Impact of estrogen receptor gene polymorphisms and mRNA levels on obesity and lipolysis – a cohort study

Background The estrogen receptors α and β (ESR1, ESR2) have been implicated in adiposity, lipid metabolism and feeding behaviour. In this report we analyse ESR1 and ESR2 gene single nucleotide polymorphisms (SNPs) for association with obesity. We also relate adipose tissue ESR1 mRNA levels and ESR1 SNPs to adipocyte lipolysis and lipogenesis phenotypes. Methods 23 ESR1 and 11 ESR2 tag-SNPs, covering most of the common haplotype variation in each gene according to HAPMAP data, were analysed by Chi2 for association with obesity in a cohort comprising 705 adults with severe obesity and 402 lean individuals. Results were replicated in a cohort comprising 837 obese and 613 lean subjects. About 80% of both cohorts comprised women and 20% men. Adipose tissue ESR1 mRNA was quantified in 122 women and related to lipolysis and lipogenesis by multiple regression. ESR1 SNPs were analysed for association with adipocyte lipolysis and lipogenesis phenotypes in 204 obese women by simple regression. Results No ESR1 SNP was associated with obesity. Five ESR2 SNPs displayed nominal significant allelic association with obesity in women and one in men. The two ESR2 SNPs associated with obesity with nominal P value < 0.01 were genotyped in a second cohort where no association with obesity was observed. There was an inverse correlation between ESR1 mRNA levels in abdominal subcutaneous (sc) adipose tissue and basal lipolysis, as well as responsiveness to adrenoceptor agonists independent of age and BMI (P value 0.009–0.045). ESR1 rs532010 was associated with lipolytic sensitivity to noradrenaline (nominal P value 0.012), and ESR1 rs1884051 with responsiveness to the non-selective beta-adrenoceptor agonist isoprenaline (nominal P value 0.05). These associations became non-significant after Bonferroni correction. Conclusion ESR1 gene alleles are unlikely to be a major cause of obesity in women. A minor importance of ESR2 on severe obesity cannot be excluded. The inverse correlation between ESR1 mRNA levels and lipolytic responsiveness to adrenoceptor agonists implies that low adipose tissue ESR1 levels attenuate catecholamine resistance in sc fat cells of obese women hereby contributing to loss of sc and gain of visceral fat. There is no evidence for a genetic impact of ESR1 on lipolysis or lipogenesis.


Background
The prevalence of obesity has reached epidemic proportions and is of great public health concern. Its metabolic complications, such as dyslipidemia and type 2 diabetes (T2D), represent a significant cost for the society. Though obesity is often triggered by life style, the importance of genetic predisposition has been clearly demonstrated [1]. Defining gene alleles that are associated with obesity will contribute to our understanding of the underlying mechanisms behind its development and potentially provide therapeutic targets.
Estrogen signalling is implicated in both central, that is regulation of food intake, and peripheral pathways protecting against adiposity. Low estrogen levels following menopause are associated with loss of subcutaneous (sc) and gain of visceral fat [2]. Estrogens signal via the transcription factors estrogen receptors α and β (ESR1, ESR2) [3]. Silencing of the ESR1 gene in mice renders an obese phenotype [4]. Both ESR1 and ESR2 are expressed in the hypothalamus, a brain area known to be involved in the regulation of appetite and satiety [5], whilst adipocytes express mainly ESR1 [6]. In the brain, ESR2 has been implicated in mediating the effects of estrogen on food intake [7]. As for peripheral effects of estrogens, hormone replacement therapy (HRT) has been shown to inhibit Epinephrine-stimulated lipolysis in human sc adipose tissue [8]. Estrogen, signalling through ESR1, lowers the lipolytic response in sc fat by increasing the number of antilipolytic α2-adrenoceptors [9].
Human genetic studies support a role for the ESR genes in regulation of body weight and certain aspects of the metabolic syndrome. Body mass index (BMI) (LOD 4.6, 4.2 and 2.8), waist circumference (LOD 3.3) and high-density lipoprotein (HDL) cholesterol levels have been linked to the chromosomal region harbouring ESR1 [10][11][12][13]. Among four investigated ESR1 polymorphisms, rs2234693 and rs9340799 were associated with waist circumference and rs1801132 with BMI in men, but not women [14]. In another study, ESR1 rs9340799, but not rs2234693, was associated with waist and BMI in Japanese women but not men [15]. Furthermore, neither rs2234693 nor rs9340799 was associated with obesity in Swedish women [16]. More recently, among 17 investigated ESR1 polymorphisms, rs6902771, rs2431260 and rs2175898 were associated with BMI in African Americans [17]. Among these same 17 ESR1 polymorphisms, rs1709183 and rs2431260 were associated with T2D in European Americans and rs1033182 in African Americans [18]. ESR2 has primarily been investigated in eating disorders. ESR2 rs928554 and rs4986938, but not rs1256049, have been associated with bulimic behavior [19] and rs1256049, but not rs4986938, with anorexia nervosa [20]. Rosenkrantz et al screened the coding region of ESR2 in probands of different weight extremes, but identified no mutations associated with the studied phenotypes [21].
We have recently reported that ESR1 mRNA expression levels in sc adipose tissue and isolated adipocytes from premenopausal women are inversely correlated with BMI [16]. However, whether ESR1 protects against obesity or is upregulated as a result of low BMI remains to be established. In this report we analyse polymorphisms covering most of the common haplotype variation in the ESR1 and ESR2 genes for association with obesity in two large cohorts of Swedish Caucasians. We also investigate if ESR1 is important for lipid turnover in human adiposity by analysing mRNA levels and gene polymorphisms for association with adipose tissue lipolysis and lipogenesis phenotypes in a third cohort of women., Table 1. To our

