Endocrinology and Metabolism

Identification of depot-specific human fat cell progenitors through distinct expression profiles and developmental gene patterns

Tamara Tchkonia, Marc Lenburg, Thomas Thomou, Nino Giorgadze, Garrett Frampton, Tamar Pirtskhalava, Andrew Cartwright, Mark Cartwright, John Flanagan, Iordanes Karagiannides, Norman Gerry, R. Armour Forse, Yourka Tchoukalova, Michael D. Jensen, Charalabos Pothoulakis, James L. Kirkland


Anatomically separate fat depots differ in size, function, and contribution to pathological states, such as the metabolic syndrome. We isolated preadipocytes from different human fat depots to determine whether the basis for this variation is partly attributable to differences in inherent properties of fat cell progenitors. We found that genome-wide expression profiles of primary preadipocytes cultured in parallel from abdominal subcutaneous, mesenteric, and omental fat depots were distinct. Interestingly, visceral fat was not homogeneous. Preadipocytes from one of the two main visceral depots, mesenteric fat, had an expression profile closer to that of subcutaneous than omental preadipocytes, the other main visceral depot. Expression of genes that regulate early development, including homeotic genes, differed extensively among undifferentiated preadipocytes isolated from different fat depots. These profiles were confirmed by real-time PCR analysis of preadipocytes from additional lean and obese male and female subjects. We made preadipocyte strains from single abdominal subcutaneous and omental preadipocytes by expressing telomerase. Depot-specific developmental gene expression profiles persisted for 40 population doublings in these strains. Thus, human fat cell progenitors from different regions are effectively distinct, consistent with different fat depots being separate mini-organs.

  • visceral fat
  • homeotic genes
  • telomerase
  • metabolic syndrome

fat tissue distribution varies considerably among individuals. Different fat depots vary in their association with diabetes, atherosclerosis, hypertension, and malignancies, with visceral fat being particularly implicated (14, 15, 33, 34). Even lean subjects that have a high ratio of central to peripheral fat are insulin resistant (28). Surgical removal of visceral fat improves insulin responsiveness in rats [although venous drainage of epididymal and perinephric depots in rodents differs from that of omental fat in humans (5)], whereas removing subcutaneous fat from hamsters has the opposite effect (51). These differences in contribution to disease and function among depots may be due to regional variation in abundance of nonadipose cell types (e.g., macrophages, endothelial cells), anatomic constraints, circulation, and innervation. Indeed, fat cells in different depots could be fundamentally the same, with differences in their properties depending on other differences among depots. Alternatively, variation in inherent properties of undifferentiated fat cell progenitors may also contribute, as new fat cells continue to appear from preadipocytes during adulthood (7, 30). Whether preadipocytes isolated from different human fat depots and maintained in culture vary in such characteristics as capacity for adipogenesis has been a matter of controversy (2, 16, 22, 26, 36, 40, 43, 48, 49). It is important to resolve this controversy, as depot-specific inherent mechanisms might provide a basis for developing drugs to target particular fat depots. Our objective was to determine whether preadipocytes from different human fat depots are distinct.



Fat tissue was obtained during intra-abdominal surgery (gastric bypass, hernia repairs, cholecystectomy) from 12 subjects [6 male, 6 female; age 38.8 ± 3.0 (means ± SE) yr, median body mass index (BMI) 41, range 21 to 60 kg/m2, fasting glucose 106 ± 9 %mg], who had given informed consent. Tissue from obese subjects was included because of the clinical relevance of defining mechanisms of fat development in this population, and lean subjects were included for comparison. The protocol was approved by the Boston University Medical Center and Mayo Clinic Foundation Institutional Review Boards for Human Research. All subjects fasted ≥10 h before surgery. Subjects with malignancies or on thiazolidinediones or steroids were excluded. Abdominal subcutaneous (outside the fascia superficialis), mesenteric, and greater omental fat was obtained from subjects in parallel.

Preadipocyte culture and differentiation.

