Common Genetic Variations in CCK, Leptin, and Leptin Receptor Genes Are Associated With Specific Human Eating Patterns

The study population was selected from the Prospect-European Prospective Study into Cancer and Nutrition (EPIC) study, a large population-based study that allowed the selection of obese individuals with specific eating patterns. The study is one of two Dutch components of the EPIC cohort, originally designed to investigate the role of nutrition factors in the occurrence of cancer (18). The cohort consists of 17,357 females aged 49–70 years at enrollment between 1993 and 1997 living in Utrecht, the Netherlands, or in the near vicinity with a Dutch ancestry. Detailed data on dietary habits, blood samples, BMI, and eating habits and physical activity (both based on a validated questionnaire) were present for all women (19,20) (Table 1). The cases were selected using the extreme discordant phenotype (EDP) approach (21) for an extreme meal size or meal frequency instead of on the broad phenotype of obesity; this selection results in a nine times–enhancement of statistical power (21).

Of the 17,357 women, a selection was made based on three criteria: BMI ≥33 kg/m²; a score on snack behavior in the top 5th percentile, based on 11 questions regarding frequency of snack consumption; and a score in the top 5th percentile of food intake, based on 28 questions using color photographs to estimate portion size. As quality control measures, women with total energy intake <500 kcal/day or total energy intake >6,000 kcal/day were excluded. An exclusion criterion was the combination of high activity levels in the top 5th percentile, combined with high food intake, based on caloric intake. Beforehand, 1,450 women were excluded because of missing data on physical activity. We did not select highly active women, as they have a reason for high energy intake either through large meals or through frequent snacking. Based on activity level, 767 women were excluded before the selection of case and control subjects.

The selection resulted in 72 women who met the criteria BMI ≥33 kg/m² and scored in the top 5th percentile of meal size, 60 women who met the criteria BMI ≥33 kg/m² and whose snack behavior was in the top 5th percentile, and 3 women who met all three criteria for a total of 135 cases. A total of 287 control subjects were randomly selected from the total cohort. The BMI of this random selection was means ± SD 25.8 ± 1.46 kg/m². Within the random control group, there were women with snack (5%), meal size (2%), or both (1%) behaviors in the top 5th percentile but with a BMI <33 kg/m². The institutional review board of the University Medical Centre Utrecht, Utrecht, the Netherlands, approved the study and all participating subjects provided written consent.

Haplotype-tagging single nucleotide polymorphisms (SNPs) were selected in the coding sequence and 100 kb up- and downstream of the genes using HapMap (Haploview 3.2) (available at http://www.gmap.net/perl/marker/marker) with r² >0.80 or Marker (22), which were available at the time of the study design. An SNP database search at the National Center for Biotechnology Information was conducted for the identification of nonsynonymous SNPs. This resulted in five haplotype-based SNPs for CCK, four SNPs for leptin, three haplotype-based and one coding SNP, and eight SNPs for the leptin receptor (seven were haplotype-based and one nonsynonymous SNP). All SNPs had a minor allele frequency of at least 15%.

All SNPs were genotyped using an allele-specific PCR. The PCR was performed using labeled primers with a 3′ locked nucleic acid, which is a modified SNP-binding nucleotide with a high binding affinity. Pooled PCR products determining different SNPs in one individual were separated by size on a 3700 capillary sequencer (ABI) and analyzed using the program Genemapper. This technique has been validated by comparing genotypes for multiple SNPs with those obtained from a Taqman system (ABI) with perfect agreement of results (data not shown).

The SNPs were tested for deviations from Hardy-Weinberg equilibrium using a Hardy-Weinberg P value cutoff of 0.01. Statistical analysis of the individual SNPs on genotype and allelic level was performed with the use of the two-sided Fisher’s exact test. Haplotypes were determined using Haploview 3.2 (22). Differences in haplotype distribution were analyzed using a two-sided Fisher’s exact test. The statistical significance level for all tests was set at P <0.05. To correct for multiple testing using the permutation analysis function of UNPHASED, 100 permutations were performed.

We genotyped 17 SNPs located in the 3′ untranslated region or coding region of CCK (n = 5), leptin (n = 4), and the leptin receptor (n = 8) in the total sample (n = 422). All SNPs showed the expected genotype proportions when tested for Hardy-Weinberg equilibrium.

The case subjects were categorized according to extreme snack behavior (n = 60) and meal size (n = 72). The genotype and allelic distributions of the 17 SNPs were compared between the two selection groups and the random control sample (n = 287); global P values were determined, and haplotypes were subsequently established.

Four of the five tested CCK SNPs showed a specific association signal with extreme meal size (P = 0.005, P = 0.00, P = 0.001, and P = 0.02) but not with extreme snack behavior (Table 2). One of eight SNPs of the leptin receptor and two of four SNPs of leptin were associated with extreme snack behavior but not with meal size (P = 0.02, P = 0.01, and P = 0.006) (Table 3). Furthermore, (corrected) global P values, derived from single SNPs, show significant associations (CCK P = 0.029, leptin receptor P = 0.009, and leptin P = 0.0198) with meal size and snacking.

