Correction: Copy number of pancreatic polypeptide receptor gene NPY4R correlates with body mass index and waist circumference

[This corrects the article DOI: 10.1371/journal.pone.0194668.].


Introduction
Excessive weight gain has become one of the major health problems worldwide. According to a report from 2013, 37% of men and 38% of women had overweight or obesity [1]. Increased body mass index (BMI) is associated with increased mortality from cardiovascular disease, type 2-diabetes and several types of cancer [2]. Increased waist circumference (WC) is associated with abdominal obesity and increased risk for metabolic complications [3].
Obesity is a complex, polygenic, and highly heritable disease. Heritability of BMI ranges between 24-80% in family studies and 47-90% in twin studies (for review see [4]). Multiple studies have demonstrated that structural differences in the genome, such as copy number variation (CNV), are associated with variation in BMI [5] and obesity [6,7]. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 A previous genetic screen of a cohort of obese German children carried out by the genetics company IntegraGen revealed an association between a genomic region on chromosome 10q11. 22 and obesity (personal communication with Dr Jorg Hager, see S1 Appendix). Among the genes located in the region, NPY4R was the strongest candidate for association with obesity and was found to have nonsynonymous single nucleotide polymorphisms (SNPs) that segregated with childhood obesity.
The copy number variable region on chromosome 10q11.22 spanning across the NPY4R, SYT15 and GPRIN2 genes has been previously described by several research groups [8][9][10][11]. The first study of CNV in this region and association with BMI reported an inverse correlation: a higher gene copy number was associated with reduced BMI in an elderly Chinese cohort [5]. A similar correlation was subsequently confirmed in a German cohort [6] and a Belgian cohort of children and adolescents with obesity and healthy adults with normal weight [12]. However, in a study of young Chinese individuals no CNV was detected in the 10q11.22 region [13].
The NPY4R gene encodes the Y4 receptor that responds to pancreatic polypeptide (PP). This gene is a strong candidate for body weight regulation because PP has been reported to be a potent appetite inhibitor [14]. There are four NPY-family receptors in humans and all of them are expressed in the brain, especially in the hypothalamic regions that are involved in the control of appetite and energy metabolism [15], as well as in the periphery (for Y4 see [16]).
PP is released from pancreatic PP-cells, previously called F cells, postprandially in proportion to caloric intake [17]. Intravenous administration of PP causes reduced energy intake in both individuals with normal weight and those with obesity [14,18]. Peripheral administration of PP decreases the hypothalamic expression of the potent hunger stimulants NPY, ghrelin and orexin, and increases anorexigenic urocortin in animal models [19].
PP affects appetite by acting through Y4 receptors in the regions that play a crucial role in energy balance, like the dorsal vagal complex, area postrema and the nucleus of the solitary tract in the brain stem [20,21], the arcuate nucleus [21], lateral hypothalamic area [22], and the paraventricular and ventromedial nuclei [21] of the hypothalamus.
To address the relationship between NPY4R copy number and obesity, we present here a study of 558 individuals with a wide range of BMIs. We investigated the associations between copy number and each of the following parameters: BMI, WC and self-reported dietary intake data from a validated questionnaire.

Study populations
The study populations included participants from the Swedish Obese Subjects (SOS) study [23], the SOS reference study (SOS-ref) [24] and the SOS SibPair study [25]. The SOS study was started in 1987 and is a prospective, controlled, intervention study involving 4047 individuals; 2010 individuals have undergone bariatric surgery and 2037 conventional treatment (matched control group). Minimum BMI for inclusion was 38 kg/m 2 for women and 34 kg/m 2 for men. Average BMI at baseline was 42.2±4.5 kg/m 2 in the surgery group and 40.1±4.7 kg/ m 2 in the control group. The SOS-ref study includes subjects from the Swedish cities Mölndal and Ö rebro. The subjects were randomly selected from a population registry to constitute a cross-sectional reference group to the SOS study [24]. The study includes 1135 subjects (46.5% men), average BMI is 25.2±3.8 kg/m 2 . The SOS SibPair study [25] consists of 732 individuals from 154 Swedish nuclear families with sibling pairs discordant for obesity, defined as a BMI difference of at least 10 kg/m 2 . BMI range was 16.9 to 57.8 kg/m 2 . The SOS SibPair study was specifically designed to study genetic aspects of obesity.
In total, 558 individuals from these three study populations were included in the present study (hereafter referred to as the study sample) with the aim to cover a wide BMI range. Measurements of BMI and WC as well as self-reported dietary intake data on food and beverage intake were available for all individuals (See S1 Table).
All procedures performed in the study involving human participants were in accordance with the ethical standards of the local and regional review boards and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed oral or written consent was obtained from all participants, according to the regional ethical review boards' guidelines and regulations. The following seven regional ethics review boards approved the study: Gothenburg, Lund (Region of Skåne), Linköping, Ö rebro, Karolinska Institute (Stockholm), Uppsala and Umeå.

