Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Obes Surg. Author manuscript; available in PMC 2010 October 18.
Published in final edited form as:
PMCID: PMC2956591

Adipocyte Accumulation of Long-Chain Fatty Acids in Obesity is Multifactorial, Resulting from Increased Fatty Acid Uptake and Decreased Activity of Genes Involved in Fat Utilization



The obesity epidemic causes significant morbidity and mortality. Knowledge of cellular function and gene expression in obese adipose tissue will yield insights into obesity pathogenesis and suggest therapeutic targets. The aim of this work is to study the processes determining fat accumulation in adipose tissue from obese patients.


Omental fat was collected from two cohorts of obese bariatric surgery patients and sex-matched normal-weight donors. Isolated adipocytes were compared for cell size, volume, and long-chain fatty acid (LCFA) uptake. Omental fat RNAs were screened by 10K microarray (cohort 1: three obese, three normal) or Whole Genome microarray (cohort 2: seven obese, four normal). Statistical differences in gene and pathway expression were identified in cohort 1 using the GeneSifter Software (Geospiza) with key results confirmed in cohort 2 samples by microarray, quantitative real-time polymerase chain reaction, and pathway analysis.


Obese omental adipocytes had increased surface area, volume, and Vmax for saturable LCFA uptake. Dodecenoyl-coenzyme A delta isomerase, central to LCFA metabolism, was approximately 1.6-fold underexpressed in obese fat in cohorts 1 and 2. Additionally, the Kyoto Encyclopedia of Genes and Genomics pathway analysis identified oxidative phosphorylation and fatty acid metabolism pathways as having coordinate, nonrandom down-regulation of gene expression in both cohorts.


In obese omental fat, saturable adipocyte LCFA uptake was greater than in controls, and expression of key genes involved in lipolysis, β-oxidation, and metabolism of fatty acids was reduced. Thus, both increased uptake and reduced metabolism of LCFAs contribute to the accumulation of LCFAs in obese adipocytes.

Keywords: Obesity, DNA, Microarrays, Dodecenoyl-coenzyme A delta isomerase, Fatty acid transport, Pathways


Obesity is, de facto, the increased deposition of long-chain fatty acids (LCFA), principally in the form of triglycerides (TG), in adipose and other tissues. Although LCFA were once believed to enter cells exclusively by passive diffusion across plasma membranes, the fact that TG accumulate in specific sites suggested that cellular LCFA uptake involves specific, regulatable mechanisms. Our laboratory [18] and others (e.g., [915]) have established that cellular LCFA uptake occurs by two distinct processes, of which diffusion is the minor component. At the LCFA concentrations typically found between meals, 80–95% of total cellular LCFA uptake is via a saturable, regulatable, facilitated transport process [7, 8]. Studies in animals and patients indicate that regulation of adipocyte LCFA uptake is an important control point for body adiposity [1621]. However, as in the liver [2224], many additional processes also contribute to LCFA and TG accumulation. Therefore, to understand the totality of the processes leading to obesity, it is important to define the global pattern of gene expression in adipose tissue. Published data suggest that at least 50 distinct genes are potentially involved in the establishment and maintenance of obesity [22]. The aim of the present study was to evaluate the multiple processes determining fat accumulation in adipose tissue from obese patients. Accordingly, we began to collect omental adipose tissue from patients undergoing bariatric surgery for the treatment of morbid obesity, and we proceeded with the analysis of an initial, small cohort of samples as a “trial run.” This study yielded a novel pathophysiologic finding, providing both a definite “proof of concept” and a valuable illustration of the translational research that can be performed via the combined efforts of bariatric surgeons and basic scientists. We subsequently achieved both technical and biological validation of the original results in an ongoing study of a second, larger cohort of patient samples. Most importantly, we were able to validate the observed changes in key genes from the first cohort by using both an alternative microarray platform and quantitative real-time polymerase chain reaction (qRT-PCR) on samples from the second cohort.

Materials and Methods


The initial study population (cohort 1) consisted of six patients undergoing clinically indicated abdominal laparoscopic surgical procedures at the Mount Sinai Medical Center (New York, NY, USA) who consented to the removal of an omental fat sample during surgery for studies of LCFA transport. Three of the patients (two males, one female) were obese and were undergoing bariatric surgical procedures related to their obesity. Despite obesity and the fact that none was on medications that influence glucose metabolism, none of the subjects had elevated fasting blood glucose concentrations. The other three patients (two males, one female) were nonobese and were undergoing various other clinically indicated laparoscopic procedures. None was diabetic, had a significant chronic inflammatory disease or malignancy, or was on medications likely to influence glucose metabolism, and—again—none had elevated fasting blood glucose. Cohort 2 consisted of seven obese female bariatric surgery patients and four nonobese female controls, all undergoing laparoscopic surgical procedures at either the Weill Cornell or Columbia Presbyterian campuses of New York Presbyterian Hospital. The protocol, consent documents, and procedures for these studies were approved by the Institutional Review Boards (IRBs) of the Mount Sinai School of Medicine (cohort 1) or the Weill Cornell and Columbia University Medical Centers (cohort 2).

Physiologic Studies


9,10-[3H]-Oleic acid (OA) was purchased from NEN Life Science Products, type I collagenase for adipocyte isolation from Sigma (St. Louis, MO, USA), fatty acid free bovine serum albumin (BSA) from Boehringer Mannheim (Indianapolis, IN, USA), and human insulin-specific and human leptin RIA kits from Linco Research (St. Charles, MO, USA).

Isolation of Adipocytes

Suspensions of human adipocytes were prepared by collagenase digestion of omental fat samples [2, 9, 16]. Suspensions were maintained at room temperature in Dulbecco’s modified Eagle’s medium (DMEM) for up to 3 h until warmed to 37°C for use [16], and met the established viability criteria [2, 9, 16]. Isolated adipocytes were sized by direct light microscopy at ×100, using an eyepiece reticle with which cell diameters were measured in arbitrary units (1 U=9.6 μ). The corresponding mean cell surface areas and cell volumes were calculated as described [25].

LCFA Uptake Studies

The initial rate of [3H]-OA uptake by omental adipocytes was determined by rapid filtration as described [1, 2, 16]. Briefly, cell suspensions with known cell counts, in 100 μl of DMEM, were added to 240 μl of DMEM containing 500 μM BSA and varying [3H]-OA concentrations and incubated for 0–30 s at 37°C. At four specified time points, uptake was stopped [1, 2], the cells were filtered and washed on the filters, and the filters with the cells were counted by liquid scintillation spectrometry [2, 16]. Adipocyte [3H]-OA uptake is linear over the initial 30 s of incubation [2, 16]. The slopes of the cumulative uptake vs time curves, representing initial uptake velocity (V0), were calculated from four samples obtained in triplicate over this portion of the curve by linear regression.

