Association of dopamine D2 receptor and leptin receptor genes with clinically severe obesity

Authors


  • Funding agency: Funding was provided by the UCLA Department of Medicine.

  • Disclosure: The authors have no competing interests.

Correspondence: Catherine L. Carpenter (ccarpenter@mednet.ucla.edu)

Abstract

Objective

The brain reward circuits that promote drug abuse may also be involved in pleasure seeking behavior and food cravings observed in severely obese subjects. Drug addiction polymorphisms such as the TaqI A1 allele of the dopamine D2 receptor (DRD2) are associated with cocaine, alcohol, and opioid use, but few studies have linked DRD2 to food craving. Other genes such as the leptin receptor gene (LEPR) and mu-opioid receptor gene (OPRM1) that affect appetite and pleasure centers in the brain may also influence food addiction and obesity. The three genes together may function synergistically.

Design and Methods

To evaluate associations between candidate genes, food craving, overeating, and BMI, we administered questionnaires including Power of Food Scale and Food Craving Inventory, conducted anthropometric measures, and collected blood from patients undergoing weight-loss treatment. Questionnaires and DNA specimens were collected for 80 participants.

Results

Participants were mostly female (74%) and Caucasian (79%), with an average age of 53 years old. Mean BMI for all participants was 43 kg/m2 and was significantly associated in a linear fashion with Food Craving Inventory scores (P=0.0001) and Power of Food (P=0.02). The DRD2 TaqI A1 allele was significantly associated with BMI (P=0.04), while LEPR Lys109Arg and OPRM1 A118G variants were not. We stratified DRD2 by LEPR and OPRM1, and observed a significant interaction (P = 0.04) between DRD2 and LEPR, and a marginally significant interaction (P=0.06) between DRD2 and OPRM1.

Conclusion

Genes associated with addictive behavior and appetite control may therefore, in combination, markedly influence development of clinically severe obesity.

Introduction

Obesity is a complex phenotype resulting from genetic susceptibility and environmental influence. One of the central questions emerging in obesity research is whether overeating and food craving can be evaluated in relationship to addictive behavior. Development of an obese phenotype could be linked, for instance, to behavioral patterns such as obsession with eating particular foods, pleasure seeking as a function of fullness, and dysregulation of appetite.

Environmental change, in combination with genetic susceptibility, is most likely responsible for the increase in obesity prevalence observed over the past three decades [1]. Recent prevalence trends in obesity (BMI ≥ 30.0 kg/m2) among US adults have leveled off, although grade 3 obesity (BMI ≥ 40.0 kg/m2) continues to increase, particularly for men [2].

Evidence from several lines of research has shown that obese individuals behave differently than normal weighted individuals with regard to food stimuli and reward [3-6]. Among obese subjects compared to normal weighted subjects, there is greater activation of the left anterior insula, a reward center in the brain, in anticipation of a food reward [5]. Previous studies have shown that the dopamine and leptin signaling pathways may be involved in response to food-related stimuli. Pictures of food may elicit a decrease in dopamine receptor availability reflecting dopamine binding to the receptor. With declining receptor availability, there is an increase in the desire for food [3]. Obese individuals have also been shown to have a heightened sensitivity to the palatability of food and once the food is consumed, they experience a reduced reward which in turn, could drive them to overeat as a compensatory reaction [4, 5]. And finally, leptin replacement has been shown to decrease brain activation in response to food cues in genetically leptin-deficient adults [6], further suggesting that once food is consumed and leptin levels rise in healthy individuals, interest in food cues are diminished. These mechanisms may not operate, however, in obese individuals who lack sensitivity to leptin. Among leptin insensitive individuals, brain activation to food cues may remain, even after eating.

Genetic polymorphisms associated with addiction and appetite control may influence overeating and food craving. While many studies have linked the dopamine D2 receptor gene (DRD2) to classical addictions such as cocaine [7], alcohol [8, 9], nicotine [10], and opioids [11], fewer studies have shown associations between the DRD2 and the development of obesity [12-14]. Dopamine receptor availability is decreased in obese individuals [15], and this may be the mechanism by which the dopamine receptor gene influences the development of obesity.

Other genetic polymorphisms could also play a role in food craving and overeating. The leptin receptor gene (LEPR), responsible for encoding the leptin receptor that binds to leptin in target tissue, plays an important role in appetite control and metabolism [16]. Lys109Arg is one of several LEPR single nucleotide polymorphisms (SNP) studied in relationship to overweight and obesity [17]. Mixed results have been observed for Lys109Arg depending on the method of body composition assessment and choice of study population, but there is some evidence suggesting an association with leptin levels in overweight and obese women [18].

