DISCLOSURE STATEMENT: the authors have nothing to disclose Clinical Trials ID NCT

Transcription

1 J Clin Endocrin Metab. First published ahead of print February 17, 2009 as doi: /jc Role of adipocyte serum amyloid A in paracrine cross-talk between adipocytes and macrophages in cholesterol efflux Christine Poitou 1,2,3,, Adeline Divoux 1,2,, Aurélie Faty 4, Joan Tordjman 1,2, Danielle Hugol 5, Abdelhaim Aissat 6, Mayoura Keophiphath 1,2, Corneliu Henegar 1,2,3, Stéphane Commans 4 and Karine Clément 1,2,3. 1 Inserm, U872 team7, Nutriomique, Cordelier Research Center, Paris, 75006, France; 2 University Pierre et Marie Curie-Paris-6, UMRS, Paris, 75006, France ; 3 Assistance Publique-Hôpitaux de Paris (AP-HP), Endocrinology and Nutrition Department, Pitié Salpêtrière hospital, Paris, 75013, France, France; 4 GlaxoSmithKline Laboratory, Research Centre metabolic Pathways CEDD GlaxoSmithKline, Les Ulis, France; 5 Assistance Publique- Hôpitaux de Paris (AP-HP), Cytopathology department, Hotel Dieu Hospital, Paris, F-75004; 6 Assistance Publique-Hôpitaux de Paris (AP-HP),, Surgery department, Hotel Dieu Hospital Paris, F C.P and A.D contributed equally to this work. Short title: Serum Amyloid A promotes cholesterol efflux in human adipocytes Key terms: SAA, obesity, adipose tissue, macrophages, cholesterol, ABCA1 Précis: In human, SAA is related to adipocyte size, macrophage infiltration and cholesterol efflux suggesting a role in the cross-talk between hypertrophied adipocyte and macrophages. Corresponding author: Christine Poitou M.D, PhD Address: INSERM Nutriomique U872 (Eq 7), University Pierre et Marie Curie-Paris6, Cordelier Research Center, 15 rue de l'école de médecine, Paris, France Phone: FAX: christine.poitou-bernert@psl.aphp.fr DISCLOSURE STATEMENT: the authors have nothing to disclose Clinical Trials ID NCT Number of words (text): 3725, number of words (abstract): 237 Total number of tables and figures: 7, supplementary data: one table Copyright (C) 2009 by The Endocrine Society 1

2 ABBREVIATIONS: ABCA1, ATP-binding cassette sub-family A AcM, activated monocytes A-SAA, acute phase serum amyloid A AT, adipose tissue ATM, adipose tissue macrophage DEX, dexamethasone FDR, false discovery rate GO, gene ontology hmads, human multipotent adipose derived stem hscrp, high sensitivity C-reactive protein IL, interleukin KEGG, Kyoto encyclopedia of genes and genomes LXR, liver X receptor MCP-1, monocyte chemotactic protein 1 owat, omental white adipose tissue PBMC, peripheral blood mononuclear cell scwat, subcutaneous white adipose tissue stnfr-ii, soluble tumor necrosis factor receptor II SVF, stroma vascular fraction TNF, tumor necrosis factor 2

3 ABSTRACT Context: Acute phase serum amyloid A (A-SAA) is secreted by hepatocytes in response to injury and is regulated by proinflammatory cytokines. In obese humans, adipocytes are also a major contributor to circulating A-SAA levels. Objective: We aimed to investigate the role and regulation of A-SAA in human adipose tissue (AT). Design: An approach combining microarrays and the FunNet bioinformatic tool was applied to human AT fractions (i.e. Adipocytes vs. Stroma Vascular Fraction, SVF) to hypothesize genes and functions related to A-SAA. Experiments with human AT from 37 obese subjects and human multipotent adipose derived stem (hmads) cells were used to confirm the microarray-driven hypotheses. Results: Microarray analysis highlighted the relationship between A-SAA and SVF inflammatory genes and between A-SAA and adipocyte-expressed ATP-binding cassette (ABC) transporters. We confirmed that SAA protein is expressed in subcutaneous AT of obese subjects (N=37, BMI = 49.3±1.5 kg/m 2 ) and showed that SAA protein expression correlated with adipocyte size (R= 0.44, p = ), macrophage infiltration (R=0.61, p=10-4 ) and ABCA1 (ABC sub family A1) protein expression (R=0.43, p= ). Interleukin-1β, Tumor Necrosis Factor α and human AT macrophages conditioned medium significantly induced A-SAA secretion (from 2.6 to 7.6 fold) in hmads cells. Recombinant SAA induced cholesterol ABCA1-dependent efflux from hmads adipocytes by 4.3- fold in a dose-dependent manner. Conclusion: This work provides original insight suggesting that A-SAA is a player in the dialogue between hypertrophied adipocytes and macrophages through its regulation of adipocyte cholesterol efflux. 3