Computations and Fitting of Kinetic Data

The unbound oleate concentration ([OAu]) in each test solution was calculated from the OA/BSA molar ratio (ν) [26], using the LCFA/BSA binding constants of Spector et al. [27]. Our rationale for the use of these particular binding constants rather than several alternatives [2830] has been reported in detail previously [8].

Based on prior analyses (e.g., [58]), measurements of initial oleate uptake velocity at values of ν from 0.25 to 2.0 were fitted to the sum of a saturable and a nonsaturable function of the corresponding [OAu], according to the equation:


in which UT([OAu]) is the experimental measurement of uptake, in picomoles per second per 50,000 cells, at the stipulated [OAu]; Vmax and Km are the maximal uptake velocity of the saturable oleic acid uptake component and the value of [OAu] at one half the maximal uptake velocity, respectively; and k is the rate constant for nonsaturable uptake [2, 7, 8, 19, 21]. Data fitting was via the SAAM II version of the Simulation, Analysis and Modeling (SAAM) program [31], modified for execution on a laptop PC computer [32]. Prior studies documented that, under the specific conditions employed in the current studies, V0 and derived parameters such as Vmax are measures of trans-membrane transport, largely unmodified by premembrane phenomena such as rate-limiting dissociation from BSA and the effects of the pericellular unstirred water layer on substrate availability at the cell surface [33] or of intracellular binding or metabolism [1]. Studies in which an increase in Vmax was preceded by an increase in adipocyte size early in the development of obesity [19] and a decrease in Vmax preceded by a reduction in adipocyte size during leptin-induced weight loss [17] established that changes in Vmax did not simply reflect changes in cell volume.

Statistical Considerations

Values for physiologic variables are reported as the mean ± standard deviation calculated according to standard methods of descriptive statistics [34]. The significance of differences between groups was assessed with Student’s two-tailed t test with α≤0.05 being considered significant.

Gene Expression Studies

Tissue Collection

Omental fat samples were collected at the time of laparoscopic surgery under IRB-approved consents. Samples were divided and 1–2 g of tissue from each biopsy was placed in RNAlater (Invitrogen, Carslbad, CA) at −80°C for long-term storage.

Isolation of Total RNA

The fat samples were thawed and then homogenized in 15 ml of TRIzol (Invitrogen). After standard phase separation and RNA isolation, the pellet was resuspended in water, RLT lysis buffer, and ethanol for RNA cleanup and on-column DNase treatment (Qiagen). Eluted RNA consistently had A260/A280 ratios >2.0. The integrity of the total RNA was verified by the presence of robust 18S and 28S peaks in the BioAnalyzer electropherograms (Agilent).

Microarray Target Labeling and Hybridization

Biotin-labeled cRNAs were generated by established procedures. In brief, 2 μg of total RNA were used for synthesizing ds cDNA. This was incubated with biotin-labeled 11-UTP in an in vitro transcription reaction. cRNA was purified by RNeasy columns (Qiagen) and quantified by UV spectrophotometry at 260 nm. The size distribution of the biotin-labeled cRNA was verified by capillary electrophoresis (Bioanalyzer, Agilent). Ten micrograms of fragmented cRNA was hybridized overnight on CodeLink Human 10K microarrays (cohort 1) or Human Whole Genome microarrays (cohort 2). Hybridized cRNAs were detected by streptavidin–Cy 5 fluor (GE Healthcare).

qRT-PCR Validation

Expression of seven genes found to be underexpressed by microarray analysis in cohort 2 was examined in the same samples by qRT-PCR. Individual gene PCR primers were designed using Primer 3 software (v.0.4.0) at (Table 1). Selection criteria included the Tm of approximately 60°C and PCR product lengths between 150 and 250 bp. First-strand cDNAs were synthesized from total RNA samples using the TaqMan Reverse Transcription Reagent Kit (Applied Biosystems) with oligo(dT) as primers. PCRs were performed on the 7300 Real-Time PCR System (Applied Biosystems) with the SYBR® GREEN PCR Master Mix (Applied Biosystems) in a total volume of 50 μl containing 500 ng cDNA as detailed in the manufacturer’s guidelines. PCR conditions were: cycle 1 at 50.0°C for 2 min, cycle 2 at 95.0°C for 10 min, followed by 40 cycles (two step; 95.0°C for 0.25 min, then 60.0°C for 1.00 min). Each PCR was performed in duplicate. BBS4, MTIF2, PGBD3, and PAICS were selected as control genes based on their robust expression, with minimal differences between the normal and obese omental fat samples in the cohort 1 analysis. The means of their expression levels were used to normalize the expression of target genes DCI-SP1, 3-hydroxyacyl-coenzyme A dehydrogenase (ECHD), alcohol dehydrogenase 1A (ADH1A), adenosine triphosphate (ATP) synthase mitochondrial F1 (ATP5D), cytochrome c oxidase IV (COX4I1), cytochrome c-1 (CYC1), and NADH dehydrogenase Fe–S (NDUFS7) in all samples. The average fold change (AFC) was computed by using the average difference in the ΔCt between each test gene and the mean of the four control genes for each sample, i.e., AFC = 2−(average Δ ΔCt.

Table 1
Primer sequences for qRT-PCR studies

Data Analysis

Microarray spot detection was performed with the GenePix Series B Scanner (Axon Instruments) and spot quantitation was performed using CodeLink Expression Analysis v4.1 (cohort 1) or CodeLink Expression Analysis v5.0 (cohort 2). Key quantitation parameters are described briefly below. Local background subtraction is carried out on the individual spot intensities, followed by a scaling of each array individually based on the overall array intensity. After median normalization, the negative control threshold is calculated using a set of negative control probes as described [35].

Gene Ontology and KEGG Pathway Analysis

Median-normalized gene expression values were then imported into the GeneSifter gene expression analysis suite (Geospiza, Seattle, WA, USA). Array data for significantly differently expressed genes were overlaid onto ontological pathways (; [36]) and the Kyoto Encyclopedia of Genes and Genomics (KEGG) pathways (; [37]) using the GeneSifter software. The ontological and KEGG pathway analyses provide data on individual genes in the context of that gene’s role in described biological/biochemical pathways. A pathway was considered significantly altered from the control gene expression profile if its z score was <−2 or >2; z scores were calculated in GeneSifter as:


where R is the total number of genes meeting the selection criteria, N is the total number of genes measured, r is the number of genes meeting the selection criteria with the specified gene ontology (GO) term, and n is the total number of genes measured with the specific GO term [38]. The z scores with an absolute value of ≥2.0 are considered to indicate significantly altered regulation of the pathway compared with controls. The meaning of the z score depends on the context of the reported score. When reported as a z up score, a positive z score ≥2 indicates that a significant number of genes in the list of differentially expressed genes are upregulated in the experimental group in that particular pathway. Conversely, a negative z up score of −2 or less is also significant and indicates that fewer than expected genes are overexpressed in the pathway. For z down scores, the interpretation is as follows: a positive z down score indicates that more genes than expected are underexpressed and a negative z down score indicates that fewer than expected genes are underexpressed in the pathway.



