点击显示 收起
1 From the Department of Medicine, Karolinska Institutet, Karolinska University Hospital, Huddinge, Stockholm, Sweden (ID, KL, EAN, IA, and PA); the Department of Molecular Medicine, Karolinska Institutet, Karolinska University Hospital, Solna, Stockholm, Sweden (KL); the Department of Biosciences, Karolinska Institutet, Novum, Stockholm, Sweden (JL); and the Institute of Preventive Medicine, Danish Epidemiology Science Centre, Copenhagen University Hospital, Copenhagen, Denmark (CV and TIAS)
2 NUGENOB is the acronym for the project "Nutrient-Gene Interactions in Human Obesity—Implications for Dietary Guidelines." Internet: www.nugenob.org. 3 Supported by grants from the Swedish Research Council, Swedish Diabetes Association, Swedish Heart and Lung Foundation, Novo Nordic Foundation, Swedish Society for Medical Research, Tore Nilsson Foundation, Swedish Medical Society, Swedish Cancer Society, Diabetes Research and Wellness Foundation, and the European Community. JL at the Affymetrix core facility was supported by the Wallenberg Consortium North. 4 Reprints not available. Address correspondence to P Arner, Karolinska Institutet, Karolinska University Hospital, Huddinge, M63, SE-141 86 Stockholm, Sweden. E-mail: peter.arner{at}medhs.ki.se.
ABSTRACT
Background: The effect of energy restriction and macronutrient composition on gene expression in adipose tissue is not well defined.
Objective: The aim of the study was to investigate the effect of different low-energy diets on gene expression in human adipose tissue.
Design: Forty obese women were randomly assigned to a moderate-fat, moderate-carbohydrate diet or a low-fat, high-carbohydrate hypoenergetic (–600 kcal/d) diet for 10 wk. Subcutaneous adipose tissue samples were obtained before and after the diet period. High-quality RNA samples were obtained from 23 women at both time points, and these samples were hybridized to microarrays containing the 8500 most extensively described human genes. The results were confirmed by separate messenger RNA measurements.
Results: Both diets resulted in weight losses of 7.5% of baseline body weight. A total of 52 genes were significantly up-regulated and 44 were down-regulated as a result of the intervention, and no diet-specific effect was observed. No major effect on lipid-specific transcription factors or genes regulating signal transduction, lipolysis, or synthesis of acylglycerols was observed. Most changes were modest (<25% of baseline), but all genes regulating the formation of polyunsaturated fatty acids from acetyl-CoA and malonyl-CoA were markedly down-regulated (35–60% decrease).
Conclusions: Macronutrients have a secondary role in changes in adipocyte gene expression after energy-restricted diets. The most striking alteration after energy restriction is a coordinated reduction in the expression of genes regulating the production of polyunsaturated fatty acids.
Key Words: Adipose tissue • mRNA • gene • hypoenergetic diet • macronutrient • obesity
INTRODUCTION
Energy-restricted diets have a central role in reducing the fat mass of obese subjects (1). The major metabolic or signaling pathways that regulate lipid depletion of fat cells in response to energy restriction are unknown. However, signal transduction pathways involved in lipolysis and acylglycerol synthesis, enzymes regulating lipolysis or lipogenesis, and lipid-specific transcription factors could be involved (2). Although the macronutrient content of food may have marked effects at the cellular level, it remains to be established to what extent the macronutrient composition and the energy deficit per se contribute to the lipid depletion of fat cells after hypoenergetic diets (3).
Gene expression profiling with the use of microarrays is a powerful tool for identifying the molecular pathways responsible for metabolic regulation and was recently used to detect new important signaling pathways involved in glucose metabolism in insulin-resistant human skeletal muscle (4, 5). The role of diets in gene expression profiles in adipose tissue remains to be established. From a clinical point of view, human studies are of interest because the regulation of function of human and rodent fat cells differs in many aspects (6).
In the present study, we compared the effect of 2 energy-restricted diets that differed in macronutrient composition but not in energy content. The focus of the study was the gene expression profile in human subcutaneous adipose tissue, which is the largest adipose depot in the body (7). We evaluated the role of adipose tissue–expressed genes and pathways in controlling fat loss in response to energy-restricted diets by hybridizing individual adipose tissue complementary DNA (cDNA) samples obtained before and after 10 wk of energy restriction to arrays containing >8500 of the best described human transcripts.