4 INTRODUCTION Serum amyloid A proteins (SAA) comprise a family of small and differentially expressed proteins. Acute-phase SAA (A-SAA) consists of two closely related isoforms, SAA1 and SAA2, inducible in response to proinflammatory stimuli (1). Additional members of the SAA family are SAA3 and SAA4. SAA3 is a pseudogene in human and SAA4 is constitutively expressed. During acute inflammation, hepatic production of A-SAA increases times. As an acute phase protein, A-SAA classically contributes to the inflammatory process defense to aggression. In vitro experiments have shown that A-SAA induces chemotaxis of immunologic cells (like T cells, monocytes, neutrophils, mastocytes) and cytokine production such as interleukins 1 β (IL-1β) and 8 (IL-8) (2). The proposed role of A-SAA is that of an apolipoprotein, which could alter the structure and function of inflammatory HDL particles to help removing excess cholesterol released from damaged tissues (3). In the context of obesity-related low-grade inflammation, the level of circulating A-SAA remains moderately high (>10mg/L), about 6- fold higher than in patients with normal weight (4, 5) or than in obese subjects experiencing weight loss (6, 7). Observational studies showed that a chronically elevated level of circulating SAA is a marker of peripheral and coronary arterial disease. Experiments suggest that A-SAA, involved in cholesterol uptake and efflux, could be involved in the development of atherosclerosis (8). Comparing the level of gene expression of A- SAA in human tissues has shown that adipose tissue (AT) is a major site of A-SAA production in obesity (6, 7, 9), in contrast with mice where expression is predominantly hepatic and the isoform expressed in AT is SAA3. In humans, A-SAA is secreted by AT at higher levels than leptin. The production of A-SAA by enlarged AT might participate to the growing pool of circulating SAA in obesity (5, 9). Within AT, the A-SAA gene is predominantly expressed and produced by mature adipocytes and not by cells of the stroma vascular fraction (SVF) (9). Neither the relationship between A-SAA protein and AT parameters (e.g. adipocyte biology, inflammation, etc.), nor the regulation and role of human A-SAA has been extensively examined in vivo. In human subcutaneous tissue, Yang et al have shown that recombinant SAA promotes an increased basal lipolysis in adipose cells and stimulates the production of inflammatory cytokines such as interleukins 6 and 8 (IL-6, IL-8) and monocyte chemoattractant protein-1 (MCP-1) from SVF cells (9). To gain further insight into the potential role and regulation of A-SAA in human AT, we have explored both the relationship between A- SAA and macrophages, and the role of A-SAA in cholesterol efflux from adipocytes. RESEARCH DESIGN AND METHODS Subjects Adipose tissue samples were obtained from 37 morbidly obese subjects (BMI 49.3±1.5 kg/m²) involved in a gastric surgery program, prospectively recruited in the Department of Nutrition, Center of Reference for Medical and Surgical Care of Obesity, Pitié-Salpêtrière Hospital (Paris, France). Patients were excluded if they had evidence of acute or chronic inflammation or infectious diseases. None of the subjects were taking medication for type 2 diabetes. Subjects were weight stable for 3 months prior to surgery. Venous blood samples were collected in the fasting state for determination of routine biochemical parameters and evaluation of adiponectin, IL-6 and high sensitivity C-reactive protein. Clinical and biological parameters are shown in Table 1. Paired subcutaneous white adipose tissue (scwat) and omental adipose tissue (owat) samples were obtained from the same individuals during the surgery. A portion of each sample was fixed in 4% paraformaldehyde and the rest was immediately frozen in liquid nitrogen and stored at -80 C for total RNA preparation. The clinical investigation promoted by Assistance/Publique Hôpitaux de Paris was approved by the Ethics Committees (Paris). Informed consent was obtained from all severely obese subjects. Gene profiling in SVF and isolated adipocytes from scwat 4

5 To gain insights about the putative role of A- SAA in human AT, we applied the recently published bioinformatics FunNet tool ( (10, 11) to previously generated microarray data (12). Differences in gene expression between isolated adipocytes and SVF cells were examined in scwat in another group of overweight women (n = 9, BMI 27.9±6.8 kg/m 2 ). The procedure was described in (12). We focused our present analysis on the A-SAA transcripts (SAA1 and SAA2) which are part of genes significantly over-expressed in adipocytes (FDR < 5%). Noteworthy, sequence similarity for SAA1 and SAA2 microarray probes was greater than 97%; therefore, we could not distinguish between these two isoforms. The Spearman s rank correlation coefficient was used to identify transcripts whose expression profiles were similar to that of the A-SAA (i.e. statistically coexpressed genes), among those displaying a significant level of differential expression in mature adipocytes or in the SVF. A threshold of gene co-expression significance, corresponding to an absolute value of a Spearman s R 0.85, was determined by maximizing a scale-free topology criterion as previously described in (13). Immunomorphological analysis of adipose tissue in obese subjects Samples fixed in paraformaldehyde were dehydrated, paraffin embedded and sectioned (thin sections of 5μm). Sections were stained with hematoxylin and eosin. Immunohistochemical detection of HAM56 (1:100; Dako Cytomation, Trappes, France), SAA (1:50; Dako Cytomation, Trappes, France), and ABCA1 (1:50; abcam, Paris, France) were performed with the avidin-biotin peroxidase method (14). Prior to staining for SAA and ABCA1, nonspecific sites were blocked using an anti-rabbit IgG (1:30, 1h at room temperature). HAM56 is a monoclonal antibody selected in alveolar macrophages pool. According to manufacturer s recommendations, HAM56 labels human macrophages and endothelial cells but not lymphocytes or granulocytes. Adipocytes staining positive for SAA or ABCA1, and cells positive for HAM56 were counted in 10 randomly chosen areas in each processed slide at 40 magnification by two independent researchers. The mean of positive cells was expressed as a percentage of the total number of adipocytes. Adipocyte cell size was determined using Perfect Images 7.4 software. The diameter of adipocytes was measured in 2 randomly chosen areas (representing approximately 400 to 600 adipocytes) in each processed slide at 10 magnification. Adipocyte sizes reported are the mean of the two fields. For the double staining of SAA/HAM56 or SAA/ABCA1 on the same slide, the protocol described above was applied for each antibody. SAA staining was revealed using diaminobenzidine and blocked by the reagent Linblock (AbCys, Paris). Then HAM56 or ABCA1 staining were revealed using Histoprime (AbCys, Paris). Quantitative Real Time-PCR (qrt-pcr) We assessed gene expression changes using primers, TAQMAN probes and 18S ribosomal RNA for normalization, obtained from Applied Biosystems (Foster City, CA, USA), as described (15). Because of the very high degree of homology between inducible A-SAA isoforms, designed primers detected both SAA1 and SAA2 mrna. Primer sequences used for A-SAA and ABCA1 detection are shown in Table 2. Preparation of human blood monocytederived macrophages and conditioned medium (AcM) Blood from female patients was processed for plasma blood mononuclear cells (PBMC) isolation as previously described in (16). Macrophages were incubated in RPMI-1% FBS for 24 h with 100 ng/ml lipopolysaccharides (LPS from E. Coli 0127:B8, Sigma St Louis, Mi, USA) prior collecting the medium (AcM). Control medium was RPMI-1% FBS kept at 37 C for 24h in the absence of macrophages. Preparation of Adipose Tissue Macrophages (ATM) and conditioned medium Isolation of ATM from human SVF was performed as described (17). Isolated SVF cells were obtained from scwat biopsies, as described in (16). The bead-coupled CD14+ cells were maintained for 24h in RPMI-1% FBS to obtain ATM conditioned medium, which was kept at -80 C before use. 5

