Reduction of Macrophage Infiltration and Chemoattractant Gene Expression Changes in White Adipose Tissue of Morbidly Obese Subjects After Surgery-Induced Weight Loss

Seventeen Caucasian morbidly obese women undergoing laparoscopic Roux-en-Y bypass (19) and seven lean women (BMI 21.95 ± 0.83 kg/m², aged 34 ± 2.25 years, adipocyte mean size 62.4 ± 2.28 μm) were enrolled at Hôtel Dieu Hospital, Paris, France. Clinical and biological data for obese subjects were recorded at their peak of weight and 3 months after gastric bypass (Table 1). Fat mass was determined by biphotonic absorptiometry (DXA; Hologic, Waltham, MA). ScWAT biopsies were performed by a surgeon after local anesthesia (1% xylocaine) of the periumbilical area. A portion of each scWAT biopsy was fixed for light and electron microscopy, as described below, and immediately frozen for RNA extraction and analysis. To study the effects of weight loss on scWAT gene expression, we processed 10 high-quality RNAs from scWAT biopsies of the 17 collected before/after bypass by microarray technique (Table 1). We examined differences in gene expression between isolated SVFs and adipocytes from scWAT specimens of previously described overweight subjects (women, n = 9, BMI 27.9 ± 6.8 kg/m²) (14). The clinical investigations were performed according to the Declaration of Helsinki and approved by the ethics committees of Hôtel Dieu (Paris, France). Signed informed consent was obtained from all subjects.

ScWAT specimens were fixed and processed as described (14). Immunohistochemical detection of HAM56 (1:200; Dako Cytomation, Trappes, France), CD68 (1:100), and IL10 (1:100) (Santa Cruz Biotechnologies, Heidelberg, Germany) were performed with the avidin-biotin peroxidase method (20). Dewaxed sections (5 μm thick) were processed as previously described (14). Processed slide images were acquired by a microscope-camera system (Nikon, Champigny sur Marne, France), and 400 cells were measured. Macrophage number was determined in all samples by two independent observers. Four different randomly chosen areas within the processed slides labeled with HAM56 antibody were analyzed. Only HAM56+ entire cells were counted, and the mean number of positive macrophages was expressed as percentages of mean adipocyte number.

Small scWAT fragments collected from all obese patients before and after weight loss were fixed in 2% glutaraldehyde in 0.1 mol/l phosphate buffer (pH 7.4), postfixed in 1% OsO4, dehydrated in ethanol, and embedded an Epon-Araldite mixture (Epon; Multilab Supplies, Fetcham, U.K.) (Araldite; Fluka Chemie, Buchs, Switzerland). Thin sections were obtained with an MT-X ultratome (RMC, Tucson, AZ), stained with lead citrate, and examined with a Philips CM10 transmission electron microscope (Royal Philips Electronics, Eindhoven, the Netherlands).

To compare adipocyte with SVF-overexpressed genes, we digested scWAT specimens of nine overweight female healthy patients (mean BMI 27.9 ± 6.8 kg/m²), as previously described (14). Adipose tissue–derived macrophages included among SVF cells were isolated as reported (12).

Total RNA from scWAT was prepared using the RNeasy RNA Mini Kit (Qiagen, Courtaboeuf, France). One microgram of total RNA from each sample preparation was amplified using the MessageAmp RNA kit (Ambion, Austin, TX), and 3 μg aRNA was Cy-dye labeled using the CyScribe first-strand cDNA labeling kit (Amersham Biosciences, Orsay, France) (21 [available at http://cmgm.stanford.edu/pbrown/protocols/index]). aRNA extracted from each scWAT sample before surgery was labeled with Cy3 dye, while the aRNA from each scWAT sample obtained 3 months after surgery was labeled with Cy5 dye. The hybridization, washing, and scanning procedures were performed as described (14). Several quality cross-checks (for total RNA quality, aRNA quality, dye incorporation efficiency, etc.) and microarray “dye swap” experiments were performed.