The opioid receptor system, which regulates intake of palatable food, is another potential pathway involved in food craving and overeating. Along with palatable food, drugs such as heroin, morphine, alcohol, and cannabinoids, interact with the opioid system, providing evidence for common neural substrates between food and drug reward with regard to the brain's opioid system [19]. The mu-opioid receptor gene (OPRM, located on chromosome 6q24) is one of the four genes whose protein products bind to endogenous opioids. In a study comparing heroin addicts to healthy controls, the most prevalent SNP in addicts was identified as A118G, a nucleotide substitution predicting an amino acid change at a putative N-glycosylation site [20]. In a targeted study of food addiction among obese subjects, the functional A118G polymorphism was found to be associated with binge eating [13].

The evidence for candidate genes influencing addictive behavior in general, and possibly appetite dysregulation and overeating, in particular, prompted us to examine whether these genes might also be related to body composition among a population of patients who are clinically severely obese. Our primary purpose was to evaluate associations between BMI and genetic polymorphisms related to craving (DRD2), pleasure-seeking (OPRM1), and appetite control (LEPR), in a cross-sectional investigation among patients undergoing weight loss treatment. Our secondary purpose was to evaluate the association between BMI and food craving and overeating behaviors.

Methods and Procedures

Study population and recruitment

All patients were recruited via a clinical weight-loss treatment program, the UCLA Risk Factors for Obesity Program (RFO) housed in the UCLA Center for Human Nutrition. Over a 7-month period beginning in October, 2009, all patients undergoing weight-loss treatment at the RFO clinic were invited to participate in the “Food Addiction" study. Out of the 356 patients undergoing treatment during the study period, we recruited 93 participants, with complete questionnaires and DNA specimens collected for 80 participants. The Institutional Review Board of the University of California at Los Angeles in accordance with assurances filed with and approved by the U.S. Department of Health and Human Services, approved the study, and participants provided written informed consent.

Data collection

All participants were given study questionnaires to complete during one of their visits to RFO. Study investigators were available to answer any questions that participants had about the questionnaires. The study questionnaire included an assessment of health-related risk factors and standard clinical assessments designed to measure dependence on, or addiction to, nicotine (Fagerstrom scale for nicotine dependence) [21], alcohol [22], illicit drugs [23], and the Community Epidemiology Survey of Depression designed to measure depression in community-based samples [24]. Addictive food behaviors were measured using the Food Craving Inventory [25] and the Power of Food Scale [26].

The Food Craving Inventory, a validated 28-item questionnaire frequently used in studies of eating behavior, is designed to measure frequency of overall food cravings as well as cravings for specified foods [25]. Responses to specific food cravings were summarized into four independent subscales: high fats (fried chicken, gravy, sausage, hot dogs, fried fish, corn bread, bacon, steaks); sweets (cakes, cinnamon rolls, ice cream, cookies, chocolate, donuts, candy, brownies); carbohydrates/starches (sandwich bread, rice, biscuits, pasta, pancakes/waffles, rolls, cereal, potato); and fast food fats (pizza, French fries, hamburger, chips). An average of the four subscales formed the general food craving score [25, 27].

The Power of Food Scale is designed to measure psychological temptation to eat in environments where foods are relatively abundant, accessible, palatable, and affordable [28]. The 15-item scale assesses influence of food on behavior and cognition, measures individual differences in motivation to consume palatable food [29], and has been previously used to evaluate overeating in relationship to OPRM1 [30]. Responses to the 15-item scale were summarized into an overall Power of Food measure according to a one factor solution [25, 28].