Subjects and clinical evaluation
Samples 1 and 2 were recruited for the purpose of studying genes underlying susceptibility to obesity, Table 1. In sample 1, the lean subjects were subjects > 45 years old and had never reached BMI > 25 kg/m 2 whereas the obese subjects had BMI > 30 kg/m 2 at < 20 years of age or morbid obesity, i.e. BMI > 40 kg/m 2 . These recruitment criteria resulted in more young obese than lean subjects, Table 1.
The aim of selecting subjects with an extreme BMI phenotype in sample 1 was to enrich for a genetic impact on obesity [22]. This was also the purpose of recruiting young obese adults, since early onset of this disorder is believed to have a stronger genetic component due to reduced time of environmental impact [22]. The size of sample 1 provided 90% power to detect an allele with a frequency of 20% among controls and odds ratio (OR) ≥ 1.5 to develop obesity assuming a threshold P value of 0.05 and dominant impact on obesity [23]. 59 subjects in sample 1 were diagnosed with T2D according to self report, but otherwise the subjects were healthy and free of medication according to self-report. 381 subjects in sampe 1 were also included in our previously reported analysis of two ESR1 SNPs in obesity [16].
Sample 2 had less stringent inclusion criteria for obesity and leanness in comparison with sample 1. Sample 2 comprised healthy non-obese subjects > 25 years old with BMI < 25 kg/m 2 and obese individuals with BMI > 30 kg/ m 2 , Table 1. 197 of the obese subjects in sample 2 had hypertension, 116 T2D, and 50 dyslipidemia according to self report. Differences in recruitment criteria resulted in phenotypic differences between sample 1 and 2, Table 1. The ratios between men and women were similar among obese and lean in both samples (17-21%).
Sample 3 comprised obese and lean otherwise healthy women recruited with the purpose of studying fat cell function, Table 1. Lipolysis and lipogenesis were investigated in 204 obese women in sample 3. 204 subjects provide approximately 75% power to detect an allele with a frequency of 20% and R 2 0.04 assuming a threshold P value of 0.05 and dominant impact on adipocyte phenotypes [23]. ESR1 mRNA levels in adipose tissue were quantified in the 122 women of sample 3 for which RNA was available. 99 women in sample 3 who met the criteria for obesity in sample 1 were included in both samples.
All subjects above were Caucasians and at least second generation Swedish. Obese subjects were recruited from an outpatient center for treatment of obesity or through local advertisement. All lean subjects were recruited through local advertisement. All subjects came to the laboratory in the morning after an overnight fast. A venous blood sample was obtained for extraction of genomic DNA and determination of serum insulin, as well as plasma glucose, triglycerides, total cholesterol, and HDL cholesterol as described [24,25]. Insulin resistance index HOMA IR (homeostasis model assessment) was calculated as fasting serum insulin (μU/ml) × fasting plasma glucose (mmol/l)/22.5 [26]. The ethical committee of the Karolinska University Hospital (Southern Campus) approved the study. It was explained in detail to each participant and her/his consent was obtained.

