Association between LEPR, FTO, MC4R, and PPARG-2 polymorphisms with obesity traits and metabolic phenotypes in school-aged children

Purpose Evaluate the relationship of leptin receptor (LEPR) rs1137101, fat mass obesity-associated (FTO) receptors 9939609, melanocortin-4 receptors (MC4R) rs2229616 and rs17782313, and proliferator-activated receptor-gamma (PPARG) rs1801282 with clinical and metabolic phenotypes in prepubertal children. Research question What is the effect of polymorphisms on clinical and metabolic phenotypes in prepubertal children? Methods A cross-sectional descriptive study was performed to evaluate anthropometric features, percentage body fat (%BF), biochemical parameters, and genotype in 773 prepubertal children. Results FTO rs9939609 was associated with an increase in body mass index (BMI) and BMI z-score (zBMI). MC4R rs17782313 was associated with a decrease in BMI and +0.06 units in zBMI. LEPR, and PPARG-2 polymorphisms were associated with decreases in BMI and an increase and decrease units in zBMI, respectively. The homozygous SNPs demonstrated increases (FTO rs993609 and MC4R rs17782313) and decreases (LEPR rs1137101, PPARG rs1801282) in zBMI than the homozygous form of the major allele. In the overweight/obese group, the MC4R rs17782313 CC genotype showed higher average weight, zBMI, waist circumference, waist-circumference-to-height ratio, and waist-hip ratio, and lower BMI, mid-upper arm circumference, calf circumference, and %BF (P< 0.05). FTO rs9939609 AT and AA genotypes were associated with lower triglycerides (P < 0.05). Conclusions We showed that MC4R rs17782313 and FTO rs9939609 were positively associated with zBMI, with weak and very weak effects, respectively, suggesting a very scarce contribution to childhood obesity. LEPR rs1137101 and PPARG-2 rs1801282 had weak and medium negative effects on zBMI, respectively, and may slightly protect against childhood obesity.


Introduction
Obesity prevalence has increased during the past century and the World Health Organization (WHO) estimates that the number of overweight/obese young children will reach 70 million in 2025 [1]. Obesity is a chronic disease with multifaceted etiology [2,3]. Socioeconomic changes during the last decades have contributed to these phenomena, including the increased availability of high-fat foods and generalized adoption of sedentary lifestyles. Furthermore, there is evidence that genes play an important role in the rise of obesity [2,3]. Heritability is estimated to account for 40-90% of the population adiposity variation. Seemingly, the presence of single nucleotide polymorphisms (SNPs) offer a protective factor in the development of noncommunicable diseases, such as obesity related diseases [4][5][6]. With the development of high-throughput genotyping techniques, new approaches such as genome-wide linkage and genome-wide association studies (GWAS) have been used to understand genetic influences in obesity [7]. However, the majority of identified SNPs have unknown biological functions and some of these studies yielded contradictory results, suggesting a need for further examination into the functions of identified SNPs related to obesity.
Fat mass and obesity-associated (FTO) variant rs9939609 was the first [8] locus to be positively associated with obesity-related phenotypes [8][9][10][11][12]. FTO is highly expressed in the hypothalamus and liver, appears to function in the central nervous system, and may have a role in energy balance, food intake regulation, and adipogenesis [13,14]. There may be cross-talk between the FTO protein and leptin, an adipose-derived cytokine implicated in food intake regulation, energy and glucose homeostasis, lipid metabolism, and reproductive function [15]. The influence of leptin on body weight control is mediated by binding to the long isoform of its receptor (LEPR-b), which stimulates gene transcription by activating cytosolic signal transducer and transcription (STAT) proteins [15]. Recent evidence suggests that the LEPR-b-STAT3 signaling pathway may be involved in FTO regulation by restricting energy in the hypothalamus [16] and there is evidence that the leptin receptor (LEPR) variant rs1137101 is positively associated with obesity [17][18][19].
Leptin acts with hypothalamic receptors to induce satiety by inhibiting the orexigenic neuropeptide Y (NPY)/agoutirelated peptide (AgRP) neuronal activity and stimulating the anorexigenic proopiomelanocortin (POMC)/amphetaminerelated transcript (CART) neurons [20]. POMC is cleaved into melancortins and is processed to form the α-melanocyte hormone, which exerts catabolic activity via melanocortin-4 receptors (MC4R) to generate a feeling of fullness to suppress appetite [20]. MC4R is a G-protein-linked receptor widely expressed in the hypothalamus and central nervous system, implicated in energy homeostasis and glucose and lipid metabolism [21]. The MC4R variant rs17782313 was the second gene that was positively associated with common obesity traits [22][23][24][25]. By contrast, the MC4R variant rs2229616 is negatively associated with obesity [26,27].
Peroxisome proliferator−activated receptor-gamma (PPARG) is another gene that has an important role in obesity. PPARG is a member of the nuclear hormone superfamily, which is involved in adipocyte differentiation and glucose metabolism [28]. There are evidences that PPARG deficiency results in increased leptin levels [28]. The PPARG rs1801282 variant is positively associated with obesity and has been extensively examined in epidemiological studies [29].
The aim of the present study was to assess the independent contributions of LEPR (rs1137101), FTO (rs9939609), MC4R (rs2229616 and rs17782313), and PPARG-2 (rs1801282) polymorphisms for clinically overweight or obesity phenotypes and endocrine-metabolic traits in prepubertal children.

