Introduction

Many neuropeptides affecting food intake and/or energy expenditure have been described to date [1–4]. The recent identification of hunger and satiety peptides, as well as their receptors, has increased the interest in appetite regulation, particularly for the treatment of eating disorders, from anorexia to overeating and obesity [2].

Peptide YY (PYY) is a member of a gene family that also includes neuropeptide Y (NPY) and pancreatic polypeptide (PPY). The three peptides have differential patterns of production: PYY is released by the l-cells of the gastrointestinal tract in response to food intake [5, 6] and inhibits both pancreatic and gastric secretion. NPY has been localised to both brain and peripheral neurons and PPY is localised to the pancreas [7].

The human PYY gene was first cloned by Hort et al. [8] and, based on the gene structures of PYY, NPY and PPY, it was concluded that PYY and NPY are the result of a gene duplication event and a further tandem duplication event gave rise to the PPY gene.

Two forms of PYY are found in the gut and circulation: PYY1–36 and PYY3–36 [9–11]. Batterham et al. provided the first evidence that PYY3–36 is a mediator of postprandial satiety in rodents [6] and humans [12] through actions at the arcuate nucleus via NPY2 receptor (NPY2R). Since then many studies have established the role of PYY in decreasing food intake in both animals [13–18] and humans [19–21]. However, contradictory evidence has also been presented, suggesting that PYY3–36 does not decrease food intake in rodents, further questioning its role as a potential anti-obesity drug target [22, 23].

The NPY2R is a pre-synaptic inhibitory receptor that is abundantly expressed in the arcuate nucleus [24]. It has been established that mice lacking the NPY2R gene have an increased body weight and food intake [25].

There is suggestive evidence of linkage of abdominal subcutaneous fat to the chromosomal region 17q21.1 (LOD = 2.24), which is in close proximity to where both the PYY and PPY genes are located [26].

Plasma levels of PPY have also been found to be decreased in human morbid obesity [27, 28] and increased in cases of anorexia nervosa [27–29]. PPY is reported to inhibit food intake and stimulate energy expenditure following its peripheral administration [30]. It has been shown that children with Prader–Willi syndrome have decreased secretion of PPY [31] and that peripheral administration of PPY leads to a decrease in food intake in rodents [32]. PPY administration in humans also leads to a sustained decrease in both appetite and food intake [33]. However, it has been demonstrated that low levels of PPY in obese children normalise after weight loss, suggesting that levels of PPY may reflect the overweight status rather than cause it [34]. These observations indicate that PPY may be involved in the regulation of energy balance and possibly the aetiology of obesity.

Three previous published studies have focused on both the PYY and NPY2R genes [35–37], while two further studies have focused on PYY variants alone in association with type 2 diabetes [38, 39] and overweight [39]. In British white subjects two common single nucleotide polymorphisms (SNPs) in exon 2 (585T > C and 936T > C) of the NPY2R gene were in strong linkage disequilibrium (LD). It was found that men homozygous for the rarer variant had significantly lower BMI (kg/m2) (p = 0.017) and WHR (p = 0.013) and also higher NEFA levels (p = 0.01) [36]. Variations in PYY and NPY2R gene were also found to be associated with obesity in male Pima Indians. It was found that two variants in the 5′ flanking region of the PYY gene and an intron 3 variant (rs162430) were associated with severe obesity (p = 0.001). Similar to the British result there was a significant association for the two common SNPs in exon 2 of the NPY2R gene with severe obesity (p = 0.002) [35]. The study in the Swedish men concentrated on the previously associated NPY2R gene variant 585T > C (rs1047214) and the PYY gene Arg72Thr variant (rs1058046). No association was found for the PYY gene variant; however, the common TT genotype of the 585T > C variant was found to be protective against obesity in the Swedish men (p = 0.002) [37]. Less is known about rare variants in these appetite-regulating genes and their role in predisposition to obesity. So far only one study has shown an association between a rare variant (Q62P) in the PYY gene and susceptibility to obesity [40]. This study also reported that a common variant in the PYY gene (R72T) was associated with obesity (p = 0.02) [40] and the same variant (R72T) has also been associated with overweight in a Danish study group [39].