Questionnaire data collection
All participants in SOS, SOS-ref and SOS SibPair completed a semi-quantitative dietary questionnaire on habitual food and beverage intake covering the last three months [26]. The subjects specified intake frequency of standard portions of different foods. The dietary questionnaire included 51 questions and had been validated against a 4-day food record [26,27] in groups of normal weight, overweight as well as adults with obesity. From the dietary questionnaire, total energy and macronutrient intake were calculated as well as energy intake from 8 carbohydrate-and / or fat-rich food groups (spreads, sandwiches, desserts, fruits, non-alcoholic beverages (excluding milk) salty snacks, candy, and buns and cakes).

Droplet digital PCR
We used droplet digital PCR (ddPCR) in order to study the copy number variation of the NPY4R gene (See S1 Table). Fluorescently labelled target (NPY4R) and reference (RPPH1) assays were designed according to the guidelines from Bio-Rad Laboratories.
DNA was digested with BstXI restriction enzyme (10U/μl, ThermoScientific, Cat#: ER1021) in Buffer 0 for 1 hour at 55˚C, followed by 20 min at 80˚C. A 20 μl mixture of 2×ddPCR Supermix for Probes (Bio-Rad, Cat#: 186-3010), forward and reverse primers for target and reference assay (final concentrations of 900nM each), probes for both assays (final concentrations of 250nM each) and 15 ng of digested DNA was emulsified with Bio-Rad Droplet Generator Oil (Bio-Rad, Cat#: 186-3005) in a Bio-Rad QX100 TM Droplet Generator (Bio-Rad, Cat#: 186-3001) according to the manufacturer's instructions. The droplets were then manually transferred to a 96-well plate (Eppendorf, Cat#: 951020362) and heat-sealed with Easy Pierce sealing foil sheets (Thermo Fisher Scientific, Cat#: AB-0757). Polymerase chain reaction was performed in a Bio-Rad C1000 thermal cycler (Bio-Rad, Cat#: 185-1197) with the following cycling parameters: 10 min at 95˚C (1 cycle), 30 s denaturation at 94˚C and 1 min annealing and extension at 58˚C (40 cycles), followed by 10 min at 98˚C and a hold at 12˚C. All steps had a ramp rate of 2˚C/s. After the PCR, droplets were analysed using a Bio-Rad QX100 Droplet Reader (Bio-Rad, Cat#: 186-3001). Fluorescent data from each well were analysed with Quan-taSoft software (v1.3.2), where copy number was calculated based on Poisson distribution [28].
We have tested the reliability of our copy number results by using the EIF2C1 reference assay (which now is one of the Bio-Rad recommended reference assays for copy number analysis using ddPCR) on a randomly selected subset of samples, including samples with non-integer copy number.

Data analysis and statistics
For genotype frequency distribution, copy number data was binned to the closest integer (e.g. 2 = 1.5-2.49). All normally distributed data was analysed using Pearson correlation and linear regression. β-coefficient was used to quantify the association and adjusted R 2 was used to assess the quality of the model. We used backward stepwise regression, where all independent variables were added to the regression model and then removed one by one, in order to optimize the model. Non-normally distributed questionnaire data was analysed using Mann-Whitney U-test. False discovery rate analysis was performed in order to correct for multiple testing. To account for known sex-related differences in BMI [29] and also to investigate any potential sex differences in the impact of CNV on body weight, we analysed our data for men and women together and separately.

