We hypothesized that in adipose tissue, molecular processes regulate and respond to changes in adipose tissue mass independent of diet, sex, or the mechanism of obesity. To identify transcriptional patterns that correlate with body mass, we used oligonucleotide microarrays to catalogue gene expression levels in the parametrial or epididymal adipose tissue from two dozen mice whose body mass and adiposity varied due to diet, sex, and mutations in genes affecting energy homeostasis. We examined six experimental groups of 20-week-old C57BL/6J mice: (a) lean C57BL/6J female mice, (b) lean C57BL/6J male mice, (c) moderately obese C57BL/6J male mice with diet-induced obesity, (d) moderately obese female B6.Cg Ay/+ mice, (e) severely obese female B6.V Lepob/ob mice, and (f) severely obese male B6.V Lepob/ob mice. Each group contained four mice. As expected, body mass varied widely, with a range of 19.4–68.4 g and a mean of 44.6 ± 17.9 g.
For each transcript represented on the array, we computed Kendall’s τ rank-based correlation statistic as a measure of the correlation between the expression data and body mass for the entire sample of mice. To avoid potential problems of non-normality and sensitivity to outliers, we chose this nonparametric approach over the standard Pearson correlation coefficient (46
). The P
values corresponding to each test of correlation were computed exactly. With this approach, we identified 1,304 transcripts that were significantly correlated with body mass after controlling the false discovery rate at no more than 0.03. Supplemental Table 1 available online (http://www.jci.org/cgi/content/full/112/12/1796/DC1) lists all transcripts whose expression in perigonadal adipose tissue correlated with body mass.
We annotated each transcript whose expression correlated with body mass using the Gene Ontology Consortium (http://www.geneontology.org/) and Mouse Genomics Informatics (MGI) (http://www.informatics.jax.org/) databases. Analysis of the 100 transcripts that correlated most closely with body mass (i.e., had the lowest P
values) revealed three groups of functionally related genes that were coordinately regulated. Thirty percent of the 100 most significantly correlated transcripts encoded proteins characteristically expressed by macrophages, such as the CSF-1 receptor (τ statistic = 0.60, P
= 4.5 × 10–5
) and the CD68 antigen (τ statistic = 0.75, P
= 2.9 × 10–7
). Twelve percent encoded mitochondrial proteins such as succinate dehydrogenase complex, subunit B, iron sulfur (Ip) (τ statistic = –0.76, P
= 2.2 × 10–7
), and ubiquinol–cytochrome c
reductase subunit (τ statistic = –0.65, P
= 7.4 × 10–6
) (Figure ), and 6% encoded lysosomal proteins (Supplemental Table 2, http://www.jci.org/cgi/content/full/112/12/1796/DC1). The expression of all of the macrophage and lysosomal transcripts correlated positively with body mass, while the expression of each mitochondrial transcript was negatively correlated with body mass. Using quantitative RT-PCR we confirmed the expression profile of five genes (colony-stimulating factor 1 receptor
, and Mcp1
) in each of the 24 samples and found excellent agreement between the microarray and RT-PCR expression data (mean Pearson correlation coefficient = 0.91, microarray versus RT-PCR expression; Supplemental Table 3, http://www.jci.org/cgi/content/full/112/12/1796/DC1). The correlation of body mass with the expression of multiple genes characteristic of macrophages suggested that the macrophage content of adipose tissue was positively correlated with adiposity. To identify and quantitate macrophages within adipose tissue, we immunohistochemically stained sections for the F4/80 antigen, a marker specific for mature macrophages (Figure ; ref. 43
). We calculated the average adipocyte cross-sectional area and the percentage of F4/80-expressing cells in the perigonadal, perirenal, mesenteric, and subcutaneous inguinal adipose tissue depots from Ay/+
female, lean male, and diet-induced obese (DIO) male mice.
Figure 1 Adipose tissue transcripts whose abundance was correlated with body mass in mice. The expression of more than 12,000 transcripts in parametrial and epididymal adipose tissue was monitored in C57BL/6J mice whose body mass varied secondary to sex, diet, (more ...)
Figure 2 Adipose tissue macrophages in mice with varying degrees of adiposity. Immunohistochemical detection of the macrophage-specific antigen F4/80 (black arrows) in perigonadal adipose tissue from C57BL/6J mice: (a) lean female, (b) Ay/+ female, (c (more ...)
We examined the relationship between average adipocyte cross-sectional area and the percentage of adipose tissue cells expressing F4/80. Average adipocyte cross-sectional area (Figure ) was a strong predictor of the percentage of F4/80-expressing cells for each depot. Body mass was also a strong predictor of the percentage of F4/80-expressing cells for each depot (data not shown). The regression coefficients or slopes that describe the linear relationship between average adipocyte cross-sectional area and percentage of F4/80-expressing cells were comparable across mesenteric, perigonadal, and perirenal depots (Table ). Although the slopes corresponding to adipocyte cross-sectional area for the subcutaneous depot were smaller than (and fell outside the 95% confidence intervals of) the slopes for the mesenteric and perigonadal depots, this difference was entirely attributable to the fact that the data from three B6.V Lepob/ob mice fell below the line relating adipocyte area to macrophage content in the other animals (Table ). Except for leptin-deficient (ob) subcutaneous adipose tissue, the relationship of adipose tissue macrophage accumulation to adipocyte size was similar in all depots studied (Figure ).
