hMADS cells differentiate into adipocytes and osteoblasts, as shown in Figure 1A. Gene expression of representative markers of adipocyte (adiponectin) and osteoblast differentiation (alkaline phosphatase) are reported and in agreement with previously published data 14 15 16 17 . The terminal differentiation of hMADS cells into adipocytes and osteoblasts is also illustrated by the typical cellular morphology as shown in Figure 1B.

In order to analyze the secretome at early step of adipogenesis and osteogenesis of hMADS cells, secretion media at day 3 of differentiation into adipocytes and osteoblasts were compared with those of cells at day 0. Cells were washed 3 times with PBS and incubated for 6 h with serum-free culture medium, containing 0.1 μg/ml of transferrin without differentiation inducers. At the end of this incubation period, media were harvested, filtered and concentrated. Very low cell death or cell lysis occurred during the incubation period, as checked by microscopic analysis and lactate dehydrogenase activity measurements [see Additional file 1]. Four μg of each protein sample were loaded onto one-dimensional SDS-PAGE and stained with Bio-Safe Coomassie Stain. A representative gel is reported in Figure 1C. The excised bands from the SDS-PAGE were cut, trypsin digested and analyzed by mass spectrometry. The list of the identified proteins is reported in Table 1. Further details about the peptides identified for each protein are available in Table 1S [see Additional file 2].

Seventy three proteins were classified according to their function and localization, determined by the Meta search engine Bioinformatic Harvester. As reported in Figure 2A and Table 1, a large amount of identified proteins (52 %) corresponded to "classical" secreted proteins characterized by a signal peptide. Among them, we found several components of the extracellular matrix (fibronectin, collagens etc...), proteases, serine protease inhibitors (serpins such as PAI-1, pigment epithelium derived factor (PEDF) and Plasma protease C1 inhibitor) and other proteins involved in collagen catabolism, protein folding/turnover and growth regulation. In differentiating osteoblasts at day 3, periostin, described previously to be associated with osteoblast differentiation 23 was identified, whereas its presence was not detected in secretion media of differentiating adipocytes. A high percentage of proteins (37 %), devoid of signal peptide, was also detected into the secretion medium. These proteins were labelled as extracellular since they have been previously described in the extracellular compartment. Some of these proteins were metabolic enzymes such as α-enolase (ENO1) and pyruvate kinase, cytoskeletal components or heat shock/chaperone proteins. Similar findings from literature have been reported (see references in Table 1) strengthening actually their presence in the extracellular medium of various cellular models as well as in the murine/human bloodstream. The remaining proteins (11 %) did not exhibit a signal peptide (in: intracellular) and, to the best of our knowledge, have not been so far described extracellularly and did not represent artefacts as cell lysis was very low [see Additional file 1]. These proteins are only putatively secreted and indicated by an asterisk in Table 1. As illustrated by the Venn diagram in Figure 2B, the secretion of 21 proteins resulted to be similar under the three culture conditions, while 28 proteins were specifically secreted under adipogenic condition compared to 3 proteins under osteogenic condition.

Finally, hMADS cell secretome was clustered in 8 main groups, as indicated in Table 1 and Figure 3. These clusters include: proteases, protease inhibitors, extracellular matrix (ECM) components, anti-inflammatory/anti-oxidant proteins, metabolic enzymes, cytoskeletal components, heat shock/protein folding proteins and other proteins. These clusters illustrated the diversity of hMADS cell secretome, thus suggesting that some of these proteins could be involved in the regulation of the adipocyte/osteoblast balance.

Mass spectrometry identification data were validated for 8 candidates, selecting them from different clusters, i.e. 1 protease, 2 protease inhibitors, 2 ECM components, 1 anti-inflammatory/anti-oxidant protein, 1 metabolic enzyme and 1 heat shock protein. Western blots or zymograms were employed to supplement the identification by mass spectrometry and to determine the expression profile among three different sets of conditions (day 0 and day 3 for adipogenesis or osteogenesis). For all the candidates, the results of immunoblotting/zymogram and a semi-quantification of the bands are reported in Figure 4.

