated upregulation of p21 Cip1 . Since the so far investigated pathways did not sufficiently explain reduced cell proliferation upon loss of ITGAV, we investigated levels of cyclin-dependent kinase inhibitor 1A (p21 Cip1 ), which has been shown to be stabilized upon disruption of filamentous actin (F-actin) 32 . Interestingly, we observed an up to 4-fold upregulation of p21 Cip1 mRNA and protein levels in ITGAV-KD (p = 0.0072) but not in ITGA5-KD cells (Fig. 5A,B). As it is known, that the members of the tumor suppressor protein p53 family 33 are regulators of p21 Cip1 , we determined levels of p53 and p73. As shown in Fig. 5A, neither p53 nor p73 levels increased in ITGAV-depleted cells, suggesting alternative induction of p21 Cip1 in ITGAV-and ITGA5-KD cells.

Loss of ITGAV promotes adipogenic differentiation of ASC. Previous studies have already shown
that expression of ITGA5 counteracts adipogenic differentiation 34 . Moreover, pericellular fibronectin levels strongly decrease during adipogenesis, essentially connecting fibronectin signaling to the differentiation process 34,35 . Since ITGAV levels declined, but were not as strongly repressed as ITGA5 in adipogenesis, we functionally tested whether loss of ITGAV induced a similar phenotype in differentiating ASCs as loss of ITGA5. Vice versa, we performed gain of function experiments by constitutively expressing transgenic ITGAV-or ITGA5 in these cells to elucidate whether inhibition of differentiation is mediated by integrin expression. Transduced cells were subjected to adipogenesis and the formation of lipid droplets and expression of classical adipocyte marker genes were analyzed. Microscopy and flow cytometry analysis of LipidTOX Green stained lipid droplets revealed that loss of ITGA5 increased the numbers of lipid droplet containing cells to up to twofold. Similarly, depletion of ITGAV enhanced adipogenic differentiation as well, but the increase was more moderate. Vice versa, transgenic expression of ITGA5 or ITGAV reduced the numbers of well differentiated cells up to 60% (ITGAV-KD, p = 0.032) and 50% (ITGA5-KD, p = 0.030), respectively (Fig. 6B). In agreement, mRNA levels of adipocyte marker genes peroxisome proliferator-activated receptor γ (PPARG), lipid droplet associated protein 2 (PLIN2), adipocyte specific fatty acid binding protein (FABP4) and adiponectin (ADIPOQ) evidenced the phenotypic observations. PPARG showed significantly lower levels in ITGAV-OE cells than in control cells (p < 0.0001) and significantly higher levels in ITGA5-KD cells (p = 0.0117). Although not significant in all groups, levels of these marker genes in KD-cells clearly exceeded those in Ctr-shRNA transduced cells in all instances, corresponding well with the morphologic phenotype. Therefore, this might hint for enhanced activation of adipogenesis regulating signaling in these cells. In summary, the data demonstrate that loss of ITGAV and ITGA5 induced a similar but not necessarily the same differentiation phenotype, thus arguing for slightly distinct differentiation signaling mediated by these integrin subtypes.

Pharmacological inhibition of ITGAV/B3 and ITGAV/B5 by cilengitide mimics loss of ITGAV in ASC.
