Background

Obesity (as characterized by excess body fat) is an established risk factor for cardiovascular and metabolic diseases. Indeed, each unit increase in the body mass index (BMI) increases the risk of hypertension by a factor of 5 and the risks of coronary artery disease and stroke by a factor of 3.6 [1]. It is noteworthy that 80% of people with type 2 diabetes are obese [2]. Moreover, non-alcoholic steatohepatitis, impotency, infertility, and several types of cancer have all been linked to obesity [1]. Overall, obesity increases the risk of premature death. Obesity has reached epidemic proportions in modern society [3]. In 2008, more than 500 million adults worldwide were considered to be clinically obese [4]. However, the prevalence of obesity and overweight differs from one region of the world to another, with 33.5% in the USA [5], 23.1% in Canada [6], between 5.1% and 32.4% in Europe (with a west/east/south gradient) [7], 14.9% in Morocco [8], 29.6% in Tunisia [9] and 21.2% in Algeria [10]. Obesity is increasingly prevalent in many African countries and other developing countries undergoing nutritional transitions as a result of urbanization, demographic transitions and the adoption of Western lifestyles [11].

Common obesity is caused by the interaction of multiple genetic and environmental factors; with a heritability ranging from 40 to 70% [12]. Genetic approaches have improved our understanding of the biological bases of obesity. To date, genome-wide association studies (GWASs) have identified and confirmed 32 BMI loci in European populations [13]-[19]. The replication of association signals in independent populations is a mandatory approach for characterizing gene-disease relationships [20]. Because initial and replication studies have been mainly reported in populations of European descent, the challenge remains to extend the studies to other populations [20]. Moreover, as each individual single nucleotide polymorphism (SNP) exerts a moderate genetic effect and thus explains a small proportion of the total BMI variation, the analysis of the combined effect of sets of variants (by calculating a genetic predisposition score (GPS)) [21] appears necessary and also useful in moderate sample size population samples which are underpowered compared to GWASs.

In the present study, we assessed whether a GPS for BMI developed from Europeans is a valid score in a North African population. Although a few association studies have been performed in this part of the world [22]-[26], no one examined the combined set of established BMI loci. Therefore, we tested whether the combination of variants robustly associated with BMI in many European populations also influence anthropometric traits and obesity risk in an Algerian population sample (provided by the ISOR study).

Methods

The ISOR (InSulino-résistance à ORan) study

The cross-sectional population-based ISOR study was performed from 2007 to 2009. The study’s objectives and procedures were approved by the independent ethics committee at the Algerian National Agency for the Development of Health Research (since renamed as the Thematic Agency of Research in Health Sciences). The ISOR study was a population-based, cross-sectional study of a representative sample of 787 subjects (378 men and 409 women, aged between 30 and 64) recruited from within the city of Oran (north-west of Algeria). Subjects were selected at random from social security rolls. All subjects consented freely to participation in the study. Details of the studies have been described elsewhere [27].

A questionnaire on lifestyle (physical activity, tobacco use and alcohol intake), personal and family medical histories, current medication, socio-economic status and educational level was completed during a face-to-face interview. The ISOR questionnaire lifestyle was an adapted version of the one used in the STEP study (investigation Stepwise led in Algeria in 2003) which was validated by the WHO. The level of physical activity was defined in quartiles as “none”, “low”, “medium” and “high” after summing exercise scores for sporting activities, walking, housework and physical activity at work.

The anthropometric measurements included height, body weight, waist and hip circumferences. Height and weight were measured while the subject was barefoot and lightly dressed. The BMI was calculated according to the Quetelet equation. Normal weight was considered for BMI < 25 kg/m2, overweight was considered for 25 ≤ BMI < 30 kg/m2 and obesity was considered for BMI < 30 kg/m2.

Genomic DNA was extracted from white blood cells by using the Stratagene® kit (Agilent Technologies, Les Ulis, France), according to the manufacturer’s protocol.

SNP selection and genotyping

We selected 32 SNPs known to be associated with BMI [19] within or near the following genes: FTO, MC4R, TMEM18, GNPDA2, BDNF, NEGR1, SH2B1, ETV5, MTCH2, KCTD15, TFAP2B, NRXN3, FAIM2, SEC16B, RBJ-ADCY3-POMC, GPRC5B, MAP2K5-LBXCOR1, QPCTL-GIPR, TNNI3K, SLC39A8, FLJ35779-HMGCR, LRRN6C, TMEM160, FANCL, CADM2, PRKD1, LRP1B, PTBP2, MTIF3-GTF3A, RPL27A-TUB, NUDT3-HMGA1 and ZNF608.