Matrix metallopeptidase 2 (MMP2), secreted protein, acidic, cysteine-rich (SPARC) and ENO1 were secreted by hMADS cells under all conditions, confirming the pattern of detection obtained by SDS-PAGE-mass spectrometry analysis (Fig. 4A–C, to be compared with Table 1). However, no significant difference was observed under the three conditions for these candidates. Zymographic analysis of the secretion media revealed a prominent band of lysis at 66 kDa corresponding to the reported active form of MMP2 (Fig. 4A). Addition of EDTA, an inhibitor of gelatinase activity, led to the absence of lytic bands (data not shown), thus confirming the specificity of MMP activity. Pigment epithelium derived factor (PEDF) was highly expressed in day 3 adipocytes compared to day 0 and day 3 osteoblasts (Fig. 4D) in agreement with mass spectrometry identification.

For 78 kDa glucose-regulated protein precursor (GRP78), transforming growth factor-beta-induced protein (BIGH3), pentraxin 3 (PTX3) and PAI-1, we confirmed the pattern of detection reported in Table 1, with significant differences of expression between day 0, day 3 adipocytes and day 3 osteoblasts, as shown by Western blot and histogram semi-quantifications (Fig. 4E–H).

Given the pattern of expression at the early step of adipogenesis and osteogenesis of hMADS cells, PAI-1 appeared as one of the most promising candidates for investigating further the adipocyte/osteoblast balance. PAI-1 mRNA levels were similar to those of protein levels (Fig. 5A). Since PAI-1 is a key regulator of the conversion of plasminogen to plasmin, negatively acting on two plasminogen activators, namely uPA (urinary plasminogen activator) and tPA (tissue type plasminogen activator) 24 , we decided to evaluate uPA and tPA levels in hMADS cell secretome. ELISA experiments were performed on secretion media from hMADS cells at day 0 and day 3 under adipogenic and osteogenic differentiating conditions after 6 h of incubation. As reported in Figure 5B, consistent with the immunoblotting data, PAI-1 levels strongly decreased in differentiating adipocytes as compared to day 0. By contrast, PAI-1 was secreted by day 3 osteoblasts, at albeit the same levels when compared to day 0. Concerning tPA measurement, we did not detect tPA in day 3 differentiating adipocytes, while a significant increase of tPA levels was observed in day 3 osteoblasts as compared to day 0 (Fig. 5C). By contrast, uPA remained undetectable in the secretion media of hMADS cells under the three conditions analyzed.

Enzymatic activities of uPA and tPA in 6 h serum-free conditioned media of hMADS cells at d0 and day 3 of differentiating adipocytes and osteoblasts were then evaluated by casein/plasminogen zymographic analysis. As reported in Figure 5D, tPA was mostly secreted by day 3 osteoblasts, as observed by the lytic band at 63 kDa, while uPA was undetectable (expected band at 48 kDa), thus supporting ELISA results. The higher band at around 110 kDa may correspond to complexes between plasminogen activators and PAI-1 as classically reported 25 . Incubation of the gel with amiloride, which selectively abrogates uPA-dependent enzyme activities, did not affect tPA-dependent lytic bands (data not shown). Furthermore, we demonstrated that hMADS cells expressed uPA receptor (UPAR) mRNA as reported by semi-quantitative RT-PCR analysis (Fig. 5E). As these data actually suggest the involvement of plasminogen activation in the regulation of the fine balance between adipogenesis and osteogenesis, this hypothesis was next examined in a pathophysiological situation where such an imbalance occurs.

Disequilibrium of the adipocyte/osteoblast balance is associated with the development of osteoporosis after ovariectomy. It is well established that in the bone of ovariectomized (ovx) mice, adipocytes develop at expense of bone formation 26 27 . In order to evaluate whether osteoporosis affects PAI-1 in bone, we analyzed the expression of PAI-1 mRNA in the humerus of ovx compared to sham mice. Ten weeks after surgery, as previously described 26 mice increased fat mass as shown by leptin levels (Fig. 6A). As shown in Figure 6B, the levels of PAI-1 mRNA decreased significantly in the humerus of ovx compared to sham mice, thus pointing out at PAI-1 modulation in the adipocyte/osteoblast balance as a potential marker of osteoporosis.