6 Adipocyte differentiation of multipotent adipose-derived stem cells (hmads) Adipocyte differentiation of multipotent adipose-derived stem cells isolated from human AT (hmads) was performed as previously described (18) with the following modifications: at post-confluence (designated day 0), cells were induced into adipogenic differentiation through DMEM/Ham F12 media supplemented with 0.86μM insulin, 0.2nM triiodothyronine (T3), 10μg/mL transferrin, 1μM dexamethasone (DEX), 100μM isobutyl-methylxanthine, and 200nM GW7845, a PPARγ agonist. Three days later, the medium was changed to DMEM/Ham F12 media supplemented with 0.86μM insulin, 0.2nM T3 and 10μg/mL transferrin only. This media was then changed every two day and cells were used from day 10 post-induction. Recombinant human IL-1β, TNFα and IL-6 were purchased from Peprotech (Rocky Hill, NJ, United States). DEX was from Sigma (St- Louis, MI, United States). hmads adipocytes were treated for 24h with or without IL-1β (20ng/mL), TNFα (20ng/mL), IL-6 (20ng/mL) and DEX (100nM) or a combination of these hormones. hmads cells were also treated with either control medium, AcM or ATM medium. Following 24h, culture medium was recovered and A-SAA was assayed. A-SAA quantification in culture media Human A-SAA (BioSource, Camarillo, California, United States) was measured with ELISA kits according to manufacturer s instructions. Cholesterol efflux Human apolipoprotein A-I (apoa1), LDL and HDL cholesterol were purchased from Calbiochem. Recombinant human SAA (corresponding to human SAA1α) was from PeproTech (Rocky Hill, NJ, United States). The cholesterol efflux assay was performed as previously described (19). Cholesterol loading of hmads adipocytes was performed with [ 3 H] cholesterol labelled LDL ( 3 μci/ml; 50μg/mL) for 24 hours. Cholesterol efflux was induced to various acceptors after a 24h incubation of the cells with or without the liver X receptor (LXR) agonists GW3965 (1μM) and a 2h incubation with or without Probucol (10μM) (Sigma, St Louis, Mi, USA). After 6h, media were collected and radioactivity measured by liquid scintillation counting. Cholesterol efflux was calculated as the percent of total [ 3 H] cholesterol released into the medium after subtraction of values obtained in the absence of cholesterol acceptor. Statistical analyses Data are expressed as means ± SEM. General statistical analysis was performed with JMP statistics software (SAS Institute, Cary, NC). Correlations were examined by the nonparametric Spearman s rank correlation test. A p < 0.05 was considered statistically significant. A Wilcoxon nonparametric test was used to compare mean values for data concerning A-SAA secretion and cholesterol efflux in hmads. RESULTS Relationship between SAA protein and adipocyte size in human AT depots We previously showed that A-SAA gene expression was higher in scwat than in owat. We examined if this is also the case for SAA protein expression using paired fat depots from 37 obese subjects. In agreement with gene expression findings, immunohistochemistry revealed that the percentage of SAA positive adipocytes was higher in scwat than owat (10.9±1.3% vs. 7.0±1.1%, p = ) (Figure 1A). Because hypertrophy may effect AT SAA expression, we assessed the relationship between adipose cell diameter and percentage of adipocytes expressing SAA in both owat and scwat. For the 37 obese subjects, adipocyte size was normally distributed in both depots. As expected, the mean adipocyte cell diameter in scwat was significantly higher than in owat (75.2 ± 1.2 μm vs 68.7 ± 1.6 μm) (Figure 1B). We observed a significant positive correlation between adipocyte cell diameter and the percentage of adipocytes expressing SAA proteins in both WAT depots (R = 0.44, p = in scwat; R = 0.49, p = in owat) (Figure 1C). More specifically, in scwat the percentage of adipocytes expressing SAA was estimated at 2.4% for adipocytes whose size was less than 60 μm and 9.4% for hypertrophic adipocytes (> 90 μm) (Figure 1D). Results did not change when separating our cohort in females (n=27) and males (n=10). The percentage of SAA positive fat cells was not related to clinical metabolic data such as 6