After assessing the importance of ITGAV and ITGA5 for the fate of mesenchymal progenitor cells isolated from human fat, we aimed to study the clinical feasibility of our observations. Therefore, we compared the effects of ITGAV-KD and the pharmacological inhibition of ITGAV/ITGB3 and ITGAV/ITGB5 complexes using ITGAV heterodimer inhibitor cilengitide, a molecule 36 originally developed for the treatment of glioblastoma. Cilengitide was added to proliferating or differentiating ASCs each time the medium was changed. The effect on proliferating cells was similar to that observed in ITGAV-KD cells, showing decreased cell proliferation (Fig. 7A). ITGAV and ITGB3 protein levels decreased after administration of 10 μ M cilengitide, corresponding to about 1/5 of a therapeutic dose 37 in a normal weight patient (Fig. 7B). At this concentration, TAZ and survivin levels were repressed, correlating with our observations in ITGAV-depleted cells. Interestingly, neither ITGA5 counter-upregulation nor p21 Cip1 accumulation was observed in cilengitide treated cells. In addition to the cell proliferation inhibiting effects, cilengitide promoted adipogenesis (Fig. 7C,D). Higher concentrations (20 μ M, 50 μ M) even enforced the phenotype, but also induced alterations in cell morphology and increased cell clustering indicating possible cell-toxic effects. At the molecular level, PPARG, PLIN1 and FABP4 were strongly induced in adipogenesis following cilengitide treatment, correlating well with the microscopic phenotype. In summary, pharmacological inhibition of ITGAV heterodimers phenotypically and molecularly corresponds to the loss of ITGAV and promotes adipocyte differentiation of ASCs.

Discussion
The aim of the study was to identify essential regulators of matrix-cell interactions in ASCs that control cell fate and discriminate them from mature adipocytes. By systematically analyzing the expression of a wide panel of alpha integrins, which specifically recognize and bind ECM substrates, we elaborated expression profiles for sorted CD34 + /CD90 + /CD31 − /CD45 − ASCs enriched from the SVF and primary adipocytes isolated from human subcutaneous tissue. Our data confirmed findings of previous in-vitro studies showing that integrins such as ITGA2, ITGA4 16 , ITGA5 and ITGA6 38 are differentially expressed in ASCs and adipocytes. In addition we also unraveled new adipogenesis-specific integrin regulations. For instance, ITGA7 was markedly upregulated in differentiated adipocytes. Similar to ITGA6, ITGA7 specifically binds to laminin, a substrate that is an important ECM component of adipocytes containing large fat vacuoles 16,34 . Since antibody-mediated inhibition of ITGA6 does not alter the differentiation phenotype 34 , ITGA7 might be responsible for laminin-dependent signaling in differentiating preadipocytes. Furthermore we found that RGD-motif and LDV-recognizing integrins were consistently repressed during adipogenesis, whereas laminin and collagen receptors as well as the leukocyte specific integrins were upregulated in adipocytes. From these data it is tempting to conclude that the in-vivo niche of ASCs is rich in matrix substrates containing RGD-motifs like fibronectin or vitronectin and is poor in laminin subtypes, as ASCs indeed express RGD-binding integrin subtypes at high levels but hardly express laminin recognizing integrins. However, the complexity of ECM architecture and the availability of distinct substrate isoforms that additionally mediate mechanical cues might dampen this simplified view.
Among RGD receptors, ITGA8 mRNA expression was lowest in ASCs, but as ITGA5 and ITGAV were highly expressed in ASCs and strongly repressed upon differentiation, we focused on these integrins for functional analyses. ITGA5 has previously been shown to be repressed upon differentiation in 3T3-L1 preadipocytes 34 , whereas this effect could not be observed in in-vitro differentiated hMSC from the bone marrow 16 . To our knowledge, ITGAV has not yet been linked to adipogenic differentiation, but was shown to be important in the neural invasion of malignant tumors 39 , tissue fibrosis 40 and neoangiogenesis 41,42 . Although mRNA levels of ITGAV were clearly lower in differentiated adipocytes, small amounts of ITGAV protein could be detected in patches between adipocytes in-vivo, suggesting an important role in maintaining tissue architecture and the three-dimensional ECM of adipose tissue.
Knockdown of ITGAV but not ITGA5 reduced of cell proliferation. It has recently been demonstrated that ITGAV is required for the maturation of focal adhesions, whereas ITGA5 controls early attachment and assembly of nascent adhesions. Our data might postulate a functional role of ITGAV in cell cycle 43 . ITGAV depletion was associated with a p53-independent induction of cell cycle regulator CDKN1A, the gene encoding p21 Cip1 protein, suggesting that ITGAV expression might be required for proliferation following cellular adhesion. In contrast, although knockdown of ITGAV impaired adhesion, it only marginally increased apoptosis due to anoikis, suggesting functional redundancy to ITGA5 or other integrins.