SUBJECTS AND METHODS
Subjects
The women participated in a European multicenter study termed Nutrition, Genes, and Obesity (NUGENOB; Internet: www.nugenob.org) within which the interaction between hypoenergetic diets and genes is examined. The present study was performed at the local center at Karolinska University Hospital, Huddinge, Sweden, and included the first 40 women out of the 100 subjects (men and women) recruited at the hospital. There was no overlap between these 40 subjects and those investigated by Viguerie et al (8) as part of the NUGENOB project. All women were obese according to World Health Organization criteria [body mass index (BMI, in kg/m2) of 31–48] and were aged between 21 and 49 y. The women were otherwise healthy and did not take regular medication, except for one woman who was treated with thyroid hormone for goiter (treatment was not changed during the investigation). All women except one were Scandinavian. Five women were postmenopausal. The results were not altered in an important way when the postmenopausal women were excluded from analysis. The 40 women were randomly assigned to either a low-fat, high-carbohydrate diet (n = 20) or a moderate-fat, moderate-carbohydrate diet (n = 20) for 10 wk. The study was approved by the committee on ethics of the hospital and was explained in detail to each participant. Informed consent was obtained.
Study design
Both diets provided 600 kcal less than individually estimated energy requirements based on resting energy expenditure, which were measured by use of a ventilated hood system at baseline and multiplied by a factor for physical activity level (1.3). The target macronutrient compositions of the 2 diets were as follows: for the low-fat diet, 20–25% of total energy from fat, 15% from protein, and 60–65% from carbohydrate, and for the moderate-fat diet, 40–45% of total energy from fat, 15% from protein, and 40–45% from carbohydrate. The randomization was done at the coordinating center of the NUGENOB project in Copenhagen. The subjects were given instructions relating to the dietary targets based on an education system consisting of isoenergetic interchangeable units (9) and were requested to abstain from alcohol consumption. Dietary instructions were reinforced weekly. A 3-d weighed-food record of 2 weekdays and 1 weekend day was performed before the study and during the last week of intervention. One-day weighed-food records were completed in the 2nd, 5th, and 7th weeks. The dietary records were analyzed by using a food-nutrient database that contained all these data.
The study examinations were performed before and after 10 wk of dieting at 0700–0800 after the subjects had fasted overnight. Body height and weight were measured on calibrated scales. A venous blood sample was obtained for analysis of plasma insulin by double-antibody radioimmunoassay (Insulin RIA 100; Kabi-Pharmacia, Uppsala, Sweden) and plasma glucose by an enzymatic spectrophotometric technique with an autoanalyzer (COBAS FARA; Roche Diagnostics, Basel, Switzerland). Thereafter, an abdominal subcutaneous fat specimen (1.5–2 g) was obtained by needle aspiration under local anesthesia as described previously (10).
Adipose tissue
Two pieces (1 g and 300 mg) of adipose tissue were frozen in separate portions in liquid nitrogen and stored at –70°C for subsequent microarray and real-time quantitative polymerase chain reaction (RT-qPCR) measurements, respectively. The remaining tissue was immediately used for collagenase isolation and determination of fat cell volume of isolated fat cells as described previously (10).
RNA extraction
Frozen subcutaneous adipose tissue from each subject was crushed under liquid nitrogen, and total RNA was prepared by using the RNeasy mini kit (Qiagen GmbH, Hilden, Germany). RNA samples for RT-qPCR were treated with RNase-free DNase (Qiagen GmbH). The RNA concentration was measured spectrophotometrically. The quality of the RNA was determined by using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Only samples containing completely pure RNA were used in the microarray or RT-qPCR analysis.