Clinical variables

Our participants were undergoing treatment for weight loss and any medication use may influence potential associations with body composition and polymorphic variants. We constructed variables for medication type and extent of use to address potential confounding influences. Medication type included summarizing current usage of anti-hypertensive medication to control blood pressure (hydrochlorothiazie, lisinopril, atenolol, furosemide, metoprolol succinate), hypoglycemic medication to control diabetes (metformin, glucophage, insulin, novolog, glyburide, aetas, cozaar, januvia), nonsteroidal anti-inflammatory medications (naproxen, enbrel, singulair, meloxicam, colchicine, allopurinol, celebrex, zyrtec, combivent, prevacid, NSAIDS), anti-anxiety/anti-depression medication (paxil, prozac, lorazapan, lamictal, lexapro, trazadone, adderall, budeprion, cymbalta, ativan, bupropion, lyrica, strattera, desvenlafaxine, topamax, xanax), cholesterol-lowering medications (crestor, lipitor, simvastatin, red yeast rice, pravachol, lorastatin, tricor, tenical), and other medications used to treat cardiovascular disease (baby aspirin, digoxin, coumadin, enalapril, digoxin, effexor). The medication frequency variable was constructed to summarize whether singular or multiple medications influenced body composition associations with polymorphic variants. The frequency variable consisted of three categories: no medication; one medication only; or more than one medication.

Anthropometric measurements

Study participants were weighed without shoes upon intake into the clinic using a doctors' scale. Height was taken using a stadiometer. BMI was calculated using height and weight. Weight gain or loss was calculated as the difference between weight taken at intake into the clinic and the latest time point within the study period that patients participated. We computed percent weight change by taking the difference and dividing by the intake weight.

Genotyping

DNA isolation

Three milliliters of peripheral blood were drawn into a BD Vacutainer K2 tube that contained EDTA (BD Biosciences, San Jose, CA, USA). Genomic DNA was extracted from blood using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich, St. Louis, MO, USA) according to manufacturer's instructions. Briefly, 200 μl of blood was treated with Proteinase K followed by RNase for 2 min before lysing cells in 55°C purified H2O for 10 min. DNA was precipitated with 100% ethanol and bound to a spin column. After two washes, DNA was eluted from the column with TE buffer.

TaqMan polymerase chain reaction genotyping

Genotyping was performed using SNP Genotyping Assays for DRD2 (rs1800497), LEPR (rs1137100), and OPRM1 (rs1799971) (Applied Biosystems, Foster City, CA, USA) on a StepOnePlus Real-Time PCR Machine (Applied Biosystems). Briefly, real-time PCR was performed with 25 ng DNA in a 20-μl reaction containing TaqMan Genotyping Master Mix and SNP genotyping assay. Reactions were heated for 95°C for 10 min following by 40 cycles of 92°C for 15 s with 60°C for 1 min. Alleles were determined using the allelic discrimination plot in the StepOne software genotyping program. Each sample was performed in quadruplicate.

Study identification

Personal identifiers were de-linked from genotyping results, so that only numeric codes associated with demographic data, questionnaire results, polymorphic classification, and body composition measures were available for statistical analysis.

Statistical analysis

Data were entered in Microsoft Access® with means and frequencies computed for all variables. ANOVA models were constructed to measure the association between polymorphic variants, questionnaire measures of addictive food behaviors, and continuous BMI variables. Allele frequencies and tests for Hardy–Weinberg Equilibrium were conducted using SAS GENETICS. We constructed gene–gene interaction models by creating product terms between each variant, including the product term along with separate terms for each variant, and evaluating the associations with continuous BMI outcome variables. Linear regression models measured interrelationships between continuous variables (BMI and food behavior measures). All statistical analyses were computed with SAS version 9.2 (Statistical Analysis System, 2008, Cary, NC, USA). Means are reported with standard deviations. Two-sided P-values are reported. Statistical significance is noted when P < 0.05.

Results

Allele frequencies for candidate genes

Frequencies for alleles comprising the three genes, DRD2 (rs1800497), LEPR (rs1137100), and OPRM1 (rs1799971), were summarized for the study population. Frequency of DRD2 TaqI A1 allele was 24%, Lys109Arg A allele was 75%, and allele frequency of OPRM1 A118G D was 16%. LEPR, DRD2 TaqI, and OPRM1 allele frequencies did not depart from Hardy–Weinberg equilibrium (P > 0.05).

Study population results

Table 1 contains frequency distributions for the overall study population. Females comprised 74% of the study population. Most of the study sample was Caucasian (79%). Average age was 53 years. Participants, on the average, had an average BMI of 43.3 kg/m2. Excessive alcohol drinking and smoking was infrequent with 17% drinking more than 1 drink per day and 3% ever having smoked.