Results
Basic characteristics of the study sample including age, weight, height, BMI and WC are summarized in Table 1.
Droplet digital PCR analysis demonstrated that NPY4R gene copy number varied from 2 to 8 (2.10-7.51 in men and 2.35-7.95 in women) with 4 being most common in BMI 25.0 kg/ m 2 (average 4.26±0.61 for men and 4.09±0.62 for women). In the study group, 84% had 4 or 5 copies (average 4.71±0.80 for men and 4.53±0.81 for women) ( Table 2).
In men and women combined, a positive correlation between NPY4R copy number and BMI was found (r = 0.206, p = 4.85×10 −7 ). It was also found in women only (r = 0.267, p = 2.65×10 −7 ) (Fig 1), whereas no statistically significant correlation between NPY4R copy number and BMI was found in men (r = 0.098, p = 0.075). Linear regression analysis demonstrated that each additional copy corresponds to an increase of 2.60 kg/m 2 in BMI (p = 3.32×10 −7 ) in women. Waist circumference correlated with NPY4R copy number in men and women together (r = 0.189, p = 3×10 −6 ) and in women only (r = 0.256, p = 8×10 −7 ), but not in men (r = 0.055, p = 0.21) (Fig  2). An increase of one copy was associated with 5.67 cm increase in WC (p = 1×10 −6 ) in women. We observed no correlation between NPY4R copy number and age, neither for the whole study group, nor for men and women separately.
Total energy intake, energy percent of macronutrients and energy intake from 8 food groups in relation to NPY4R copy number are presented in Table 3. No correlations were observed between total energy intake or energy intake from specific food groups and NPY4R copy number, neither in the entire study group, nor in men or women separately. No correlations were found between energy percent from carbohydrates, protein or fat and NPY4R copy number in the entire study group or in men or women separately. Total energy intake adjusted to body weight had a strong negative correlation with the NPY4R copy number in the whole study (r = -0.199, p = 1×10 −6 ) sample and women separately (r = -0.239, p = 4×10 −6 ). Each additional copy was associated with a decrease of 3.09 kcal/kg in the whole study sample (p = 2×10 −6 ) and a decrease of 3.54 kcal/kg in women separately (p = 8×10 −6 ). There was no statistically significant correlation between the total energy intake adjusted to body weight and the NPY4R copy number in men.

NPY4R CNV in the study sample
Droplet digital PCR has recently emerged as the most accurate way for absolute DNA copy number quantification [30]. Here we report copy number state of the NPY4R gene in 558 adult Swedish individuals from the SOS, SOS Ref and SOS SibPair cohorts, representing a wide range of BMIs. We investigated the relationship between the CNV and BMI, WC and selfreported energy intake. We found that the copy number of NPY4R varies from 2 to 8 copies per genome. Our results demonstrate a positive correlation between NPY4R copy number and both BMI and WC for the entire study group and for women only. The findings we describe here differ from previous studies with respect to normal NPY4R copy number, the copy number distribution, and its association with body weight. Several of these studies reported that copy number loss  in this genomic region was associated with weight gain. Such negative correlation has been described for an elderly Chinese cohort of 597 individuals [5], a German cohort of 3255 individuals [6] and a Belgian cohort of 622 individuals [12]. In contrast, a study of 12 females with Rett syndrome found a positive correlation of NPY4R copy number with weight gain [31]. A genome wide association study (GWAS) of obesity-related CNVs reported that three out of 430 individuals with obesity were "carrying this CNV", whereas none of the 379 controls with the normal weight did [10]. A study of 799 young Chinese individuals could not detect CNV of NPY4R, neither in subjects with obesity nor in subjects with normal weight [13]. Genetic differences between populations could be one of the reasons for differences between our results and the previous findings, since four of the previous studies have been performed in Asian cohorts [5,9,10,13]. However, our study of NPY4R copy number in a subset of samples from the 1000 Genomes Project (S2 Table) shows no striking copy number difference between samples from Asian and Caucasian populations. Thus, we suggest that these differences may be due to an incorrect assumption about the normal copy number of the NPY4R as well as methodological and cohort differences between our and earlier studies. Methodologically, most of the previous studies were based on SNP-arrays, aCGH and RT-PCR-based methods that require a reference copy number (most commonly set at 2 copies per genome) [5,9,13] or depend heavily on relative fluorescence data quality [32]. In contrast, we used the ddPCR method, which allows for more precise quantification of target nucleic acid [30] and that has been validated for absolute copy number determination [28,33]. It is equally [34] or more reliable than other molecular methods of copy number determination [35], depending on the copy number distribution and the complexity of the region. Incorrectly chosen reference copy number or inappropriate choice of reference gene in PCR-based copy number determination methods represent sources of errors in CNV-studies. Therefore, we have used two reference assays: RPPH1 and EIF2C1. We initially chose RPPH1 as a reference gene because it is a well-known single-copy gene [36]. EIF2C1, a newer and now recommended reference assay for CNV detection, was used as a control in a subset of the samples. We found no differences in the results obtained with RPPH1 compared with EIF2C1.
Selection of study populations is an important factor in obesity studies. Our study population consists of Swedish adults of both sexes and covers a wide range of BMIs (for details see Table 1). When studying BMI, it is important to take into consideration both the sex and age distribution. Age differences were accounted for using age as a covariate in the linear regression analyses. To account for sex-related differences in BMI, we analysed men and women together as well as separately.