7 fasting glycaemia, insulinemia, HDLcholesterol and total cholesterol (data not shown). Functional profiles of transcripts coexpressed with A-SAA expression in scwat To gain insight into the role of A-SAA within scwat, we examined the relationships between A-SAA gene expression and groups of genes and functions characterizing the isolated adipocyte and SVF. Using previously published microarray gene expression data in which 9 independent female subjects were studied (10), we found that A-SAA was present among the 6,587 genes over-expressed in isolated adipocytes compared to the SVF (13.4-fold). 127 genes belonging to the adipocyte fraction correlated with the A-SAA expression profile (R>0.85), and 92 genes over-expressed in the SVF correlated with A- SAA gene expression. In this latter gene list, 26 among the 27 genes were independently described to characterize the transcriptomic signature of macrophages by Svennson et al. (20) (supplementary data). Figure 2A shows the most significant KEGG themes among annotated genes, ranked by their functional coverage of the transcriptional domain. The most significant functions characterizing the genes over-expressed within the SVF and highly correlated with A-SAA were related to inflammatory and immune processes (cytokine-cytokine receptor interaction, leukocyte transendothelial migration, and antigen processing and presentation). The enriched functions characterizing genes of the adipocyte fraction that correlated with A-SAA expression included ABC transporters, olfactory transduction, ether lipid metabolism and retinol metabolism (Figure 2A), where ABC transporter function was the most enriched functional theme. Furthermore, ABC transporter function was also found to occupy a central position in the functional network generated using genes that co-express with A- SAA in adipocyte fraction (Figure 2B). Using this functional approach in AT fractions, we generated hypotheses regarding the relationships between A-SAA and inflammatory pathways, especially with monocytes/macrophages functions, and the putative association between A-SAA and cholesterol transport. A-SAA and inflammatory factors in human adipose cells Because of the aforementioned statistical association between A-SAA gene expression and inflammatory related-functions in the SVF of AT, we first examined the links between inflammatory characteristics of WAT and SAA in tissues from our obese cohort (n=37). In agreement with our previous observation, we observed in this new group of subjects that the number of infiltrated macrophages (i.e. percentage of HAM56 positive cells) was significantly higher in owat than in scwat (24.8% ± 2.1 vs 15.1 ± 1.4, p < 0.05) (Figure 3A). In addition to the close proximity between adipocytes expressing SAA and adipocytes with macrophages crown-like structures observed by immunohistochemistry (Figure 3B), we observed that the percentage of SAA positive cells significantly correlated with the percentage of macrophages in both obese scwat (R = 0.61, p = 10-4 ) and owat (R = 0.46, p = ) (Figure 3C). To explore the possibility of a biological link between human adipocytes expressing SAA and macrophages, we evaluated whether the production of SAA in WAT was subject to regulation by proinflammatory mediators known to be produced by macrophages. We used hmads adipocytes to test the induction of A-SAA secretion by pro-inflammatory factors and their different combinations. Treating hmads cells with IL- 1β, TNFα, or a combination of the two significantly induced A-SAA secretion by 4.6- fold, 2.6-fold, and 7.6-fold, respectively (Figure 3D). Contrary to the in vitro published studies performed in hepatic cells, IL-6 did not have synergistic effect with IL1-β and/or TNFα on A-SAA secretion (data not shown). No significant effect was observed with undifferentiated hmads (data not shown). Next, we found that ATM medium and AcM significantly induced hmads A-SAA secretion 3.6 to 5.8-fold (Figure 3D). Altogether these observations suggest that the inflammatory environment provoked by macrophage secretions markedly increased the production of A-SAA by differentiated human adipocytes. Role of SAA1 in cholesterol efflux A-SAA is known to increase cholesterol efflux, both dependently and independently of ABCA1, in different cell types such as macrophages and hepatocytes; however no 7

8 data exists in adipocytes. Firstly, to confirm the potential association between A-SAA and ABCA1 identified by the microarray data, we explored A-SAA and ABCA1 gene and protein expression. In a subject s subgroup, we observed by qrtpcr that ABCA1 mrna expression was significantly associated with A-SAA mrna expression (R = 0.7, p<3.10-4, n=16), thus confirming the microarray observations. Using immunohistochemistry, we found that ABCA1 protein was present in adipocytes of obese scwat (Figure 4A). Some adipocytes tested completely negative for ABCA1, whereas others were strongly positive. A positive correlation was found between the percentage of adipocytes expressing ABCA1 and the percentage of adipocytes expressing SAA (R = 0.43, p = ). To illustrate this correlation, we performed a double immunohistochemical staining for SAA and ABCA1 on the same slide. We showed that the same adipocytes express-saa and ABCA1 (Figure 4B). We also observed that macrophages located around adipocytes in crown-like structures were also strongly positive for ABCA1. Secondly, we assessed the effect of recombinant SAA1 on ABCA1-mediated cholesterol efflux from hmads cells (Figure 5A). Recombinant SAA1 increased cholesterol efflux by 4.3 fold in differentiated hmads compared to ApoA1, while no effect was found in undifferentiated hmads. This effect in differentiated hmads was dose-dependent and present even at low doses of SAA1 (1μg/ml) (Figure 5B). Cholesterol efflux is increased when using the LXR-specific agonist GW3965, a known inducer of ABCA1 expression (Figure 5A). Probucol, a well known inhibitor of ABCA1 efflux, decreased markedly the cholesterol efflux (Figure 5C). These results confirm the specific contribution of ABCA1 in SAA1 mediated cholesterol efflux in human adipocytes. DISCUSSION Human obesity associates with profound morphological and biological changes in AT. The cellular modifications include increased adipocyte cell size (i.e. hypertrophia) (21) and accumulation of inflammatory cells (i.e. macrophages, lymphocytes), whose role remain to be established (22). The cross-talk between adipocytes and cellular components of the SVF has been shown to modify their respective functions (16, 23, 24). The present study provides novel insight suggesting that A- SAA might be a player in this paracrine dialogue. In agreement with our previous gene expression findings (5), we showed in vivo that SAA protein is strongly related to adipocyte hypertrophy; particularly that SAA is more abundant in scwat which is significantly more hypertrophic than owat (25). Jernas et al also showed that A-SAA gene expression is 18.7 times higher in large adipocytes (>90 μm) versus small adipocytes (<60 μm) (26). Adipocyte hypertrophia associates with metabolic changes (i.e. increased lipolysis, insulino-resistance) and is a condition favoring the secretion of adipokines such as IL-6, IL-8 and MCP-1 (27). In contrast to these proinflammatory factors, synthesized in abundance by cells of the SVF during obesity, SAA is predominantly produced by large adipocytes, thereby suggesting a specific role. Using a hypothesis generating approach combining microarrays and bioinformatics analysis, we found that A-SAA adipocyte gene expression was co-expressed with SVF genes that are related to inflammatory processes associated with macrophages and adipocyte genes related to cholesterol metabolism, suggesting an association between A-SAA, AT inflammation and adipocyte cholesterol metabolism. To first confirm the relationship between A- SAA and the inflammatory environment in human AT, we performed a series of tissue and culture experiments using hmads cells (18). In vivo, SAA staining highly correlated with macrophage infiltration in both WAT depots in obese subjects, suggesting a link between SAA production and the presence of AT macrophages. In addition we showed an induction of human adipocyte A-SAA by culture media obtained from macrophages derived from activated monocytes or ATM. A cocktail containing the pro-inflammatory molecules TNF-α and IL-1β led to an increase in SAA protein expression. Interestingly this in vitro induction is obtained only in the presence of a glucocorticoid. Thorn et al have studied A-SAA regulation in HepG2 hepatoma and KB epithelial cell lines and showed that both SAA1 and SAA2 genes were induced by TNFα and IL-6, where TNF-α induction of SAA1, but not SAA2, was enhanced by 8