Study design and participants
This descriptive cross-sectional study was part of a larger project (Nutritional, Biochemical, and Genetic Study of an Overweight and Obese Child Population in the Southern Region) approved by the Directorate General of Health, the Ministry of Science and Education of Portugal and by the Ethics Committee of the Hospital Garcia de Orta, according to the principles of Helsinki Declaration. The project was conductd from January 2009 to June 2013 in a population of prepubertal children (based on Tanner stage) recruited from 87 public schools in Lisbon and the Tagus Valley metropolitan region. Initially, 5989 subjects were initially recruited based on the assessment of anthropometric measurements, bioelectrical impedance, biochemical and genetic analysis.
To be included in the study, children should have completed nine years old during the ongoing school year, an inclusion criterion that reduced the initial population to 5577 children. The exact chronological age in days was calculated as the date of examination minus the date of birth. Children who transferred to another school prior to completing the minimum required measurements were excluded, further reducing the population to 5514 eligible children. Children whose parent did not provide written informed consent consent or withdrew consent for venous blood sampling, and those who self-reported as not fasting at the time of blood collection were excluded from the study, further reducing the sample size to 1496 children. For polymorphism analysis, 773 children were enrolled (Fig. 1). To address the possibility of self-selection bias, we compared anthropometric and biochemical data between selected and non-selected participants. No significant differences were found (data not shown).

Anthropometric and bioelectrical impedance analysis
Clinical assessments were performed in the schools under the supervision of two pediatric consultants. All anthropometric measurements (weight, height, BMI, BMI z-score (zBMI), waist circumference (WC), hip circumference (HC), waist-hip ratio (WHR), waist-circumference-to-height ratio (WHtR), mid-upper arm circumference (MUAC), calf circumference (CC), percent body fat (%BF), percent skeletal muscle (%SM), and resting metabolic rate (RMR)) were obtained from barefoot participants dressed with lightweight clothing using methods described previously [30]. zBMI was determined using the least mean squares method [30]. Children were categorized as normal weight (control group) or overweight/obese (case group) per the World Obesity/Policy and Prevention standards [formerly International Obesity Task Force (IOTF)] [31].

Biochemical analysis
Participants were instructed to fast for 12 hours before venipuncture for blood sampling in the morning at school. Blood samples were refrigerated at approximately 5°C until transferred to our center immediately processed for serum separation, frozen at −80°C on the same day, and stored until further analysis. The following serum biochemical parameters were assessed: total cholesterol (TC), highdensity lipoprotein (HDL-c), low-density lipoprotein (LDLc), triglycerides (TG), apolipoproteins A1 (Apo 1) and B (Apo B), glucose, creatinine, total proteins, ferritin, serum insulin and leptin. The homeostasis model assessment of insulin resistance (Homa-IR) was calculated from glucose (mg/dl) and insulin (µU/ml) using the following formula: Homa-IR = (insulin (µU/ml) × glucose (mg/dl))/405. All assessments were made using methods previously described