Fat tissue was digested in a collagenase buffer (43). Digests were filtered, centrifuged, and treated with an erythrocyte lysis buffer (23). Cells were plated at 4 × 104 cells/cm2 in 1:1 Dulbecco’s modified Eagle’s medium-Ham’s F-12 with 7.5% FBS and antibiotics. After 18 h (during which no replication occurs), cultures were trypsinized until 95% of cells were detached and replated. Plating medium was changed every 2 days until confluence. Newly confluent cultures were subcultured at a 1:2 ratio 5–7 times, as indicated below and in the figure legends, before study. These procedures yield essentially pure preadipocyte populations (43), with macrophages being rare (<5 per 106 cells by microscopy), irrespective of depot origin. Using these approaches, we found equivalent recovery among depots determined by adding known numbers of preadipocytes to weighed fat tissue aliquots before digestion compared with fat tissue aliquots of equivalent weight without added preadipocytes (30). Macrophage marker mRNA abundance (ADAM8, CD11b, CD68, F4/80, macrophage inflammatory protein-1α, monocyte chemoattractant protein-1) did not differ consistently between subcutaneous and omental primary preadipocytes or telomerase-expressing strains derived from single preadipocytes (see below), nor did endothelial (platelet endothelial cell adhesion molecule-l, CD46, VEGF receptor 2, von Willebrand, Tie-1, Tie-2) or stem cell (including Nanog, CD117, Sox2, ABCG2) markers (Supplemental Table S1; see AJP Endocrinol Metab website).1 For differentiation, preadipocytes were treated with plating medium (without serum) enriched with dexamethasone, insulin, triiodothyronine, ciglitazone, antibiotics, and methylisobutylxanthine (removed after 2 days) (43). Medium was changed every 2 days for 30 days. For the final 2 days, cells were maintained in plating medium without serum. Serum was also removed for 2 days from confluent undifferentiated preadipocytes before studies.

Preparation of telomerase-expressing clones.

Abdominal subcutaneous and omental preadipocytes were isolated from 2 female subjects (aged 42 and 50 yr). After the cells had undergone seven population doublings, they were transduced with a retrovirus (9) containing the plasmid, pBABE-hTERT-Puro (12), that expresses human telomerase reverse transcriptase (hTERT) driven by the Moloney murine leukemia virus long-terminal repeat promoter and a puromycin resistance gene. The cells were grown in a puromycin-containing medium and followed daily to ensure selection of colonies arising from single puromycin-resistant cells. The three abdominal subcutaneous and omental stably transduced clones capable of achieving confluence the fastest were selected from a total of 38 subcutaneous and 42 omental clones. Telomerase activity was verified using a telomere repeat amplification protocol (29).

Microarray analysis.

RNA was isolated from preadipocytes by the TRIzol method (11). Using a poly(dT) primer incorporating a T7 promoter, double-stranded cDNA was synthesized from 10 μg of total RNA using a Superscript cDNA Synthesis kit (Invitrogen, Carlsbad, CA). Biotin-labeled cRNA was generated from the double-stranded cDNA template through in vitro transcription with T7 polymerase using an RNA transcript labeling kit (Enzo Diagnostics, Farmingdale, NY). Biotinylated cRNA was purified using RNeasy affinity columns (Qiagen, Valencia, CA) and fragmented in 40 mM Tris-acetate, pH 8.1, 100 mM KOAc, and 30 mM MgOAc for 35 min at 94°C to 35–200 bases. cRNA (10 μg) and controls (Affymetrix, Santa Clara, CA) were hybridized to Affymetrix Human Genome U133A and U133B GeneChip arrays (U133A arrays only for the hTERT studies), washed, and stained according to the Antibody Amplification for Eukaryotic Targets protocol (Affymetrix). The arrays were scanned at 488 nm using a G2500A GeneArray Scanner (Agilent, Palo Alto, CA). Images were quantified and linearly scaled using GeneChip Operating Software 1.1 to a mean hybridization intensity of 500 units (GCOS; Affymetrix). Data were deposited in the GEO database (http://www.ncbi.nlm.nih.gov/geo/; accession no. GSE1657).

Real-time PCR.