Microarray
A prerequisite for total RNA samples to be included in the microarray was a total yield of 8 µg RNA per sample and a nondegraded pattern in the Agilent analysis of RNA quality described above. Because of this, we performed a microarray for only 13 women in the moderate-fat diet group and 10 women in the low-fat diet group who were examined before and after the diet period. This represents 60% of the total number of subjects. According to our previous experience with human adipose tissue, 1 g of sample yields 8 µg of high-quality RNA in 70% of the cases. Therefore, the expected success fraction of 2 separate samples from the same individual is 50%, which corresponds with the results obtained. Microarray hybridization was performed by using tools obtained from Affymetrix (Affymetrix Inc, Santa Clara, CA). Eight micrograms of total RNA per subject was used in the standard protocol from Affymetrix to label targets. These targets (biotinylated complementary RNA) were hybridized to the Human Genome Focus arrays from Affymetrix containing probe sets for 8793 transcripts including controls. Thus, a total of 46 array hybridizations were performed. After probing and scanning (using the standard protocols from Affymetrix), signal values from Affymetrix software MAS 5.0 were further analyzed by using the methods described below. Quality control was performed and fulfilled the criteria for chip hybridization suggested by the Tumor Analysis Best Practices Working Group (11).
Verification with real-time quantitative PCR
From each RNA sample, 1 µg was reverse transcribed to cDNA by using the Omniscript RT kit (Qiagen GmbH) and oligo(dT) primers (Invitrogen, Tåstrup, Denmark). In a final volume of 25 µL, 5 ng of cDNA was mixed with 2X SYBR Green PCR master mix (Eurogentec SA, Ougrée, Belgium), fluorescein (Bio-Rad Laboratories, Sundbyberg, Sweden), and gene-specific primers (Invitrogen, Tåstrup, Denmark), which are specified in Table 1. The primer pairs were designed to reach over exon-intron boundaries. They were selected to yield a single amplicon on the basis of dissociation curves and analysis by agarose gel electrophoresis. RT-qPCR was performed in an iCycler IQ (Bio-Rad Laboratories Inc, Hercules, CA). The thermal cycler details were as follows: 10 min at 95°C followed by amplification of the cDNA for 40 cycles with melting for 20 s at 95°C, annealing for 20 s at 55–60°C depending on the primer pair, and elongation for 20 s at 72°C. All samples were run in triplicate.
View this table:
TABLE 1. Primers for real-time quantitative polymerase chain reaction1
The cycle threshold (Ct) values corresponding to 0.2 ng/µL in a standard curve for the target genes and the reference gene [GAPDH (glyceraldehyde-3-phosphate dehydrogenase)] were used as "calibrators" and were subtracted from all other Ct values for the target genes and reference gene. The Ct values were then normalized by using GAPDH as an internal standard. The CV for GAPDH Ct values was 5–6% before and after both diets, and the Ct values for GAPDH showed a <1% increase after the diet. Therefore, GAPDH was considered to be a valid reference gene in these experiments. Samples for mRNA analysis from the subjects involved in the microarray study were available before and after diet for 21 cases (termed RNA1). Corresponding samples from those not involved in the microarray study were available for 15 cases, termed RNA2. In all, 36 of the 40 women were subject to mRNA quantification.
Expression of 4 genes was measured by using TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA) and the following TaqMan Gene Expression Assays: uroporphyrinogen III synthase (UROS) (Hs00958459_m1); aminolevulinate, delta-, dehydratase (ALAD) (Hs00765604_m1); aminolevulinate, delta-, synthase 1 (ALAS1) (Hs00167441_m1); and hydroxymethylbilane synthase (HMBS) (Hs00609293_g1). GAPDH was used as the reference gene and mRNA amounts were quantified by using the formula above, except that for each gene the sample with the highest Ct values was used as the calibrator.
Analysis of microarray gene expression data
From a statistical point of view, there is a risk of false-positive and false-negative results because of multiple comparisons. Diet-induced changes in the expression of individual genes were analyzed by using the significance analysis of microarrays method (SAM), and changes in the expression of whole biological pathways were analyzed by using MAPPFINDER (12, 13). SAM adjusts for the multiple comparisons caused by the presence of thousands of genes on the microarrays. The software assigns a score to each gene on the basis of change in gene expression relative to the SD of repeated measurements (12). For genes with scores greater than an adjustable threshold, SAM uses permutations of the repeated measurements to estimate the false discovery rate. In the joint analysis of all 23 samples subjected to array, and in the moderate-fat subgroup, the number of expected false-positive up- or down-regulated genes was set to 2 according to the SAM criterion. In the low-fat subgroup, it was set to 3. Five hundred permutations were used in the numerical analysis. Paired comparison was applied, ie, we analyzed gene expression for each individual before and after intervention. Signal values from all transcripts on the microarrays were used in the analysis.