The demographic and clinical characteristics of the patients are summarized in Table 2. The initial study population (cohort 1) consisted of three obese and three nonobese subjects, with two males and one female in each group. Cohort 2 consisted of seven obese and four nonobese female patients. The average ages of the obese patients in both cohorts and control patients in cohort 1 were similar. The nonobese patients in cohort 2 were younger. Virtually by definition, the body mass index (BMI) was significantly greater in the obese patients than in the control patients in both cohorts. Values for obese and nonobese subjects in the two cohorts were very similar. In additional studies in cohort 1, the fasting plasma leptin concentration was significantly higher in obese than in control subjects. Although the mean fasting blood glucose, plasma insulin, and serum cholesterol and TG were all higher in the obese patients than in the controls, only the difference in glucose levels between the obese and nonobese patients of cohort 1 achieved statistical significance.

Table 2
Demographic and clinical characteristics of study patients

Adipocyte Sizes

Adipocyte size measurements and the results of adipocyte LCFA uptake studies (cohort 1 only) are presented in Table 3. As expected, adipocytes from obese patients were appreciably larger than those from nonobese controls in both cohorts. Mean cell diameters were 1.7–1.8 times larger, surface areas were 3.0–3.3 times larger, and cell volumes were 5.1–6.0 times larger in obese than in nonobese patients of both cohorts.

Table 3
Adipocyte measurements and fatty acid uptake kinetics

LCFA Uptake Kinetics

Computer fits of the LCFA uptake curves in the six study subjects in cohort 1 are illustrated in Fig. 1. As in a series reported previously [16], there was no overlap whatsoever in the curves from obese compared with nonobese subjects. Both the Vmax for saturable LCFA uptake and the rate constant (k) for nonsaturable uptake were increased in adipocytes from the obese patients. The increase in Vmax was statistically significant, while increases in k and in the ratio of Vmax to cell surface area were not. As with some of the biochemical values and measures of adipocyte size, all of the comparisons between obese and nonobese subjects paralleled those reported in our larger, earlier series [16]. Failure of differences between groups for some parameters to achieve statistical significance in the present study results principally from the small sizes of the groups comprising cohort 1.

Fig. 1
Fatty acid uptake curves. [3H]-OA uptake by omental adipocytes was studied in three obese (dashed curves) and three normal-weight individuals (solid curves). The curves represent the computer fits to the sum of a saturable plus a nonsaturable function ...

Microarray Analysis

In cohort 1, the expression of approximately 10,000 human genes and expressed sequence tags (ESTs) was measured using Codelink Human 10K microarray in each sample of omental fat from obese individuals (n=3) and normal-weight donors (n=3). In cohort 2, Codelink Human Whole Genome microarrays queried the expression of ~50,000 genes and ESTs in each sample. Thus, samples in cohort 2 were evaluated for the expression of approximately five times as many genes and ESTs as those in cohort 1. Median-normalized expression values were analyzed using the GeneSifter (Geospiza, Seattle, WA, USA) software suited for the identification of differentially expressed genes and of KEGG pathways with significantly altered gene expression.

Measures of Quality Control of Expression Data

In the cohort 1 studies, a log–log plot of the means of obese (ordinate) vs normal (abscissa) expression values for each data point is seen in Fig. 2a (complete data set) and b (enlarged section enclosed by the magenta square). The limited data scatter on either side of the line of identity indicates the overall equivalence of the two data sets, while the statistically significant genes (t test, no correction for multiple testing) indicates the tightness of the data scatter for individual genes. A second indication of this overall equivalence is a comparison of the four mitochondrial ribosomal genes (L12, L38, L42, or S7) on the chips (data not shown). None of these four “housekeeping genes” demonstrates significant differences in expression between the two sample sets, showing the overall metabolic equivalence between the obese and normal fat depots. A third measure of overall equivalence can be seen in the corresponding quartile plots (Fig. 2c: obese [left] and normal [right]), which reveal no systematic differences between the data sets. These three global measures of gene expression indicate that there are no systematic differences between the two data sets, suggesting that valid biological conclusions may be drawn if differences in expression are detected for particular individual genes. Highly similar quality control comparisons were obtained from the cohort 2 samples.

Fig. 2
Comparisons of obese vs normal omental fat mRNA expression (cohort 1). The expression levels of approximately 9,200 genes were queried by microarray. a Log–log plot of all genes assayed in this study. Means of each data point are presented as ...

Identification of Individual Genes that are Differentially Expressed in Obese Fat

Pairwise comparisons between the three obese and the three normal samples in cohort 1 were performed with minimum fold changes in expression, followed by standard t tests and corrections for multiple testing. Using criteria of an expression difference of ≥1.5-fold and a p value of ≤0.05, we identified 166 differentially expressed genes and ESTs in the cohort 1 analyses. However, after applying the Benjamini and Hochberg correction for multiple testing [39], only one gene from this set, dodecenoyl-coenzyme A delta isomerase (3,2-trans-enoyl-coenzyme A isomerase; DCI) demonstrated a statistically significant difference between the two groups in cohort 1, being underexpressed 1.6-fold in obese fat (Fig. 3a). This gene encodes a member of the hydratase/isomerase superfamily. The protein encoded is a key mitochondrial enzyme involved in β-oxidation of unsaturated fatty acids. It catalyzes the transformation of 3-cis- and 3-trans-enoyl-CoA esters arising during the stepwise degradation of cis-, mono-, and polyunsaturated fatty acids to the 2-trans-enoyl-CoA intermediates. For the cohort 2 samples, the results were nearly identical for DCI expression differences between obese and normal omental fat samples (Fig. 3b).

Fig. 3
Expression of DCI mRNA in normal vs obese omental fat samples. DCI mRNA expression in obese fat is reduced vs control fat. a In cohort 1, DCI mRNA in omental fat depots from normal and obese individuals demonstrates a 1.59-fold underexpression in obese ...