Real-time PCR analysis was used to analyze expression of genes in undifferentiated subcutaneous, mesenteric, and omental preadipocytes from 10 subjects (5 nonobese males: age 34.2 ± 4.4 yr, BMI 27.5 ± 1.8 kg/m2; 5 obese females: age 39.4 ± 5.0 yr, BMI 48.9 ± 3.2 kg/m2). Total RNA was extracted from 5th passage primary cultures from each depot with Trizol (Invitrogen). For exclusion of genomic DNA, 10 μg of total RNA were treated with DNase I (Ambion, Austin, TX) for 1 h at 37°C. RNA integrity and quality was assessed by electrophoresis. One microgram of RNA was reverse-transcribed into cDNA using a Taqman One-Step RT-PCR kit (Applied Biosystems, no. 4309169) in 100-μl reaction mixture. The real-time PCR amplification was performed in 20-μl reaction mixture containing the cDNA along with primers for each transcript (Supplemental Table S2) and SYBR Green PCR master mix (Applied Biosystems, no. 4329593T). Human TATA box-binding protein (TBP) was used as an endogenous control and was detected using a dual-labeled fluorogenic probe (5′-FAM/3′-MGB probe, Applied Biosystems, no. 4333769F). mRNA and TBP mRNA levels were quantified using a fluorogenic 5′-nuclease PCR assay (24) with a GeneAmp 5700 sequence detection system (ABI/PerkinElmer). Duplicate reactions of each standard or sample were incubated for 2 min at 50°C, denatured for 10 min at 95°C, and subjected to 40 cycles of annealing at 55°C for 20 s, with extension at 60°C for 1 min followed by denaturing at 95°C for 15 s. Amplicon numbers were determined by comparison with a standard curve generated using five 10-fold dilutions of plasmids (10 pg to 1 fg) containing TBP cDNA. mRNA levels are shown as number of amplicons/104 TBP amplicons.

Data analysis.

For array analyses, fold change was calculated using the average signal from groups. Statistical significance of expression differences was calculated using a one-factor ANOVA (NIA Array Analysis Tool; http://lgsun.grc.nia.nih.gov/ANOVA). This method uses a Bayesian estimate of the within-group variance based on a prior distribution obtained from the variance of many transcripts at similar expression levels. To correct probabilities of differential expression for multiple hypothesis testing, we used the false discovery rate (FDR) method (6) that estimates the proportion of type I errors within a group of probe sets meeting a significance cutoff. The FDR is the quotient of the number of genes expected at a given significance cutoff under the null hypothesis of no differential expression over the number of genes detected at that cutoff. Because the Bayesian ANOVA is a single-factor analysis, we identified genes that vary as a function of depot in preadipocytes and adipocytes separately, combined both analyses, and then calculated the P value threshold that results in an FDR <0.05. To assess similarities and differences among gene expression profiles, we used hierarchical clustering (Spotfire DecisionSite; Spotfire, Somerville, MA) and, to test sample relatedness, principal component analysis (also performed in Spotfire DecisionSite) (3, 21, 32). Genes were annotated using NetAffx (Affymetrix) and Resourcerer (TIGR, Rockville, MD) databases. Overrepresentation of lipid metabolism and developmental genes among genes that differed among depots was determined by the Fisher exact test.


Preadipocytes from different depots have distinct expression profiles. Because the abundance of visceral relative to subcutaneous fat is a key component of the metabolic syndrome, abdominal subcutaneous, mesenteric, and omental preadipocytes were compared. The latter two depots account for most visceral fat, which has been considered as a single entity in many studies. We compared expression profiles generated from primary undifferentiated and differentiated preadipocytes from the three depots (18 array sets). Results were confirmed in subcutaneous and omental preadipocyte strains (6 array sets) made by expressing hTERT as well as by real-time PCR in additional subjects. For the primary culture array studies, preadipocytes from the three depots were cultured in parallel for five population doublings to reduce effects of any regional variation in influences extrinsic to preadipocytes, such as nutrient supply or innervation (n = 3 female subjects; age 32.7 ± 2.7 yr, BMI 47.8 ± 1.9 kg/m2). In genome-wide arrays, 44,928 probe sets of undifferentiated preadipocytes, and cells that had differentiated for 1 mo, sequence-specific hybridization was detected in at least one sample for 25,928 probe sets that were used in subsequent analyses. The effect of this filter is to remove hybridization intensities for transcripts that are not reliably measured (which we expect is due to transcripts for these genes not being expressed to a physiologically significant extent in adipose tissue). Adipogenesis affected expression of 5,783, or 22%, of these probe sets (with a Bayesian ANOVA P < 0.01 for a difference in hybridization intensity between the undifferentiated and differentiated samples), which corresponds to an FDR <0.05. 2,118 of these probe sets with differential hybridization varied by more than twofold during adipogenesis.

We identified 920 transcripts that differed among depots (Supplemental Table S3). These transcripts were identified by first testing for significant differences among depots (Bayesian ANOVA P < 0.001 corresponding to an FDR <0.05) in the 25,928 probe sets with sequence-specific hybridization in at least one sample. We then applied a secondary filter of the genes that passed this significance threshold to select those that varied at least twofold among depots (to exclude genes with small expression differences) either in differentiated or undifferentiated preadipocytes. Of these 920 transcripts, 503 varied among undifferentiated preadipocytes as a function of depot and 462 varied among differentiated adipocytes. Forty-five transcripts varied in both.