The MAPPFINDER software was used to rank the impact of the GenMAPP pathways as well as Gene Ontology biological function terms during energy restriction (13-15). Briefly, MAPPFINDER assigns to each analyzed pathway, or gene set, a z score that is based on the percentage of genes in each pathway or set that meet a user-defined criterion for change in expression. Criteria for change were significant gene expression present in 26 samples, paired t test (two-sided) for change in expression of an individual gene P 0.05, and mean signal fold change >10% from baseline. Ten percent fold change was selected because this was the lowest change considered significant by SAM.
Hierarchical clustering of genes that changed their expression during the 10-wk period was performed by using GENE WEAVER software (Affymetrix Inc). MICROARRAY SUITE 5 (Affymetrix) was used to select genes for this analysis.
Statistical methods
Values are means ± SDs. Student’s unpaired t test (two-sided) was used to compare clinical, adipose tissue, and blood chemistry phenotypes, as well as dietary intake between groups. The same test was used to determine whether expression of individual genes differed between groups. The one-sample t test (two-sided) was applied to test whether changes in clinical and blood chemistry phenotypes differed from zero. Paired t tests (one- or two-sided) were applied to determine whether dietary intake and expression of individual genes changed from baseline during the intervention. The Kolmogorov-Smirnov and Shapiro-Wilk tests ensured normally distributed variables. Clinical phenotypes were analyzed by using SPSS software version 11.0 (SPSS Inc, Chicago, IL). The expression of individual genes was analyzed by using STATVIEW version 5.0.1 (SAS Institute Inc, Cary, NC).
RESULTS
Clinical findings
All women completed the dietary intervention. Of the 40 obese women randomly assigned to a diet group, we performed microarray experiments on 10 women in the low-fat, high-carbohydrate group and 13 women in the moderate-fat, moderate-carbohydrate group (see the "Microarray" section in Subjects and Methods). The clinical data on this subset of individuals are reported. BMI, body weight, fat cell size, age, and fasting plasma concentrations of insulin and glucose at entry and the changes in these variables during the intervention did not differ significantly between the groups (Table 2). The decrease in body weight was 7.5% in both groups. There was no significant difference in self-reported, baseline dietary intake between the 2 groups of 10 and 13 women. The total amount of energy from fat during the interventions was within the targeted 40–45% in the moderate-fat group (42 ± 3%) and close to the targeted 20–25% in the low-fat group (28 ± 4%; P = 0.07 for the change in fat as a percentage of energy from baseline). During the intervention, carbohydrate, protein, and fiber intakes were higher in the low-fat group than in the moderate-fat group. Values were 51 ± 4% and 40 ± 3% of energy (P < 0.001), 21 ± 1% and 19 ± 1% of energy (P < 0.001), and 19 ± 4 and 13 ± 2 g/d (P < 0.001), respectively. The reduction in energy intake did not differ significantly between the 2 groups (556 ± 183 and 503 ± 161 kcal/d for the low-fat and moderate-fat diet groups, respectively). The ratio of saturated to monounsaturated to polyunsaturated fatty acids was 2:2:1 in the habitual diet and in the 2 intervention diets. Self-reported dietary intake corresponded with 1-d measured intake in the 2nd, 5th, and 7th weeks of the study (values not shown).
View this table:
TABLE 2. Baseline and 10-wk measurements in women participating in the microarray study1
Clinical variables were also investigated in all 40 subjects. The initial clinical profile, dietary data, and the changes in BMI, body weight, body fat, fat cell volume, and plasma concentrations of glucose and insulin in response to the intervention were not significantly different between the whole group and the subgroup of subjects involved in the microarray experiment (values not shown). Thus, the 2 diets were equally effective in the whole group as in the subjects participating in the microarray experiments. We therefore concluded that the latter subgroup was representative of all women undergoing dietary treatment.
