Obesity and associated comorbidities (such as cardiovascular diseases, type 2 diabetes, musculoskeletal disorders and some cancers) represent a major public health problem that, in recent years, has reached epidemic proportions. In Mexico, obesity is the most common nutritional disorder of children, and the prevalence of this disease has increased alarmingly over the past decade: recent data from ENSA 2006 (‘Encuesta Nacional de Salud’, i.e. The National Health Census) reported an increase in overweight and obesity from 18% in 1999 to 26% in 2006, in young Mexican children (between 5 and 11 years old) 1 . Overweight in childhood or adolescence is a risk factor for overall mortality in adulthood 2 .

The prevalence of obesity varies across populations and it is noteworthy that the Mexican population has been disproportionally affected. Although environmental changes (linked to the ‘westernization’ of ways of life) can explain the increase in prevalence of obesity on a global level, individual variation of body mass index (BMI) persists in the same environment, and the heritability of this trait is very high. Indeed, heritability estimates of BMI range between 40% and 70% according to studies 3 . The high genetic susceptibility to obesity (and both insulin resistance and type 2 diabetes) in the Mexican population may be ascribed to the American Indian heritage 4 .

To date, genome wide association studies (GWAS) and linkage studies, mostly performed in European adult populations, have identified more than 50 loci associated with obesity or BMI 5 6 7 8 9 10 . However, replication attempts have yielded inconsistent outcomes 11 12 13 . In particular, replication in other populations is not obvious, and several factors (such as ethnic differences in linkage disequilibrium patterns, ethnic-specific associations, gene × environment interactions) may puzzle the picture. Furthermore, very few studies have been performed in children presenting with severe obesity in non-European populations.

In the present study, we aimed to perform a follow-up replication study including sixEuropean obesity-associated genetic variants, in Mexican children.

From an initial sample of 1,685 children (between 6 and 12 years old), we extracted 949 lean and 514 obese children (see clinical characteristics in Table 1) for the case–control study, and 1,171 nonobese children (with BMI < 95^th percentile) for the study of metabolic quantitative traits including BMI, fasting serum insulin, fasting plasma glucose, total cholesterol and triglycerides. As expected, obese children exhibited higher values of insulin and cholesterol levels (P ≤ 0.001) compared to lean children. Insulin resistance was found in 45% of obese children.

Data are means ± standard deviation.

*P-value according to Student’s t test comparing lean and obese children.

BMI, body mass index; pc, percentile.

In both studies, we genotyped six SNPs that are known to be associated with risk of obesity in European populations: rs17782313 near MC4R 6 ; rs2815752 near NEGR1 7 ; rs7754561 near ENPP1 8 ; rs1805081 in exon 6 of NPC1 9 ; rs10938397 near GNPDA2 7 and rs1421085 in intron 1 of FTO 10 .

We identified a significant contribution of GNPAD2 rs10938397 to risk of obesity (G risk allele [in European populations]; odds ratio_{[95% confidence interval]} [OR] = 1.30_[1.11;1.53]; P = 1.34 × 10^-3; Table 2). Furthermore, we found nominal association between risk of obesity and both ENPP1 rs7754561 (G risk allele; OR = 0.84_[0.72;0.97]; P = 0.020; Table 2) and MC4R rs17782313 (C risk allele; OR = 1.40_[1.06;1.83]; P = 0.016; Table 2). This last MC4R SNP also showed a nominal effect on BMI (C risk allele; β_{[standard error]} = 0.41_[0.16] kg/m²; P = 0.012; Table 3). Moreover, we identified a nominal association between NEGR1 rs2815752 and BMI (A risk allele; β = 0.24_[0.10] kg/m²; P = 0.019; Table 3). Of note, we did not find any significant association between risk of obesity (or BMI variation) and either NPC1 rs2815752 or FTO rs1421085 (Tables 2 and Table 3).

*We noted the risk-alleles reported in European populations.

RA, risk-allele; RAF, risk-allele frequency; OR, odds ratio; CI, confidence interval; P, P-Value.

Data for fasting serum insulin and triglycerides were logarithmically transformed before statistical analysis. Units are: BMI in kg/m², fasting serum insulin in μU/mL, fasting plasma glucose in mmol/L, total cholesterol in mg/dL and triglycerides in mg/dL.

*We noted the risk-alleles reported in European populations.

RA, risk-allele; SE, standard error; β, effect-size; P, P-value; BMI, body mass index.