Lipolysis and lipogenesis investigation
In the morning after an overnight fast a sc fat biopsy was obtained from the abdominal area by needle biopsy [27]. One part of the adipose tissue pieces was immediately frozen in liquid nitrogen for subsequent RNA analysis.
Another piece was digested with collagenase to isolate fat cells and mean fat cell weight was determined [28]. Lipolysis in isolated fat cells was investigated as described [29,30]. Briefly, cell suspensions were incubated in the absence or presence of increasing concentrations of noradrenaline, the non-selective β-adrenoceptor agonist isoprenaline, or the α2-adrenoceptor selective agonist clonidine. At the end of incubation an aliquot of the medium was removed for analysis of glycerol release, which was used as an index of lipolysis.
Lipogenesis was investigated by determining the uptake of radio labeled glucose into lipids as described [31]. Briefly, diluted suspensions of fat cells were incubated in the absence or presence of increasing concentrations of crystalline human insulin (10 -15 -10 -6 mol/l) for 2 h at 37°C. Incubation was terminated by addition of sulphuric acid and radioactivity incorporated into lipids determined.
The concentration (log mol/l) of hormone or agonist causing half maximum effect was determined using logarithmic conversion of each concentration response curve. This value was converted to its negative form (pD 2 ), which reflects hormone sensitivity. The maximum effect or responsiveness of hormone or agonist was determined as glycerol release or glucose incorporation into lipids at the maximum effective hormone concentration.

SNP selection and genotyping
Genotype information for ESR1 and ESR2 and regions approximately 10 000 base pairs up-and downstream of these genes were downloaded from HAPMAP in May 2005 [33]. Genotype data for the population of individuals of European ancestry were visualized using Haploview [34]. We initially selected tag-SNPs defining all haplotypes with frequency > 5% from the block-by-block tags displayed in the HAPLOVIEW "Haplotypes" window. However, sometimes these tag-SNPs were exchanged to other SNPs on the same haplotype due to difficulties in designing genotyping assays. In addition, in regions not covered by common haplotypes we aimed to select one common SNP (allele frequency > 5%) every 5 000 base pair. No non-synonymous SNPs full-filling these criteria were detected in the ESR genes. In selecting between different SNPs we prioritized (1) Golden-gate validated assays, and (2) SNPs with high score according to Illumina, which indicate that the designed Illumina genotyping assays are highly likely to work [35]. Besides the HAPMAP SNPs we genotyped the ESR2 SNPs rs928554 and rs4986938, which previously have been associated with feeding behavior [19]. Samples 1 and 3 were genotyped using the Illumina technology at the SNP technology platform in Uppsala [36,35] except SNPs rs928554 and rs4986938 that were genotyped by RFLP as described [19]. Sample 2 was genotyped using matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (SEQUE-NOM Inc., San Diego, California) as described [37]. Primers can be provided on request. The genotype call rate for all genotyping platforms was ≥ 97% and the accuracy was 99.99% according to duplicate analysis of, on average, 2% of the total genotypes. Hardy-Weinberg equilibrium (HWE) calculations were performed to ensure that each marker was within population equilibrium.

Statistical analyses
Association between ESR1 mRNA levels and lipolysis and lipogenesis measures was assessed by multiple regression with age and BMI as additional independent variables. The Finetti software [38] using Pearson's goodness-of-fit Chi 2 was employed to test for allelic association between single SNPs and obesity. ANCOVA with age and BMI as covariates were used to analyze differences in insulin resistance and blood lipid phenotypes between geno-types. HOMA IR , Pl-HDL cholesterol, and Pl-triglycerides were ln-transformed before analysis to become normally distributed. For genetic analysis of adipocyte lipolysis and lipogenesis, logarithm transformed phenotypes were used as independent quantitative variables in linear regression models. Genotypes were indicated as character (nominal) with three levels, two different homozygous and heterozygous. There was no association between the adipocyte phenotypes and age or BMI in our sample and these variables were therefore not included in the regression model. Haplotypes were estimated and analyzed with Haploview [34]. LD was calculated as D'. We used HAPMAP Caucasian data to define haploblock limits. Association between haplotypes and obesity status was evaluated by Chi 2 . Haplotypes with frequency < 5% and individuals with > 50% missing genotypes were excluded in the analysis. We performed 10 000 permutations in Haploview to adjust P values to multiple tests.