Only spots with a regression coefficient >0.6 and a signal intensity >50% above background in both Cy5 and Cy3 channels were retrieved from the Stanford Microarray Database (available at http://genome-www5.stanford.edu/). Genes with significant changes in expression were identified using the significance analysis of microarrays (SAM) procedure (available at http://www-stat.stanford.edu/∼tibs/SAM/) (22). We used a selection threshold yielding the lowest median estimate of 0.05 false-positive genes (false discovery rate [FDR] of 5%). We also checked that all significant genes were sorted by SAM with 10% FDR when comparing the profiles of gene expression before and after weight loss in obese subjects. Hierarchical clustering of the significant genes was computed for the ratios after normalization. The clusters were visualized by Tree-View (23).

Functional profiling of differentially regulated genes in adipocytes and the SVF was performed based on the Gene Ontology Consortium (GO [available at http://www.geneontology.org]) terms (24). Using an “in-house” automated discriminatory annotation procedure, we identified GO categories that annotate genes differentially expressed in adipocytes and the SVF. The detailed procedure of this method, implemented into an algorithm called FunCluster (C. Hengar et al., unpublished data), is available online as supplementary data (available at http://corneliu.henegar.info/FunCluster.htm). We clustered together all related GO categories, based on sharing of a significant number of annotated genes. The gene annotation procedure was performed separately for each of the three available GO ontologies: molecular function, biological process, and cellular component. The biological process annotations presented in this article were the more exhaustive and immediately comprehensive ones. Molecular function–, cellular component–, and KEGG (Kyoto Encyclopedia of Genes and Genomes)-based annotations are available as supplementary online data. The P value of gene enrichment significance was calculated for each of the identified clusters by using a unilateral Fisher’s exact test adjusted for multiple testing errors with the Benjamini and Hochberg FDR correction approach. The clusters were ranked based on the statistical significance of the gene space coverage.

The SVF gene list was crossed with results for human macrophage predominantly expressed gene selection to estimate the contribution of macrophage gene expression in adipose tissue–derived SVF cells. Human macrophages from 15 healthy nonobese donors were differentiated and cultured as described (25). Total cellular RNA was isolated from differentiated macrophages using the RNeasy kit (Qiagen, Hilden, Germany). Target preparation was performed according to the manufacturer’s instructions (Affymetrix, Santa Clara, CA). Aliquots of the target preparation (15 μg copy RNA) were hybridized to individual U133A-DNA microarrays and then washed, stained, and scanned (GeneArray scanner G2500A; Hewlett Packard) according to procedures developed by the manufacturer (Affymetrix). Scanned output files were visually inspected for hybridization artifacts and then analyzed with MAS-5 software (Affymetrix). Sixty-eight nonmacrophage U133A expression profiles were obtained from the Symatlas database (available at http://symatlas.gnf.org/SymAtlas/) (26). Cancer tissues, fetal tissues, and cell lines present in the Symatlas database were excluded from analysis. The signal of each gene was normalized to the mean signal intensity of each DNA microarray (27). The average signal for each gene was calculated from the 15 macrophage and the 68 duplicate nonmacrophage expression profiles. The level of macrophage expression of each gene analyzed was ranked, with genes having maximal expression in macrophages ranked 1 and genes having minimal expression in macrophages ranked 69, as previously detailed (28).

We validated the changes in gene expression on RNA from whole scWAT, isolated adipocytes, the SVF, and macrophages by reverse transcription and real-time quantitative PCR (RTqPCR) (21). 18S ribosomal RNA amplification (Ribosomal RNA Control TaqMan assay kit; Applied Biosystems, Foster City, CA) was used as normalization control. The primers and TaqMan probes for mRNA were obtained from Applied Biosystems.

General statistical analysis was performed with JMP statistics software (SAS Institute, Cary, NC). For genes analyzed by RTqPCR, significant differences were determined by Wilcoxon nonparametric paired test (before/after surgery; adipocyte vs. SVF) and Student’s t test with unequal variance (obese vs. lean subjects). Correlations were examined by the nonparametric Spearman’s rank correlation test. P < 0.05 was the threshold for significance.

