Recent evidence suggests that brain-derived neurotrophic factor (BDNF) regulates food intake and the control of body weight. A common polymorphism in human BDNF, Val66Met (single-nucleotide polymorphism database (dbSNP) no. rs6265), impairs intracellular trafficking, resulting in the reduced secretion of BDNF. Several European studies have indicated that Val66Met is associated with BMI. In this study, we examined the association of the Val66Met polymorphism with BMI in Koreans (n = 20,270) from three independent epidemiological cohorts. All three studies observed a consistent association of this polymorphism with BMI, and their combined analysis demonstrated a robust correlation (β = −0.17 ± 0.03 and P = 5.6 × 10−8). We also examined the effect of smoking on the link between Val66Met and BMI. The association of Val66Met with BMI was statistically significant only in the smoking group, reflecting a possible interaction between smoking and the BDNF polymorphism for BMI. Thus, we have confirmed BDNF as a genetic risk factor for BMI in an Asian population and hypothesize that the Val66Met mutation influences individual differences in BMI. In addition, smoking might interact with BDNF Val66Met to modulate BMI.
Obesity is a major health problem in industrialized and developing countries, leading to increased mortality and morbidity from type 2 diabetes, metabolic syndrome, coronary heart disease, stroke, cancer, liver and gallbladder disease, sleep disorders, and osteoarthritis (1). The prevalence of obesity in adults and adolescents has increased in recent decades in Asia (2). Similarly, the rate of death from any obesity-related cause has developed a J-shaped association with BMI in Korea (3).
Recent evidence suggests that brain-derived neurotrophic factor (BDNF) influences food intake and the control of body weight (4). Lapchak and Hefti first proposed a function for BDNF in food intake (5), and Pelleymounter et al. observed that infusion of BDNF in the lateral ventricle of adult rats induced severe, dose-dependent appetite suppression and weight loss, implicating BDNF as having anorexigenic properties (6). Heterozygous mice that have been engineered for targeted disruption of BDNF (+/−) experience chronic hyperphagia and develop age-dependent obesity gradually throughout life (7,8).
Gunstad et al. first linked the Val66Met polymorphism (rs6265) of BDNF with BMI in healthy adults (9). 66Met homozygotes had a lower BMI than Val66Met heterozygotes and 66Val homozygotes. Shugart et al. recapitulated this association in two studies of British women (10). A genome-wide association study also identified the Val66Met allele as a significant variant that correlated with BMI (11). Yet, Friedel et al. failed to observe such a relationship in a study of extremely obese persons and underweight controls (12).
Smoking is a well known weight-loss factor. The weight loss in the smoking group was also found in the Korean population (3). Lang et al. linked the Val66Met polymorphism and smoking addiction in healthy, unrelated adults, suggesting that carriers of the 66Met allele of BDNF are more vulnerable to a smoking addiction (13). Moreover, the lower plasma concentration of BDNF was reported in the smoking group (14), implying the relationship among the Val66Met polymorphism of BDNF, BMI, and smoking.
In this study, we examined the effects of the Val66Met polymorphism of BDNF on BMI in 20,270 unrelated adults from three independent epidemiological studies in Koreans to validate this association in the Asian population. Additionally, we investigated whether smoking habit affected the association of the Val66Met with BMI to study the interaction between them.
Methods and Procedures
This study used populations from the KARE, Health2, and Urban Epidemiology Studies. Each study obtained permission from the respective ethical committees, and all participants provided informed consent for participation in the study.
The KARE study recruited participants aged 40–70 years in 2001 from the Ansan and Ansung regions; the Health2 study from 2004 comprised participants aged 40–70 years from rural areas of the KangWon-do and Jeolla-do provinces; and the Urban Epidemiology Study recruited participants aged 40–69 years in 2004 from the community health care centers in the Seoul and Busan metropolitan areas, Korea. The KARE and Health2 studies were reported by Cho et al. (15).
BMI was calculated as weight (kg)/height2 (m2). Body fat percentage (BF%) was measured only in the KARE study and was obtained by bioelectrical impedance analysis using InBody 3.0 (Biospace, Seoul, Korea).
Data on cigarette smoking were available only in the KARE and Health2 studies. Based on self-reported information on the questionnaire, subjects were sorted into three groups: Never-, Quit- and Current-smoking.
Val66Met single-nucleotide polymorphism (SNP) (rs6265) genotypes were obtained from various SNP array platforms in each study. Genotypes in the KARE study, Urban Epidemiology study, and Health2 study were obtained from the Affymetrix Genome-Wide Human SNP array 5.0, the Affymetrix Genome-Wide Human SNP array 6.0, and GoldenGate Custom Genotyping on a BeadArray (Illumina, San Diego, CA), respectively. The genotypes were clustered into three groups (Supplementary Figure S1) in all platforms.
The genotyping quality control criteria have been reported in a previous GWAS study (15). Briefly, 500 ng of genomic DNA was genotyped using the Affy 5.0 and 6.0 SNP arrays for the KARE and Urban cohorts, and Bayesian Robust Linear Modeling using the Mahalanobis Distance genotyping algorithm was used for genotype calling (16). Samples with high missing genotype call rates (≥4%), high heterozygosity (>30%), inconsistencies in gender, and any kind of tumor were excluded from subsequent analyses, as were related or identical individuals whose estimated identity-by-state value was high (>0.80) (17). The GoldenGate genotyping assay was evaluated with regard to accuracy by duplicating up to 2.5% of samples from the Health2 study, generating a concordance rate of 99.7% for the Val66Met genotype.