ESR genotyping results
To investigate if ESR1 and ESR2 alleles contribute to susceptibility to obesity, polymorphisms covering the common variation in these genes were genotyped in sample 1, Table 2. For ESR1, 6 out of 29 (21%) Illumina genotyping assays failed, which is higher than the expected 10% failure rate. The majority represented Golden-gate validated assays indicating, according to the supplier, that the failure is due to interference between different SNP assays under the employed multiplex conditions. LD between ESR gene SNPs is shown in Figure 1 and 2, in which haploblock limits according to Caucasian HAPMAP data have been labeled. Genotyped ESR1 SNPs built 25 haplotypes, Table 3. These SNPs unambiguously identified 67% (20/ 30) of the common (> 5%) haplotypes in the ESR1 region in HAPMAP, representing 75% of the common haplotype variation in ESR1 according to HAPMAP data. Remaining ten ESR1 haplotypes in HAPMAP could in our sample not be separated from another haplotype, that is they were merged into five haplotypes. For ESR2, all genotyped assays were called successfully although one SNP, rs7154455, displayed a low genotyping call rate, 83%, Table 2. Thus there was 100% coverage of the common Caucasian ESR2 haplotypes in HAPMAP, Figure 2. All SNPs were in HWE.

Analysis of ESR SNPs and haplotypes in obesity
No ESR1 SNP displayed significant allelic association with obesity, but one SNP, rs2813544, was associated with obesity in analysis of homozygous subjects only, nominal P value 0.05, Table 4. Two ESR2 SNPs, rs7154455 and rs3020450, displayed allelic association with obesity in the joint analysis of women and men, nominal P value 0.0003-0.01, Table 4. In permutation test, the association between rs7154455 and obesity remained significant (P value 0.0022), whereas the result for rs3020450 became borderline significant (P value 0.064). Both SNPs were associated with obesity in women, but not in men. For ESR2 SNPs rs1152582, rs1271572 and rs1269056, allelic associations with obesity were observed in women, nominal P value 0.03-0.05, but not in joint analysis of women and men, Table 4. ESR2 SNP rs4986938 was associated with obesity in men only, nominal P value 0.05. rs7154455 and rs3020450 were genotyped in sample 2 where no association with obesity was observed, Table 4. Nor was there any association with obesity in analysis of pooled samples 1 and 2 (results not shown). No ESR SNP was associated with HOMA IR , Pl-cholesterol, Pl-HDL cholesterol, and Pl-triglycerides with P < 0.0125, and did thus not remain significant after Bonferroni correction for analysis of several SNPs (results not shown).
In sample 1, no ESR1 haplotype was associated with obesity, Table 3. One common ESR2 haplotype, TCCC (frequency 29%) at SNPs rs3020450-rs1271572-rs1887994-rs7159462 was associated with obesity, nominal P value 0.024, Table 3. TCCC was captured by SNP rs3020450. rs3020450 was genotyped in sample 2, where we were unable to confirm the association observed in sample 1.

Relation of ESR1 SNPs and mRNA levels to adipocyte lipolysis and lipogenesis phenotypes
In 122 women with a large variation in BMI, part of sample 3, there was an inverse correlation between ESR1 mRNA levels in abdominal sc adipose tissue and adipocyte basal lipolysis as well as responsiveness to noradrenaline and more selective adrenoceptor agonists, Figure 3. In multiple regression, including BMI, age, and ESR1 mRNA levels as independent variables, ESR1 levels were an independent factor regulating basal lipolysis (P = 0.018), as well as responsiveness to noradrenaline (P = 0.013), Table 5. Association between ESR1 mRNA levels and lipolytic responsiveness was stronger for the α2adrenoceptor selective agonist clonidine (P = 0.009), than for the non-selective β-adrenoceptor agonist isoprenaline (P = 0.045), Table 5. There was no association between ESR1 mRNA levels and sensitivity of lipolysis to adrenoceptor agonists, nor between ESR1 mRNA levels and measures of lipogenesis. We next analysed ESR1 SNPs for impact on adipocyte lipolysis and lipogenesis in the 204 obese women of sample 3, for which these phenotypes had been investigated. No SNP was associated with basal lipolysis (results not shown). One ESR1 SNP, rs532010, was associated with lipolytic sensitivity to noradrenaline, nominal P value 0.012, Table 6. Another ESR1 SNP, rs1884051, was associated with responsiveness to the non-selective beta-adrenoceptor agonist isoprenaline, nominal P value 0.05. These SNP associations became non-significant after Bonferroni correction for analysis of several SNPs. No ESR1 SNP was associated with lipogenesis (results not shown).