9 glucocorticoids in both cell lines (28). In THP- 1 cells, A-SAA secretion was induced with IL- 1β in the presence of dexamethasone (29). For the first time, we demonstrate in human adipocytes that, in the presence of a glucocorticoid, A-SAA is regulated by TNF-α and IL-1β synergistically; molecules predominantly secreted by non-fat cells (30). IL-6, one of the molecules most secreted by non-fat cells in AT (30) and described to induce A-SAA production by hepatic cell lines (31), has no effect in this model. However its role in vivo cannot be excluded. Our observations suggest a biological link between human adipocytes expressing A-SAA and macrophages, strengthening recent results obtained in vitro with recombinant SAA1 (9). Yang et al demonstrated that A-SAA has a proinflammatory role. More specifically, the addition of A-SAA protein in a media together with cells of the SVF significantly increased the release of IL-6, IL-8 and MCP-1, a chemokine involved in macrophage recruitment, from adipocytes (9). These results agree with our suggestion that A-SAA has a role in the adipocyte-macrophage dialogue. Whether adipocyte hypertrophy is a primary event promoting the increased synthesis of A- SAA, and then macrophages recruitment, or the increased secretion of A-SAA by large adipocytes results from the stimulation of proinflammatory mediators produced by macrophages is currently unknown. Under such circumstances, what could be the role of A-SAA in the adipocyte-macrophage dialogue? Our experiments aimed to address primarily the relationship between A-SAA and adipocyte expressed genes. Within the adipocyte itself, A-SAA was co-expressed with genes coding for proteins involved in metabolism, particularly ABC transporters. Circulating A-SAA is an apolipoprotein component of HDL and was previously shown to have a role in ABCA1-dependent cholesterol efflux (3). In our study, ABCA1 protein is expressed in human scwat and SAA expression correlated with ABCA1. We subsequently confirmed that recombinant SAA1 protein affects ABCA1-mediated cholesterol efflux in human adipocytes. This effect was also observed when using lower amounts of SAA1 (1μg/ml), which is more physiologically relevant in obesity. The contribution of other cholesterol transport mechanisms (e.g. the scavenger receptor class B type 1) cannot be excluded; however, no correlative association between A-SAA and such receptors were observed in our microarray dataset. Similarly to cleaning mechanisms during acute inflammation, A-SAA could be a mediator of cholesterol transfer and capture by the macrophages in large necrotic adipocytes. In this regard, it has been observed that hypertrophic adipocytes enable the aggregation of macrophages on their surface and that the vast majority (>90%) of macrophages in obese mice and humans are found around dying adipocytes (32). Macrophages may thus engulf adipocyte lipid remnants that could include cholesterol molecules. As such, it will be interesting to establish co-culture models of macrophages and hmads cells in order to investigate cholesterol transfer between them. In summary, our findings suggest that A-SAA has, amongst others, a localized action in the WAT of obese subjects. In vivo SAA production is associated with macrophage infiltration and with production of the adipocyte ABCA1 transporter. In hmads, A- SAA is regulated by IL1-β and TNF-α and by ATM and SAA1 is able to induce ABCA1- dependent cholesterol efflux. One limitation of this study is that we did not analyze the involvement of A-SAA on cholesterol efflux in mature adipocytes derived from human WAT, but the fragility of mature human adipocytes in culture and the duration of cholesterol efflux experiments make such studies difficult. Nevertheless, the relationships highlighted in the present work suggest that A-SAA is a mediator of cholesterol efflux in human AT and, therefore, an important player in the cross-talk between hypertrophic adipocytes and macrophages. Acknowledgement This work was supported by INSERM, Paris6 University, the Assistance/Publique des Hôpitaux de Paris (P050318) and a grant from the European Community seven th framework program; ADAPT Adipokines as Drug to combat Adverse effects of excess Adipose Tissue project (grant agreement Number ). We thank Florence Marchelli and Christine Baudoin, who contributed to the clinical database constitution at the Centre de Recherche en Nutrition Humaine-Ile de France and Dr David Mutch for critical suggestions and editing modifications. 9