Genotyping
Genomic DNA was extracted from a whole peripheral blood sample using the MagNA Pure Compact Nucleic Acid Isolation Kit (Roche Diagnosis, 03730964001) and the MagNA Pure Compact Instrument (Roche Diagnostics GmbH, Germany) per the manufacturer's instructions. DNA samples were stored at −20°C until use. Real-time polymerase chain reaction (PCR)was performed in 96-well plates on an automated LightCycler 480 Real-Time PCR System (Roche Diagnostics, Vienna, Austria) using Light-Cycler FastStart DNA Master Hybprobe (Roche Diagnostics, Berlin, Germany) and LightSNPs (rs1137101 LEPR, rs9939609 FTO, rs2229616 and rs17782313 MC4R, and rs1801282 PPARG-2; TIB Molbiol Synthese labor, Berlin, Germany). The initial step for the allelic discrimination genotyping assay protocol included preincubation at 95°C for 4 min, followed by 45 cycles of denaturation at 95°C for 15 s, and annealing, extension, and detection for 40 s at 60°C. To assess genotyping reproducibility, 10% of the sample was double-genotyped for all SNPs. Concordance rates >99% were obtained for the five tested SNPs. For negative control, Sterile PCR-grade H 2 O was used for the negative control.

Statistical analysis
SPSS Statistics (IBM, version 24, Armonk, NY, USA) was used for all statistical test procedures. Unless otherwise indicated, variables in tables are means ± standard deviations. Multiple comparisons were made by pairwise t-tests with Dunn-Bonferroni adjustment. Pearson's chi-square statistic was utilized to examine allele/genotype distribution differences across categories and to test for Hardy-Weinberg equilibrium. Phenotype mean differences for each SNP genotype were obtained against the homozygous wild type genotype. Genotype and allele effects on BMI (kg/m 2 ) and zBMI for each SNP were calculated according to their formulae as described in Falconer [32]. Major allele dominance type was evaluated using the relationship between dominance and additive effects. Effect size for polymorphic allele on zBMI was graded according to the following cutoffs: very weak effect (|x|<0.05), weak effect (0.05<|x|<0.2), medium effect (0.8>|x|>0.2), and strong effect (|x|>0.8). A two-sided P < 0.05 determined the significance level for statistical analysis.