In contrast to our findings in ASCs, proliferation inhibiting effects of ITGA5 have been described previously for a variety of cell types such as keratinocytes 44 or muscle cells 45 . Although transgenic expression of ITGA5 increased proliferation in ASCs, the effect of RNAi mediated ITGA5 depletion on proliferation was not significant during the first six days after transduction. At later time-points proliferation levels approximated the level of proliferation in ITGAV-KD cells (Supplementary Fig. 2). These findings suggest that integrin signaling varies depending on the cell type. Under cell culture conditions used in this study the effect of transgenic ITGA5 expression might not be as significant as it might be for cells that are additionally cultured in the presence of exogenous RGD-containing substrates. On the other hand, this may be valid only for very early effects as adherent cells immediately begin to produce their own extracellular matrix upon attachment 46 .
To search for responsible intracellular signaling mechanisms for the phenotype we screened for activation of the ERK1/2 pathway as ERK1/2 signaling was previously shown to regulate actin polymerization 47 . In our experiments ERK phosphorylation at Thr202/Tyr204 did not depend on the differential expression of individual integrins, suggesting that ITGA5 and ITGAV signaling are not mediated by distinct activation of the MAPK-pathwaya finding that also holds true for the analysis of the AKT pathway. Contrarily, we found that the major Hippo pathway mediator TAZ but not YAP was significantly repressed in ITGAV-KD cells and TAZ repression went along with a reduction in survivin expression. Since Hippo signaling is involved in fundamental processes like proliferation arrest, contact inhibition, sensing of mechanical cues and cytoskeletal changes 14 , it is not surprising that this pathway is affected by ITGAV depletion. Although transgenic survivin did not rescue ITGAV-cells from reduced growth, the data strongly suggest -to our knowledge for the first time -a functional link between the Hippo pathway and integrin signaling. Whether the classical Hippo pathway cascade via LATs1/2 activation is involved or TAZ downregulation is due to another mechanism is the subject of current investigations.
The strongest effect of ITGA5/ITGAV knockdown was an increase in differentiation. Vice versa we observed adipogenesis inhibiting effects of transgenic ITGA5/ITGAV expression. In agreement with other authors 35,48 who described a negative impact of fibronectin signaling on adipogenesis before, our experiments show significantly showing ORO stained lipid droplets (red) and Hoechst stained nuclei (blue), scale bar: 50 μ m. (E) Cell lysates of day 14 differentiated ASC exposed to 10 μ M cilengitide or control treatment (DMSO) were subjected to immunoblotting for expression analysis of the adipocyte specific markers FABP4 and PLIN1. As loading controls of the depicted immunoblots cropped images of total protein stains acquired before blotting are shown.
Scientific RepoRts | 6:28889 | DOI: 10.1038/srep28889 enhanced differentiation of ASCs upon loss of ITGA5 or ITGAV, mimicking the absence of fibronectin signaling. Vice versa, overexpression of both integrins strongly impaired adipogenesis. Although depletion of ITGAV decreased cell proliferation and slightly increased cell death at early time points after infection of proliferating cells, ITGAV-KD cells subjected to adipogenesis effectively underwent differentiation without any signs of cell death. In fact, cells plated to undergo differentiation have to be seeded at high density to induce a proliferation stop by cell contact inhibition. From this observation, we conclude that loss of ITGAV primarily affects proliferating ASCs and has no effect on the viability of arrested cells.