Materials and methods details and annexes to figures are available online at http://corneliu.henegar.info/FunCluster.htm.

The morphology of scWAT was examined by light microscopy before (T0) and 3 months after (3M) bypass. Macrophage infiltration was observed in scWAT of all obese subjects (Fig. 1A, T0). The morphology of macrophages in scWAT parenchyma differed markedly between obese and lean subjects. In lean individuals, single isolated macrophages surrounded endothelial trees, sometimes reaching small capillaries between adipocytes (annex to Fig. 1, see online supplementary data). In obese subjects, macrophages spread into the parenchyma of scWAT (Figs. 1A–D). In some areas, macrophages surrounded mature adipocyte cytoplasm in typical “crowns” (Figs. 1A–D). They exhibited triglyceride molecules in their cytoplasm and resembled foam cells (Figs. 1A–D). Transmission electron microscopy analysis confirmed the presence of macrophages very close to adipose cell membranes. Macrophage projections were directed to regions of adipocyte cytoplasm very rich in lipofuscin inclusions, typical of injured cells (annex to Fig. 1). The percentage of macrophages was significantly higher in scWAT of obese patients than lean subjects: 22.6 ± 4.3% (95% CI 16.8–28.3) vs. 1.4 ± 0.6% (0.95–1.8) in obese vs. lean subjects (P < 0.001) (Fig. 1I). All macrophages tested positive for HAM56, a marker of mature macrophages (Fig. 1I). Macrophages stained for CD68, a marker of the phagocytic activity (Fig. 1C). Before surgery, no staining was found for the anti-inflammatory cytokine IL10 (Fig. 1D).

Significant correlations were found between the number of macrophages and both BMI (ρ = 0.48, P = 0.0009) and mean adipocyte size (ρ = 0.53, P = 0.0008). A significant positive correlation was found between macrophage number and patient age (ρ = 0.677, P < 0.05). The associations between BMI and adipocyte size and macrophage number remained significant after age adjustment.

A significant decrease in plasma concentrations of leptin, insulin, and C-reactive peptide (CRP), as well as two acute-phase proteins serum amyloid A (SAA) and orosomucoid (ORO), were observed after bypass (Table 1). Three months after weight loss, a significant reduction of infiltrating macrophages (−11.63 ± 2.3% [95% CI −8.43 to −14.8], P = 0.01) (Figs. 1I–K) was observed. The tissue localization of the remaining macrophages was substantially different: macrophages were near blood vessels, and crowns had disappeared (Fig. 1 3M, E–H). The remaining macrophages still stained for HAM56 and CD68 (Figs. 1F–G). To test the hypothesis of phenotypic change from proinflammatory (M1) to anti-inflammatory (M2) macrophages after rapid weight loss, we examined the staining of scWAT-infiltrating macrophages for the most potent anti-inflammatory cytokine, IL10. Staining was almost absent in T0 macrophages but was strong in 3M remaining macrophages (Fig. 1H). Change in macrophage number was significantly correlated with changes in plasma levels of two acute-phase proteins, SAA (ρ = 0.71, P < 0.05) and ORO (ρ = 0.69, P < 0.05). A borderline significant correlation was observed for change in CRP variation (ρ = 0.58, P ≤ 0.09). No significant correlation was observed between macrophage number and clinical and metabolic parameters, neither with fasting insulin levels and quantitative insulin sensitivity check index nor with changes in insulin sensitivity after weight loss (annex to Table 1).