Table 1. Characteristics of the study population
VariableCharacteristicNPercent
  1. aQuestionnaire responses.
  2. b(weight at study − intake weight)/intake weight.
  3. cDopamine D2 receptor gene.
  4. dLeptin receptor gene.
  5. eOpioid receptor, mu 1 gene.
GenderFemale5973.8
Male2126.3
Race/ethnicityAsian22.5
African American78.8
Hispanic810.0
Caucasian6378.7
SmokeaEver smoke23.1
Never smoke6397.9
Drink alcohola1+ drinks per day1116.9
<1 drink per day5483.1
  MeanSD
Food Craving Inventorya
Overall craving acore9.95.2
Power of Food Scalea
Overall score36.310.5
Age53.315.5
BMI (kg/m2)43.312.5
Percent weight changeb20.127.4
DRD2c Taq1 AA1/A122.5
A1/A23543.8
A2/A24353.7
LEPRd Lys109ArgA/A4556.3
A/G3037.5
G/G56.2
OPRM1e A118GD/D11.3
D/N2430.0
N/N5568.7

We constructed medication type and extent of usage variables to address whether medications might influence the association between genotype and BMI. Mean BMI according to medication type is presented in Table 2. Medication frequency (no medications, mean BMI = 40.4 kg/m2; one medication, mean BMI = 43.4 kg/m2; multiple medications, mean BMI = 44.8 kg/m2) was not associated with BMI (P=0.36). Use of hypoglycemic medications to control diabetes (P=0.14), nonsteroidal anti-inflammatory medications (P=0.10), cardiovascular disease medications (P=0.10), anti-anxiety/anti-depressants (P=0.17), and cholesterol-lowering medications (P=0.43) were not associated with mean BMI. BMI was significantly different between those patients who used anti-hypertensive medications and those who did not (P=0.02), with users of anti-hypertensive medication averaging a BMI of 49.7 (SD = 14.33), and nonusers averaging 41.5 (SD = 11.4).

Table 2. Mean BMI according to medication use
Medication groupUserNonuserP-valuea
NMeanS.D.NMeanS.D.
  1. aP-value derived from Students' T-test.
  2. bAnti-hypertensive medications: hydrochlorothiazie, lisinopril, atenolol, furosemide, metoprolol succinate.
  3. cAnti-anxiety/anti-depressant medications: paxil, prozac, lorazapan, lamictal, lexapro, trazadone, adderall, budeprion, cymbalta, ativan, bupropion, lyrica, strattera, desvenlafaxine, topamax, xanax.
  4. dCardiovascular disease: baby aspirin, digoxin, coumadin, enalapril, digoxin, effexor (antianginal, antiarrhythmic).
  5. eCholesterol lowering medications: crestor, lipitor, simvastatin, red yeast rice, pravachol, lorastatin, tricor, tenical.
  6. fHypoglycemic (diabetes) medications: metformin, glucophage, insulin, novolog, glyburide, aetas, cozaar, januvia.
  7. gNonsteroidal anti-inflammatory medications: naproxen, enbrel, singulair, meloxicam, colchicine, allopurinol, celebrex, zyrtec, combivent, prevacid, NSAIDS.
Anti-hypertensiveb1849.5714.336241.5111.400.02
Anti-depressantc2347.0716.475741.8110.270.17
Cardiovascular diseased (antianginal, antiarrhythmic)2740.449.545344.7913.600.10
Cholesterol loweringe1040.3613.817043.7512.340.43
Hypoglycemicf1148.518.316942.5012.890.14
Nonsteroidal anti-inflammatoryg2447.7517.445641.439.210.10

We evaluated whether BMI was associated with food craving (Food Craving Inventory) and motivation to overeat (Power of Food Scale). We plotted BMI according to both food behavior measures (see Figures 1 and 2), and evaluated their interrelationship by computing linear regression models. There was a linear association between BMI and mean Food Craving Score (model P-value = 0.0001), with a moderate model fit (r2 = 0.26) (Figure 1). The Power of Food Scale, on the other hand, was less significantly associated with increasing BMI (P=0.02), and had a weaker model fit (r2 = 0.12). Both food behavior measures increased in a positive direction in relationship to increasing BMI. Power of Food scores were linked to Food Craving scores with a correlation coefficient of 0.51 (P<0.001).

Figure 1.

Black diamonds indicate Food Craving Inventory score by BMI.

Figure 2.

Black diamonds indicate Power of Food score by BMI.

Table 3 summarizes BMI according to genetic polymorphism. Mean BMI was significantly higher among participants who carried one or more copies of the TaqI A1 alleles (mean BMI = 46.4 kg/m2) from DRD2 compared to participants with no A1 alleles (mean BMI = 40.7 kg/m2) (P=0.04). The relationship remained when we restricted our analysis to participants with only one copy of the TaqI A1 allele. Associations between BMI and polymorphic variants from the other two candidate genes, LEPR and OPRM1, were not apparent in our sample. We evaluated whether percent-weight change from time of entry into weight loss treatment until the last time seen during the study period might be related to allelic variation and determined that none of the polymorphisms were associated with weight change (Table 3).