In our study, a total of 16 common SNPs (minor allele frequency ≥0.05) were analysed: three SNPs in PYY, two SNPs in PPY, seven SNPs in the NPY2R gene and a further four SNPs in the 10-kb region between PYY and PPY. These were examined for association with severe obesity in 557 French white obese adults and 515 obese children with a total of 1,163 non-obese control subjects. A total of nine rare variants were also examined in 700 obese children (from 479 families) and the non-obese/non-diabetic control subjects (n = 1,163). If the rare variant was not present in the control subjects and only in the obese children then it was subsequently followed up in the family of the obese children to test for co-segregation with obesity.

Subjects and methods

Informed written consent was obtained from all the subjects before participation. The Ethical Committee of Hotel Dieu in Paris and CHRU in Lille approved the genetic study.

Subjects

Initial SNP screening was performed in 105 morbidly obese (BMI ≥ 40 kg/m2) adults, 56 morbidly obese children with z score of BMI (ZBMI) ≥ 2 and an early onset of obesity. The average age of onset of childhood obesity was 4.0 ± 2.7 years and the maximum age of onset was 18 years. The 97th percentile was used as the threshold of obesity for the French children in accordance with the European Childhood Obesity Group recommendations [41]. For the case-control study, 557 (131 males, 426 females) unrelated adult morbidly obese subjects were used and 515 (226 males, 277 females, sex data missing for 12 subjects) unrelated obese children were used (age < 18 years with a BMI > 97th percentile for age and sex). The controls used for the study were all non-obese and non-diabetic (n = 1,163); further details are provided in the Electronic Supplementary Material (ESM).

Fasting blood glucose was used to define NGT with Gly0 < 5.6 mmol/l used to define the non-diabetic state for all control subjects. A total of 700 obese children (341 males, 359 females) from 479 distinct pedigrees were analysed for the co-segregation analysis of the rare variant; these children had a BMI > 97th percentile (ZBMI = 3.98 ± 2.57 kg/m2, age at examination = 11.36 ± 3.15 years, age of onset of obesity = 3.91 ± 2.57 years).

Phenotypes

Phenotypes were measured as previously described [42]; further details are available in the ESM.

PCR and sequencing

Primers were designed for all three genes using the Primer 3 program [43] to include 1 kb upstream from the first exon and 1 kb downstream from the last exon covering the entire gene. All primer sequences are available on request. There were a total of eight primer pairs designed for PYY, 13 for NPY2R and 10 for PPY. PCR and sequencing was carried out as previously [44].

Genotyping

Genotyping of the majority of SNPs was performed using the Sequenom MassARRAY system [45] as previously [44]. Two NPY2R SNPs (rs17376798, rs6857530) were genotyped using the fluorescent 5′ nuclease Taqman method on an ABI PRISM 7900 HT Sequence Detection System. The conditions for Taqman reaction were as follows: 50°C for 2 min, 95°C for 10 min, 95°C for 15 s, 60°C for 1 min; for 50 cycles. Taqman MGB probes were from Applied Biosystems, Foster City, CA, USA.

Statistical analysis

None of the SNPs significantly deviated from Hardy–Weinberg equilibrium (p > 0.05). All the SNPs had a genotyping success rate of above 75%, with an average rate of 88% overall. Furthermore, we also genotyped 100 severely obese subjects by sequencing and this showed 99% concordance between the results for Sequenom and sequencing. LD was estimated using the expectation–maximisation algorithm implemented in GOLD [46]. Statistical analyses of SNPs (chi-squared test and quantitative trait analysis) were performed using SPSS software (version 11.0, SPSS, Inc., Chicago, IL, USA). Linear regression was used to assess the effects of the genetic variable on a given phenotype. Differences between the genotypes in a co-dominant model were tested with one-way ANOVA for all the quantitative traits studied. However, if any genotype groups had a sample size below 30, the Kruskal–Wallis test was used. Differences between two groups of genotypes were tested using a parametric t test. Haplotype frequencies were determined and compared between the groups of cases and controls using UNPHASED [47]. In consideration of the number of statistical tests carried out, a simple correction for multiple testing of SNPs in LD with each other was carried out using the method proposed by Nyholt [48].