Inverse correlation between body weight, waist circumference and NPY4R copy number
In our study sample, both BMI and WC exhibited positive correlation with the gene copy number. Our results in women demonstrated that for each additional copy of NPY4R the BMI increased with 2.60 kg/m 2 . In comparison, in men and women each FTO risk-allele increases BMI by 0.34-0.46 kg/m 2 [37]. The analyses of the correlations in the separate sexes showed a correlation in women but not in men. Studies indicate that there are sex-specific genetic factors contributing to obesity development [29]. Our finding of a correlation between NPY4R copy number and both BMI and WC in women, but not in men, adds support to the idea that different genes contribute to the variation in BMI in women and men.
We addressed the question whether dietary intake correlates with NPY4R copy number by analysing self-reported food intake. We observed no correlation between NPY4R copy number and total energy intake or energy percent of macronutrients, which may indicate that NPY4R influences body weight through metabolic pathways, rather than food intake. Alternatively, the lack of associations between NPY4R copy number and dietary intake may be due to the large variation in daily energy need in our study sample that consists of individuals with BMI ranging from 17-58 kg/m 2 . There is also a well-known misreporting of self-reported food intake data affecting large population-based samples [38] as well as sub-groups of different BMIs [39,40]. Under reporters of energy intake are more often obese and overestimation of energy intake are more often seen in normal weight groups [38]. Also depending on BMI, misreporting of energy intake can be specific so that social desirable foods (i.e. high-fat, sugar-rich foods and beverages) are more often underreported by subjects with obesity and low-fat, fibrerich foods are more often over-reported by normal weight and underweight groups [38,41]. The present study includes subjects covering a large BMI-interval. The questionnaire used in the present study has been found to be equally valid and reproducible for individuals with obesity as well as individuals with normal weight [26] and has previously been used in normal weight, overweight and obese study groups [23,40,42].
We have investigated the relationship between total energy intake adjusted to body weight and NPY4R copy number. We found a negative correlation for the whole study sample and women separately. Taking into account the role of Y4 and PP in mediation of satiety [14,18], such negative correlation would mean that individuals with more NPY4R copies experience more satiety. At the same time, we see that individuals with more NPY4R copies have higher BMI and WC. Such seemingly contradictory findings might be explained by differences in energy metabolism between individuals with low and high NPY4R copy number. It is also important to mention that the role of PP (and hence, its receptor Y4) in food intake and energy metabolism regulation is no so univocal. In a long-term study of hormonal levels after weight loss in adults [43], the level of satiety hormones was significantly decreased while the levels of hunger hormones increased significantly, as if trying to bring the individuals back to higher body weight. Surprisingly, the level of PP increased after weight loss as if it were a hunger-inducing signal. The same effect was observed in a study of weight loss in obese children [44]. In the light of aforementioned possible misreporting in self-reported food/energy intake and the complex role of the PP/Y4 system we think that these results should be interpreted with great care.
We observed that NPY4R copy number is not always an integer. Therefore, we binned the data to whole numbers (±0.5) for genotype frequency estimation. Variation in the data due to ddPCR methodology is unlikely given the precision of the method [28] and consistency of non-integer results despite troubleshooting [45] and replication. We instead suggest that noninteger copy number may be a result of somatic mosaicism in peripheral blood DNA. Previously reported analysis of somatic mosaicism in tissues of healthy donors have revealed CNV of multiple regions, some spanning genes [46]. Also, mosaicism of inversion polymorphisms has been demonstrated in peripheral blood [47]. The investigated rearrangements were more abundant in adults than in new-borns, suggesting gradual accumulation of frequent postnatal rearrangements [47].

Conclusions
We found a positive correlation between NPY4R copy number and BMI and WC in women only, suggesting that the role of NPY4R in body weight may be more complex than previously thought. Thus, our results also lend support to the belief that there are different genetic contributors to BMI variation in men and women.
Our findings of a positive NPY4R copy number correlation with BMI is opposite to what we expected, based on previous studies of NPY4R and its ligand PP. This invites further investigation of their specific roles in appetite regulation and energy metabolism.
Supporting information S1