Gene expression by microarray
In total, 3746 genes were present on >23, ie, one-half, of the arrays according to MICROARRAY SUITE criteria. Genes specifically expressed in the immune system, such as CD4, CD8, immunoglobulins, and co-stimulatory molecules, whose expression would indicate contamination with blood vessel tissue, were not detected on the arrays. Ninety-six genes with a statistically significant change in gene expression in either subgroup or the pooled sample were identified, excluding 6 genes that gave such weak specific signals on the arrays that they were scored as absent on 40–46 microarrays by the software. Thus, a significant change in expression was observed for 2.5% of the genes present in more than one-half of the analyzed samples. Fold change (expression after diet divided by before diet) among these genes varied in the up-regulated group between 1.10 and 1.46 and in the down-regulated group between 0.35 and 0.90, respectively. These genes are depicted in Table 3 (52 genes up-regulated) and Table 4 (44 genes down-regulated). The expected number of false positives was 2 genes out of 84 significant genes in total in the whole population; 2 genes out of 29 significant genes in the moderate-fat, moderate-carbohydrate subgroup; and 3 genes out of 10 significant genes in the low-fat, high-carbohydrate subgroup.
View this table:
TABLE 3. Genes up-regulated by diet1
View this table:
TABLE 4. Genes down-regulated by diet1
The pattern of response to the 2 diets was almost identical. Thus, expression of the same genes was up- or down-regulated in response to both the moderate- and the low-fat diet except for 3 genes. For the latter, however, no significant difference in fold change between diet groups was observed. In particular, the quantitatively most up-regulated (ratio >1.25) or down-regulated (ratio <0.75) genes were essentially the same for the 2 diets. Therefore, all subsequent analyses were done on data from both groups combined.
Among the genes appearing to be up- or down-regulated by the energy restriction, none encoded lipolytic enzymes or enzymes involved in acylglycerol formation, lipid-specific gene transcription, or signal transduction systems related to lipolysis or lipogenesis except for the p55 regulatory subunit of phosphoinositide-3-kinase (PIK3R3). The ratio of PIK3R3 expression after compared with before was 0.76. PIK3R3 is involved in insulin signaling (16). The expression of 4 genes known to be of importance in the regulation of body fat stores according to studies in knock-out mice were significantly altered. Both estrogen receptor-1 and metallothionein-1 null mice develop obesity (17-19); these genes were up-regulated in response to energy restriction (ratios of 1.26 and 1.29, respectively). Loss of stearoyl-CoA desaturase-1 (SCD1) or cell-death-inducing DFFA-like effector A (CIDEA) protects mice against obesity (20, 21); these genes were down- and up-regulated, respectively, in response to energy restriction (ratios of 0.55 and 1.46, respectively). Transferrin, which was recently implicated in obesity in a mouse model (22), was down-regulated by energy restriction (ratio of 0.80).
Individual mRNA measurements
To confirm the microarray results, we performed individual mRNA measurements by using RT-qPCR for 7 of the genes that showed a marked change in expression after dieting. We studied samples from the subjects who were included (RNA1) and those who were not included (RNA2) in the array separately and combined (Table 5). In all groups, the results were the same. The results for the RNA2 samples were confirmed with repeated cDNA syntheses with almost identical results (values not shown). We were able to confirm the change in mRNA expression for all except one gene, ß-1,4-galactosyltransferase 6 (B4GALT6). Therefore, we measured B4GALT6 in only one RNA cohort. The inability to confirm the change in expression for B4GALT6 was not surprising because the SAM analysis assumed a few false-positive significant changes in gene expression. A careful examination of the microarray data showed that, in most cases, the signal values for B4GALT6 were low. This was not the case for the other genes in Table 5. Thus, the microarray data with B4GALT6 might be uncertain.
View this table:
TABLE 5. Individual messenger RNA measurements
Pathway, biological function, and cluster analysis
All 23 individuals were analyzed together because no diet-specific effect on gene expression was observed in the results obtained from the SAM analysis. A total of 330 genes met the MAPPFINDER criteria, and of those, 81 were not linked to a map. Thus, only 249 genes could be used to calculate the results. The z score was based on analysis of 6113 of the Human Genome Focus array genes annotated in the MAPPFINDER database. Pathways or Gene Ontology biological function terms with z scores >3 are depicted in Table 6. The most affected terms were heme biosynthesis and porphyrin biosynthesis. The same 4 genes met the criteria in both terms: ALAD, ALAS1, HMBS, and UROS. However, none of these genes showed up as being significantly changed in their expression in the SAM analysis. Expression of these 4 genes was quantified by using RT-qPCR in all 23 individuals used in the pathway analysis. A significant difference in gene expression was obtained for HMBS only [fold change after compared with before low-energy diet: 0.91 ± 0.24 arbitrary units (AU), P = 0.036]. For the other 3 genes, the fold changes were as follows: ALAD, 1.10 ± 0.63 AU; ALAS1, 1.06 ± 0.26 AU; and UROS, 1.07 ± 0.17 AU. Because only the change in HMBS expression was significant in the individual RT-qPCR measurements, and no fold change was >10%, we did not analyze the heme or porphyrin biosynthesis pathways further.