Results
A final sample of 773 Portuguese school children (381 boys and 392 girls) with a mean age of 9.81 years were characterized according to clinical and biochemical parameters. Significant gender-related differences were observed; relative to the values in girls, boys were significantly taller and presented higher mean values of %SM, RMR, HDL-c, Apo A1, glucose, and creatinine, and lower mean values of % BF, LDL-c, TG, Apo B, TC/HDL, LDL/HDL, Apo B/Apo A1, TP, and leptin (P < 0.05) ( Tables 1 and 2). Study subjects were stratified by zBMI according to the IOTF category as normal weight (69.9%, n = 540) or overweight/obese (30.1%, n = 233). Several clinical and biochemical features became progressively poorer as BMI increased, with overweight/obese children presenting significantly higher anthropometric, bioimpedance (except for %SM), and biochemical parameters (except for TC, LDL/ Apo B, creatinine, and ferritin) (P < 0.05) ( Tables 1 and 2).
The frequencies of wild type (major) and polymorphic (minor) alleles and the genotypes for overweight/obese and Age in days presented here as age in years. Distributions (mean ranks) vary between groups with different letters: subscript a, subscript b (P < 0.05) c Distributions (mean ranks) and medians are the same between groups (P > 0.05) d Distributions are different between groups (P < 0.05). Test of significance adjustment was performed using the Dun-Bonferroni correction e n = 773, 381, 392, 540, and 233 for overall, M, F, normal weight and overweight/obese f n = 661, 329, 332, 468, and 197 for overall, M, F, normal weight and overweight/obese g n = 392, 195, 199, 278, and 116 for overall, M, F, normal weight and overweight/obese BF body fat, BMI body mass index, CC calf circumference, F female, HC hip circumference, M male, MUAC mid-upper arm circumference, RMR resting metabolic rate, SM skeletal muscle, WC waist circumference, WHR waist-hip ratio, WHtR waist-circumference-to-height ratio, zBMI BMI z-score normal weight children are presented in Table 3. No statistical differences were found in allele/genotype frequencies between overweight/obese (case) and normal weight (control) subjects for all polymorphisms (P > 0.05). The absence of a positive association was extensive, even when the polymorphism effect was analyzed by gender (Table S1) (P > 0.05). Genotype frequencies in both overweight/obese and normal weight were in accordance with Hardy-Weinberg equilibrium, except for the LEPR polymorphism in the control group (P = 0.0058) ( Table 3).
To indirectly account for the childhood obesity propensity of each SNP, association analyses between the selected SNPs and zBMI/BMI were conducted (Table 4). Additive effects (half of the divergence between major and minor allele, homozygous outcome) of each selected SNP on zBMI resulted in the homozygous alleles for these selected SNPs were, on average +0.12 (FTO rs9939609), +0.34 (MC4R rs17782313), −0.30 (LEPR rs1137101) and −2.24 (PPARG rs1801282) zBMI units different than major allele homozygous. Conversely, the dominant effects of all SNPs for zBMI are in the opposite direction of the additive ones, reflecting a recessive zBMI inheritance pattern for these SNPs. LEPR rs11371101 and PPARG-2 rs1801282, on average, reduced zBMI / BMI (kg/m 2 ) by −0.09/−0.25 and −0.15/−0.21, respectively, with an average decreasing  Age in days was converted into age in years to compare groups. Distributions (mean ranks) vary between groups with different letters: subscript a, subscript b (P < 0.05) c Distributions (mean ranks) and medians are the same between groups (P > 0.05) d Distributions are different between groups (P < 0.05). Test of significance adjustment was performed using the Dunn-Bonferroni correction e n = 330, 160, 170, 211, 119 for overall, male, female, normal weight, overweight/obese, respectively Apo A1 apolipoprotein A1, Apo B apolipoprotein B, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, TC total cholesterol, TG triglycerides, TP total proteins effect of change to minor allele (AECME) of −0.48 and −2.34 kg/m 2 on BMI, and −0.17 and −1.68 on zBMI, respectively. By contrast, FTO rs9939609 allele contributed, on average, an increase in our sample BMI by +0.06 kg/m 2 and zBMI by +0.03, with an increase in AECME of +0.12 kg/m 2 and +0.06 for BMI and zBMI, respectively. The MC4R rs17782313 allele, on average, reduced BMI by −0.12 kg/m 2 , with a decrease in AECME of −0.57 kg/m 2 on BMI, and accounted for an increase in effect of +0.06 zBMI units with an associated increase in zBMI AECME of +0. 26. In general, LEPR and PPARG-2 polymorphisms had weak and medium negative (decreasing) effects on zBMI, respectively, whereas FTO rs9939609 and MC4R rs17782313 had very weak and weak positive (increasing) effects, respectively. However, the effect of PPARG-2 polymorphism was determined by screening only two polymorphic homozygous for zBMI / BMI (kg/m 2 ), due to the low relative frequency of the PPARG-2 polymorphic allele in the study population (9%). LEPR and PPARG-2 major alleles (wild type) showed overdominance and partial dominance for zBMI,while the major alleles showed complete dominance for zBMI outcome in FTO rs9939609 and MC4R rs17782313 (Table 4).
We performed an association analysis of anthropometric traits, %BF, and biochemical parameters (dependent variables) on the three different genotypes (independent variable) of the selected genes using zBMI case-control groups (Tables 5 and 6). Significant differences were detected between anthropometric parameters and %BF in the LEPR polymorphism, where homozygous polymorphic normal weight children had significantly lower %BF than wild type homozygous and heterozygous (P < 0.05) ( Table 5). Significant associations were observed for the MC4R rs17782313 CC genotypes in the overweight/obese group, with significantly higher mean scores for weight, zBMI, WC, WHR, and WHtR, and significantly lower mean scores for BMI, MUAC, CC, and %BF (P < 0.05) ( Table 5). Higher mean scores in anthropometric parameters (height, BMI, zBMI, WC, HC, WHR, WHtR, MUAC, and CC) and %BF were also found for the FTO rs9939609 AA genotype in the overweight/obese group, whereas lower mean values of BMI, zBMI WC, WHtR, and %BF were observed for the LEPR rs1137110 GG genotype; however, these differences were not statistically significant (P > 0.05).
Regarding the biochemical parameters, children with LEPR rs1137110 AG and GG genotypes in the normal weight group had significantly higher glucose levels than children with AA genotype (P < 0.05). Children with FTO rs9939609 AT and AA genotypes in the overweight/obese group had significantly lower TG levels (P < 0.05) ( Table  6). For the same polymorphism, normal weight subjects with the AA genotype had significantly higher levels of HDL-c (P < 0.05) ( Table 6).