Results

Sequencing

The locations of all the SNPs are shown in Fig. 1.

Fig. 1
figure 1

Schematic diagram of PYY, PPY and NPY2R genes showing SNPs identified by sequencing. a chromosome 17q21; b chromosome 4q31. All the SNPs shown were identified by sequence analysis, exons are shown as boxes; filled box regions represent the coding region; patterned box regions represent the non-coding region; numbers represent the exons. SNPs in italics represent the rare variant (frequency < 0.05) or those in LD with other variants

Full size image

PYY gene

Seven variants were detected by sequencing in the PYY gene (Table 1). Four of these variants were common, i.e. had a minor allele frequency of ≥0.05 and three were rare. Initial LD was calculated between these SNPs and rs2070592 in the 5′ UTR and rs1058046 in exon 3 in high LD (r 2 = 0.955, D′ = 0.985, p < 0.001) (data not shown), so only the latter was selected for genotyping.

Table 1 Sequencing results for NPY2R, PPY and PYY genes
Full size table

PPY gene

Four variants were identified (Table 1): two common and two rare. Figure 1 shows the location for all PPY variants.

NPY2R gene

A total of 12 variants were detected by sequencing in the NPY2R gene (Table 1). Eight of the variants were common and four rare. High LD was observed between the two common variants in exon 2 (r 2 = 0.942, D′ = 0.985, p < 0.001) so only one (rs1047214) was selected for genotyping.

Genotyping

The SNPs selected for genotyping in the obese cases and controls are shown in Table 1. SNPs that were in high LD were not selected for further genotyping. Four additional SNPs were selected directly from dbSNP (http://www.ncbi.nlm.nih.gov/SNP/) within the 10-kb region between the PYY and PPY genes. These were frequent (minor allele frequency > 0.38) and on average 2 kb apart. Pairwise LD was calculated for all the common variants (frequency > 0.05) for NPY2R SNPs at chromosome 4q31 and for PYY and PPY SNPs at chromosome 17q21 region; the results are shown in Fig. 2a and b. For a relative risk of 1.2 it is estimated that we have an 84% power to detect an association with severe obesity in our population using 1,163 controls and our combined 1,072 cases. The statistical power is reduced when analysing adult and child cases separately; however, for a slightly raised relative risk of 1.4 we still have 68% power to detect an association with severe obesity at p < 0.05 [49].

Fig. 2
figure 2

Graphical representation of LD for NPY2R, PPY and PYY genes. a Pairwise LD measures (D′) for common SNPs spanning the NPY2R gene on chromosome 4q31. b Pairwise LD measures (D′) for common SNPs spanning the PPY, 10-kb region and PYY gene on chromosome 17q21. All SNPs are positioned according to NCBI chromosome location

Full size image

PYY and PPY genes

LD was calculated over the ∼12-kb region containing both the PYY and PPY genes. The average D′ across the region was 0.74. Strong LD was observed between the PPY variants rs231471 and rs231473 (D′ = 0.99). Similarly, complete LD was observed for PYY SNPs rs162431 and rs1058046 (D′ = 1).

We examined whether the PYY and PPY SNPs were associated with obesity in both severely obese adults (BMI ≥ 40 kg/m2) and children (age < 18 years with a BMI > 97th percentile for age and sex). It was observed that one C/T variant in intron 3 (rs162430) of the PYY gene was associated with childhood obesity (p = 0.04), but this was not found to be associated with severe adult obesity (p = 0.84) (Table 2). The minor T allele associated with obesity, with a frequency of 13% in the child cases compared with 10% in the controls (p = 0.03, odds ratio [OR] = 1.32 [95% CI 1.04–1.67]) (Table 2).