Genotyping was performed using KASPar technology (KBioscience, Hoddesdon, UK). The genotyping of the ZNF608 rs4836133 SNP failed. The genotyping success rates of the FANCL rs887912 and CADM2 rs13078807 SNPs were too low (75% and 76%, respectively) to be considered in the analyses. The genotyping success rate of the 29 other SNPs was at least 95%. So 29 SNPs were considered in the final analyses.

Statistical analysis

Statistical analyses were performed with SAS software (version 9.1, SAS Institute Inc., Cary, NC, USA). Deviation from the Hardy-Weinberg equilibrium was tested using a X2 test and the threshold for statistical significance was set to p ≤ 0.0017 (with Bonferroni correction for 29 independent SNPs at α = 0.05).

The GPS was obtained as previously described [28]. Briefly, a weighting method was used to calculate the GPS on the basis of 29 genotyped SNPs. The genotypes were coded as 0, 1, or 2 according to the number of copies of the effect allele. Each SNP was weighted according to its relative effect size (i.e. the β coefficient); in order to measure the effect of each SNP on BMI with greater accuracy and precision, β coefficients were derived from Speliotes et al. [19]. The GPS was calculated by multiplying each β-coefficient by the number of corresponding risk alleles and then summing the products. Because this produced a score out of 8.24 (twice the sum of the reported β-coefficients), all values were divided by 8.24 and multiplied by 58 (number of alleles) to make the GPS easier to interpret: each point of the GPS corresponds to one risk allele. When calculating the GPS, missing genotype data were replaced with the average allele count for the corresponding SNP. However, 47 individuals with missing genotypes for more than 3 SNPs were excluded from the GPS analyses.

Intergroup comparisons of means were performed with a general linear model using an additive genetic model. A Pearson X2 test was used to compare groups in terms of genotype and allele distributions.

Odds ratios (ORs) were obtained by multivariate logistic regression analyses with an additive genetic model.

For BMI and obesity risk, the covariables were age, gender, smoking status and physical activity. For waist circumference and hip circumference, the covariables were age, gender, smoking status, physical activity and (in some models) BMI.

For individuals associations, after Bonferroni correction, only individuals associations with a p value below 0.0017 were considered to be statistically significant (i.e. 0.05 divided by 29 polymorphisms).

For combined association (GPS analyses); statistical significance was set to p ≤ 0.05. Power calculations for individual SNP analyses were performed using Quanto v1.2.4 (http://biostats.usc.edu/Quanto.html) using a one-sided p value at 0.05.

Concerning the GPS analysis, the statistical power of detecting a significant association with BMI was 98% (calculated with the pwr R package developed by Stéphane Champely).

Results

Characteristics of the study subjects

The ISOR study includes 340 (43.2%) normal weight subjects, 280 (35.6%) overweight subjects and 167 (21.2%) obese subjects (Additional file 1).

Distributions and individual associations with anthropometric traits

All 29 SNPs conformed to Hardy-Weinberg equilibrium (Table 1). When we compared allelic frequencies of these 29 loci between the ISOR study and the study of Speliotes et al., the allelic distributions differed for 5 SNPs (Table 1). For example, for SH2B1 rs7359397, the T allele frequency was 0.15 and 0.40 in the ISOR and Speliotes et al.’ studies, respectively.

Table 1 Distributions of the 29 genotyped SNPs in the ISOR study
Full size table

Power calculations indicated that, based on results from Speliotes et al. [19], the statistical power of our study to detect significant associations between individual SNPs and BMI was below 44%. Nevertheless, the RBJ rs713586 and RPL27A rs4929949 SNPs were nominally associated with BMI (β = 0.62 ± 0.25, p = 0.01, and β = 0.67 ± 0.24, p = 0.006, respectively); and 23 of the 29 tested SNPs were directionally consistent with the results reported in the original GWAS on BMI (Additional file 2). This result was greater than that expected by chance (Binomial test, p = 0.0009).

Combined associations with anthropometric-related traits and obesity risk

The 29 SNPs were used to calculate the GPS, which was normally distributed (mean: 25.7 ± 3.7 alleles; range: 14.2 to 37.6 alleles). We observed significant associations between the GPS and BMI, waist circumference and hip circumference (Table 2). The mean [95% CI] allele effect of the GPS was +0.15 [0.06 - 0.24] kg/m2 (p = 0.001) for BMI, +0.26 [0.03 - 0.49] cm (p = 0.02) for waist circumference and +0.22 [0.04 - 0.40] cm (p = 0.02) for hip circumference. We did not detect a statistically significant association between the GPS and the waist-to-hip ratio (p = 0.37). Associations with waist circumference and hip circumference were no longer statistically significant after further adjustment for BMI. Of note, when the analyses did not take into account missing genotypes, similar results were obtained for BMI, waist circumference, hip circumference and the waist-to-hip ratio (data not shown).