To highlight coordinately regulated genes, these depot-dependent transcripts were analyzed by correlation mapping (52) to cluster genes with similar profiles across subjects, depots, and differentiation states (Fig. 1). For example, the cluster around peroxisome proliferator-activated receptor (PPAR)γ2 and CCAAT/enhancer-binding protein-α (C/EBPα) (Fig. 1, graphs, bottom right) contains transcripts linked to expression of these adipogenic transcription factors (including aP2, adipsin, sterol regulatory element-binding protein-1, perilipin, hormone-sensitive lipase, apolipoprotein E, lipoprotein lipase, and glycerol-3-phosphate dehydrogenase). Genes involved in lipid metabolism [Gene Ontology (GO) database classification GO:06629] were concentrated in this cluster with C/EBPα and PPARγ2. In comparing expression across different fat depots, this cluster was noted to be most coordinately regulated during adipogenesis in abdominal subcutaneous, less so in mesenteric, and least in omental cells. This was associated with more extensive lipid accumulation in subcutaneous than omental differentiating preadipocytes, with mesenteric cells being intermediate (as in Ref. 43). These findings are consistent with earlier reports of greater expression of C/EBPα, PPARγ2, and other markers of adipogenesis in differentiating subcutaneous than omental primary preadipocytes (1, 36, 43). Furthermore, genes involved in lipid metabolism were overrepresented among the genes that varied among depots (Supplemental Table S4). Sixty-six out of 153 lipid metabolism transcripts varied among depots representing 7.2% (66/920) of depot-dependent transcripts [compared with 0.6% (153/25,928) of all probe sets for which sequence-specific hybridization was detected (P < 10−42; Fisher’s exact test)]. This regional variation in coordinately expressed transcripts implicates a contribution of upstream regulators to the distinct phenotypes of preadipocytes from different fat depots. This suggests that undifferentiated progenitors from different fat depots are distinct, setting the stage for regional variation in expression of adipogenic regulators, their downstream targets, and other genes.

Fig. 1.

Variation in fat depot-dependent gene expression variation occurs in clusters of genes whose expression is correlated. After 5 population doublings, abdominal subcutaneous, mesenteric, and omental undifferentiated (U) and differentiated (D) preadipocytes were subjected to genome-wide expression analysis. The 920 transcripts with expression levels that varied among depots in undifferentiated or differentiated preadipocytes were ordered identically along both axes by hierarchical clustering. Top: pair-wise Pearson correlation coefficients for each transcript using expression data from all depots. Positive correlations are shown in red and negative correlations in blue. Bottom: in the maps the order of the transcripts on each axis is the same as the top, but correlations were calculated for each transcript using only expression data from preadipocytes from a single depot. The expression profile of each transcript is shown to the right of the depot-specific correlation plots. Transcripts known to play a role in either development or lipid metabolism are shown to the left of each correlation plot. The symbols for genes with profiles most similar to peroxisome proliferator-activated receptor-γ and CCAAT/enhancer-binding protein-α are presented in the cutout. The full names, accession numbers, and hybridization intensities of all 920 genes are in Supplemental Table S3.

To test this possibility, we compared expression profiles of undifferentiated preadipocytes isolated from different fat depots (Fig. 2A). Abundance of 503 transcripts varied both significantly and at least twofold among undifferentiated preadipocytes from different depots (Table 1 and Supplemental Table S5). Unexpectedly, the major difference in gene expression between samples is one that distinguishes the omental preadipocytes from the subcutaneous and mesenteric preadipocytes. The similarity of abdominal subcutaneous to mesenteric preadipocytes was apparent when these genes were used in a principal component analysis to examine overall variability in expression of these genes specifically among samples (Fig. 2B). As expected by virtue of the fact that these genes were identified as being differentially expressed between the depots by ANOVA, the samples from within a depot were all relatively similar to each other. The different pattern of gene expression in the mesenteric compared with omental preadipocytes was also apparent by linear regression (Supplemental Table S6). Thus, mesenteric preadipocytes have expression patterns distinct from omental cells, indicating that visceral fat is not homogeneous.

Fig. 2.