View this table:
TABLE 6. MAPPFINDER analysis of pathways and biological function terms1
Next in rank were the Gene Ontology biological function terms fatty acid metabolism, carboxylic acid metabolism 2, organic acid metabolism 2, and metabolism. A detailed analysis of these terms showed that their rank in each case was above all caused by genes in the fatty acid degradation and fatty acid synthesis subgroups meeting the predefined criterion for change in expression. The effect on fatty acid turnover was further supported by the observation that fatty acid degradation and synthesis were the highest ranked GenMAPP pathways, next to the heme biosynthesis pathway.
Regarding fatty acid degradation, 6 genes formed a pathway from triacylglycerol breakdown by lipoprotein lipase to production of acyl-CoA. These were long-chain fatty acid CoA ligase 1 and 2, carnitine palmitoyl transferase 1, carnitine-acyl carnitine translocase, and long-chain acyl-CoA dehydrogenase (13–26% change from baseline values). However, this pathway is uncertain because none of the genes were among those showing a significant change in the SAM analysis presented in Tables 3 and 4. Two more genes from Tables 3 and 4 could be added to the pathway, namely, acyl-CoA binding protein (ratio 0.83) and malic enzyme 1 (ratio 0.75), which play important roles in acyl-CoA transport between the mitochondria and the cytoplasm (20).
Fatty acid synthesis according to GenMAPP contained several genes fulfilling the SAM criterion. Therefore, fatty acid synthesis was investigated in detail, as shown in Figure 1. All relevant genes involved in the final steps of synthesis of polyunsaturated fatty acids were markedly down-regulated after dieting. The pathway started with fatty acid synthase (FASN; ratio of 0.64), which catalyzes the formation of unsaturated fatty acids from malonyl-CoA and acetyl-CoA (23). The next step in this pathway is desaturation to monounsaturated fatty acids, which is catalyzed by stearoyl-CoA desaturase-1 (ratio of 0.55) (24). The final step is desaturation to polyunsaturated fatty acids, which is catalyzed by fatty acid desaturase 1 (FADS1) and 2 (FADS2) (ratios of 0.57 and 0.40, respectively) (25). It is noteworthy that these genes were 4 of the 7 most down-regulated genes in the microarray, all of which showed a significant change. GENE WEAVER produced no obvious clusters, probably because of the small sample size and the rather small number of genes undergoing a change in expression.
FIGURE 1.. A metabolic pathway markedly down-regulated after the energy-restricted diet according to microarray gene expression analysis of human subcutaneous adipose tissue. Values within parentheses indicate the mean (±SD) ratio of gene expression after compared with before the energy-restricted diet. CoA, coenzyme A.
DISCUSSION
Energy-restricted diets are commonly used to treat obesity. The effects of such diets on gene expression are not well known. We therefore examined the gene expression profile of subcutaneous adipose tissue in women who were randomly allocated to either of 2 low-energy diets with very different carbohydrate and fat contents but similar total energy contents. With arrays each covering 8800 of the best-described human genes, we performed an individual analysis of 23 women before and after the diet and were able to confirm the validity of these arrays with individual RT-qPCR measurements.
Twenty women from each of the 2 dietary interventions were recruited, but for technical reasons, we performed the microarray hybridizations for only 13 subjects in the moderate-fat, moderate-carbohydrate diet group and 10 subjects in the low-fat, high-carbohydrate diet group. There is no reason to believe that the women participating in the array were not representative of the whole group. First, clinical data were essentially the same in the overall group and in the subgroup. Second, RT-qPCR measurements gave the same type of results in the overall group and the subgroup.