Lipolysis-Related Gene Expression in Obese Fat

The initial step in the release of fatty acids from triacylglycerol stores is their hydrolysis via hormonally regulated lipolysis. Two key genes in this process, adenylate cyclase 6 (Fig. 4a, b) and adenylate cyclase-activating peptide receptor 1 (Fig. 4c, d), are both underexpressed in the obese fat samples in both cohorts. Adenylate cyclase 6 encodes a membrane-associated enzyme that catalyzes the formation of the second messenger cyclic adenosine mono-phosphate (cAMP). Adenylate cyclase-activating polypeptide 1 receptor type I encodes a membrane-associated receptor protein which mediates diverse biological actions of adenylate cyclase-activating polypeptide 1 and is positively coupled to adenylate cyclase. The fact that both of these key genes in the adenylate cyclase signaling cascade are underexpressed in obese omental fat suggests that this tissue may demonstrate reduced responses to physiologic stimulation by lipolytic hormones.

Fig. 4
Two adenylate cyclase genes involved in hormonal stimulation of lipolysis are underexpressed in obese omental fat. Adenylate cyclase 6 is a membrane-associated enzyme and catalyzes the formation of the secondary messenger cAMP. This gene was 1.78-fold ...

KEGG Pathway Analysis

To identify a larger group of differentially expressed genes for inclusion in the pathway analysis, we selected all genes in a pairwise comparison with an uncorrected p value ≤0.05, without correction for multiple testing, since the pathway analysis itself applies a second-level statistical filter. Using this uncorrected p value criterion, we identified 612 differentially expressed genes and ESTs between obese and normal fat from the list of 10,000 that were queried in cohort 1. This list of differentially expressed genes was subjected to KEGG pathway analysis for the identification of biological pathways with significantly altered gene expression in obese vs normal fat, as indicated by significant z up or z down scores. The complete set of downregulated KEGG pathways found in cohort 1 is presented in Table 4. No pathways with z up scores had enough differentially expressed genes to be considered biologically significant.

Table 4
KEGG pathways with significant z down scores

Of the pathways with significant z down scores and a sufficient number of genes involved to suggest biological relevance, oxidative phosphorylation and fatty acid metabolism, both directly related to energy metabolism and fatty acid biosynthesis and degradation [37], stand out. Of the genes in each pathway that demonstrate differential expression in cohort 1, all are underexpressed in obese fat samples relative to the samples from normal-weight individuals (eight genes in the oxidative phosphorylation pathway and six genes in the fatty acid metabolism pathway) in 12 of 14 instances with a fold change of ≥1.2. Results for these 12 genes in cohort 2 were generally very similar to those described below (Tables 5 and and66).

Table 5
Downregulated genes in oxidative phosphorylation pathway
Table 6
Downregulated genes in fatty acid metabolism pathway

The individual genes that were underexpressed in the oxidative phosphorylation pathway in cohort 1 with a fold change ≥1.2 are presented in Table 5. They include two H+ transporting mitochondrial ATP synthases, one lysosomal H+ transporting ATPase, two cytochrome c genes, and two dehydrogenases (ubiquinone and flavoprotein). Of the hundreds of proteins that make up the various enzymatic and electron transport complexes found in the inner mitochondrial membrane, the expression of the genes encoding seven of these proteins is downregulated in obese omental fat. Five of these genes encode proteins that are key regulatory enzymes or transport molecules, and these are spread across all five of the large complexes comprising the electron transport chain. As a result, even minor downregulation of each of these, when factored together, can result in an important functional difference in the production of ATP by complex V. The functional organization of these downregulated genes and their coordinate regulation has been brought into focus by the novel pathway analysis employed in this study.

In complex I, NADH dehydrogenase ( is the first enzyme of the complex that catalyzes the transfer of electrons from NADH to coenzyme Q. In complex II, succinate dehydrogenase ( reduces succinate to fumarate. In complex III, ubiquinol-cytochrome-c reductase ( is the heme-containing component of the cytochrome bc1 complex, which accepts electrons from Rieske protein and transfers electrons to cytochrome c in the mitochondrial respiratory chain. In complex IV, COX4I1 ( is one of the nuclear-encoded polypeptide chains of cytochrome c oxidase, the terminal oxidase in mitochondrial electron transport. In complex V, ATPase, H+ transporting, mitochondrial F1 complex, delta subunit (ATP6V1H, activates the ATPase activity of the enzyme and couples ATPase activity to proton flow.

Genes underexpressed with a fold change ≥1.2 in obese fat samples from cohort 1 that are components of the fatty acid metabolism pathway are presented in Table 6. These genes include acyl-coenzyme A oxidase, alcohol dehydrogenase 1A, DCI, another isomerase (peroxisomal D3, D2-enoyl-CoA isomerase), and hydroxyacyl-coenzyme A dehydrogenase.


The expression of seven genes from either the oxidative phosphorylation or fatty acid metabolism pathways which were found to be underexpressed by microarray analysis in cohort 1 was examined in the cohort 2 samples by qRT-PCR. The genes are ECHD, ADH1A, DCI, ATP5D, CYC1, NDUFS7, and COX4I1 (Fig. 5). All seven of these genes were underexpressed in cohort 2 as assessed with the Whole Human Genome microarray. Thus, there was complete concordance between the Human Whole Genome array and qRT-PCR with regard to the direction of regulation of expression of these seven genes. Also, for the seven genes queried, the direction of difference in expression between obese and controls is also consistent between the two sample sets (cohort 1 and cohort 2), since these seven genes are all underexpressed in the obese fat samples. These findings are consistent with a generally reduced energy metabolism in obese fat.

Fig. 5
Biological and technical validation of key gene expression results in cohort 1 by alternative microarray and in cohort 2 by qRT-PCR. Key genes which were underexpressed in obese omental fat in cohort 1 were selected for repeat testing in cohort 2 samples. ...


As already noted, obesity is—virtually by definition—the increased deposition of LCFA, principally in the form of TG, in adipose and other tissues. That this deposition is highly selective and that its extent and turnover differ even between different adipose tissue depots (e.g., [4043]) indicate that there is a complex underlying pathophysiology that involves far more than the passive, unregulated diffusion of LCFA—the essential building blocks of TG—across cell membranes.

Indeed, many processes may contribute to TG accumulation: increased LCFA uptake or synthesis, increased LCFA conversion to TG, or increased uptake of preformed TG from lipoproteins or, alternatively, decreased TG or LCFA removal by decreasing TG lipolysis, decreased LCFA and TG excretion as components of VLDL, or decreased LCFA oxidation [18, 2224, 4446]. These multiple processes involve multiple genes, e.g., for plasma membrane and intracellular LCFA transporters; for enzymes of fatty acid or TG synthesis; for receptors, enzymes, and other proteins associated with the import and/or hydrolysis of preformed lipoprotein TG, such as the LDL receptor, hepatic lipase, and lipoprotein lipase; genes associated with VLDL synthesis, assembly, and export; and proteins and enzymes of fatty acid oxidation; not to mention numerous transcription factors and other regulatory genes. It would be desirable to assay all of these processes simultaneously. While the key genes involved in many of these processes have been identified, for other processes, including cellular LCFA uptake, considerable uncertainty remains [47, 48]. Several of the processes, such as cellular LCFA uptake and oxidation, can be assayed directly. While direct quantitation of others is more difficult, a first-order approximation may be obtained via RNA expression studies.