In our study of quantitative traits, we found a significant association between NPC1 rs1805081 and fasting serum insulin levels (A risk allele; β = −0.10_[0.03] μU/mL; P = 1.21 × 10^-3; Table 3). We also identified an effect of the obesity risk allele of MC4R rs17782313 on increased fasting plasma glucose levels (C risk allele; β = 0.36_[0.11] mmol/L; P = 1.47 × 10^-3; Table 3). A trend of association was found between FTO rs1421085 and both total cholesterol levels (C risk allele; β = 3.70_[1.72] mg/dL; P = 0.032; Table 3) and fasting serum insulin levels (C risk allele; β = −0.06_[0.04] μU/mL; P = 0.064; Table 3).

From our study based on obese and nonobese Mexican children, we found a significant contribution of the minor allele of GNPDA2 rs10938397 to increased risk of obesity. The association between GNPDA2 and BMI had firstly been identified in a meta-analysis of several European GWASs performed by the GIANT consortium 7 . Subsequently, the association signal has been confirmed in other populations: in adults from East-Asia 18 19 and in Chinese children 20 . To our knowledge, the present study is the first confirmation of the GNPDA2 association signal with risk of obesity in the Mexican population.

Furthermore, we found nominal associations between risk of obesity or BMI and the following SNPs: ENPP1 rs7754561, MC4R rs17782313 and NEGR1 rs2815752. Of note, the risk alleles for obesity or increased BMI were the same between Europeans and Mexicans, except for the risk allele of ENPP1 rs7754561 (in Europeans) that showed a protective effect in Mexicans. Although the association between both MC4R and NEGR1 and risk of obesity has been confirmed in a plethora of studies and populations, the ENPP1 association signal with obesity is more controversial 21 . To our knowledge, no other studies demonstrate a protective role of the risk allele of ENPP1 rs7754561. Recently, it has been shown that ENPP1 overexpression in human adipocyte cell lines resulted in defective adipocyte maturation 22 . If confirmed in other Mexican populations, the protective effect of the ENPP1 variant may be due to a loss-of-function of the protein.

We did not find any significant contribution of either NPC1 rs2815752 or FTO rs1421085 to obesity or BMI variation. Therefore, the FTO association signal with obesity which was found in Mexican adults by Villalobos-Comparán and colleagues 23 , was not confirmed in children. Thus, the obesogenic effect of FTO would occur later in the Mexican population than in the Europeans 9 , but as we lack some statistical power, additional genetic studies on Mexican children would be needed.

Importantly, we identified a significant effect of the risk allele of NPC1 rs1805081 on decreased fasting serum insulin levels. NPC1 encodes Niemann Pick disease type C1 protein that mediates intracellular cholesterol trafficking via binding of cholesterol to its N-terminal domain. Very recently, Jelinek and colleagues have reported that Npc1 haploinsufficiency developed abnormal metabolic features (including hyperinsulinemia) and increased susceptibility to weight gain in mice 24 . We could suspect that the variant may have a gain-of-function effect in Mexican children. Of note, we did not report any association between the NPC1 variant and obesity or BMI variation, and this association signal remains quite controversial according to population studies 20 25 26 .

Furthermore, we identified an effect of the risk allele of MC4R variant and increased fasting plasma glucose levels. To our knowledge, no previous study has shown this association. However, two recent large meta-analyses of GWAS identified a significant association between the MC4R locus and type 2 diabetes risk, in European and Asian populations 27 28 . Of note, our present association with fasting plasma glucose levels remained significant after adjustment for BMI (data not shown). Altogether, these findings would suggest a potential effect of MC4R polymorphisms on decreased pancreatic beta-cell function.

In summary, in our sample of Mexican children, we replicated four European obesity-related genes (GNPDA2, and nominally, NEGR1, MC4R and ENPP1) in the same direction of effect as previous findings, except for ENPP1. Interestingly, we found two novel association signals: between NPC1 variant and fasting serum insulin levels, and between MC4R variant and fasting plasma glucose levels. These findings should deserve some confirmatory studies in other Mexican populations.

It is noteworthy that these well established genetic associations with obesity explain very little of the genetic risk for pediatric phenotype, suggesting the existence of additional loci whose number and effect size remain unknown, which guarantee intense additional investigations in the near future.

β: Effect-size; BMI: Body mass index; CI: Confidence interval; GWAS: Genome-wide association study; HOMA-IR: Homeostasis model assessment of insulin resistance; OR: Odds ratio; P: P-value; pc: Percentile; RA: Risk-allele; RAF: Risk-allele frequency; SE: Standard error; SNP: Single nucleotide polymorphism.

The authors declare no competing financial interests.

AMB performed the genotyping, and contributed to statistical analyses and the writing of the paper; AB wrote the paper; DM, JGM, PF and MC revised the paper and contributed to discussion; LY and AB performed the statistical analyses; AMB, MKK, CA, EC, EPL, JMM, MEG, JPR, SFH, JGM and MC contributed to recruitment and clinical data of the Mexican children. All the authors approved the final version of the submitted draft.