The secretome of murine and human adipocytes has been characterized 19 20 21 22 . However, these studies primarily focused on the late events of adipogenesis. Concerning osteoblast secretome, a single detailed characterization has been performed so far by Xiao and collaborators, who analyzed the extracellular matrix vesicle proteome of mineralizing osteoblasts and identified proteins at late steps of osteogenesis of MC3T3-E1 cells 28 .

Herein, proteomic analysis has been used for the first time to compare the secretome of hMADS cells differentiating into adipocytes and osteoblasts and to gain insights into the adipocyte/osteoblast balance. In order to delineate the initial secretory events of adipogenesis and osteogenesis, we performed a 6 h short term secretion which likely limits the number of secreted proteins and cell lysis. A total of 73 proteins was identified and classified into 8 clusters. The largest cluster corresponds to extracellular matrix proteins, thus confirming the secretome characteristic of our fraction. In addition, more than 50% of hMADS cell secretome is represented by proteins with signal peptide. Cytoskeletal proteins and heat shock and folding proteins are also largely represented with 14 and 12 proteins, respectively. Among the heat shock/folding proteins, two chaperones, e.g. HSP60 and HSP70 were found. These findings are reinforced by recent observations describing the presence of HSP60 and HSP70 in the bloodstream. HSP60 circulating levels in diabetic patients have been found associated to cardiovascular diseases 29 , while elevated HSP70 circulating levels are related to systemic inflammatory reaction in patients with heart failure following acute myocardial infarction 30 . In the metabolic enzyme cluster, 9 enzymes were also found, including glyceraldehyde dehydrogenase (GAPDH) and phosphoglycerate kinase (PGK). According to literature, extracellular GAPDH, also found in rat serum, inhibits spreading of COS-7 cells 31 , while PGK is secreted by tumor cells and its activity is regulated by hypoxia 32 . These enzymes might reflect increased metabolic requirements, but they could also exert independent and still undiscovered functions. Anti-oxidant proteins, like Thioredoxin reductase 1 (TXNRD1), represent another cluster of proteins secreted by hMADS cells. TXNRD1, that does not exhibit a signal peptide, has been described in rat as a circulating protein 33 .

Eleven percent of secreted proteins from hMADS cells is however represented by intracellular proteins, presence of which has not been reported up to now outside the cells. A modest contamination by uncontrolled release of intracellular proteins cannot be excluded despite the fact that no cell lysis was detected. Although cell death can take place in culture and is unavoidably accompanied by a subsequent release of cytosolic components, it should be pointed out that cell viability was not affected [see Additional file 1]. Among non-artifactual explanations, it is worth pointing out that this phenomenon could implicate exosome secretion, as most of the identified proteins were reported to be present in exosomes. Exosomes are membrane vesicles that originate from the cell membrane and are released extracellularly 34 . Their occurrence has been also described in the bloodstream and in different tissues in vivo 35 , suggesting their participation in physiological and/or pathological processes. To date, the exact function of exosomes in vivo is not completely understood. A large number of cell types, including reticulocytes, immune cells, platelets, epithelial cells, dendritic cells and others (see 34 and 36 for detailed reviews) appears to secrete exosomes either as a mechanism to extrude some proteins and/or to communicate between cells and/or to play immune functions. Typically, exosomes contain chaperones, cytoskeletal proteins (moesin, desmoplakin), elongation factors and several enzymes (ENO1), as described in literature through proteomic analysis of the secretion media from different sources as well as from analysis by flow cytometry and Western blot 34 37 . Therefore intracellular proteins recovered outside hMADS cells may be a mere reflection of events taking place within the cytosol at the time of exosome formation. Furthermore, 4 (Q14019, Q15293, P50990 and P30153) out of 8 candidates, known to be intracellular, have a score higher than a fixed threshold when analyzed using software predicting a non-classical protein secretion 38 , it is likely that these proteins are secreted by non-traditional mechanisms and did not arise from cell lysis. Clearly, these proteins can only be considered as putatively secreted and their actual secretion should be assessed.

Mass spectrometry data have been validated for 8 selected candidates. MMP2, SPARC and ENO1 displayed no significant differences in terms of quantification by immunoblotting or zymogram when comparing the three culture conditions.