Table 2 Genotype distribution of common PYY and PPY SNPs in obese adults and children
Full size table

NPY2R gene

The average LD (D′) observed across the NPY2R gene was 0.80. Strong LD was observed between the 5′ flanking region SNPs compared with the exon 2 SNP (Fig. 2b). SNP G-6503A was in strong LD with rs6857530 (D′ = 0.99), rs6857715 with rs6857530 (D′ = 0.99) and rs12507396 with rs2234759 (D′ = 0.97).

One variant (rs6857715) in the NPY2R gene was found to be significantly associated with both severe adult obesity and childhood obesity (p = 0.002 and p = 0.02). The minor T allele was significantly associated with severe obesity in adults (OR = 1.4 [95% CI 1.1–1.6], p = 0.0005) and children (OR = 1.2 [95% CI 1.0–1.4], p = 0.02) (Table 3). Pooling data from obese children and obese adults vs controls slightly increased the significance of association (OR = 1.3 [95% CI 1.1–1.5], p = 0.0004).

Table 3 Genotype distribution of common NPY2R SNPs in obese adults and children
Full size table

Haplotype analysis was performed for the common variants using UNPHASED for all the obese cases and controls for both NPY2R and the PPY gene, and the 10-kb region and PYY gene separately. There was no association with obesity for haplotypes above a frequency of 5% for NPY2R or the PPY, 10-kb region and PYY with obesity. Association with obesity was observed for haplotypes of low frequency (<5%) but these may be artefacts of the haplotype prediction process (data not shown).

Quantitative trait analysis

Quantitative analysis of BMI and WHR was performed for all 16 common SNPs in three different groups: 557 severely obese adults, 662 severely obese children and 546 Supplementation en Vitamines et Minéraux Antioxidant study (Suvimax) subjects (largest set of controls). No association was found with BMI and WHR in the Suvimax control subjects for any of the tested SNPs (p > 0.05); similarly no association was found in the severely obese adults. However, a T to C variant in exon 2 (rs1047214) of the NPY2R gene was found to be significantly associated with an increase in WHR in severely obese children (p = 0.005), and this was stronger when analysed using a recessive model (TT + TC: WHR = 0.92 ± 0.08 vs CC: WHR = 0.95 ± 0.09) (p = 0.001). Further analysis by sex demonstrated that it was still associated with an increase in WHR in males (p = 0.01) and females (p = 0.05) (Table 4). Lastly, age of onset of obesity was also analysed as a quantitative trait in the obese children and no association was found with the studied common variants (data not shown).

Table 4 Association studies of SNP rs1047214 in the NPY2R gene with BMI and WHR
Full size table

Rare variants

Eight out of the nine observed rare variants were present in non-obese controls (n = 1,163) and children (data not shown). One NPY2R gene rare variant (C-5641T) was not found in the non-obese control subjects (n = 1,163). The C-5641T variant was only observed in obese children belonging to one family when a total of 700 obese children (from 479 families) were analysed (Fig. 3). Four subjects carrying the variant T allele were severely obese (the mother [BMI = 47.46 kg/m2] and three offspring [ZBMI = 3.30, 2.40 and 2.91, respectively]), while the father and his mother carrying the C-5641C genotype were lean (BMI = 24.26 and 25.28 kg/m2, respectively).