Table 3. Average BMI and mean percent weight change according to genetic polymorphism
Genetic polymorphismAllelic variantNMean BMISDP-valueMean percent weight changedSDP-value
  1. aDopamine D2 receptor gene.
  2. bLeptin receptor gene.
  3. cOpioid receptor, mu 1 gene.
  4. d(weight at study − intake weight)/intake weight.
DRD2a TaqI AA1/A1 or A1/A23746.414.5 12.19.3 
A2/A24340.79.90.0412.711.00.80
LEPRb Lys109ArgA/A4545.214.0 12.19.3 
A/G or G/G3540.910.00.1312.911.30.74
OPRM1c A118GN/N5542.411.1 12.910.1 
D/D or N/D2545.515.10.3111.410.50.58

We attempted to address confounding influences through stratified analysis and covariate adjustment. However, our small sample size limited our capacity to conduct stratified analyses. In particular, we had insufficient power to evaluate genotype associations with BMI according to gender (data not shown). In addition, we studied participants who were seeking treatment for weight-loss and therefore the observed associations could have been influenced by treatment-related covariates such as medication. Mean BMI was shown to be significantly different between users and nonusers of anti-hypertensive medication (see Table 2). We adjusted our genotype–BMI association models for anti-hypertensive medication and associations between DRD2 (P=0.04), LEPR (P=0.12), OPRM1 (P=0.29), and BMI, did not appreciably change.

We evaluated mean Food Craving and Power of Food scores according to the polymorphic variants and stratified the mean scores by median BMI (40.0 kg/m2). None of the polymorphic variants were associated with mean Food Craving or Power of Food scores for the overall study sample or for the sample stratified by median BMI (data not shown).

To examine whether obesity-related pathways may work in combination, we evaluated whether BMI was more closely associated with multiple variants, as compared to each separate variant (see Table 4). Mean BMI evaluated at LEPR polymorphic variants according to OPRM1 did not produce sizeable differences compared to each variant alone (data not shown). However, when we evaluated mean BMI for the DRD2 variants stratified by LEPR, we observed a much higher mean BMI (50 kg/m2) for individuals who carried one or more copies of DRD2 Taq1 A1 alleles and the AA variant of LEPR Lys109R, compared to individuals carrying all other combinations of alleles (BMI = 41 kg/m2) (data not shown). The degree of interaction between DRD2 and LEPR was significant with a P-value of 0.04. We also evaluated whether mean BMI according to DRD2 significantly varied when we stratified by OPRM1. The degree of interaction between DRD2 and OPRM1 was marginal (P=0.06), with mean BMI equal to 50 kg/m2 among participants carrying one or more copies of the TaqI A1 allele and heterozygous for A118G, while BMI equaled 42 kg/m2 among individuals carrying all other combinations of variants.

Table 4. Mean BMI for LEP-Ra Lys109Arg and OPRM1b A118G genotypes stratified by DRD2c Taq1 A
 DRD2 Taq1 A
A1/A1 or A1/A2A2/A2Interaction model P-value
NMean BMISDNMean BMISD
  1. Leptin receptor gene.
  2. Opioid receptor, mu 1 gene.
  3. Dopamine D2 receptor gene.
LEP-R Lys109Arg
A/A2150.015.52441.011.1 
A/G or G/G1641.712.01940.38.30.04
OPRM1 A118G
N/N2143.611.83441.610.9 
D/D or N/D1650.017.3937.43.20.06

Discussion

We studied a clinically severe obese sample of individuals who were undergoing weight loss to evaluate associations between food craving, overeating, genetic polymorphisms known to impact addiction, and body composition. With regard to polymorphic variants of DRD2, average BMI was significantly higher among those who carried one or more copies of the TaqI A1 allele, compared to individuals who had no copies of the allele (P=0.04). Mean BMI evaluated according to the other genetic variants, LEPR Lys109R and OPRM1 118G was not significant. When we stratified DRD2 by LEPR, we found a significant gene–gene interaction (P=0.04). Participants who carried one or more copies of TaqI A1 and were homozygous for the Lys109R variant had a significantly higher BMI than carriers of all other alleles. Mean BMI values for the OPRM1 variant stratified by DRD2 produced a marginally significant interaction (P=0.06).