MMP2, a zinc-dependent endopeptidase, involved in the degradation of the ECM, is actively secreted by adipocytes 39 and osteoblasts 40 . The presence of MMP2 in hMADS cells reinforces the concept that ECM remodelling represents a crucial event during the early steps of differentiation. As observed in this study, SPARC has been already described as a secreted molecule in both adipocytes 41 and osteoblasts 42 . SPARC is reported to be up-regulated in the adipose tissue of different models of murine obesity 41 and its circulating levels correlate with body mass index in humans 43 . A potential interest in further studying SPARC in the adipocyte/osteoblast balance relies upon the fact that bone marrow stromal cells from SPARC-null mice tend to form more adipocytes than cells from wild type mice 44 . ENO1 is mainly described for its role as an intracellular glycolytic enzyme, however its presence extracellularly has been already reported in the secretion medium of 3T3-L1 adipocytes 21 , as well as in other cellular models 45 . ENO1 is a multifunctional protein and a putative plasminogen receptor, since it has been reported as a cell surface protein 46 . Therefore, a potential role of ENO1 in the plasminogen cascade cannot be excluded.

Concerning the five other candidates (PEDF, GRP78, BIGH3, PTX3 and PAI-1), relevant differences of expression between day 0 and day 3 of adipogenesis and osteogenesis have been found. PEDF has been reported to be preferentially secreted by 3T3-L1 preadipocytes as compared to fully differentiated adipocytes 20 . However, Zvonic et al. 22 have described an increase in PEDF expression during adipogenesis of primary cultures of human adipose-derived stem cells, in agreement with our observations in hMADS cells when comparing hMADS cells at day 0 and after complete adipocyte differentiation (data not shown). PEDF could represent a valuable candidate for the adipocyte/osteoblast balance given its expression also in differentiating hMADS osteoblasts. A putative role for PEDF as mediator of angiogenesis and in matrix remodelling of the bone has been recently postulated 47 .

Several molecular chaperones have been found in the secretome of hMADS cells, including calumenin, calreticulin, protein disulfide isomerase 6, heat shock proteins and GRP78. As GRP78 controls intracellular protein transport 48 , its up-regulation in hMADS differentiating adipocytes and osteoblasts as compared to day 0 could reflect an increased demand in protein folding or a protection against endoplasmic reticulum stress.

BIGH3 is reported to be down-regulated during the differentiation of murine osteoblasts and to play negative effects on the early stages of osteogenesis 49 . In hMADS cells, BIGH3 protein levels increase in differentiating osteoblasts as compared to day 0, however a decrease during osteoblast terminal differentiation was observed (data not shown). Interestingly, BIGH3 is also released by differentiating adipocytes at day 3. So far, no report has described the secretion of BIGH3 by adipocytes or adipose tissue, whereas BIGH3 is involved in distinct cellular functions, such as cell growth, tumorigenesis, wound healing, apoptosis and migration 50 . Thus a role for BIGH3 in adipogenesis remains to be established.

PTX3 is a member of the pentraxin family that we described to be secreted by adipocytes and to respond to inflammatory stimuli such as TNFα 51 . In differentiating hMADS cells, PTX3 appears up-regulated during both adipogenesis and osteogenesis. The presence of PTX3 in osteoblasts has never been reported; we speculate that, under conditions of stress or inflammation, PTX3 could represent a potential target of the adipocyte/osteoblast balance that might deserve further studies.