Fig. 3
figure 3

Co-segregation of rare NPY2R variant C-5641T in an obesity-enriched pedigree. Circles, female subjects; boxes, male subjects; black boxes or black circles, affected individuals within the pedigree; empty circles or empty boxes, unaffected individuals

Full size image

Discussion

Over the last few years much emphasis has been placed on the role of PYY, PPY and NPY2R in relation to obesity [6, 12, 16, 22, 28, 33] in both humans and animal models. From the literature it has been demonstrated that NPY2R is a crucial receptor in appetite regulation. NPY2R knock-outs have been shown to increase food intake, weight and adiposity [25]. Furthermore, in other mouse models, it has been observed that NPY2R regulates production of NPY, agouti-related peptide (AgRP), proopiomelanocortin (POMC) and cocaine and amphetamine regulated transcript (CART), and these changes are associated with functional changes in food intake [50]. Abbott et al. [51] have demonstrated that blockade of the NPY2R with a specific antagonist lowers the effect of endogenous and exogenous PYY3–36 on food intake. This study supports that NPY2R plays an important role in post-prandial satiety. There has also been discussion regarding potential therapeutic manipulation based on development of Y2 agonists, exogenous administration of PYY3–36 or increased endogenous release from the gastrointestinal tract [52]. These physiological studies draw attention to the importance of both PYY and NPY2R with regard to appetite regulation.

An SNP (rs6857715) in the 5′ flanking region of the NPY2R gene was significantly associated with severe obesity in both French white adults (p = 0.002) and children (p = 0.02). The variant T allele was associated with adult obesity (OR = 1.4 [95% CI 1.1–1.6], p = 0.0005) and with childhood obesity (OR = 1.2 [95% CI 1.0–1.4], p = 0.02). It is also important to note that the association in severely obese adults remains significant even after correcting for multiple testing using the method of Nyholt [48]. When analysing each sex separately, the SNP rs6857715 was significantly associated with adult obesity in female (p = 0.003), but not in male (p > 0.05) subjects, possibly due to the greater number of female subjects (426) compared with men (131). In Pima Indians, the NPY2R SNP rs6857715 was associated with obesity [35], using a co-dominant model, but the results were non-significant (p = 0.067). However, using a recessive model it was found to be significantly associated with severe obesity (p = 0.02). These results are in agreement with ours, as using a recessive model we also demonstrate significant association with adult obesity (p = 0.006). However, from our results it was observed that the dominant model (CC vs CT + TT) for SNP rs6857715 was more significant for the obese children (p = 0.007), and this model was also found to be more significant in the obese adults (p = 0.004). This SNP was not studied in either the British white subjects [36] or the Swedish men [37].

Using the Vista Genome Browser (http://www.pipeline.lbl.gov/cgi-bin/gateway2?bg=hg17&selector=vista), the NPY2R gene region where the rs6857715 SNP is located (chr4:156486787) in humans was tested for alignment in other species (dog, zebra fish, chicken, frog, fugu, rat, mouse and cow). It was found that the SNP was conserved as a ‘C’ nucleotide only in the cow.

No known transcription factor binding sites could be identified using either Genomatix (http://www.bfam.bio.wzw.tum.de/partners/genomatix) or Tess (http://www.cbil.upenn.edu/tess) that overlapped with rs6857715. A possible explanation for this association could be that this variant is in LD with the true functional variant. In our results the rs6857715 was in LD with the following SNPs: rs17376798 (D′ = 0.833), G-6503A (D′ = 0.859), rs12507396 (D′ = 0.974), rs6857530 (D′ = 0.994) and rs2234759 (D′ = 0.946). However, these SNPs were not associated with obesity and this is probably due to differences in allele frequencies.

Unlike the previous three studies, there was no association of the NPY2R rs1047214 variant with obesity when considering all subjects and even when separating based on sex (p > 0.05). When analysing the quantitative traits of BMI and WHR for all the SNPs, it was observed that there was a significant association with SNP rs1047214 and an increase in WHR in severely obese children (p = 0.005); this was more evident in the male children compared with the females. A significant association was also observed for rs1047214 variant with an increase in weight at age 20 for the morbidly obese adult males. Thus our results support the fact that this SNP may be important in determining BMI and WHR especially in young males. However, there was no significant association detected in the case-control analysis for rs1047214 and this could be due to the fact that subjects were selected based on BMI and not WHR. We were unable to demonstrate the previously described association of rs1047214 SNP with fasting NEFA due to our study design not including this quantitative measurement.