The expression of a limited number (2.5%) of the genes in adipose tissue was changed in response to the energy restriction; the changes in expression varied from 50% up-regulation to 60% down-regulation. Only a few genes showed a change in expression that deviated >25% from the baseline value. This moderate effect on gene expression was expected because the energy restriction was mild (600 kcal deficit/d), weight loss was moderate (7.5% of initial weight), and the treatment period rather long (10 wk), which would allow for compensatory effects. It is thus possible that more marked effects could have been obtained with short-term treatment using very low calorie diets. This notion is supported by a recent microarray study of the effect of short-term, very-low-calorie diets on gene expression in subcutaneous adipose tissue of obese subjects (26). The changes in expression were markedly different from our findings. There was, above all, a variation in inflammatory genes.
The dietary data suggest that, in general, the subjects followed the food instructions closely. Self-reported dietary intake corresponds well with repeated one-day food recordings. Thus, major differences in dietary intake of lipids and carbohydrates were observed after the 2 types of intervention despite a similar total energy intake. Nevertheless, weight loss and decrease in fat cell volume induced by the intervention were on the same order of magnitude for both diets, which suggests a superior role of energy restriction for loss of lipids in fat cells in the studied women. Concerning the role of macronutrients, there was no important difference in dietary-induced changes in gene expression when the moderate-fat, moderate-carbohydrate and low-fat, high-carbohydrate diets were compared. This indicates that adipose gene expression is also influenced by the energy deficit and not the macronutrient composition of the food, at least not the fat and carbohydrate contents. Our results agree with those of Viguerie et al (8), who, for 38 investigated genes, observed no significant differences in response between low-fat, high-carbohydrate and moderate-fat, moderate-carbohydrate diets. It should be emphasized that the observed changes in mRNA expression in our study in general were modest and that a relatively small cohort was investigated. Therefore, type 2 statistical errors may have masked small differences in results between the diets. However, it is likely that such differences only relate to the magnitude of changes in gene expression and not to different expression patterns. Only 3 genes on the arrays were regulated differently by the 2 diets.
Only 2 of the genes regulated by energy restriction in our study overlapped with the 38 genes investigated by Viguerie et al (8): FASN and CIDEA, which were not regulated by diet in their study. This discrepancy is not surprising because the genetic and cultural background of the investigated subjects differed between the studies.
An unexpected finding was the lack of effect of energy restriction on common genes involved in the regulation of lipolysis or acylglycerol synthesis or signaling pathways, such as G-protein, MAP kinase, insulin signal transduction transcription factors, and lipid-specific transcription factors. The expression of PIK3R3 was 25% decreased. Although this gene is important for insulin signal transduction, it is unlikely that a small change in just one step of a complex signaling pathway has an important regulatory effect (16). As discussed, several coordinate changes in gene expression of a metabolic or signaling pathway should occur before the pathway is considered to have an important regulatory function (4, 5). The genes involved in the synthesis of polyunsaturated fatty acids fulfilled this criterion. Thus, the expression of all genes involved in the transformation of acetyl-CoA and malonyl-CoA to polyunsaturated fatty acids was decreased during energy restriction. These genes (fatty acid synthase, stearoyl-CoA desaturase 1, and fatty acid desaturase 1 and 2) were 4 of the 7 most down-regulated genes on the microarray (35–60% reduction from baseline). A third isoform of fatty acid desaturase is cloned in man (23). The functional role of this gene is not known and it was not included in the Affymetrix chip. The pathway of inhibited production of fatty acids could also be further expanded by including the somewhat uncertain findings with decreased expression of genes involved in fatty acid degradation (GenMAPP findings in Table 6 and the additional findings with acyl-CoA binding protein and malic enzyme), because these events will lead to decreased formation of acetyl-CoA.
The present data with pathway analysis strongly suggest an important role of genes involved in the production of polyunsaturated fatty acids in the regulation of lipid loss in human fat cells during energy-restricted dieting. According to the results of animal experiments, these fatty acids have a variety of effects in fat cells, including gene transcription, metabolism, cellular membrane composition, adipocyte differentiation, and signal transduction (27-29). In this respect, it is important to note that human fat cells have a significant capacity to synthesize fatty acids from glucose, although the rate of synthesis seems to be much lower than that in rodents (30). The exact role of polyunsaturated fatty acids in human adipocyte function is not known, but we can speculate. The coordinated decrease in expression during energy restriction might be of importance for depletion of adipocyte lipids. The whole pathway of genes forming polyunsaturated fatty acids from acetyl- or malonyl-CoA could also be a negative feedback regulator of obesity. Polyunsaturated fatty acids in high concentration might inhibit adipocyte differentiation or lipid accumulation. If so, decreased production could be a way to counteract adipocyte depletion during weight loss. The knockout data with stearoyl-CoA desaturase-1 support the latter idea (20). It is unlikely that a decreased supply of fatty acids to acylglycerol synthesis is a major cause of the alterations in gene expression. Polyunsaturated fatty acids except linoleic acid contribute marginally to the total fatty acids in human adipose tissue, and the linoleic acid content seems largely dependent on dietary supply (31).