Discussion
Estrogen signalling has been implicated in regulation of human adiposity [4]. There is evidence that ESR1 acts in fat cells, whereas ESR2 mediates the effects of estrogen on food intake [7,9]. We here report that ESR1 mRNA levels in adipose tissue were inversely correlated with basal lipolysis and adrenoceptor responsiveness in obese women. In sample 1, two ESR2 SNPs were associated with obesity with nominal P value < 0.01. However, these associations were not confirmed in sample 2. No ESR1 SNP displayed allelic association with obesity or lipolysis.
Although no ESR1 SNP displayed allelic association with obesity, we cannot completely exclude an impact of ESR1 alleles on female obesity since genotyped SNPs failed to capture a few common haplotypes in the region of the ESR1gene.
Two ESR2 SNPs, rs7154455 and rs3020450 were associated with obesity with nominal P value < 0.01. The call rate for rs7154455 was only 83% adding some uncertainty to this association. rs7154455 and rs3020450 were genotyped in sample 2. In sample 2, as well as in pooled analysis of sample 1 and 2, no association with obesity was observed. This suggests that the allelic association of LD (D') between ESR1 SNP and haploblocks according to Caucasian HAPMAP data Figure 1 LD (D') between ESR1 SNP and haploblocks according to Caucasian HAPMAP data.
the two ESR2 SNPs in sample 1 was spurious. However, since recruitment criteria for sample 2 were based on less stringent definitions of obesity and leanness, and included subjects with metabolic complications of obesity, we cannot exclude that the ESR2 SNPs have a modest impact on more extreme forms of uncomplicated obesity such as in sample 1. In addition, this study was not designed to study potential gene-environment interactions. This is particularly important for ESR2, which has been implicated in mediating the effects of estrogen on food intake. Furthermore, our cohort of men was too small to exclude a male-specific impact of ESR1 and ESR2 on obesity. This might explain the difference between our results and the association of ESR1 and male obesity previously reported [14].
We have reported before that ESR1 rs2234693 and rs9340799 are not associated with obesity in a cohort of Swedish women partially overlapping with the cohort studied in this project [16]. We therefore did not genotype rs2234693 and rs9340799 in this study. ESR1 rs1801132, reported to be associated with BMI in men in the Framingham Heart Study [14], is according to the Caucasian HAP-MAP data on the same haplotype as rs9397456, which was genotyped in this project. We did not genotype the three ESR1 SNP reported to be associated with BMI in African Americans [17]; however the reported BMI associated SNP rs6902771 is according to HAPMAP on the same haplotype as rs2234693, which we genotyped. We genotyped rs1709183, but not rs2431260, reported to be associated with T2D [18]. Finally, we genotyped rs928554 and rs4986938, but not rs1256049, implicated in eating disorders [19,20]. SNPs genotyped in our present project, that previously have been associated with obesity related and eating disorders, were not associated with obesity in the present project.
Human obesity is associated with increased rate of basal lipolysis in all fat depots and lipolytic resistance to catecholamines in sc fat [39]. Our finding that ESR1 mRNA levels in adipose tissue are inversely correlated with rate of basal lipolysis might suggest that low adipose tissue ESR1 levels, as observed in obesity [16], increases basal lipolysis. In addition, the inverse correlation between ESR1 mRNA levels and lipolytic responsiveness to catecholamines implies that low adipose tissue ESR1 levels among obese counteracts catecholamine resistance in sc fat cells of obese women. Catecholamine resistance among obese is specific to sc fat [39]. Attenuation of lipolytic catecholamine resistance may explain why low estro- gen signalling is associated with fat redistribution with loss of sc and gain of visceral fat [2]. In our analyses, there was no evidence for a genetic impact of ESR1 on lipolysis or lipogenesis.

Conclusion
Common ESR1 gene alleles are unlikely to contribute to obesity in women, whereas a minor importance of ESR2 on severe obesity cannot be excluded. The male subset in our sample was too small to exclude that ESRs have an    impact on obesity in men. Low ESR1 mRNA levels in adipose tissue may counteract catecholamine resistance in fat cells of obese women hereby contributing to loss of sc and gain of visceral fat [39]. In our analysis of obese women there was no evidence for a genetic impact of ESR1 on lipolysis or lipogenesis. In future research, it will be important to investigate if there are interactions between ESR2 gene alleles and environmental factors such as food intake. a) For quantitative analysis of lipolysis and lipogenesis, logarithm transformed phenotypes were used as independent quantitative variables in linear regression models. Genotypes were indicated as character (nominal) with three levels, two alternative homozygous and heterozygous. b) The concentration (log mol/l) of hormone or agonist causing half maximum effect was determined using logistic conversion of each concentration response curve. This value was converted to its negative form (pD 2 ), which reflects hormone sensitivity. c) Lipolysis is expressed as mmol of glycerol/2 h/10 7 cells.