Microarray Expression Analysis

We estimate that the number of known genes potentially involved in obesity pathogenesis is at least 50 [22]. RNA expression microarray technology is an effective approach for analyzing a large number of genes and also for identifying genes whose role in the process of interest—in this case, obesity—is unknown. Microarrays simultaneously monitor the expression of thousands of mRNAs from individual samples. In addition to providing information regarding the expression of preselected candidate genes, the high-throughput nature of microarray analysis is ideal for the identification of novel candidate genes and/or pathways responsible for the pathophysiology of complex diseases such as obesity.

We analyzed our data using a two-stage approach. The first stage was to identify individual genes whose expression was significantly different between normal and obese omental fat samples. Using a standard cutoff of 1.5-fold difference in expression, along with Benjamini and Hochberg correction for multiple testing [39], a single gene, DCI or 3,2-trans-enoyl-coenzyme A isomerase, was identified in cohort 1 as being underexpressed in obese omental fat. Its underexpression was confirmed in cohort 2 both by microarray and qRT-PCR expression analyses, providing both biological and technical validation of this result. This gene is intriguing since it is a mitochondrial enzyme involved in the β-oxidation of unsaturated fatty acids. Metabolic intermediates produced during the stepwise degradation of unsaturated LCFA enter the citric acid cycle where they contribute to ATP production by oxidative phosphorylation. Interestingly, central obesity has been positively associated with an increase in n-6 unsaturated fatty acids and inversely associated with monounsaturated fatty acids [49].

Genomic Organization of DCI

This gene is located at chromosomal locus 16 p13.3, encoding a predicted protein of 302 amino acids (ENTREZ [NM_001919]). Locus 16p13.3 has twice been linked to obesity-related factors, including BMI, both by logarithm of the odds score (LOD score) of microsatellite marker D16S510 in a study of Old Order Amish families [50] and in a genome scan of African-American families enriched for nondiabetic nephropathy [51]. These reports and the reduction of DCI expression in the current study make this an intriguing candidate gene for further study in obesity. The ~50% loss of DCI expression is consistent with functional loss of one DCI allele, which could reflect microdeletions in this gene region or single-nucleotide polymorphisms adversely affecting the splicing or stability of the DCI message.

KEGG Pathway Analysis

The power of microarray analysis is that it allows the simultaneous screening of thousands of genes in each sample, unbiased by candidate gene preselection. Recent advances in software allow changes in gene expression to be considered in the context of biological pathways. The KEGG consortium (; [37]) has established a collection of gene pathways whose gene members are known to interact as parts of a greater whole. Significant changes in pathway member gene expression, especially when the changes are coordinate (e.g., involving several sequential members of a signaling pathway) indicate that biologically meaningful alterations of pathway regulation and/or function are likely. Moreover, relatively small changes in expression of multiple genes in a pathway, each too small to achieve significance individually, can lead to biologically significant alterations in throughput along a pathway. For example, using an admittedly simplistic model, 10% downregulation of seven genes in a pathway, each likely to be overlooked by single-gene analysis, could result in a biologically significant reduction of ≥50% in throughput along the pathway.

Oxidative Phosphorylation KEGG Pathway

The oxidative phosphorylation pathway captures the energy released by the oxidation of NADH and succinate in the citric acid cycle, producing ATP via ATP synthase. Of the 51 genes in this pathway that are included on the arrays used in our study, seven are differentially expressed, and these seven are all underexpressed in obese fat. Our results show that two protein components of the ATP synthase complex are downregulated in obese fat: the G subunit of the ATP synthase F0 complex and the delta subunits of the F1 complex. These results further show that components of the electron transport chain, including cytochrome c oxidase and cytochrome c-1, plus NADH dehydrogenase and succinate dehydrogenase (subunit A), are also under-expressed. If follow-up studies demonstrate that the levels of the corresponding proteins are also diminished, it would be reasonable to expect decreased ATP production in obese fat tissues. It is important to recognize that each of the five complexes is built up of many subunits that depend on a delicate macrostructure.

Our data showing underexpression of all seven of the seven differentially expressed genes involved in oxidative phosphorylation are in remarkable agreement with similar findings in monozygotic twins that were discordant for obesity [52]. Microarray analysis of fat biopsies identified 30 genes in the oxidative phosphorylation pathway that were underexpressed in the obese subjects compared with their nonobese twins. A number of these genes were also underexpressed in our study, including ATP5L, ATP5D, ATP6V1H, and CYC1. Mustelin et al. also found members of the NADH dehydrogenase complex to be underexpressed in obese fat. Similar data have been reported by Patti et al. [53] who found a coordinated reduction in oxidative metabolism gene expression in skeletal muscle of type II diabetic Mexican Americans, related to the reduced expression of two transcription factors, nuclear respiratory factor 1 (NRF-1) and PPARγ coactivator 1, which are known to regulate the expression of genes in the oxidative phosphorylation pathway. The genes encoding these transcription factors are not present on the 10K microarrays used in the original studies. In the cohort 2 samples, NRF-1 and PPARγ coactivator 1 were essentially unchanged, indicating that another mechanism might be responsible for the consistent downregulation of the oxidative phosphorylation pathway genes in obese omental fat.

Fatty Acid Metabolism KEGG Pathway

The fatty acid metabolism pathway includes the enzymes responsible for the biosynthesis and degradation of LCFAs, leading to the production of acetyl-CoA, which is then directed into the TCA cycle. From there, oxidation of NADH and succinate produce ATP via ATP synthase in the oxidative phosphorylation pathway. Of the 37 genes probed by our arrays, six of six with differential expression were all underexpressed in obese fat. Each of these underexpressed genes plays a role in the metabolism of LCFA, and several have already been reported to play a role in obesity. Acyl-coenzyme A oxidase 1 ( expression in omental fat reportedly predicts weight loss outcome after gastric bypass surgery [54]. DCI (, grouped with 3-hydroxyacyl-coenzyme A dehydrogenase (, plays a key role in LCFA degradation (discussed above). A higher incidence of childhood obesity has been reported in children with long-chain 3-hydroxyacyl-coenzyme A dehydrogenase or trifunctional protein deficiency. Placing these children on low-fat, high-protein, and lower-carbohydrate diets resulted in lower energy intake and increased energy expenditure [55].