Most study participants had clinically severe obesity. Therefore, we may have been more likely to observe rare alleles, and allelic associations with BMI. Prevalence of alleles from the three candidate genes in our study sample exceeded those found in population samples [17, 31-33]. The prevalence of TaqI A1 allele in our sample was 24%, while in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, prevalence was 21%, and, in a meta-analysis, prevalence ranged from 10 to 22% [31, 32]. In a random sample of healthy participants in an outpatient study of OPRM1, minor allele frequency was 11%, which was 5% lower than that seen in our study sample (16% prevalence) [33]. The allele frequency of Lys109Arg that we observed in our study (75%) greatly exceeded the allele frequency (12–35%) observed in Caucasian samples reported in a recent meta-analysis of LEPR and overweight [17].

We observed a linear relationship with the Food Craving Inventory and increasing levels of BMI, and a moderate association between the Power of Food Scale and increasing BMI. We chose the Food Craving Inventory to represent the classical addictive behavior of “craving" and the Power of Food Scale to represent the addictive behavior of “pleasure seeking.” We did not, however, observe associations between the food behavior measures and the candidate genes, nor did we observe associations when we stratified for median BMI. Because of our small sample size, mean differences in food behavior scores according to polymorphism may have lacked power to observe significant associations. Also, we relied on self-report questionnaires to evaluate food behavior and the increased variability associated with self-report measures may have prevented us from detecting significant associations.

Body-weight regulation can be viewed as a feedback system that involves several components: body fat production of leptin that signals to the hypothalamus; central brain mechanisms involving neurotransmitters; and involuntary systems controlling digestion and metabolism [34]. The neurotransmitters, dopamine, GABA, norepinephrine, and serotonin are central brain signaling molecules that control food intake [35]. DRD2 availability is decreased in obese individuals in inverse proportion to their increased BMI [15], and reduced density of the DRD2 is suspected to result from having the A1 allele of TaqI polymorphism [36]. The TaqI A1 DRD2 polymorphism attenuates dopamine signaling in the dorsal striatum, a region of the brain sensitive to reward feedback, and individuals with this polymorphism may exhibit compensatory overeating behavior as a result [37]. Studies of DRD2 TaqI have not been entirely consistent, however, and some suggest that it is unlikely that the TaqI A1 polymorphism is causative for obesity [38], although most of the scanning and genetic studies suggest that the DRD2 TaqI polymorphism contributes to development of obesity [35, 37, 39].

Consistent with other studies of DRD2 and obesity [13, 14], we found a significant association between BMI and the DRD2 TaqI A1 allele among a population of severely obese individuals seeking weight loss treatment. In addition to the BMI and TaqI A1 association, we also observed a significant gene–gene interaction between TaqI and the Lys109Arg variant of LEPR, with BMI much higher among participants who carried both the TaqI A1 allele and who were homozygous for Lys109Arg. While we did not observe an independent association with Lys109Arg and BMI, these results suggest a multifactorial relationship between leptin signaling and reduced dopamine activity that was apparent in our data when polymorphisms from both genes were considered together. These findings are supported by recent pre-clinical evidence suggesting that the leptin signaling pathway may exert influence on orexin and the mesolimbic dopaminergic systems [40].

The subset of patients who participated in the study may differ from patients who did not, and therefore we may have inadvertently selected for the associations that we observed. BMI among patients seen in the RFO clinic during the study period averaged 37.8 kg/m2, while average BMI among study participants was 43.3 kg/m2, suggesting that some self-selection into the study may have occurred. That is, patients who had higher BMI measurements may have been more likely to participate because of their BMI status, although it is unclear whether participation affected the distribution of alleles, since patients were not aware of their genotypes before or after the study.

Our study was conducted in a relatively small clinical sample of obese patients, most of whom had extreme phenotypes. Therefore, estimates derived from our study may not reflect patterns in the general population, and must be viewed with caution. These results suggest that mechanisms of food craving and overeating may involve several pathways such as leptin signaling and the dopaminergic pathway, which could operate in a synergistic fashion.

In summary, we studied a clinically severe obese patient population whose average BMI was associated with TaqI A1 DRD2 allele, and, when stratified by LEPR Lys109Arg, was significantly higher among individuals carrying variants from both genes.

Acknowledgments

We appreciate the participation of our study subjects.