Table 2 Effect of the genetic predisposition score on anthropometric variables in the ISOR study (n = 740)
Full size table

To distinguish between the effects of the GPS and the effects of the covariables classically associated with BMI (age, gender, physical activity and smoking status), we compared the crude and adjusted models (Table 3). The GPS alone accounted for 1.0% of the BMI variance and the covariables accounted for 13.1% of the variance. Overall, the GPS and covariables explained 14.1% of the BMI variance.

Table 3 Effects of the genetic predisposition score and covariables on BMI in the ISOR study (n = 740)
Full size table

Next, we investigated the association between the GPS and the obesity risk. Because, overweight participants could not be considered as obese or normal weight subjects, we removed them for the analyses. We detected a significant association between the GPS and the obesity risk (OR [95%CI] = 1.11 [1.05 - 1.18], p = 0.0004). We also examined the association between the obesity risk and the GPS in quartiles (Table 4). Subjects in the highest GPS quartile (i.e. > 28.3 alleles) had a higher obesity risk than subjects in the lowest GPS quartile (i.e. < 23.2 alleles) (OR = 2.53 [1.38 - 4.65], p = 0.003).

Table 4 The obesity risk per quartile of the genetic predisposition score in the ISOR study, excluding overweight subjects
Full size table

As the rs3817334, rs987237, rs10968576, rs4929949 SNPs could be considered as outliners in terms of respect of the Hardy-Weinberg equilibrium (4.0×10-3 ≤ p ≤ 8.4×10-3), we performed all GPS analyses without these 4 SNPs and similar results were obtained (data not shown).

Discussion

Meta-analyses of GWASs have identified 32 loci as being unequivocally associated with BMI, individually or combined in a genetic predisposition score in populations of European descent [19]. In the present study, we have investigated the combined effects of 29 successfully genotyped, established BMI loci on anthropometric variables and obesity risk in a sample of the general population from north-west of Algeria (n = 740).

The prevalence of obesity in the ISOR study (21.2%) is similar to that reported in Algeria (21.2%) in the Transition and Health Impact In North Africa (TAHINA) study in 2010 [10]. The prevalence of obesity in Oran is higher than in France (14.5%) [29], Spain (13.6%) [30] and Italy (8.2%) [31], lower than in the USA (33.5%) [5], Canada (23.1%) [6] and the UK (23.5%) [32], and similar to that in Nigeria (21.2%) [33]. When compared with other urban North African populations, the prevalence of obesity in Oran is higher than in Morocco (14.9%) [8] and lower than in Tunisia (29.6%) [9].

Although the ISOR study was not sufficiently powered (<44%) to detect significant individual associations, most of the tested SNPs presented effects with the expected direction (23 of the 29 tested). We also observed that few SNPs presented significant allele frequency differences between the ISOR and Speliotes et al. [19] studies. Differences in allele frequencies may contribute to differences in disease prevalence between ethnic groups [34].

Although a few association studies on BMI or obesity have been performed in North African samples [22]-[26], no one examined the GWAS established BMI loci either individually or combined. In the ISOR study, the GPS (corresponding to the 29 established, BMI-associated SNPs’ cumulative contribution) for which we had 98% power in the ISOR study, showed a significant, positive association with BMI (p = 0.001). Each additional BMI-raising allele was associated with a mean increment of 0.15 kg/m2 (which corresponds to a weight increment of 434 g for a person measuring 1.70 m in height). Although the GPS was also significantly associated with waist and hip circumferences, these associations were mediated by BMI.

The overall known genetic susceptibility associated with the GPS explained only 1.0% of the variance in BMI, whereas the combined effect of genetic and known environmental factors accounted for 14.1%. Our results are in agreement with previous reports [19],[35]. Moreover, the GPS was associated with a 11% higher obesity risk and subjects with more than 28.3 risk alleles had a 2.5-fold higher risk of obesity. Although (i) it has been calculated that genetic factors account for between 40% and 70% of the population variation in BMI and (ii) the 29 SNPs studied here have been robustly validated as BMI-susceptible variants in GWA/replication studies, their combined effect on BMI and obesity risk was quite small. However, our results are in agreement with previous reports [16],[19],[35]-[39].

Conclusion

In conclusion, although larger samples will be needed to firmly replicate our findings, our data showed that a GPS comprising 29 BMI established loci in Europeans was associated with higher BMI and obesity risk in an Algerian population. Our findings contribute to a better understanding of the genetic susceptibility to obesity in Algeria.

Additional files