Lipolysis-Related Gene Expression

Our results showing reduced expression of key genes in the cAMP-mediated signaling of hormonally stimulated lipolysis are consistent with reports showing a blunted lipolytic response to catecholamine stimulation in obese individuals, especially in abdominal fat [56] and adipocytes from obese individuals, due in part to reduced expression of hormone-sensitive lipase [57]. However, fuller understanding of this complex system awaits ongoing transcriptome studies.

ABC Transporters KEGG Pathway

One of the genes in this pathway encodes the ATP-binding cassette (ABC), subfamily D (ALD), member 3 (ABCD3), a member of the superfamily of ABC transporters. ABC proteins transport various molecules across extracellular and intracellular membranes. ABCD3 is a member of the ALD subfamily, which is involved in peroxisomal import of fatty acids and/or fatty acyl-CoAs into the organelle. While little is known about this gene in human obesity, its upregulation in the present study is intriguing in light of its known role in intracellular LCFA transport. The possibility that increased peroxisomal LCFA import and subsequent oxidation might partly substitute for the decreased mitochondrial β-oxidation implied by our other findings merits further study.

Validation of Results

It is widely accepted that microarray results analyses require validation. This may involve biological validation in a second, independently collected set of samples and technological validation by using an alternative microarray platform or an entirely different technology such as qRT-PCR. Validation may also be accomplished by measurements of the proteins encoded by regulated genes or of biological activity. In the present study, biological validation was achieved by the study of two independently collected sample sets and technical validation by the use of two different microarray platforms and qRT-PCR.

In the early days of microarray analysis, very large sample sizes were considered essential due to concerns about the statistical issues involved in making thousands of simultaneous comparisons on a given sample. As the statistical theory developed for this issue, the numbers required have dropped dramatically. The consensus report of a major conference sponsored by all of the major government science agencies in 2006 concluded that an n=5 per group was sufficient for most microarray analyses [58]. Since that conference, numbers have dropped further. An appreciable number of microarray papers have been published in reputable journals based on studies in even smaller cohorts, including those in which n=3 per group (e.g., [5961]). The present study is well within the published 2006 guidelines.

Because different cell populations within a tissue may express particular genes at different rates, it was once considered important to fractionate tissues into purified cell populations before performing microarray analyses. This may still be useful in specific instances, but gene expression array studies have tended to move away from this since cell isolation procedures themselves can introduce major artifactual changes in gene expression. This has been especially well studied in whole blood vs peripheral blood mononuclear cell comparisons where changes in RNA expression due to experimental manipulation can seriously mask what was going on in vivo [62]. Many currently believe that cell fractionation is essential only in specific situations. We chose to perform the present studies in unfractionated adipose tissue samples. Theoretical concern that increased macrophage infiltration into obese fat might influence comparisons with results in nonobese fat seems very unlikely in a study such as this one in which the key results are downregulation of multiple biologically relevant genes in the obese fat samples. Furthermore, the expression ratio of 23 macrophage-specific genes in unfractionated obese vs nonobese omental fat averaged 1.1±0.06, strongly arguing against the possibility that the observed downregulation of DCI and related genes reflected macrophage infiltration.

Working Model of the Accumulation of Long-Chain Fatty Acids in Obese Adipocytes

Many laboratories have examined the balance between LCFA uptake, LCFA disposition, and cellular TG content in the liver as it relates to the pathogenesis of hepatic steatosis [18, 2224, 4446]. Similar considerations apply to adipocytes and the pathophysiology of obesity. In adipocytes of normal-weight subjects, allowing for meal-related diurnal variation, LCFA uptake (including facilitated transport) and degradation (including β-oxidation) or elimination (e.g., lipolysis) are in balance. Consequently, the net amount of fat in each cell is essentially constant over time, resulting in the relatively stable weights seen in nonobese subjects. In contrast, in adipocytes from obese subjects, the combination of increased facilitated LCFA uptake, shown in our uptake studies, and reduced β-oxidation, lipolysis, and LCFA metabolism, suggested by gene expression analysis, leads to accumulation of LCFAs and TG over time. The result of the observed changes in LCFA kinetics and expression of metabolic genes will be chronic accumulation of LCFA and TG, resulting in enlarged adipocytes and significant weight gain over time.

The strength of this study is the combination of the high-throughput first stage where thousands of genes are queried and the systems biology approach of the second stage where established pathways that contain a statistically significant number of genes with altered expression are identified. The combination of microarray analysis with extensive pathway filtering of identified genes is increasingly considered a state-of-the-art hypothesis-generating approach, which should guide future experimental validation. The promise for the future is that synergistic effects of drug cocktails whose components are aimed at various targets in these pathways might work at lower individual doses, with less toxicity, and in more patients than drugs that target only one of a pathway’s many gene components. To realize that goal, it will be necessary to do more such studies on larger sample sets from obese patients, including samples from different populations and classes of obesity. The interface between bariatric surgery and basic science may prove to be the optimal place to carry out this critically important translational research.


This study was supported by grants DK-52401 and DK-72526 from the National Institute of Diabetes, Digestive and Kidney Disease of the National Institutes of Health and by the Liver Disease Research Fund at Columbia University.

Contributor Information

José L. Walewski, Department of Medicine, Columbia University College of Physicians and Surgeons, New York, NY 10032, USA.

Fengxia Ge, Department of Medicine, Columbia University College of Physicians and Surgeons, New York, NY 10032, USA.

Michel Gagner, Department of Surgery, Weill Cornell Medical College, New York, NY 10065, USA.

William B. Inabnet, Department of Surgery, Columbia University College of Physicians and Surgeons, New York, NY 10032, USA.

Alfons Pomp, Department of Surgery, Weill Cornell Medical College, New York, NY 10065, USA.

Andrea D. Branch, Department of Medicine, Mount Sinai School of Medicine, New York, NY 10029, USA.

Paul D. Berk, Department of Medicine, Columbia University College of Physicians and Surgeons, New York, NY 10032, USA. Division of Digestive and Liver Diseases, Columbia University Medical Center, Russ Berrie Medical Science Pavilion, 1150 Saint Nicholas Avenue, Room 412, New York, NY 10032, USA.