The difference in results for the qualitative analysis of adult obesity and association with NPY2R rs1047214 variant could be due to a number of reasons. One reason may be the differences in the number of individuals between ours and the Pima Indian study. In the Pima Indians male-only analysis the total number of subjects was only 167 (100 obese and 67 lean subjects). In the Swedish subjects there were a total of 148 lean, 129 overweight and 226 morbidly obese males. In the current study there were 648 male subjects (131 obese and 517 lean subjects). The second reason could be the difference in ethnic origin of the subjects.

There was no association with the quantitative measures of BMI and WHR for the three obesity associated SNPs (PYY rs162430 and NPY2R rs6857715 and rs2234759), while these were found to be associated with obesity in the case-control analysis. This could be due to the fact that obesity is not just purely defined by BMI and WHR but it is a complex entity and it could be due to an influence of the variant on other obesity traits, such as feeding behaviour, for which we did not have adequate numbers for analysis.

Rare variants were also studied, and one NPY2R variant, C-5641T, was not found in the lean controls (n = 1,163). In 700 obese children from 479 distinct pedigrees it was found to co-segregate with obesity in one family and may play a potential role in predisposition to monogenic obesity. This variant is not conserved in other species (data not shown); however, in order to establish the possible functional role of this variant, the sequence containing the NPY2R rare variant was analysed using Genomatix and the T allele introduces a transcriptional factor binding site for steroidogenic factor-1, which is a member of the nuclear hormone receptor family of transcriptional regulators and acts at multiple levels of the hypothalamic-pituitary-steroidogenic organ axis to regulate the expression of genes that are important for regulated steroidogenesis. The exact mechanism of how this may influence obesity is unclear. It is necessary to study additional pedigrees carrying the C-5641T mutation to confirm its implication in monogenic forms of obesity. We did not study the previously associated PYY rare variant (Q62P) [40] as this was not found by sequencing in our subjects (n = 161).

There was only marginal association to one PYY SNP (rs162430). This intron 3 SNP was found to be significantly associated (p = 0.04) with childhood obesity but not adult obesity. The rare T allele was more prevalent in the obese children (13%) compared with the control subjects (10%) (OR = 1.32 [95% CI 1.04–1.67], p = 0.03). However, these results are not significant if corrected for multiple testing. This variant was also found to be significantly associated (p = 0.001) in Pima Indian men with severe obesity [35]. However, we found no significant association when restricting the analysis to male subjects (p > 0.05). A possible explanation of the stronger association could be high LD with two novel 5′ variants identified in the Pima Indians, which were not tested in our population as they were upstream of the 1-kb region covered in our study.

In summary, significant association was observed for a 5′ flanking variant (rs6857715) of the NPY2R gene with both childhood and adult obesity in French white subjects. There was also a significant association with an increase in WHR in severely obese children and rs1047214 variant of the NPY2R gene. One rare NPY2R variant (C-5641T) was found to co-segregate with obesity in one family, but further work is necessary in other pedigrees to confirm these results. Association was also observed for the PYY variant rs162430 with childhood obesity; however, this becomes non-significant if multiple testing correction is applied. Lastly no association was observed for variants in the 10-kb interval between PYY and PPY. To the best of our knowledge, this is the first study that has examined the genetic variants in the PPY gene in relation to obesity, in both morbidly obese adults and children. So far there is a great deal of evidence for a crucial role of the PPY gene in obesity, and it is important to examine the hypothesis that genetic variants could be responsible for variable levels of expression of PPY in obese subjects. However, our study, which was purely genetic, shows that variants in the PPY gene are not associated with severe obesity.

In conclusion, our work builds upon the previous studies and suggests that on a genetic level specific variants in the NPY2R gene are associated with obesity and further work is necessary to understand the physiology of the receptor.