Based on the finding that ITGAV knockdown is able to enhance adipogenesis as described above, we tested whether ITGAV/ITGB3 and ITGAV/ITGB5 heterodimer inhibitor cilengitide is able to mimic this phenotype. As expected, pharmacological inhibition of ITGAV mimics total loss of ITGAV expression in both cell phenotype and intracellular signaling in a dose-dependent manner. Intriguingly, although all other downstream targets investigated showed regulations similar to that found in ITGAV-depleted cells, p21 Cip1 levels were not increased following cilengitide treatment. Even longer exposure to the drug did not increase levels of this cell regulator, thus suggesting differential intracellular signaling. A possible explanation for this finding might be that shRNA-mediated ITGAV targeting affects the expression of all ITGAV heterodimers and is not restricted to ITGAV/ITGB3 and ITGAV/ITGB5, implying that upregulation of p21 Cip1 in ITGAV-KD cells might result from the lack of ITGAV/ITGB1 and ITGAV/ITGB6 heterodimers.
In summary, we identified ASC specific integrins and functional analyses demonstrated that integrins expressed in primary ASCs and adipocytes contribute to cellular processes by influencing intracellular signaling pathways. Our data also strongly suggest a negative impact of RDG motif signaling on adipogenic differentiation of ASCs via ITGA5 34 and -to our knowledge -for the first time via ITGAV.

Methods
Isolation and cell culture of human ASC. ASCs were isolated from subcutaneous abdominal fat tissue obtained from patients (mean age 36.2 ± 15.3 years; 88% female, 12% male) undergoing elective abdominoplasty. The study was approved by the Ethics Committee of the Medical University of Innsbruck (UN4368; EK 301/4.5), written informed consent was obtained from all donors and the methods were carried out in accordance with the approved guidelines. For ASC isolation, adipose tissue was washed with phosphate-buffered saline (PBS), minced into pieces and incubated with collagenase Type I (0.15% in PBS, Roche, Germany) for 1 h at 37 °C. After digestion, samples were centrifuged at 500 × g for 10 min. The upper phase containing primary in-vivo differentiated adipocytes 49 was transferred into a new tube, washed with PBS and immediately subjected to RNA-and protein-isolation. Pelleted stromal vascular fraction (SVF) was treated with erythrocyte lysis buffer (0.5 M NH 4 Cl, 0.5 M KH 2 PO 4 , 100 mM EDTA, Roth, Germany) for 10 min and spun at 500 × g for 10 min. The SVF pellet was resuspended in DMEM/F12 medium (Lonza, Austria), filtered through a 100 μ m and 40 μ m nylon mesh cell strainer (VWR, Austria) 49 , counted with a CASY TM cell counter (Schärfe System, Germany) and plated at a density of 10000 cells/cm 2 for culture in PM4 medium 50 containing DMEM/F12 (PAN Biotech, Germany) supplemented with 1 ng/ml rhFGF2, 10 ng/ml EGF (Immunotools, Germany), 50 ng/ml Insulin (Roche, Austria), 2.5% FCS and 1% Penicillin/Streptomycin (GE Healthcare, Austria). Twenty-four hours after plating non adherent cells were washed off and attached cells were assessed for multipotency. For FACS-sorting, SVF-cells were stained with CD31-FITC, CD45-FITC, CD34-PE (Biolegend, UK), CD90-APC (eBioscience, Austria) and 7AAD (BD Pharmingen, Germany) and sorted on a FACS Aria II cell sorter (Becton Dickinson, Germany). The CD34 + / CD90 + /CD31 − /CD45 − /7AAD − ASC-subfraction 51 was directly subjected to RNA-and protein-isolation.

Isolation of lipid droplet containing adipocytes by isopycnic centrifugation.
Lipid-droplet containing cells were separated from undifferentiated cells using a modified density centrifugation protocol with a 6% OptiPrep ® − 0.5% FCS/PBS gradient. The gradient was generated by mixing 3 ml of OptiPrep ® (Sigma Aldrich, Germany) with 11 ml of a 0.5% FCS/PBS solution. After centrifugation (30 min, 800 × g) at 20 °C, trypsinized, in-vitro differentiated cells were layered on top of the gradient and spun for 30 min at 800 × g. The upper phase (~0.5 ml) containing adipocytes with lipid-droplets and the cell pellet (undifferentiated cells without lipid droplets) were directly lysed in laemmli buffer and subjected to immunoblotting. All reagents were obtained from Sigma Aldrich, Germany. Germany). Positive controls (placenta, liver) and negative control slides were added to each experiment to validate the IHC staining reactions. Digital images were acquired using AxioVision microscope software linked to an AxioCamHRc camera and an AxioPlan2 microscope (Zeiss, Germany). Because the dead bodies were immediately anonymized no certificate of non-objection was needed.