Figure 3 The relationship between adipocyte size and the percentage of macrophages in adipose tissue. Average adipocyte cross-sectional area and the percentage of F4/80+ cells (macrophages) in adipose tissue depots were determined for each mouse in this (more ...)
Correlation of percentage of F4/80-expressing cells in adipose tissue with adipocyte size
Macrophages in the adipose tissue of lean mice were uniformly small, isolated, and widely dispersed among the adipocytes. In contrast, the macrophages in adipose tissue from obese animals were frequently found in aggregates. In the extremely obese animals, some of these macrophage aggregates completely surrounded adipocytes (Figure ). These aggregates resembled the macrophage syncytia characteristic of chronic inflammatory states such as rheumatoid arthritis and foreign body giant cell induction (47
We also examined the population of F4/80-expressing cells in liver and the extensor digitalis longus muscle from lean and obese Lepob/ob female mice. In liver, there was no significant difference in the number of F4/80-expressing Kupffer cells between lean and obese mice (Figure ). In muscle tissue, we observed only rare F4/80-expressing cells between myofibrils. However, muscle from both lean and obese animals was infiltrated and surrounded by adipose tissue (Figure ). Adipose tissue within muscle contained significant numbers of F4/80+ macrophages, and the percentage of F4/80+ cells within this adipose tissue was markedly increased in obese mice compared with lean mice (41% ± 4% of macrophages vs. 12% ± 2% of macrophages, respectively; P < 0.005, mean ± SD) (Figure ).
Figure 4 Macrophages in the liver and muscle of lean and obese mice. Immunohistochemical detection of cells expressing the macrophage-specific antigen F4/80 (arrows) in extensor digitalis longus muscles from C57BL/6J (a and c) Lepob/ob female and (b and d) lean (more ...)
To more fully characterize the F4/80+ cells in adipose tissue and define their cellular lineage, we used FACS to isolate and study the population of F4/80+ cells from adipose tissue. Perigonadal adipose tissue was collected from obese B6.V Lepob/ob female mice and digested with a combination of collagenase I and collagenase II. The resultant digest was centrifuged and yielded a buoyant adipocyte-enriched fraction and a pellet of SVCs. The SVCs were incubated with a fluorescently labeled anti-F4/80 antibody and were sorted by FACS into F4/80-expressing (F4/80+) and -nonexpressing (F4/80–) populations. Quantitative RT-PCR analysis of these cell populations revealed that F4/80+ cells express genes for macrophage-specific markers, including Csf1r, the CD68 antigen (Cd68), and the F4/80 antigen (Emr1). These are among the genes whose expression in our microarray expression data set correlated positively with body mass and adipocyte size.
The F4/80– population of cells expressed Emr1, Csf1r, and Cd68 at less than 2% of the levels expressed in the F4/80+ population (Figure ). Using FACS we also found that all of the F4/80+ cells coexpressed the common leukocyte antigen CD45 and the monocyte lineage marker CD11b (data not shown). F4/80+ cells did express detectable amounts of mRNA for adiponectin/ACRP30 (Acrp30), albeit at levels more than an order of magnitude lower that those found in primary adipocytes and differentiated 3T3-L1 adipocytes (Figure ). In contrast, at no timepoint during the differentiation of the preadipocyte cell line 3T3-L1 into adipocytes did we did detect significant expression of macrophage-specific genes (Csf1r, Emr1, Cd68) (Supplemental Table 4, http://www.jci.org/cgi/content/full/112/12/1796/DC1).
Figure 5 F4/80+ cells express macrophage markers. Perigonadal adipose tissue was collected from female B6.V Lepob/ob mice, digested, and centrifuged to yield a buoyant adipocyte-enriched cell population and a pellet of SVCs. The SVCs were separated into (more ...)
The accumulation of adipose tissue macrophages in direct proportion to adipocyte size and body mass may explain the coordinated increase in expression of genes encoding macrophage markers observed in our microarray expression data. However, macrophages and adipocytes express a number of genes in common, including Cd36
), and aP2
). Preadipocytes may also have phagocytic capabilities similar to macrophages (53
Tissue macrophages are derived from bone marrow precursors that migrate from the peripheral circulation. Preadipocyte populations are thought to be derived from resident mesenchymal cells. To test whether adipose tissue F4/80+ cells shared a common bone marrow origin with other tissue macrophage populations, we transplanted bone marrow from C57BL/6J mice expressing the CD45.1 leukocyte marker into 6-week-old lethally irradiated C57BL/6J mice expressing the CD45.2 leukocyte marker. After 6 weeks on a high-fat diet, 85% of the F4/80+ cells in periepididymal adipose tissue of the recipient mice were donor-derived (i.e., CD45.1+). Conversely, only about 14% of the F4/80+ cells in adipose tissue were recipient-derived CD45.2 cells (Figure ). Thus most F4/80+ cells in adipose tissue are bone marrow–derived.