Using Haploview 3.2 (22) we constructed haplotypes to determine whether specific haplotypes underlie the associations found. Haplotypes were determined for CCK, CCK_H1–H5 (rs6809785, rs7611677, rs6801844, and rs6791019) and leptin, Lep_H1–H7 (rs4577902, rs2060736, and rs4731413). A very high D′ (close to 1) (Tables 4 and 5) was detected between the haplotype SNPs within both CCK and leptin. This indicates that the haplotype data do reflect the findings of the individual SNPs (i.e., significantly associated). No haplotypes could be determined for the leptin receptor. (Tables 4 and 5) using Haploview 3.2.

CCK_H3 showed an increase in frequency in extreme portion eaters compared with control subjects, an increase from 10% (n = 37) to 17% (n = 21) (P = 0.022), suggesting that H3 is a risk haplotype for an increased meal size. The three SNPs determining this haplotype also individually showed an association with this phenotype.

In agreement with the observed association for the allelic variations for the leptin gene, the SNPs also showed at the haplotype level a specific association with extreme snacking. The most frequent haplotype (Lep_H1) was observed in the extreme snack group (58% [n = 252] vs. 68% [n = 53], P = 0.03), while a clear loss of Lep_H3 was seen (13% [n = 56] to 3% [n = 2], P = 0.01). These data suggest that there is a protective (Lep_H3) and a risk (Lep_H1) haplotype for extreme snack behavior.

We also assessed the SNPs in the total case group. Three of the individually tested SNPs, rs6791019, rs6809785 and rs11129946, of CCK showed an association (P = 0.001, P = 0.02, P = 0.03, respectively, Table 2).

Finally we checked the haplotype distribution within the random control group. Within the control group 5% (n = 15) and 2% (n = 7) displayed extreme snacking or meal size without a BMI ≥33 kg/m². The associated haplotypes found in the obese group were borderline associated in the extreme snackers (Lep_H1, P = 0.05) and not in the portion eaters (CCK_H3, P = 0.74) in the control group. The very small number of women included in these groups probably caused this. Of the total random control group, 13.5% had a BMI ≥33 kg/m², without displaying the specific eating patterns. The allelic and haplotype distributions of the CCK, leptin, and leptin receptor genes in the group with BMI ≥33 kg/m² did not differ from the distributions found in the rest of the control group.

In this study, we associated SNPs and haplotypes in the CCK gene with large meal size and in the leptin genes with high meal frequencies, two specific eating patterns within the broad phenotype of obesity.

We took an alternative approach to previous research addressing the genetic basis of obesity. Using the EDP approach we selected cases based on the extreme expression of specific traits of food intake and show that this mode of selection is useful in determining common genetic variation underlying a specific trait within a broad phenotype. Interestingly, the selected population fell into two almost nonoverlapping groups, which for a significant portion was explained by genetic variation at the three genes that were genotyped. This may not be surprising since extreme snacking may decrease meal size.

As far as we know, no association studies have been reported on the role of the CCK gene in obesity, satiety, or meal size. In this study we found associations between CCK polymorphisms/haplotypes and excessive meal size, which probably contribute to problems with satiety signaling in these carriers.

Some genetic studies have been conducted on the role of leptin genes in obesity (5,14,23,24). Only a small number of rare variants have been found explaining only a minority of the obese cases. Our findings suggest a specified role of the leptin genes in obesity, namely in extreme meal frequencies in the population sample.

The usefulness of the EDP approach is illustrated by the different frequencies of the observed Lep_H1 haplotypes in the extreme cases (68%) and the overall cases (62%) versus the control sample (58%). It is remarkable that such a common haplotype was found to be associated with extreme snacking. Further study is required to investigate the genetic history of this haplotype and its possible biological relevance for obesity.

Based on the haplotype and genotype associations, we could determine that obese carriers of variants in the CCK gene have an increased risk of eating large portion sizes, with a 60% increased risk for carriers of the CCK_H3 haplotype and that obese carriers of variants in the leptin genes have an increased risk if displaying extreme snack behavior, with a 20% increased risk for Lep _H1 haplotype carriers. However, given the high frequency of this allele in the population, the population attributable risk is very high.

Further studies are necessary to confirm our findings and to examine the effect of the genetic variants in order to be able to evaluate the general risk in the population. It will be interesting to see to what extent the found risk alleles and haplotypes contribute to eating patterns in the general population.

This work was supported by Zon-MW Grant 016.036.322.

We thank Bobby Koeleman, PhD, of the Medical Genetics Department, University Medical Center Utrecht, the Netherlands, for his advise and assistance on the statistical analysis of the work presented in this article.