Since macrophages enriched the SVF of scWAT of obese subjects and their number decreased by weight loss, we examined the expression profile of SVF genes to determine whether some of them may be involved in the mechanisms of macrophage recruitment to scWAT. To select SVF-overexpressed genes, we used results of a previously established gene expression profiling experiment performed after dissociation of adipocytes and the SVF (14). We extracted the differentially expressed genes using the SAM procedure and then searched for enriched functions that best characterized the sets of adipocyte- and SVF-related genes. With a 5% FDR, SAM isolated 10,432 cDNAs, among which 6,587 exhibited increased expression in adipocytes (ratio range 1.04–16.31), while 3,845 were overexpressed in the SVF (ratio range 1.06–8.95). GO biological process annotations were found for 3,098 genes (with Locuslink ID), with 1,743 and 1,355 overexpressed in adipocytes and the SVF, respectively (online supplementary data). We validated expression changes for 15 randomly chosen genes, 12 with increased expression in the SVF (P < 0.05), and 3 mainly expressed in mature adipocytes (P < 0.05) (Table 2). To find the set of genes and functions that best characterized the SVF, we used our in-house approach based on GO “biological process” terms. Forty significant functional clusters were found for adipocyte-overexpressed genes, while 30 significant clusters characterized SVF-overexpressed genes (online supplementary data).

Figure 2 shows the first GO-annotated clusters ranked by statistical significance. The enriched functions that best characterized the genes overexpressed in adipocytes represent their well-known metabolic and secretory properties (Fig. 2A). In contrast, the most significant functions characterizing the genes overexpressed in the SVF were related to inflammatory or immune processes (Fig. 2B). The GO terms that grouped genes in the first five significant clusters were “defense response,” “chemotaxis,” “humoral immune response,” “antigen processing,” and “inflammatory response” (Fig. 2B). Other significant functional clusters characterizing the SVF were related to “regulation of cell proliferation,” “regulation of transcription,” “protein biosynthesis,” and “extracellular matrix organization and biogenesis” GO terms (P < 0.001–0.05).

We performed a set of RTqPCRs to validate the expression levels of seven SVF inflammatory genes. As shown in Table 3, all selected genes (MCP-1, α-2–macroglobulin, tumor necrosis factor-α, CD68, IL10, IL1 receptor antagonist, and hypoxia-inducible factor-1α [HIF-1α]) were significantly overexpressed in scWAT-derived macrophages compared with isolated adipocytes.

We subsequently focused on the set of 260 inflammatory genes of SVF cells to study the modulation of expression of these genes after weight loss and to examine if candidate genes of macrophage chemoattraction might be present among them. We compared the list of these genes with the genes differentially expressed in scWAT after 3 months’ weight loss and recovered 196 significant common genes. Among them, 119 (46%) exhibited a change in expression with 10% FDR, and 90 (35%) cDNAs exhibited differential expression with 5% FDR. As shown in Fig. 3, after clustering, 49 and 41 gene transcripts were under- and overexpressed, respectively, after weight loss. The set of upregulated genes included mostly those for complement-related factors and the major histocompatibility complex. The cluster of downregulated genes was composed of genes annotated by the following terms: “defense and immune response” (IL6 and IL6R), “chemotactism” (such as CCL1, MCP-1, and CCL6), and “positive regulation of cell proliferation” or “cell-cell signaling” (CSF-3 and regulator of macrophage development and survival). Two clones of the α subunit of the HIF-1α gene printed on the array were found in this cluster of downregulated genes (Fig. 3). To estimate the contribution of macrophage gene expression in the SVF, we used expression profiles from cultured human macrophages and 68 other tissues and cell types and the ranking procedure described in research design and methods. The list of 90 inflammatory SVF-related genes was crossed with a list of genes predominantly expressed in human macrophages that has been previously reported (28): 39 of 90 SVF-related genes (43.3%) were macrophage-overexpressed and 33 genes (36.6%) downregulated. Ten of the 90 regulated genes were ranked as number one for macrophages, showing that macrophages are the major site of expression of these genes and strongly contribute to their expression in the SVF (Table 4, online supplementary data). Three of these genes have previously been found to be predominantly expressed in macrophages (28). For 27 additional genes, macrophage expression was ranked in the top 15, suggesting that macrophage expression of these genes contributes to their expression in the SVF (Table 4, online supplementary data). For the 34 genes with the lowest ranking (ranks 36–69), macrophage expression is probably limited or expression of them is due to other cell types in the SVF.