Progenitors from different human fat depots have distinct expression profiles. A: profiles of the 503 transcripts that varied specifically among abdominal subcutaneous (S), mesenteric (M), and omental (O) undifferentiated preadipocytes were organized by hierarchical clustering, which groups transcripts with similar expression profiles together (names, accession numbers, and hybridization intensities are in Supplemental Table S5). Higher expression is shown in red and lower expression in blue. B: profiles of the 503 transcripts were subjected to principal components analysis. Sample loadings for the first principal component, which describes 65% of the total variability in expression of these transcripts, demonstrate that the overall expression profile of these transcripts in mesenteric preadipocytes more closely resembles that of abdominal subcutaneous than omental cells. C: profiles of the 124 developmental genes out of the 503 transcripts were subjected to principal components analysis. Sample loadings for the first principal component, which describes 64% of the total variability in gene expression, indicate that the expression profile of developmental genes in mesenteric preadipocytes more closely resembles that of abdominal subcutaneous than omental cells.

View this table:
Table 1.

Examples of genes differentially expressed among abdominal subcutaneous, mesenteric, and omental preadipocytes

Depot-dependent patterns of developmental gene expression are inherent. Genes involved in developmental processes (GO:07275; Supplemental Table S4) constituted 18.4% of the transcripts whose expression varied among depots (169/920, P < 10−16; Fisher’s exact test for a difference in the proportion of developmental genes among depot-dependent transcripts compared with the proportion of developmental genes among all probe sets with which sequence-specific hybridization was detected). The overrepresentation of developmental genes was particularly evident in comparing transcripts that varied among depots in undifferentiated preadipocytes (124/503, 24.7%; Fisher’s exact test, P < 10−27). As with the comparison using all transcripts that varied among depots, mesenteric undifferentiated preadipocytes were more like subcutaneous than omental cells with respect to developmental gene expression (Fig. 2C).

To determine whether these regional differences in developmental gene expression may be inherent, we made preadipocyte strains from single cells from different fat depots by stably expressing hTERT. The strains were then subcultured for 40 population doublings, both to amplify the cell population and to exclude effects of such in vivo influences as hormonal or nutritional state. The hTERT-expressing cells were capable of dividing >200 population doublings and retained capacity to differentiate into fat cells (44). As anticipated from the extensive effects of hTERT expression and serial subculturing in other cell types (41, 42, 46, 54, 56), global expression profiles of the hTERT-expressing strains differed from those of primary undifferentiated preadipocytes (Fig. 3A). We analyzed those transcripts that exhibited significant regional variation among undifferentiated primary preadipocytes in the hTERT-expressing clones (the 395 probe sets on U133A arrays out of the 503 on U133A and B arrays; Fig. 2). In general, the expression of transcripts was more variable among the hTERT preadipocyte clones than we found in primary cultures, perhaps reflecting intrinsic cell-to-cell variability in gene expression in the primary cultures that was effectively fixed by immortalization with hTERT. As a result of this heterogeneity, only 45 of these 395 probe sets showed statistically significant differential expression (FDR <0.05) between subcutaneous and omental clones. However, there was a strong positive correlation (r = 0.64, P < 10−15) in the fold change of gene expression observed between primary omental and primary subcutaneous preadipocytes on the one hand and the fold change observed in comparing hTERT clones derived from these depots on the other, even after 40 population doublings. The pattern of gene expression differences among depots in primary undifferentiated cells was also apparent in the hTERT strains by principal component analysis (Fig. 3B). Intriguingly, developmental regulators constituted almost one-half of these probe sets (22 of 45) that showed significant differential expression in the hTERT clones (chi-square test, P < 10−8).

Fig. 3.

Depot-dependent patterns of gene expression are inherent. The gene expression patterns of abdominal subcutaneous and omental primary preadipocytes and human telomerase reverse transcriptase (hTERT)-expressing strains that had undergone 40 population doublings were analyzed. A: a principal components analysis of the 12,543 probe sets with sequence-specific hybridization in at least 1 sample in primary preadipocytes on Affymetrix U133A chips indicates that overall expression profiles of hTERT-expressing preadipocyte strains differ substantially from those of primary preadipocytes. B: profiles of the 395 probe sets represented on Affymetrix U133A arrays out of the 503 that varied among depots in undifferentiated preadipocytes (from Fig. 2) were compared with hTERT-expressing strains. The similar pattern of subcutaneous and omental sample loadings between the primary cultures and hTERT-expressing clones demonstrates that the differences in depot-dependent gene expression between subcutaneous and omental preadipocytes are preserved in the hTERT-expressing clones. PC1, first principal component; PC2, second principal component.