10 REFERENCES 1. Urieli-Shoval S, Linke RP, Matzner Y 2000 Expression and function of serum amyloid A, a major acute-phase protein, in normal and disease states. Curr Opin Hematol 7: Uhlar CM, Whitehead AS 1999 Serum amyloid A, the major vertebrate acute-phase reactant. Eur J Biochem 265: van der Westhuyzen DR, de Beer FC, Webb NR 2007 HDL cholesterol transport during inflammation. Curr Opin Lipidol 18: van Dielen FM, van't Veer C, Schols AM, Soeters PB, Buurman WA, Greve JW 2001 Increased leptin concentrations correlate with increased concentrations of inflammatory markers in morbidly obese individuals. Int J Obes Relat Metab Disord 25: Poitou C, Coussieu C, Rouault C, Coupaye M, Cancello R, Bedel JF, Gouillon M, Bouillot JL, Oppert JM, Basdevant A, Clement K 2006 Serum amyloid A: a marker of adiposity-induced low-grade inflammation but not of metabolic status. Obesity (Silver Spring) 14: Poitou C, Viguerie N, Cancello R, De Matteis R, Cinti S, Stich V, Coussieu C, Gauthier E, Courtine M, Zucker JD, Barsh GS, Saris W, Bruneval P, Basdevant A, Langin D, Clement K 2005 Serum amyloid A: production by human white adipocyte and regulation by obesity and nutrition. Diabetologia 48: Sjoholm K, Palming J, Olofsson LE, Gummesson A, Svensson PA, Lystig TC, Jennische E, Brandberg J, Torgerson JS, Carlsson B, Carlsson LM 2005 A microarray search for genes predominantly expressed in human omental adipocytes: adipose tissue as a major production site of serum amyloid A. J Clin Endocrinol Metab 90: O'Brien KD, Chait A 2006 Serum amyloid A: the "other" inflammatory protein. Curr Atheroscler Rep 8: Yang RZ, Lee MJ, Hu H, Pollin TI, Ryan AS, Nicklas BJ, Snitker S, Horenstein RB, Hull K, Goldberg NH, Goldberg AP, Shuldiner AR, Fried SK, Gong DW 2006 Acute-Phase Serum Amyloid A: An Inflammatory Adipokine and Potential Link between Obesity and Its Metabolic Complications. PLoS Med 3:e Henegar C, Tordjman J, Achard V, Lacasa D, Cremer I, Guerre-Millo M, Poitou C, Basdevant A, Stich V, Viguerie N, Langin D, Bedossa P, Zucker JD, Clement K 2008 Adipose tissue transcriptomic signature highlights the pathological relevance of extracellular matrix in human obesity. Genome Biol 9:R Prifti E, Zucker JD, Clement K, Henegar C 2008 FunNet: an integrative tool for exploring transcriptional interactions. Bioinformatics 24: Cancello R, Henegar C, Viguerie N, Taleb S, Poitou C, Rouault C, Coupaye M, Pelloux V, Hugol D, Bouillot JL, Bouloumie A, Barbatelli G, Cinti S, Svensson PA, Barsh GS, Zucker JD, Basdevant A, Langin D, Clement K 2005 Reduction of macrophage infiltration and chemoattractant gene expression changes in white adipose tissue of morbidly obese subjects after surgery-induced weight loss. Diabetes 54: Zhang B, Horvath S 2005 A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4:Article Hsu SM, Raine L, Fanger H 1981 Use of avidin-biotin-peroxidase complex (ABC) in immunoperoxidase techniques: a comparison between ABC and unlabeled antibody (PAP) procedures. J Histochem Cytochem 29: Clement K, Viguerie N, Diehn M, Alizadeh A, Barbe P, Thalamas C, Storey JD, Brown PO, Barsh GS, Langin D 2002 In vivo regulation of human skeletal muscle gene expression by thyroid hormone. Genome Res 12: Lacasa D, Taleb S, Keophiphath M, Miranville A, Clement K 2007 Macrophage-secreted factors impair human adipogenesis: involvement of proinflammatory state in preadipocytes. Endocrinology 148: Curat CA, Miranville A, Sengenes C, Diehl M, Tonus C, Busse R, Bouloumie A 2004 From blood monocytes to adipose tissue-resident macrophages: induction of diapedesis by human mature adipocytes. Diabetes 53:

11 18. Rodriguez AM, Elabd C, Amri EZ, Ailhaud G, Dani C 2005 The human adipose tissue is a source of multipotent stem cells. Biochimie 87: Verghese PB, Arrese EL, Soulages JL 2007 Stimulation of lipolysis enhances the rate of cholesterol efflux to HDL in adipocytes. Mol Cell Biochem 302: Svensson PA, Hagg DA, Jernas M, Englund MC, Hulten LM, Ohlsson BG, Hulthe J, Wiklund O, Carlsson B, Fagerberg B, Carlsson LM 2004 Identification of genes predominantly expressed in human macrophages. Atherosclerosis 177: Spalding KL, Arner E, Westermark PO, Bernard S, Buchholz BA, Bergmann O, Blomqvist L, Hoffstedt J, Naslund E, Britton T, Concha H, Hassan M, Ryden M, Frisen J, Arner P 2008 Dynamics of fat cell turnover in humans. Nature 453: Bouloumie A, Casteilla L, Lafontan M 2008 Adipose tissue lymphocytes and macrophages in obesity and insulin resistance: makers or markers, and which comes first? Arterioscler Thromb Vasc Biol 28: Suganami T, Nishida J, Ogawa Y 2005 A paracrine loop between adipocytes and macrophages aggravates inflammatory changes: role of free fatty acids and tumor necrosis factor alpha. Arterioscler Thromb Vasc Biol 25: Permana PA, Menge C, Reaven PD 2006 Macrophage-secreted factors induce adipocyte inflammation and insulin resistance. Biochem Biophys Res Commun 341: Cancello R, Tordjman J, Poitou C, Guilhem G, Bouillot JL, Hugol D, Coussieu C, Basdevant A, Hen AB, Bedossa P, Guerre-Millo M, Clement K 2006 Increased infiltration of macrophages in omental adipose tissue is associated with marked hepatic lesions in morbid human obesity. Diabetes 55: Jernas M, Palming J, Sjoholm K, Jennische E, Svensson PA, Gabrielsson BG, Levin M, Sjogren A, Rudemo M, Lystig TC, Carlsson B, Carlsson LM, Lonn M 2006 Separation of human adipocytes by size: hypertrophic fat cells display distinct gene expression. Faseb J 20: Skurk T, Alberti-Huber C, Herder C, Hauner H 2007 Relationship between adipocyte size and adipokine expression and secretion. J Clin Endocrinol Metab 92: Thorn CF, Lu ZY, Whitehead AS 2004 Regulation of the human acute phase serum amyloid A genes by tumour necrosis factor-alpha, interleukin-6 and glucocorticoids in hepatic and epithelial cell lines. Scand J Immunol 59: Yamada T, Wada A, Itoh K, Igari J 2000 Serum amyloid A secretion from monocytic leukaemia cell line THP-1 and cultured human peripheral monocytes. Scand J Immunol 52: Fain JN, Madan AK, Hiler ML, Cheema P, Bahouth SW 2004 Comparison of the release of adipokines by adipose tissue, adipose tissue matrix, and adipocytes from visceral and subcutaneous abdominal adipose tissues of obese humans. Endocrinology 145: Jensen LE, Whitehead AS 1998 Regulation of serum amyloid A protein expression during the acute-phase response. Biochem J 334 ( Pt 3): Cinti S, Mitchell G, Barbatelli G, Murano I, Ceresi E, Faloia E, Wang S, Fortier M, Greenberg AS, Obin MS 2005 Adipocyte death defines macrophage localization and function in adipose tissue of obese mice and humans. J Lipid Res 46:

12 Figure legends Fig 1. SAA protein expression in obese scwat and owat: relationship with adipocyte size. A. Percentage of adipocytes expressing SAA protein in scwat and owat. The value refers to the number of adipocytes that stained positive for SAA per 100 adipocytes. Mean ± SEM is shown. p = , n = 37 morbidly obese subjects, B. Mean diameter of adipocytes in scwat and owat. Mean for adipocyte diameter for all 37 subjects ± SEM is shown, where the mean adipocyte diameter for each subject was determined by measuring adipocytes. p = 10-3, n = 37 obese subjects. C. Correlation between the percentage of adipocytes expressing SAA and mean adipocyte diameter in scwat and owat. Adipocytes expressing SAA were labeled by immunohistochemistry and quantified in 37 obese subjects. Simultaneously mean adipocyte diameter was calculated in two randomly fields (x10) for each patient. R = 0.44, p = , in scwat and R = 0.49, p = , in owat. D. Percentage of adipocytes expressing SAA protein according to adipocyte size. In a x10 field, adipocytes are ranked according to the diameter range, i.e. adipocyte diameter < 60μm (white), between 60μm and 90μm (grey), and > 90μm (black). Once adipocytes were sorted by diameter, SAA positive adipocytes were counted. Data are the mean of values obtained in the scwat of 3 independent obese subjects. Fig 2. The functional profile (KEGG) of the genes co-expressed with A-SAA and the map of transcriptional interactions relating significantly over-represented KEGG themes A. KEGG categories showing a significant enrichment in functions for genes co-expressed with A- SAA. The bar plot represents the percentage of genes differentially expressed and functionally annotated in isolated adipocytes or the SVF and co-expressed with A-SAA. B. These categories were then related to construct a biological interaction map after quantifying their proximity based on the similarity in expression profiles for the annotated genes. Continuous lines indicate the strongest interactions (that is, superior to the upper quartile of their distribution), while dashed lines depict medium strength interactions (that is, superior to the median of the distribution but inferior to its upper quartile). The size of the nodes illustrating various KEGG themes is proportional to their interaction centrality in the transcriptional network using genes co-expressed with A-SAA. This figure shows that ABC transporter-general is a central theme that has an important position within the network (as illustrated by the size of the node). Fig 3. Regulation of A-SAA by inflammatory factors: relationship with human AT macrophages A. Percentage of macrophages (i.e. HAM56 positive cells) in both AT depots. Macrophages were labeled by immunohistochemistry and counted in 37 obese subjects. The mean for the 37 subjects ± SEM is indicated. This value is expressed as the number of cells stained positive for HAM56 per 100 adipocytes. p < B. Proximity of macrophages and adipocytes expressing SAA in WAT. We show immunohistochemistry performed in the scwat of a single obese female (BMI = 60.8 kg/m²). The image is representative of what is observed in the 36 other obese subjects. SAA positive cells are indicated in brown, HAM56 positive cells are shown in green and cell nuclei are in blue. The asterisk indicates SAA positive adipocytes and arrows indicate macrophages. C. Correlation between the percentage of SAA positive adipocytes and the percentage of HAM56 positive cells in scwat and owat. Serial fat tissue sections from 37 obese subjects were stained for SAA and HAM56. Values are expressed as the number of cells stained positive for either antibody per 100 adipocytes. R = 0.61, p = -4, in scwat and R = 0.46, p =4.10-3, in owat. D. A-SAA stimulation by inflammatory factors in hmads cells: The recombinant cytokines IL1-β and TNF-α and macrophage conditioned media were examined for the influence on A-SAA secretion in differentiated hmads. These experiments were performed in the presence of DEX (100nM). 12