Immunohistochemistry.
Proliferation assay. Cell proliferation was determined by resazurin-based PrestoBlue assay (Invitrogen, Germany) by measuring mitochondrial activity 52 . Cells were seeded in 96 well plates (5000 cells per well). PrestoBlue reagent was added at a concentration of 0.1 mg/ml in PM4. Cells were incubated for 15 min at 37 °C. Fluorescence was measured at 530 nm on a Tecan Genios spectrophotometer (Tecan, Germany) according to the manufacturer's instructions.
Adipogenic differentiation. Adipogenic differentiation was induced using adipogenic induction medium (AIM) based on DMEM/F12 medium supplemented with 33 μ M biotin, 1 μ M troglitazone, 250 μ M 3-isobutyl-1-methyl-xanthine (IBMX), 10 μ g/ml transferrin, 200pM T3, 100 nM dexamethasone and 500 nM insulin. After 72 h, AIM was exchanged for adipocyte differentiation medium (ADM; AIM without IBMX) and cells were cultured until day 14. To induce chondrogenic differentiation, high density seeded cells were cultured for three weeks in StemPro basal medium containing StemPro chondrogenesis supplements (ThermoScientific, Austria). Osteogenic differentiation was induced by exposing proliferating cells (5000 cells/cm 2 ) to osteogenic differentiation medium (DMEM, 10% FCS, 100 nM dexamethasone, 50 μ M ascorbic acid, 10 mM glycerol-2-phosphate) for three weeks. Cells were stained with Oil red O (ORO, adipogenesis), Alcian blue (chondrogenesis) or Alizarin red (osteogenesis). All reagents and dyes were obtained from Sigma Aldrich, Germany. For the assessment of intracellular lipid content, cells were fixed in 3% PFA for 10 min, washed with PBS and incubated with HCS LipidTOX ™ Green Neutral Lipid Stain (ThermoScientific, Austria) for 30 min. Digital images were acquired using an inverse Zeiss Axiovert 200 M microscope (Carl Zeiss, Austria) linked to an CoolSNAP FX CCD-camera and analyzed with ImageJ 1.46r Software (National Institutes of Health, USA) by counting differentiated cells per visual field.
RNA isolation and quantitative RT-PCR. RNA was isolated using TRIzol (MRC Inc. Cincinnati, OH, USA) and cDNA was synthesized using random hexamer primers and an iScript cDNA-synthesis kit (Biorad, Germany). The quantitative RT-PCR reactions were performed using the SsoAdvanced ™ Universal SYBR ® Green Supermix kit (Biorad, Germany) in a realplex-2 lightcycler (Eppendorf, Germany) using the following protocol: 95 °C for 3 min, 40 cycles at 95 °C (15 s), 60 °C (15 s), and 72 °C (10 s). Gene expression was estimated using the Mastercyler ep-realplex software version 2.2. CT values and normalized to the geometric mean of the reference genes glucuronidase beta (GUSB) and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein zeta (YWHAZ) 58 . Primers (see Supplementary Table 1) were designed using NCBI Primer-blast software, and specificity was tested by the assessing the melting curve.
To assess quality of data, descriptive statistics were performed. Data were analyzed with unpaired Student's t-test and the Mann-Whitney-U test and are presented as mean + /− SD or SEM. P-values < 0.05 were considered statistically significant. Statistical analyses were performed using Statview software (SAS Institute Inc., version 5.0.1) and Prism 5 (Graphpad Software Inc., version 5.0).