The effect of a genotype was determined by ANOVA and linear regression analysis. We calculated the effect size (β) and s.e. of the Val66Met allele on BMI and BF%. All analyses were adjusted for matching variables: age, living area, and gender. Because of the small sample sizes of Quit- and Current-smoking groups in women (<5%), the smoking-BDNF interaction tests were conducted only in men. For each smoking group, the BMI differences among three genotypes were tested by ANOVA test and linear regression analysis with controlling cohorts and age. SPSS 15.0 was used for all statistical tests.
The baseline phenotypes and Val66Met genotypes of the three studies are described in Table 1. Age, male-to-female ratio, and BMI differed significantly between the three populations, but genotype frequencies were similar and showed Hardy—Weinberg equilibrium (P > 0.05).
Table 1. Baseline phenotypes and rs6265 (Val66Met) genotypes of participants
All three studies demonstrated a consistent association of the Val66Met polymorphism with BMI; the comparisons of major allele homozygotes vs. minor allele carriers by ANOVA and of the genotypes by linear regression analysis are shown in Table 2. We also examined the association in men and women and did not detect any difference between genders. The combined analysis of the three studies indicated a robust correlation of the polymorphism with BMI (β = −0.17 ± 0.03 and P = 5.6 × 10−8).
Table 2. BMI comparison by genotype and study group
The association between Val66Met of BDNF and BF% was also analyzed in KARE subjects. 66Met allele carriers had significantly lower BF% (β = −0.27 ± 0.09, P = 0.0014) compared with 66Val carriers (Supplementary Table S1).
The cigarette smoking in Korean populations was associated with reduced BMI (combined β =−0.41 ± 0.04, P value = 2.85 × 10−23) (Table 3). However, the association of the Val66Met polymorphism with smoking habit was not found in the Korean populations.
Table 3. Smoking phenotypes and the association results with BMI and BDNF genotypes
For the further analysis, we classified the men subjects into three groups such as Never-, Quit- and Current-smoking depending on the smoking status based on the questionnaire, and then tested separately the association of Val66Met polymorphism with BMI in each group. Interestingly, no association was found in the Never-smoking group, while consistent association was found in the Current-smoking group (combined β = −0.23 ± 0.07, P value = 1.4 × 10−3) (Table 4). Noticeably, the effect size calculated in the analysis of Current-smoking group (combined β = −0.23 ± 0.07) was greater than that in the analysis of total subjects (combined β = −0.17 ± 0.03, Table 2).
Table 4. BDNF and Smoking Interactions with BMI in men
This study investigated the association between the Val66Met polymorphism of BDNF and BMI in 20,270 unrelated Koreans. The Val66Met polymorphism was significantly linked with BMI in the three independent studies, supporting the findings of European studies (9,10). Thus, BDNF is an important molecule that regulates obesity, and the functional polymorphism, Val66Met might be a strong candidate of causative variations in BDNF.
Our results indicate that the effects of the Val66Met polymorphism on BMI are smaller in Koreans compared with Europeans (Figure 1). The genetic effects that were observed by Gunstad et al., however, must be confirmed due to the low number of samples (n = 481) that was used in their association analysis (9). Additionally, this study did not distinguish between genders with regard to the link between the polymorphism and BMI (Table 2), whereas European studies indicated the difference of the association with BMI between genders (9,10). Gunstad et al. observed a statistically significant correlation between 66Met and decreased BMI in women but not in a combined analysis of men and women, indicating that 66Met has effects preferentially in women (9,10).
Chinese and Japanese studies have recently reported a link between an intergenic SNP (rs925946, −13 kb of rs6265) of BDNF and BMI (18,19), and a Filipino study noted an association of another intergenic SNP (rs492346, −23 kb of rs6265) with BMI(20). The pairwise linkage disequilibrium (r2), calculated based on International HapMap data in Asians, between rs925946 and rs6265 was 0.056, and that between rs4923461 and rs6265 was 0.642. Therefore, the association of rs925946 with BMI in Chinese and Japanese might be independent of rs6265, and that of rs4923461 in Filipinos might be caused by LD with rs6265. Also, a recent study reported that rs6265 correlates with BMI in Chinese children (21).
One novel finding of this study is the possible interaction of the Val66Met genotype and smoking habit in regard to its effects on BMI. Although the 66Met allele was reported associated with BMI (9−11) and addiction to smoking (13), its interaction with smoking in regulating obesity has not been established. This study indicated that the association of Val66Met polymorphism with BMI was statistically significant only in the Current-smoking group. Because the findings were replicated in two independent cohorts (KARE and Health2), there may be possible the gene-environmental interaction on BMI between the BDNF Val66Met and smoking habit.
The 66Met allele impairs intracellular trafficking, resulting in the reduced secretion of BDNF (22), and the plasma BDNF level was found lower in Current-smoking group than Quit- and Never-smoking groups (14). Therefore, it may be speculated that the reduced function of BDNF by 66Met allele in addition to the reduced secretion of BDNF by smoking enhances the decrease of BMI together, as supported by the greater genetic effect size in the Current-smoking group (Table 4). However, the previous report that infusion of BDNF in the lateral ventricle of adult rats induced appetite suppression and weight loss (7) contradicts this finding, awaiting further functional studies.
In conclusion, we have confirmed the association of the Val66Met polymorphism of BDNF with BMI in the Asian population—BMI is lower in carriers of a minor allele, 66Met, similar to what is observed in whites. In addition, it appears that the 66Met allele interacts with smoking to influence BMI.
This work was supported by a grant from Kyung Hee University in 2009 (KHU-20090597).