Fat cells can synthesize several monounsaturated and polyunsaturated fatty acids (27, 28, 31). At present, we do not know which unsaturated fatty acids play a regulatory role during energy restriction. Unfortunately, the amount of adipose tissue available was far too small for detailed studies on these lipids.
According to Gene Ontology, the pathway most affected by energy restriction was heme-porphyrin biosynthesis. The 4 genes contributing to this result, ALAD, ALAS1, HMBS, and UROS, displayed small (<10%) differences in gene expression by RT-qPCR, and the results were significant for HMBS only. Furthermore, there was no uniform picture: one gene was down-regulated and 3 genes were up-regulated by the low-energy diet. Thus, there is not convincing support that these pathways are involved in body weight regulation. MAPPFINDER pathway or biological term analysis is above all valuable for grouping large numbers of analyzed genes and ranking the relative importance of these groups. It was previously shown that additional information can be gained if the pathway analysis includes genes that demonstrate minor, nonsignificant changes in gene expression between groups (4). We therefore used less stringent criteria for change in gene expression and inclusion in pathways analysis than that applied in SAM when determining the 96 genes with a significant difference in gene expression on the microarrays. As it turned out, the criteria applied in the MAPPFINDER pathway analysis lead to the inclusion of genes where a change in expression could not be confirmed by RT-qPCR and to unreliable results, eg, heme-porphyrin biosynthesis. Thus, in our experiment, when evaluating the MAPPFINDER results, one needs to consider to what extent individual pathway genes are on the SAM list, which seems to select most of the genes affected by dieting.
As mentioned earlier, the expression of some genes known to be important for body weight regulation in rodents was altered by energy restriction, ie, estrogen receptor 1, metallothionein 1, CIDEA, and transferrin. This finding highlights the critical importance of mouse models to the discovery of candidate regulatory genes for human obesity. Recently, obesity was shown to be associated with changes in the expression of several inflammatory genes of adipose tissue, which occur both in adipocytes and in stromal cells of adipose tissue (32). Note, however, that we found no consistent change in expression of inflammatory genes after diet.
Another finding in the present study was that leptin gene expression was not down-regulated according to the array. These results may well reflect the fact that low levels of transcripts are not always detected as being present in the RNA preparation when using MAS 5.0 from Affymetrix. In separate RT-qPCR experiments on adipose tissue from the present cohort, however, we found that leptin mRNA was significantly decreased by 20% after diet (33).
This study was conducted on abdominal subcutaneous adipose tissue from healthy obese women. We do not know at present how the findings relate to men, other fat depots, or obesity with comorbidity.
In conclusion, it appears that macronutrients are of no or little importance for changes in gene expression in human adipose tissue of obese women after energy restriction. A moderate energy restriction has no important effect on adipocyte genes involved in the regulation of acylglycerol turnover. However, a marked effect was seen on the genes regulating the production of polyunsaturated fatty acids and on the genes shown to regulated obesity in experimental models.
ACKNOWLEDGMENTS
We thank Eva Sjölin, Katarina Hertel, Britt-Marie Leijonhufvud, Maria Johansson, and Kerstin Wåhlén for excellent technical assistance; Gun Åberg for superior dietary monitoring; and Karin Dahlman-Wright for valuable advice about estrogen receptor signal transduction.
ID and KL contributed equally to the study. PA was responsible for the study and wrote, together with ID and KL, the first version of the manuscript. ID, KL, and JL performed and evaluated the microarray. EAN performed the mRNA measurements. IA was responsible for the dietary treatment. CV and TIAS were responsible for the NUGENOB protocol. All authors contributed to the writing of the manuscript. The partners of the NUGENOB project are listed on the project’s Web site (Internet: www.nugenob.org). None of the authors of this manuscript had a relation with a company or organization that could benefit financially from the publication of the data in this manuscript.
REFERENCES