Finally, PAI-1 represents another member of the serpin family which is secreted by hMADS cells. PAI-1 is a pleiotropic molecule, exerting functional roles in wound healing, atherosclerosis, tumor angiogenesis, rheumatoid arthritis, fibrosis etc., besides its main role as regulator of fibrinolysis 52 . In addition, PAI-1 plays a role in metabolic disorders such as obesity and insulin resistance, representing a marker of metabolic syndrome 53 . A role for PAI-1 in bone remodeling has been postulated, since mice lacking PAI-1 are protected from trabecular bone loss after ovariectomy, suggesting a site-specific role for PAI-1 in bone turnover 54 . Altogether, these data indicate that PAI-1 should represent a reliable candidate for the study of the adipocyte/osteoblast balance. In hMADS cell secretome, PAI-1 signal disappears at day 3 of adipogenesis while remaining present in differentiating osteoblasts. Despite some demonstration that PAI-1 is produced by adipocytes during adipogenesis 53 , it was recently shown that adipogenesis, as such, may not induce PAI-1; rather it enhances the potential of adipocytes to respond to PAI-1 inducers 55 . This may explain why PAI-1 synthesis dropped after differentiation of human adipocytes cultured in the absence of serum, known to contain several PAI-1 inducers 56 57 . Moreover overexpression of PAI-1 by adenovirus-mediated gene transfer inhibited adipocyte differentiation 58 . Conversely, preadipocytes from PAI-1^-/- mice showed greater differentiation than those issued from wild type mice 58 . This suggests that not only PAI-1 is dispensable for adipocyte differentiation but could even be deleterious as it has been recently suggested that bone marrow-derived PAI-1 had an effect on the development of obesity through its effect in inflammation 59 .

tPA is also preferentially secreted by hMADS differentiating osteoblasts, while no detectable levels of uPA have been detected, despite the presence of UPAR mRNA. Interestingly, PAI-1 and tPA serum levels have been reported as putative non-invasive diagnostic biomarkers of idiopathic osteonecrosis of the femoral head 60 . As for circulating PAI-1, a correlation with dysbaric osteonecrosis has been also described 61 . These data strongly suggest that the plasminogen/plasmin system could represent a potential target for further investigation of the adipocyte/osteoblast balance. Even if preliminary, our data concerning a decrease of PAI-1 mRNA levels in bone of ovx as compared to sham mice clearly indicate that PAI-1 expression is modulated under conditions altering the adipocyte/osteoblast balance, such as osteoporosis.

Cell culture media were purchased from Cambrex and FCS from Dutscher S.A. (Brumath, France). EGF was a product of Euromedex (Souffelweyersheim, France). Antibodies directed against ENO1, PEDF, BIGH3 and SPARC were supplied from Santa Cruz Biotechnology (Santa Cruz, CA), Chemicon International (Temecula, CA), R&D Systems (Minneapolis, MN) and Haematologic Technologies Inc. (Essex Junction, VT), respectively. PAI-1 antibody was a kind gift from Prof. PJ Declerck, (Leuven, Belgium). Rat anti-human PTX-3 antibody was kindly provided by Prof. A Mantovani 62 . Reverse transcriptase and trypsin were from Promega (Charbonnières-les-Bains, France). PVDF membranes were from Amersham Biosciences (Orsay, France). Complete protease inhibitor cocktail was from Roche Diagnostics (Meylan, France). All the other products were from Sigma-Aldrich (Saint Quentin Fallavier, France).

The establishment and characterization of the multipotency and self-renewal of hMADS cells have already been described 14 15 16 . In the experiments reported herein, hMADS-2 cells, established from the pubic region fat pad of a 5-year old male donor, were used at passages between 16 and 35 corresponding to 35 to 100 population doublings. Cells were seeded at a density of 5000 cells/cm² in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS, 2.5 ng/ml hFGF2, 60 μg/ml penicillin and 50 μg/ml streptomycin. The medium was changed every other day and hFGF2 was removed when cells reached confluence. At day 2 post-confluence (designated as day 0) adipogenic or osteogenic differentiation was induced, as described previously 14 17 . Briefly, for adipogenic differentiation hMADS cells were induced to differentiate in the presence of DMEM/Ham's F12 media supplemented with 0.85 μM insulin, 0.2 nM triiodothyronine, 10 μg/ml transferrin, 1 μM dexamethasone (DEX) and 500 μM isobutyl-methylxanthine (IBMX). Three days later, the medium was changed (DEX and IBMX omitted) and 100 nM Rosiglitazone was added up to day 9. For osteogenic differentiation, cells were induced to differentiate in α-MEM containing 1% FCS supplemented with 10 nM 1,25-dihydroxyvitamin D3, 100 nM DEX, 10 μM L-ascorbic acid phosphate, 10 mM β-glycerophosphate and 10 ng/ml EGF. For both differentiation protocols, the media were then changed every other day and cells were used at the indicated days.