Discussion
Ethnicity and environmental factors (i.e., modifying the gene expression but not it's structure) may affect specific genetic variants under specific conditions, which may distinctly affect obesity-related phenotypes. Obesity can be associated with different metabolic phenotypes of atherogenic lipid profiles and insulin resistance and several studies have investigated the links between obesity, biochemical traits, and polymorphisms to establish possible mechanisms of action [29,[32][33][34]. This genetic information could be useful to identify children at risk, plan early interventions, and reduce the life-long burden of obesity-related diseases. However, most studies of obesity-SNP associations have yielded controversial results, and the mechanisms underlying the increased risk of obesity conferred by specific alleles remain unclear.

LEPR rs11371101
The LEPR rs11371101 variant is one of the most frequent LEPR gene polymorphisms and the most likely to have functional consequences [35]. Previous studies reported conflicting results with either positive [21][22][23] or no association [36] with obesity traits and metabolic parameters. For example, Pyrzaket et al. analyzed a cohort of 101 obese children (12−18 years old) and found that the LEPR gene variant was not associated with obesity, leptin, insulin resistance, or other metabolic abnormalities [36]. Similarly, Endo et al. verified that the LEPR Gln223Arg (rs11371101) polymorphism was not associated with obesity in 553 Japanese school children aged 9−15 years [37]. A metaanalysis of case-control studies and a systematic review also reported that there was no association between the LEPR gene polymorphism and obesity [35,38]. By contrast, Shabana and Hasnain reported that the LEPR polymorphism was associated with weight, BMI, plasma glucose levels, TC, TG, HDL-c, and LDL-c, whereas it was not associated with WC, HC, and WHR in 475 Pakistani subjects (10-78 years) [39]. Our results show no statistical association between the LEPR polymorphism,  [45]. Consistent with these results, the results of the current study suggest that this polymorphism has no effect, or eventually a recessive very weak increasing one, on zBMI, despite being present at a high frequency (46%) in the studied population.

MC4R rs2229616 and rs17782313
The MC4R variant is expressed in the central nervous system, and is part of the melanocortin pathway that controls food intake and energy homeostasis [21]. The most common coding MC4R polymorphism is MC4R rs2229616 (V103I missense variant) and it was the first described as showing no association with BMI, plasma insulin, and glucose levels in white British males [46]. Another study reported similar heterozygous frequencies in lean and obese individuals (4.2 vs. 4.5%, respectively), and found no homozygous frequencies for the polymorphic allele [26].
These results are consistent with our findings and indicate that the MC4R rs2229616 polymorphic allele is rare. By contrast, Geller et al. performed a meta-analysis of 7000 individuals and reported that the MC4R polymorphic allele was negatively associated with obesity [47], consistent with the results of other studies [26,27]. A possible mechanism underlying the protective effect of the MC4R V103I polymorphism could be an increase in energy expenditure [47]. Consistent with others, our results indicate that the MC4R rs17782313 variant is positively associated with obesity traits in overweight/obese children [22][23][24][25]. This gene variant may facilitate obesity by increasing the intake of high-energy or fatty foods or promote overeating in response to emotional eating [48]. We did not identify any differences in biochemical parameters associated with different genotypes of this polymorphism. García-Solis et al. studied 580 children (8−13 years old) and found that heterozygous subjects for this polymorphism risk allele were significantly associated with obesity but not with TC, HDL-c, or insulin levels [22]. Furthermore, Loos et al. confirmed that BMI in children (7−11 years old) was positively associated with each additional copy of the polymorphic allele, with a BMI increase of 0.10 kg/m 2 and 0.13 z-score units (P < 7.3 × 10 -6 ), twice of that observed in adults (P = 0.001) [23] and in agreement with the results of the present study (0.26 zBMI units increase for changing to minor allele). In our population, this recessive increasing effect polymorphism for zBMI yielded a weak effect size. The results from populations of African-American children remain controversial [49].