13 Values represent the average of quadruplicate determinations (mean ± SEM). Similar results were obtained in two independent experiments. p < 0.02 vs. controls C refers to controls, ATM refers to adipose tissue macrophages, and AcM refers to activated blood monocytes. Fig 4. Immunohistochemical detection of ABCA1 protein in scwat A. Presence of ABCA1 in the adipocyte membrane. Cells expressing ABCA1 were labeled by immunohistochemistry in scwat tissue from an obese female (BMI = 49.6 kg/m²). The asterisk indicates ABCA1 positive staining in adipocytes. The arrow indicates positive staining of macrophages. B. Co-localization of ABCA1 and A-SAA in adipocytes. ScWAT tissue from an obese female (BMI = 47.1 kg/m²) was stained for SAA and ABCA1. SAA positive cells are identified by the brown colour, ABCA1 positive cells by the green colour and cell nuclei by the blue colour. This picture is representative of the other 36 obese subjects. Black arrows indicate ABCA1 staining, brown arrow indicates SAA staining. Fig 5. Cholesterol efflux to SAA1 in hmads cells hmads cells were labeled with 0.2μCi/mL [ 3 H]cholesterol for 24h, equilibrated in media in the absence of cholesterol for 24h and then incubated with either recombinant SAA1 (10μg/mL), ApoA-I (25μg/mL) or human HDL at 37 C for 24h and cholesterol efflux was measured. Values represent the average of triplicate experiments (mean ± S.D.). A. Recombinant SAA1 increases ABCA1-mediated cholesterol efflux in hmads. ABCA1 expression was stimulated by the LXR agonist (GW3965). Grey bars represent undifferentiated hmads and black bars represent differentiated hmads. Statistical significance (p = ) was observed when comparing SAA1 vs ApoAI and SAA1 + GW3965 vs ApoAI + GW3965. B. The effect of SAA1 on cholesterol efflux is dose dependent. Squares and triangles squares reflect SAA effect alone and with GW3965, respectively. C. Effect of an ABCA1 specific inhibitor (probucol) on cholesterol efflux. White bars represent untreated hmads, grey bars represent hmads pre-treated with GW3965, black diagonal bars represent untreated hmads + probucol, and grey diagnonal bars represent hmads + GW probucol. p = Supplementary Table 1. List of macrophage-specific genes expressed in the SVF which correlate with SAA. This list of macrophage-specific expressed genes was obtained from a previously published work by Svensson et al. (20). 13

14 Table 1. Clinical and biochemical characteristics of morbidly obese subjects PARAMETERS VALUES Number of subjects 37 Gender (F/M) 27/10 Age (years) 42.9 ± 2.0 Weight (kg) ± 4.2 BMI (kg/m 2 ) 49.3 ± 1.5 Glucose (mmol/l) 5.56 ± 0.22 Insulin (μu/ml) 22.3 ± 3.5 Total cholesterol (mmol/l) 4.6 ± 0.2 HDL-cholesterol (mmol/l) 1.2 ± 0.1 Triglycerides (mmol/l) 1.3 ± 0.1 Adiponectin (μg/ml) 9.4 ± 1.2 Interleukin 6 (pg/ml) 4.7 ± 0.6 hscrp (mg/dl) 1.3 ± 0.1 Data are indicated as means ± SEM Table 2. Primers used for SAA1+2 and ABCA1 gene expression detection Genes SAA1+2 ABCA1 Primers CGAGCATGGAAGTATTTGTCTG / CGGGACATGTGGAGAGCCT CCACGCTGGGATCACTGTA/ GCCTGCTAGTGGTCATCCTG 14

15 Figure 1 % OF ADIPOCYTES EXPRESSING SAA A scwat owat % OF ADIPOCYTES EXPRESSING SAA C scwat ADIPOCYTE SIZE ( m) % OF ADIPOCYTES EXPRESSING SAA D <60 > 90 ADIPOCYTES SIZE ( m) B scwat owat % OF ADIPOCYTES EXPRESSING SAA owat ADIPOCYTE SIZE ( m) ADIPOCYTE SIZE ( m)

16 Figure 2 A Up-regulated Transcripts KEGG Down-regulated Transcripts Retinol metabolism Ether lipid metabolism 28.6% Olfactory transduction 28.6% ABC transporters - General 28.6% 14.3% Allograft rejection 17.6% Autoimmune thyroid disease 17.6% Type I diabetes mellitus 17.6% Antigen processing and presentation 17.6% ECM-receptor interaction 17.6% Cell adhesion molecules (CAMs) 23.5% Leukocyte transendothelial migration 23.5% Cytokine-cytokine receptor interaction 41.2% Transcriptional domain coverage (%) B Autoimmune thyroid disease Cell adhesion molecules (CAMs) Antigen processing and presentation Retinol metabolism Cytokine cytokine receptor interaction ECM receptor interaction ABC transporters General Allograft rejection Ether lipid metabolism Olfactory transduction Type I diabetes mellitus Leukocyte transendothelial migration

17 % OF HAM56 POSITIVE CELLS A scwat owat C % OF ADIPOCYTES EXPRESSING SAA scwat % OF HAM56 POSITIVE CELLS D SAA (ng/ml) C IL1 TNF IL1 +TNF ATM AcM B X40 % OF ADIPOCYTES EXPRESSING SAA owat % OF HAM56 POSITIVE CELLS Figure 3 X100

18 A B X20 X40 X100 X100 Figure 4

19 Figure 5 A C SPECIFIC CHOLESTEROL EFFLUX (%) SPECIFIC CHOLESTEROL EFFLUX (%) ApoAI + SAA (5 g/ml) B SPECIFIC CHOLESTEROL EFFLUX (%) SAA ( g/ml)