1. Stremmel W, Berk PD. Hepatocellular influx of [14C]oleate reflects membrane transport rather than intracellular metabolism or binding. Proc Natl Acad Sci U S A. 1986;83(10):3086–90. [PubMed]
2. Schwieterman W, Sorrentino D, Potter BJ, et al. Uptake of oleate by isolated rat adipocytes is mediated by a 40-kDa plasma membrane fatty acid binding protein closely related to that in liver and gut. Proc Natl Acad Sci U S A. 1988;85(2):359–63. [PubMed]
3. Sorrentino D, Stump D, Potter BJ, et al. Oleate uptake by cardiac myocytes is carrier mediated and involves a 40-kD plasma membrane fatty acid binding protein similar to that in liver, adipose tissue, and gut. J Clin Invest. 1988;82(3):928–35. [PMC free article] [PubMed]
4. Stremmel W. Uptake of fatty acids by jejunal mucosal cells is mediated by a fatty acid binding membrane protein. J Clin Invest. 1988;82(6):2001–10. [PMC free article] [PubMed]
5. Nunes RMIL, Sorrentino D, Berk PD. Oleate uptake by isolated hepatocytes consists of two components, each driven by the unbound oleate concentration. Proceedings of the 3rd International Congress, Mathematical Modelling of Liver Excretory Process; Tokyo: Juntendo University Press; 1990. pp. 312–6.
6. Stump DD, Nunes RM, Sorrentino D, et al. Characteristics of oleate binding to liver plasma membranes and its uptake by isolated hepatocytes. J Hepatol. 1992;16(3):304–15. [PubMed]
7. Berk PD, Stump DD. Mechanisms of cellular uptake of long chain free fatty acids. Mol Cell Biochem. 1999;192(1–2):17–31. [PubMed]
8. Stump DD, Fan X, Berk PD. Oleic acid uptake and binding by rat adipocytes define dual pathways for cellular fatty acid uptake. J Lipid Res. 2001;42(4):509–20. [PubMed]
9. Abumrad NA, Perkins RC, Park JH, et al. Mechanism of long chain fatty acid permeation in the isolated adipocyte. J Biol Chem. 1981;256(17):9183–91. [PubMed]
10. Abumrad NA, Park JH, Park CR. Permeation of long-chain fatty acid into adipocytes. Kinetics, specificity, and evidence for involvement of a membrane protein. J Biol Chem. 1984;259(14):8945–53. [PubMed]
11. Glatz JF, van Nieuwenhoven FA, Luiken JJ, et al. Role of membrane-associated and cytoplasmic fatty acid-binding proteins in cellular fatty acid metabolism. Prostaglandins Leukot Essent Fatty Acids. 1997;57(4–5):373–8. [PubMed]
12. Luiken JJ, Turcotte LP, Bonen A. Protein-mediated palmitate uptake and expression of fatty acid transport proteins in heart giant vesicles. J Lipid Res. 1999;40(6):1007–16. [PubMed]
13. Luiken JJ, Glatz JF, Bonen A. Fatty acid transport proteins facilitate fatty acid uptake in skeletal muscle. Can J Appl Physiol. 2000;25(5):333–52. [PubMed]
14. Kampf JP, Kleinfeld AM. Fatty acid transport in adipocytes monitored by imaging intracellular free fatty acid levels. J Biol Chem. 2004;279(34):35775–80. [PubMed]
15. Kleinfeld AM, Kampf JP, Lechene C. Transport of 13C-oleate in adipocytes measured using multi imaging mass spectrometry. J Am Soc Mass Spectrom. 2004;15(11):1572–80. [PubMed]
16. Petrescu O, Fan X, Gentileschi P, et al. Long-chain fatty acid uptake is upregulated in omental adipocytes from patients undergoing bariatric surgery for obesity. Int J Obes (Lond) 2005;29(2):196–203. [PubMed]
17. Fan X, Bradbury MW, Berk PD. Leptin and insulin modulate nutrient partitioning and weight loss in ob/ob mice through regulation of long-chain fatty acid uptake by adipocytes. J Nutr. 2003;133(9):2707–15. [PubMed]
18. Bradbury MW, Berk PD. Lipid metabolism in hepatic steatosis. Clin Liver Dis. 2004;8(3):639–71. xi. [PubMed]
19. Berk PD, Zhou SL, Kiang CL, et al. Uptake of long chain free fatty acids is selectively up-regulated in adipocytes of Zucker rats with genetic obesity and non-insulin-dependent diabetes mellitus. J Biol Chem. 1997;272(13):8830–5. [PubMed]
20. Berk PD, Zhou SL, Bradbury M, et al. Regulated membrane transport of free fatty acids in adipocytes: role in obesity and non-insulin dependent diabetes mellitus. Trans Am Clin Climatol Assoc. 1997;108:26–40. [PubMed]
21. Berk PD, Zhou S, Kiang C, et al. Selective up-regulation of fatty acid uptake by adipocytes characterizes both genetic and diet-induced obesity in rodents. J Biol Chem. 1999;274(40):28626–31. [PubMed]
22. Berk PD. The master’s perspective: regulatable fatty acid transport mechanisms are central to the pathophysiology of obesity, fatty liver, & metabolic syndrome. Hepatology. 2008;48:1362–76. [PMC free article] [PubMed]
23. Verna EC, Berk PD. Role of fatty acids in the pathogenesis of obesity and fatty liver: impact of bariatric surgery. Semin Liver Dis. 2008;28(4):407–26. [PubMed]
24. Goldberg IJ, Ginsberg HN. Ins and outs modulating hepatic triglyceride and development of nonalcoholic fatty liver disease. Gastroenterology. 2006;130(4):1343–6. [PubMed]
25. Di Girolamo M, Mendlinger S, Fertig JW. A simple method to determine fat cell size and number in four mammalian species. Am J Physiol. 1971;221:850–8. [PubMed]
26. Wosilait WD, Nagy P. A method of computing drug distribution in plasma using stepwise association constants: clofibrate acid as an illustrative example. Comput Programs Biomed. 1976;6:142–8. [PubMed]
27. Spector AA, Fletcher JE, Ashbrook JD. Analysis of long-chain free fatty acid binding to bovine serum albumin by determination of stepwise equilibrium constants. Biochemistry. 1971;10:3229–32. [PubMed]
28. Bojesen IN, Bojesen E. Binding of arachidonate and oleate to bovine serum albumin. J Lipid Res. 1994;35:770–8. [PubMed]
29. Richieri GV, Anel A, Kleinfeld AM. Interactions of long-chain fatty acids and albumin: determination of free fatty acid levels using the fluorescent probe ADIFAB. Biochemistry. 1993;32:7574–80. [PubMed]
30. Rose H, Conventz M, Fischer Y, et al. Long-chain fatty acid-binding to albumin: re-evaluation with directly measured concentrations. Biochim Biophys Acta. 1994;1215:321–6. [PubMed]
31. Berman M, Weiss MF. Users’ manual for SAAM. Washington: US Government Printing Office; 1967.