PPARG-2 rs1801282
PPARG isoforms 1 and 2 are transcription factors that activate adipocyte differentiation and mediate the expression of specific fat cell genes [28]. However, PPARG-2 rs1801282 may not be associated with obesity and type 2 diabetes mellitus [50]. Although we did not detect significant associations, the results showed a trend toward reduced mean BMI in heterozygous overweight/obese children than in homozygous wild type. Furthermore, our results also suggest the presence of a recessive medium decreasing effect on zBMI, although there was a low frequency (9%) of this minor allele in the studied population, which is a limitation to this finding. This is consistent with a study of 194 premenopausal Caucasian Portuguese females, which also found no significant differences in BMI between the control and case groups for this polymorphism [51]. However, a meta-analysis concluded that PPARG variants contributed to human adiposity variation and predisposition for obesity [29]. These inconsistencies may suggest that PPARG polymorphisms may be labeled differently in different ethnic populations, or that there are dissimilar gene-environment interactions.

Conclusion
We did not detect statistical associations between LEPR rs11371101, FTO rs9939609, MC4R rs2229616, and PPARG-2 rs1801282 polymorphisms and most obesityrelated phenotypes and metabolic parameters. Possible explanations could be low statistical power, low carrier frequency, or moderate sample size for some variants. Age differences and genetic environment background could also explain the effect of genes influence in a trait at different developmental stages or the same genes may have a larger impact on a trait as it develops. A longitudinal study may potentially disclose this point, allowing an exploration of the life course genetic associations with clinical and biochemical parameters. Despite these limitations, our data identifies the BMI and zBMI effects of genetic traits that are likely related to obesity, although with modest impact in younger ages, which is in agreement with other authors [9,23]. This study demonstrates that it is possible to detect and measure the influence of genetic variants on clinical and metabolic characteristics in childhood, reinforcing the concept that there is an important interaction between genes and environment (even if the role of environmental cues may not have much impact in such younger ages) in the development of excessive weight gain and its related complications.
The current study also collected data of weight and height since child birth (Personal Child Health Record) until 9 years of age, which allowed us to determine the adiposity rebound (AR) [52] and permitted us to conclude that over 50% of children (data not shown) had an AR prior to the age of 6 years, suggesting that negative environmental factors (e.g., nutritional) are already present in early ages which may explain the high rate of overweight/obesity in our population. Another strength of this study is the characterization of five gene SNPs that were cross-matched with an extensive panel of anthropometric and biochemical parameters. To our knowledge this is the first study trying to establish an association between clinical, metabolic phenotypes and LEPR rs11371101, MC4R rs2229616 and rs17782313 and PPARG-2 rs1801282 in Portuguese children and the first association between biochemical parameters dependent from obesity and FTO rs9939609 in the same population.
Finally, this study showed that MC4R rs17782313 and FTO rs9939609 were positively associated with zBMI, with weak and very weak effects, respectively, suggesting a very scarce contribution to childhood obesity at this age. LEPR rs1137101 and PPARG-2 rs1801282 had weak and medium negative effects on zBMI, respectively, and may slightly protect against childhood obesity. Considering that, in our prepubertal children, the impact on obesity of the SNPs of the genes included in this study is very modest, so we think that, at this age, a clinical application is not justified. Therefore we recommend further research on this topic, with longitudinal design studies or cross-sectional studies including children at a more advanced stage of development taking in account the impact of environmental factors (specially nutritional and physical activity).