Among the inflammatory genes modulated by weight loss (Fig. 3), four genes known to be involved in macrophage tissue attraction and regulation clustered together: MCP-1, CSF-3, and PLAUR. These genes play roles in chemotaxis and maintenance of circulating monocytes in tissue, developing an inflammatory state (29–31). The hypoxia-inducible gene HIF-1α was also located in this cluster. As shown in Fig. 4, levels of expression of MCP-1, CSF-3, HIF-1α, and PLAUR genes were significantly higher in scWAT of obese than in that of lean women (P < 0.05). After weight loss, we confirmed the significant decrease in expression of all these genes (Fig. 4). In the RTqPCR validation procedure, IL10 and IL1 receptor agonist genes, two anti-inflammatory cytokines, and the regulatory cytokine gene IL12A were upregulated after surgery-induced weight loss (annex to Fig. 4).

Macrophage infiltration of scWAT may participate in the low-grade inflammation associated with human obesity (10,11). We demonstrate that weight loss is associated with a reduction in the number and modified distribution of macrophages infiltrating scWAT of obese subjects. We confirmed that the increase in macrophage infiltration is significantly correlated with BMI and adipocyte size (10,12,13). Nevertheless, we better characterized specific histological features of infiltrating macrophages in WAT. In lean women, the rarely observed macrophages were dispersed among mature adipocytes and located near blood vessels. They probably correspond to the number of resident macrophages in scWAT that may eventually respond to infections. In scWAT of obese subjects, macrophages formed typical crowns, completely surrounding single adipocytes. They contained triglyceride particles in cytoplasm and resembled foam cells or phagocytes (32–35). This feature was never observed in lean subjects. On transmission electron microscopy, macrophage projections frequently held adipocyte membranes that contained cytoplasm lipofuscin inclusions, a typical feature of stressed, aged, and useless cells (36,37). We found no evidence of apoptosis. Altogether, morphology, distribution, and the staining pattern observed suggest a phagocytic activity of macrophages directed toward adipocytes in human scWAT. The potential capacity of macrophages to phagocyte adipose cells and the mechanisms involved warrant evaluation.

After weight loss (3 months), we observed a reduced number of macrophage cells and a disappearance of crowns. Despite substantial amelioration of fasting insulin and insulin sensitivity after surgery, assessed by quantitative insulin sensitivity check index, we failed to detect a significant correlation between these parameters and the number of scWAT-infiltrating macrophages. The paired correlation study performed had a weak statistic power due to the sample size (n = 17), and we did not detect significant correlations for insulin sensitivity or fasting insulin levels in these extremely obese subjects. The number of scWAT macrophages remained elevated 3 months after bypass, despite a general trend toward a decrease in the entire macrophage population. Reduction in the number of macrophages in scWAT may nevertheless decrease obesity comorbidities via improvement of low-grade inflammation. The correlations between reduction of macrophage number and decrease in two major acute-phase protein levels (ORO and SAA) suggest the existence of a relationship or of a cross-talk between scWAT-infiltrating macrophages and the systemic inflammatory state sustained by mature adipocytes (8). A larger cohort is needed for the borderline significant correlation observed for CRP variations. Noteworthy, the microarray subgroup had a higher CRP level than the whole bypass group. As described above, they were chosen for the good rate of RNA extraction and quality from scWAT specimens, and a higher CRP level was observed by chance. RTqPCR validations performed on scWAT samples from patients with various CRP levels were homogeneous, suggesting that CRP levels did not influence gene expression.

The important role of the SVF in immunity and inflammatory processes (38) was confirmed by our automated procedure of annotation. At least 30–50% of SVF inflammatory genes were highly sensitive to weight loss. This observation agrees with our previous results in less obese subjects after a 4-week very-low-calorie diet (14). We show here that decreased expression of inflammatory markers is associated with decreased macrophage number and increased expression of molecules with anti-inflammatory properties such as IL10 (39) by resting macrophages. These findings strongly support the existence of a phenotypic M1/M2-like shift of macrophages and provide additional clues to improvement of the inflammatory status of scWAT by modification of macrophage number and phenotype. Many of the SVF-overexpressed genes and their functions correspond to recognized properties of macrophages. Of the 90 SVF-enriched inflammatory genes, at least 39% were expressed at high levels in macrophages compared with other tissues and cell types (28). Enrichment of macrophages in the SVF is probably the principal reason for the inflammatory state observed in scWAT of morbidly obese subjects. The contribution of a small number of other immune/nonimmune cells to the SVF gene expression profile cannot be excluded. Preadipocytes exhibit close similarities in gene expression to macrophages and may convert to macrophages in inflammatory microenvironments (40–42).