32. SAAM II user guide. Seattle: SAAM Institute; 1998.
33. Sorrentino D, Robinson RB, Kiang CL, et al. At physiologic albumin/oleate concentrations oleate uptake by isolated hepatocytes, cardiac myocytes, and adipocytes is a saturable function of the unbound oleate concentration. Uptake kinetics are consistent with the conventional theory. J Clin Invest. 1989;84:1325–33. [PMC free article] [PubMed]
34. Snedecor GW, Cochran WG. Statistical methods. 6. Ames: Iowa State University Press; 1967.
35. Ramakrishnan R, Dorris D, Lublinsky A, et al. An assessment of Motorola CodeLink microarray performance for gene expression profiling applications. Nucleic Acids Res. 2002;30(7):e30. [PMC free article] [PubMed]
36. Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet. 2000;25(1):25–9. [PMC free article] [PubMed]
37. Kanehisa M, Goto S, Kawashima S, et al. The KEGG resource for deciphering the genome. Nucleic Acids Res. 2004;32:D277–80. [PMC free article] [PubMed]
38. Doniger SW, Salomonis N, Dahlquist KD, et al. MAPPFinder: using Gene Ontology and GenMAPP to create a global gene-expression profile from microarray data. Genome Biol. 2003;4(1):R7. [PMC free article] [PubMed]
39. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B. 1995;57:289–300.
40. Fried SK, Kral JG. Sex differences in regional distribution of fat cell size and lipoprotein lipase activity in morbidly obese patients. Int J Obes. 1987;11(2):129–40. [PubMed]
41. Edens NK, Fried SK, Kral JG, et al. In vitro lipid synthesis in human adipose tissue from three abdominal sites. Am J Physiol. 1993;265(3 Pt 1):E374–9. [PubMed]
42. Ostman J, Arner P, Engfeldt P, et al. Regional differences in the control of lipolysis in human adipose tissue. Metabolism. 1979;28 (12):1198–205. [PubMed]
43. Russell CD, Petersen RN, Rao SP, et al. Differences in adiponectin protein expression: effect of fat depots and type 2 diabetic status. Horm Metab Res. 2002;34:650–4. [PubMed]
44. Chirieac DV, Chirieac LR, Corsetti JP, et al. Glucose-stimulated insulin secretion suppresses hepatic triglyceride-rich lipoprotein and apoB production. Am J Physiol Endocrinol Metab. 2000;279 (5):E1003–11. [PubMed]
45. Diehl AM. Lessons from animal models of NASH. Hepatol Res. 2005;33(2):138–44. [PubMed]
46. Yamaguchi K, Yang L, McCall S, et al. Inhibiting triglyceride synthesis improves hepatic steatosis but exacerbates liver damage and fibrosis in obese mice with nonalcoholic steatohepatitis. Hepatology. 2007;45(6):1366–74. [PubMed]
47. Kampf JP, Parmley D, Kleinfeld AM. Free fatty acid transport across adipocytes is mediated by an unknown membrane protein pump. Am J Physiol Endocrinol Metab. 2007;293(5):E1207–14. [PubMed]
48. Su X, Abumrad NA. Cellular fatty acid uptake: a pathway under construction. Trends Endocrinol Metab. 2009;20(2):72–7. [PMC free article] [PubMed]
49. Garaulet M, Pérez-Llamas F, Pérez-Ayala M, et al. Site-specific differences in the fatty acid composition of abdominal adipose tissue in an obese population from a Mediterranean area: relation with dietary fatty acids, plasma lipid profile, serum insulin, and central obesity. Am J Clin Nutr. 2001;74:585–91. [PubMed]
50. Hsueh WC, Mitchell BD, Schneider JL, et al. Genomewide scan of obesity in the Old Order Amish. J Clin Endocrinol Metab. 2001;86:1199–205. [PubMed]
51. Freedman BI, Langefeld CD, Rich SS, et al. A genome scan for ESRD in black families enriched for nondiabetic nephropathy. J Am Soc Nephrol. 2004;15(10):2719–27. [PubMed]
52. Mustelin L, Pietiläinen KH, Rissanen A, et al. Acquired obesity and poor physical fitness impair expression of genes of mitochondrial oxidative phosphorylation in monozygotic twins discordant for obesity. Am J Physiol Endocrinol Metab. 2008;295(1):E148–54. [PubMed]
53. Patti ME, Butte AJ, Crunkhorn S, et al. Coordinated reduction of genes of oxidative metabolism in humans with insulin resistance and diabetes: potential role of PGC1 and NRF1. Proc Natl Acad Sci U S A. 2003;100(14):8466–71. [PubMed]
54. Kim K, Perroud B, Espinal G, et al. Genes and networks expressed in perioperative omental adipose tissue are correlated with weight loss from Roux-en-Y gastric bypass. Int J Obes. 2008;32:1395–406. [PubMed]
55. Gillingham MB, Purnell JQ, Jordan J, et al. Effects of higher dietary protein intake on energy balance and metabolic control in children with long-chain 3-hydroxy acyl-CoA dehydrogenase (LCHAD) or trifunctional protein (TFP) deficiency. Mol Genet Metab. 2007;90(1):64–9. [PMC free article] [PubMed]
56. Buijs MM, Burggraaf J, Wijbrandts C, et al. Blunted lipolytic response to fasting in abdominally obese women: evidence for involvement of hyposomatotropism. Am J Clin Nutr. 2003;77:544–50. [PubMed]
57. Langin D, Dicker A, Tavernier G, et al. Adipocyte lipases and defect of lipolysis in human obesity. Diabetes. 2005;54(11):3190–7. [PubMed]
58. Shi L, Reid LH, Jones WD, et al. The MicroArray Quality Control (MAQC) project shows inter- and intraplatform reproducibility of gene expression measurements. Nat Biotechnol. 2006;24:1151–61. [PMC free article] [PubMed]
59. Bardag-Gorce F, Oliva J, Dedes J, et al. Chronic ethanol feeding alters hepatocyte memory which is not altered by acute feeding. Alcohol Clin Exp Res. 2009;33(4):684–92. [PMC free article] [PubMed]
60. Trentin L, Giordan M, Dingermann T, et al. Two independent gene signatures in pediatric t(4;11) acute lymphoblastic leukemia patients. Eur J Haematol. 2009 in press. [PubMed]
61. Ukena SN, Koenecke C, Geffers R, et al. T helper type 2 differentiation is associated with induction of antibacterial defense mechanisms in blood lymphocytes of patients with sarcoidosis. Immunol Invest. 2009;38(1):49–66. [PubMed]
62. Kennedy L, Vass JK, Haggart DR, et al. Hematopoietic lineage transcriptome stability and representation in PAXgene collected peripheral blood utilising SPIA single-stranded cDNA probes for microarray. Biomarker Insights. 2008;3:403–17. [PMC free article] [PubMed]