Figure 6 F4/80+ cells in adipose tissue are bone marrow–derived. Adipose tissue was collected and SVCs were isolated 6 weeks after lethal irradiation and bone marrow transplantation. The SVCs were incubated with APC-conjugated anti-F4/80 (F4/80-APC (more ...)
The primary regulator of macrophage development and survival is CSF-1 (also known as M-CSF). Mice that carry a homozygous missense mutation in the Csf1
) are relatively macrophage deficient (43
). As a consequence of tissue macrophage deficiency, these mice develop a complex recessive phenotype of osteopetrosis, central blindness, and infertility (56
). This spontaneously arising mutation has been maintained on a mixed genetic background (C57BL/J × C3Heb/FeJ-a/a × CD1) but recently has been backcrossed onto the FVB/NJ strain (57
In subcutaneous and parametrial adipose tissue, the mean adipocyte size of the FVB/NJ Csf1op/op mice (311 ± 71 μm2) was smaller though not significantly different from that of control FVB/NJ Csf1+/+ mice (476 ± 326 μm2) (P = 0.37). However, FACS analysis showed that the fraction of F4/80+ SVCs was significantly greater in adipose tissue from control mice than from macrophage-deficient mice. In the depots studied, the fraction of F4/80+ cells in Csf1op/op mice was only 34% of that in Csf1+/+ mice (P < 0.01) (Figure ). Together these data suggest that the F4/80+ cells identified in adipose tissue are CSF-1–dependent, bone marrow–derived adipose tissue macrophages.
Figure 7 Macrophage-deficient FVB/NJ Csf1op/op mice are also deficient in F4/80+ cells in adipose tissue. SVCs were isolated from subcutaneous and perigonadal adipose tissue of macrophage-deficient (FVB/NJ Csf1op/op) and control (FVB/NJ Csf1+/ (more ...)
In most tissues, macrophages are a significant source of proinflammatory molecules. To determine whether adipose tissue macrophages express any molecules implicated in obesity-associated complications, we isolated three cell populations from the parametrial adipose tissue of three obese B6.V Lepob/ob mice: (a) an adipocyte-enriched population, (b) a stromal vascular macrophage F4/80+ population, and (c) an F4/80– stromal vascular population. Following isolation, RNA was extracted from each population, and the expression of three proinflammatory genes (Tnfa, Nos2, and Il6) was determined by quantitative RT-PCR. Of the three adipose tissue cell populations, the F4/80+ adipose tissue macrophages were the predominant source of TNF-α expression. Expression of iNOS was detectable at significant levels in each fraction, though it was highest in the macrophage fraction and thus it is likely that a significant portion of the iNOS expression in adipose tissue is derived from macrophages. In contrast, IL-6 was highly expressed in all three factions (Figure ).
Figure 8 Adipose tissue macrophages express proinflammatory factors. Perigonadal adipose tissue was collected from female B6.V Lepob/ob mice, digested, and centrifuged to yield a buoyant adipocyte-enriched cell population (gray bars) and a pellet of SVCs. The (more ...)
Based on these results in mice, we assessed the relationship of BMI and adipocyte size to macrophage abundance in abdominal subcutaneous adipose tissue of humans. Using quantitative RT-PCR we determined the relative expression levels of the macrophage expressed gene Cd68 in human abdominal subcutaneous adipose tissue. We found that both BMI (r2 = 0.43, P < 0.01; Figure ) and average adipocyte cross-sectional area (r2 = 0.46, P < 0.05; data not shown) were significant predictors of Cd68 expression. Cd68 expression was also significantly higher in the obese (BMI > 30 kg/m2; Cd68 expression = 1.67 ± 0.93 arbitrary units) compared to the lean subjects (BMI < 30 kg/m2; Cd68 expression = 0.69 ± 0.39 arbitrary units, P < 0.02).
Figure 9 CD68 expression in human subcutaneous adipose tissue. Subcutaneous adipose tissue samples were aspirated from the subcutaneous abdominal region of human subjects whose BMIs ranged from 19.4 to 60.1 kg/m2. CD68 transcript expression was measured by quantitative (more ...)
Using an antibody that recognizes the CD68 antigen we performed immunohistochemical analysis on these human samples and calculated the percentage of CD68 expressing cells (Figure ). Both BMI (r2 = 0.83, P = 0.0004; data not shown) and average adipocyte cross-sectional area (r2 = 0.86, P = 0.0002) were strong predictors of the percentage of CD68-positive cells (Figure ).