We tested the 96 developmental genes that are differentially expressed between depots in primary preadipocytes on U133A chips for differential expression in the hTERT strains. Twenty-two probe sets (20 genes) were differentially expressed at an FDR threshold of 0.05 (Bayesian ANOVA, P < 0.001 for a difference in hybridization intensity between hTERT strains derived from preadipocytes isolated from subcutaneous and omental depots corresponding to an FDR <0.05; Fig. 4A). These genes were not selected with reference to the direction of differences in expression among depots (i.e., if a transcript was more abundant in subcutaneous than omental primary cultures but more abundant in omental than subcutaneous hTERT strains, it would still have been included). Despite this, the ratios of subcutaneous to omental expression of these genes were highly correlated in comparing the primary preadipocytes with hTERT strains (r2 = 0.94). Many of these genes are involved in segmentation and other processes during embryonic development, as well as regulation of cell replication and differentiation. Since the hTERT cell strains had been derived from single preadipocytes, contamination of the primary cultures with other cell types did not account for differences in expression patterns among depots. Also, since the hTERT cells had undergone 40 population doublings before analysis, there appears to be cellular memory for depot-specific expression of these developmental genes.

Fig. 4.

Depot-dependent developmental gene expression patterns are heritable. A: the expression profiles of 20 developmental genes are shown here organized by hierarchical clustering in primary preadipocytes and hTERT-expressing strains. Higher expression is shown in red and lower expression in blue. The direction and magnitude of differences in expression between the subcutaneous and omental strains were remarkably similar to those between subcutaneous and omental primary cultures (r2 = 0.94) despite the strains having been in culture for 40 population doublings. *In several cases, multiple probe sets for a single gene indicated differential expression. B: expression profiles were confirmed by real-time PCR of RNA from undifferentiated subcutaneous and omental primary preadipocytes from 5 nonobese males (age 34.2 ± 4.4 yr, BMI 27.5 ± 1.8 kg/m2) and 5 obese females [age 39.4 ± 5.0 yr, BMI 48.9 ± 3.2 kg/m2; 3 were those included in the array analyses (indicated by black bars)]. The majority of these genes show significant differential expression between depots by RT-PCR (P < 0.0025 by Student’s t-test corresponding to a Bonferroni-corrected P < 0.05), as indicated by gray boxes to the right of the heat map. Means ± SE are shown in Supplemental Table S7.

We used real-time PCR to examine the expression of these 20 developmental genes in primary subcutaneous and omental cultures from additional subjects [5 lean males and 5 obese females (including the 3 subjects in whom the array studies were done)] that had been subcultured five times. Expression of these genes differed as much or more among depots by PCR analysis, as had been observed by microarray analysis [r2 = 0.71 for the obese females (P < 0.0001) and r2 = 0.70 for the lean males (P < 0.0001); Fig. 4B and Supplemental Table S7]. Expression did not differ significantly in preadipocytes from the lean males compared with obese females. Thus, the differential gene expression observed by microarray analysis was found in additional subjects of both sexes, and irrespective of leanness or obesity, using an independent method. The prominence of developmental genes among transcripts with depot-dependent expression profiles indicates that fat cell progenitors from various regions are effectively different types of cells.


More than 900 genes that vary significantly among subcutaneous, omental, and mesenteric fat depots more than twofold were identified through genome-wide expression analysis in cultured human preadipocytes and adipocytes. Many of these genes are involved in lipid metabolism and development. In differentiating preadipocytes, expression of the adipogenic transcription factors C/EBPα and PPARγ was higher in subcutaneous than omental cells, as reported previously (2, 36, 43), possibly contributing to differences in expression of the lipid metabolism genes. In the present study, we focused on undifferentiated preadipocytes to address whether they differ among depots, potentially predisposing the fat tissue that develops from them to exhibit distinct properties. In the undifferentiated preadipocytes, differences among depots in expression of several of the developmental genes was confirmed without exception in hTERT preadipocyte strains and by real-time PCR in additional lean and obese male and female subjects, arguing for the general nature of our findings. The persistence of distinct developmental regulator gene expression profiles for 40 population doublings indicates that fat cell progenitors from different human fat depots are indeed inherently different.