Lactate dehydrogenase (LDH) activity has been performed as described previously 63 . Briefly, the activity has been measured in the culture media and compared to that obtained from cells treated with a detergent in order to release the whole intracellular LDH activity.

Eight weeks old C57Bl/6J mice were subjected either to bilateral ovariectomies (ovx) from the dorsal approach or to sham surgery during which the ovaries were exteriorized but replaced intact by the operator (Charles River, L'Arbresle, France). Mice were housed in local animal facility for a period of 10 weeks following surgery to await the development of osteoporosis. Animals were sacrificed with carbon dioxide according to the guidelines of the local animal care and experimentation committee. RNA from humerus was extracted using Totally RNA™ kit according to manufacturer's instructions (Ambion, Courtaboeuf, France).

Conditioned media for each condition (24 ml, corresponding to 8 dishes of 100 mm diameter) were collected on ice, centrifuged and filtered to remove cell debris, if any, and supplemented with complete protease inhibitor cocktail. Twelve ml of the samples were concentrated by ultra-filtration (Millipore, Centricon, 5 kDa cut-off). Protein concentrations were determined using Bio-Rad Protein Assay reagent (Bio-Rad, Marnes-la-Coquette, France) and were: 0.12 ± 0.04, 0.15 ± 0.06; 0.12 ± 0.02 μg/μl for day 0, day 3 adipocyte and day 3 osteoblast respectively.

One-dimensional polyacrylamide gel electrophoresis (SDS-PAGE) was performed using Bio-Rad mini-Protean system. Gels were stained with Bio-Safe Coomassie Stain (Bio-Rad) for 2 h at room temperature and rinsed with water, according to manufacturer's instructions.

After Coomassie staining, the whole lanes were manually excised into 10 bands which were processed for tryptic digestion according to a standard protocol. Briefly, after a washing step with 50% ethanol, proteins were reduced and alkylated with 10 mM DTT and 50 mM iodoacetamide, respectively. The spots were washed twice with 50% ethanol, shrunken with acetonitrile (ACN) and then subsequently digested with 100 ng of trypsin in 50 mM ammonium bicarbonate, 15% ACN. The peptides were extracted with 0.5% TFA, 50% ACN then 100% ACN. LC-MS/MS mass spectrometry was performed with a Surveyor system coupled with an LCQ DECA XP ion trap mass spectrometer (ThermoQuest, San Jose, CA). Peptides were separated on a C18 column (300 μm × 10 cm, Hypersil, ThermoQuest) at 5 μl/min with 1 to 40 % ACN gradient in 0.1 % formic acid. A 3 kV voltage was applied on the micro-ESI needle and the automatic acquisition was set as previously described 64 .

Fragmentation spectra acquired by Xcalibur 1.3 (ThermoElectron Corp.) were searched with Sequest (Bio-Works version 3.3, Thermo Electron Corp.) against the Swiss-Prot protein database (235 673 entries, uniprot_sprot.fasta file downloaded on October 2006 from 65 ). Sequest parameters were: i) peptide mass tolerance of 1.7 Da, ii) parent ion masses treated as monoisotopic, iii) fragmentation ion masses treated as monoisotopic with 1.0 Da mass tolerance, iv) trypsin (KR, strictly enzymatic specificity at the both ends) and two miss cleavages, v) a 57.0 Da static modification on cysteines accounted for alkylation and a 16.0-Da variable modification on methionine accounted for oxidation.

Sequest results were filtered with the following requirements: i) at least 2 peptides were required for protein identification, ii) peptides identified as first candidate, iii) Xcorr > 1.7, 2.2 and 3.3 for mono-, di- and tri-charged peptides, iv) P value up to 0.001 for peptide. The filters' relevance was checked with different tryptic digests as control (HSP90, albumin, lactoglobulin) and using the different SDS-PAGE molecular weight markers; filters automatically removed all the proteins from Bioworks results sheet with the exception of the exact identification. To address the database redundancy issue, only identifications with highest scoring and corresponding to human protein references were selected. Protein classification has been performed using annotation from Bioinformatic Harvester 66 and Swiss-Prot Protein database 67 .