The mechanisms responsible for macrophage recruitment and activation in human hypertrophic scWAT are unknown. Macrophages derived from bone marrow precursors and circulating monocytes may infiltrate scWAT (10,12). Paracrine, autocrine, and endocrine signals as well as mechanical modifications (i.e., hypertrophy/hyperplasia) of WAT cells may also play roles in such mechanisms. Several chemokines, such as MCP-1 and CSF-3, but in addition other hitherto undefined factors, may be involved in macrophage recruitment into scWAT (5,31,43,44). Some chemotactic factors (and/or their receptors) exhibited decreased expression after weight loss and clustered together (Figs. 3 and 4). Levels of expression of chemotactic molecules (MCP1, CSF3, and PLAUR genes) in scWAT of morbidly obese subjects were significantly higher than in that of lean subjects. The overexpression of HIF-1α in morbid obesity and the decrease in its expression after weight loss also indicate the contribution of local scWAT hypoxia to macrophage movement, as suggested (8). Tissue and cellular hypoxia is a well-known cause of macrophage attraction and retention, particularly in tumors and atherosclerotic plaque (30). The hypoxic state is relatively common in obese patients (8,45). Among the clusters of genes downregulated by weight loss, we found, together with HIF-1α, several genes known to be targets of this transcription factor (such as IL6, super oxide dismutase 2, C-X-C chemokines, and PLAUR) (46,47), suggesting that “codownregulation” of these molecules occurs during weight loss. Progression of adiposity could, together with worsening of local microhypoxia and persistent stimulation of chemotactic cytokines, enable macrophage infiltration and maintenance. Such a role was suggested for MCP-1 and CSFs (48,49). This hypothesis warrants further evaluation, in particular the analysis of the modification of these biomolecules at the protein level.

In conclusion, we show that the improvement of the inflammatory profile of morbidly obese patients after weight loss is associated with reduced number of infiltrating inflammatory cells in scWAT and transcriptional modulation of a cocktail of immune and inflammatory biomarker genes in WAT, some of which might play roles in macrophage aggregation. Increased knowledge of the mechanisms by which these inflammatory processes occur in scWAT and cause associated comorbidities is required.

This work was supported by Institut National de la Santé et de la Recherche Médicale (INSERM) Avenir contract, the Direction de la Recherche Clinique/Assistance Publique-Hopitaux de Paris, the Programme Hospitalier de Recherche Clinique (AOR02076 and clinical research contract no. 97123), and the Alfediam Association (clinical research contract 2002/2003). R.C. and S.T. are funded by INSERM, ADR Paris 6, Saint-Antoine, Paris, and Conseil Régional de l’Ile-de-France. R.C. received a grant from the Association Française d’Etude et de Recherche sur l’Obésité/Roche (France). Funding of the EA3502 team was provided by the Servier Research Institute and the Benjamin Delessert Institute, Paris 6 University, Paris, France. P.-A.S. received grants from Swegene and AFA.

The authors thank Dr. Patrick Levy, Dr. Philippe Sellam, and Dr. Marianne Cerceau (Surgery Department, AP/HP, Hôpital Hôtel Dieu, Paris, France) for performance of surgery biopsies. We also thank Prof. Thierry Molina and Dr. Mohib Morcos for support at the Jacques-Delarue Department of Anatomio-Pathology (Hôpital Hôtel Dieu, Paris, France) and Audrey Sicard, Annie Le Gall, Madeleine Gouillon, and Geneviève Bonhomme for their helpful technical assistance. We thank Dr. Michèle Guerre-Millo, Dr. Danièle Lacasa, and Dr. Cecile Lubrano for critical reading of this manuscript.