Our findings support the contention that omental fat is fundamentally different from subcutaneous fat. We were surprised to find that fat cell progenitors from the two main visceral depots, omental and mesenteric, are also distinct, with gene expression profiles of mesenteric being much more like subcutaneous than omental undifferentiated preadipocytes. Interestingly, differences in lipid synthesis and lipolysis between omental and mesenteric fat have been found, at least in obese women (17, 19). We found differences between mesenteric and omental preadipocytes in capacities for replication and adipogenesis (43, 45). These findings raise the possibility that the distribution of visceral fat between the mesenteric and omental depots has clinical consequences. Thus, not all subjects with visceral obesity may have the same risk profiles.

Beyond giving rise to new fat cells, fat cell progenitors are an important cell type in their own right, accounting for 15 to 50% of the cells in fat tissue and actively producing paracrine factors, hormones, and metabolic signals in a manner distinct from that of differentiated fat cells (31). Remarkably, developmental gene products were prominent among transcripts that exhibited regional variation among undifferentiated preadipocytes. Some of the developmental regulators that vary among depots, for example, homeobox family members, regulate differentiation, replication, and apoptosis in adipocytes and other cell types (8, 13, 39, 47). These cell-dynamic properties vary among preadipocytes isolated from different fat depots; abdominal subcutaneous preadipocytes are capable of more extensive replication and adipogenesis and are less susceptible to apoptosis than omental preadipocytes (1, 36, 37, 43, 45). Another developmental gene differentially expressed in abdominal subcutaneous and omental preadipocytes, pregnancy-associated plasma protein A (PAPPA), has a potentially direct influence on fat tissue growth. PAPPA is a protease inhibitor that cleaves insulin-like growth factor (IGF)-binding protein-4, an inhibitor of IGF-I (35). These fat depot-dependent differences in developmental gene expression may contribute directly to regional variation in fat tissue function. They may indicate that upstream mechanisms, for example, histone acetylation or methylation patterns or other epigenetic processes, could cause or contribute to regional differences in both developmental gene expression and fat tissue function.

Consistent with the differences in developmental gene expression we found in undifferentiated human preadipocytes, differences in expression of homeobox family members and engrailed 1 [En1; a patterning gene (27)] have been found in comparisons between human subcutaneous and omental whole fat tissue (50), between human subcutaneous and visceral (depot not specified) whole fat (20), and between fat depots in rats (8, 20). This was the case even though fat tissue is subject to effects of hormones, paracrine factors, local variations in circulation and innervation, and other factors in vivo that are extrinsic to adipocytes and even though whole fat contains multiple cell types (with adipocytes accounting for around one-third of cells) whose abundance varies among depots. Indeed, 77% of all transcripts in previous microarray analyses of human whole subcutaneous and omental fat tissue gene expression (50), including developmental genes, varied in the same direction as in the preadipocytes in our study (P < 10−30; chi-square test). These regional differences in developmental genes in human fat were subsequently confirmed in microarray analyses of whole rat fat (20). On the basis of these array studies in rat fat tissue, these authors selected certain developmental regulators for RT-PCR analysis in subcutaneous and visceral whole human fat tissue. Region-specific variations in HOXA4, A5, C8, and C9, as well as En1, were reported in these two studies of whole human fat tissue (20, 50), consistent with our findings in human fat cell progenitors. Since we found regional differences in developmental gene expression in hTERT strains made from single cells, the varying mixtures of different cell types that are present in different fat depots or short-term primary cultures of rodent stromal vascular digests cannot solely account for the regional variation in developmental genes observed in these studies. Our finding that these regional differences in developmental regulators persist for many cell generations in pure cultures establishes that patterns of key developmental gene expression are, in fact, heritable in fat cell progenitors and human preadipocytes from different fat depots are distinct cell types.

Factors extrinsic to preadipocytes and adipocytes, including hormones, drugs (sex steroids, glucocorticoids, HIV protease inhibitors, thiazolidinediones), vascular supply, anatomic constraints, innervation, and presence of other cell types (e.g., macrophages or endothelial cells), likely contribute to fat depot-specific characteristics (4). Depot-specific, cell-autonomous mechanisms may act in concert with such extrinsic factors to result in regional heterogeneity. For example, inherent regional differences in expression of receptors, signaling, effector, and processing pathway components likely contribute to depot-specific effects of hormones, drugs, and nutrients that might be present at similar concentrations throughout the body. Regional differences in chemokine production by preadipocytes and fat cells may in turn contribute to heterogeneity in abundance of other cell types among fat depots. Sex, obesity, or other factors that may impact the cellular composition and paracrine microenvironment of adipose tissue (17–19, 53, 55) could combine with inherent regional differences to influence variation in function among depots.