Total RNA was isolated with TRI-Reagent kit (Euromedex, France) according to the manufacturer's instructions. Reverse transcriptase reactions and semi-quantitative PCR assays were performed as already described 14 51 . PCR products (25 cycles) were analyzed by 1.5% agarose gel electrophoresis visualized by ethidium-bromide staining.

qRT-PCR assays were run on an ABI Prism 7000 real-time PCR machine (PerkinElmer Life Sciences). Reactions were performed according to manufacturer's instructions using SYBR green master mix (Eurogentec, France). The expression of selected genes was normalized to the expression of the TATA-binding protein (TBP) encoding gene. Gene expression was quantified using the comparative-delta Ct method.

Primer sequences and annealing temperatures are reported in Table 2S [see Additional file 3].

Equal amounts of proteins (2 μg) were denatured, reduced and separated on 12% polyacrylamide-SDS gels. Proteins were transferred in 25 mM Tris, 192 mM glycine and 20% ethanol onto PVDF membrane. Blots were blocked for 30 min with TBS (10 mM Tris-HCl pH 7.5 and 150 mM NaCl) plus 0.1% Tween 20 (TTBS buffer) containing 5% dried milk powder (blocking buffer) and then hybridized in the same buffer with specific primary antibodies (0.2 μg/ml) at 4°C. After overnight incubation, the blots were washed twice in TTBS and incubated 1 hr at room temperature in blocking buffer using the appropriate secondary antibodies (Sigma). After 5 washes in TTBS, immunoreactive proteins were visualized using the ECL chemiluminescence's detection kit (Amersham) according to the manufacturer's instructions.

Proteins with gelatinolytic activity were identified by electrophoresis in 10% SDS-PAGE containing 1 mg/ml gelatin. Two μg of secretion media were directly loaded on gels under nonreducing conditions. After electrophoresis, proteins were renatured by washing twice with 2.5% Triton X-100 in 50 mM Tris-HCl pH 7.4. The gels were then incubated at 37°C for 16 h in an activation buffer (50 mM Tris-HCl pH 7.4, 5 mM CaCl₂ and 0.2 M NaCl) and stained with Coomassie Brilliant Blue R-250 (in 40% methanol and 10% acetic acid). Destaining was performed in the same buffer in the absence of the dye. Migration of proteins was compared with that of a prestained molecular weight marker. EDTA (15 mM) was added to the incubation buffer in parallel gels, in order to inhibit cation-dependent enzymatic activities.

Two μg of conditioned media were analyzed on 10% polyacrylamide gels containing 1 mg/ml alpha-casein and 10 μg/ml plasminogen (Sigma), under nonreducing conditions. After the electrophoretic run, gels were rinsed once in 100 mM glycine buffer with 2.5% Triton-X 100 for 45 min and then incubated overnight at 37°C in 100 mM glycine buffer, 15 mM EDTA, pH 8.0. Identical gels were incubated in the above buffers containing 500 μM amiloride, in order to discriminate uPA from tPA bands. Gels were then stained with a 0.25% Coomassie Blue/10% acetic acid/40% methanol solution for 1 hr, followed by destaining in a 40% methanol/10% acetic acid mixture. Caseinolytic activity resulting from plasminogen activation was visualized by white lytic bands on a blue background after conversion of plasminogen to plasmin by uPA and tPA. Molecular weights were calculated from the position of prestained markers that were subjected to electrophoresis in parallel lanes. No lytic bands were observed in plasminogen-free gels.

Supernatant PAI-1 antigen was assayed using ELISA's specific for human PAI-1, as previously described 68 . PAI-1 ELISA detects latent and active forms of human PAI-1 and PAI-1 complexes. The assay is insensitive to PAI-2. The amount of uPA and tPA antigen was measured with the Immunobind uPA ELISA kit (American Diagnostic, Greenwich, CT) and the commercially available kit Asserachrom tPA (Diagnostica Stago, France) according to the instructions of the manufacturer. Inactive and active forms of plasminogen activators are all recognized by the ELISA kit. Plasma concentration of leptin was measured by ELISA using EIA kit (SPI-Bio, France).

Data are expressed as mean values ± standard deviation (SD). qRT-PCR and ELISA data and quantifications of immunoblot/zymogram experiments were analyzed using Student's t-test. Statistical significance was assumed at p level <0.05.