The mechanism through which the cell-autonomous properties of preadipocytes arise is uncertain. One possibility is that the microenvironment in fat depots may confer heritable characteristics to fat cell progenitors during early development or later in life and could attract circulating cell types that influence or contribute to the progenitor pool, such as macrophages or fibrocytes (25, 53, 55). The local microenvironment may be able to modify these heritable characteristics. For example, obesity has been reported to modulate homeotic gene expression patterns (20). However, more than 40% of cells in fat tissue in obesity can be macrophages from the circulation, whereas only 3–12% of cells in fat from the lean are macrophages (53, 55). The substantial differences in abundance of these cells in the obese compared with lean individuals would be anticipated to impact apparent developmental gene profiles in whole fat tissue and could account for regional differences in developmental gene expression reported in whole fat tissue from lean compared with obese subjects. We found, despite these possible in vivo influences, that regional differences in developmental gene expression persisted for 40 population doublings in the hTERT expressing preadipocyte strains derived from single cells. Furthermore, similar regional differences in developmental gene expression were found in primary preadipocytes from lean males and obese females, in whom the local microenvironment is likely very different (e.g., differences in macrophage abundance, inflammatory cytokine levels, metabolites, sex steroids). Thus, inherent properties of fat cell progenitors are remarkably robust, and the extensive changes in fat tissue cellular composition that occur with obesity may contribute to the differences in developmental gene expression that have been observed in whole fat tissue from lean compared with obese subjects.

Fat cell progenitor populations are heterogeneous. Preadipocytes may be plastic, with capacity to develop not only into fat cells, but also macrophages, endothelial cells, or other cell types (10, 38). It is not clear whether there are differences in proclivity of progenitors from different fat depots to convert into other cell types. However, markers of these other cell types did not vary among the primary or hTERT-expressing preadipocyte cultures from different fat depots (Supplemental Table S1). Therefore, this is not likely to account for the regional differences we found. We recently found two preadipocyte subtypes, one capable of more extensive replication and adipogenesis than the other, the abundance of which varies among fat depots (45). Both subtypes from subcutaneous fat had similar developmental gene expression profiles (data not shown). Thus it is unlikely that regional variation in preadipocyte subtype abundance accounts for the differences in developmental gene expression we found among depots. Coupled with the regional differences in expression of developmental genes that remained apparent in hTERT strains derived from single preadipocytes, these observations indicate that regional differences in fat cell progenitor properties cannot be fully explained by variation in cell subpopulations. To reduce the chance that culture effects could contribute to apparent regional variation, we were careful to isolate preadipocytes from different depots within the same subjects at the same time and to culture the cells in parallel under identical conditions. We feel that any differences in selection of preadipocytes with distinct properties among depots because of our culture methods are unlikely to explain the regional variation we found for the following reasons. Similar patterns of developmental gene expression were found in early (5 population doublings) compared with later (40 population doublings) cultures. Regionally distinct patterns of developmental gene expression were found in hTERT strains derived from single preadipocytes. Differences among cultures from different depots in cell types other than preadipocytes were excluded by measuring markers, and we used methods that result in similar preadipocyte recoveries among depots. Differences among depots in preadipocyte subtype abundance did not appear to be the cause of differences in developmental gene expression. Thus we feel the regional differences we found cannot be readily ascribed to technical issues.

Fat is not a homogeneous organ. Our findings that fat depot origin affects both developmental gene profiles and the changes in gene expression that occur during adipogenesis support the view that depot-specific patterns of adipose cell gene expression predispose to regional heterogeneity in fat tissue function, tying fundamental developmental processes to the genesis of regional differences. The various fat depots appear to be, in effect, separate mini-organs. Perhaps regional heterogeneity in other mesenchymal tissues, such as bone or muscle, is also related to local differences in the inherent properties of the progenitors from which they arise.


This work was funded by National Institutes of Health Grants AG-13925 (J. L. Kirkland), DK-56891 (J. L. Kirkland), AG-23960 (J. L. Kirkland), and DK-46200 (Adipocyte Core).


We are grateful for administrative support by J. Armstrong and A. O’Brien. pBABE-hTERT-Puro was provided by R. Weinberg. E. Kokkotou provided help with real-time PCR studies.


  • 1 The Supplemental Material for this article (Supplemental Tables S1, S2, S3, S4, S5, S6, and S7) is available online at http://ajpendo.physiology.org/cgi/content/full/00202.2006/DC1.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract