Relative Contributions of Selected Genetic and Lifestyle Factors to Inter-Individual Variations in Serum Lipid and Apolipoprotein Levels

Serum lipid and apolipoprotein profiles are influenced by both genetic and environmental factors. However, the relative effects of these factors on the phenotypic variations remain unclear. In this study, the relative contributions of selected genetic and lifestyle factors to inter-individual variations in serum lipid and apolipoprotein levels were estimated by using the multiple regression model in a rural Japanese population. Four restriction fragment length polymorphisms (RFLPs) were tested with Xba I and EcoR I at the apolipoprotein B (apo B), and Msp I and Sac I at the apo Al-CIII gene loci. The contribution of individual RFLP to serum total cholesterol ranged from 0.08% to 1.60%; serum low density lipoprotein cholesterol, 0.06% to 1.69%; triglycerides, 0.04% to 0.89%; apo B, 0.06% to 1.99%; high density lipoprotein cholesterol, 0.05% to 2.59%; apo AI, 0.54% to 2.73%; apo AII, 0.12% to 1.96% and apo CIII, 0.13% to 0.48%. These percentages were almost the same as or a little lower than those of some lifestyle variables - dietary factors (Keys dietary score, and energy-adjusted intake of carbohydrate, fiber and n-3 fatty acids), smoking, alcohol consumption and physical activity-to the serum traits. J Epidemiol, 1995 ; 5 : 187-196.

The level of serum lipid or apolipoprotein at any given time is a result of interactions of genetic endowment with lifestyle over the entire lifespan. For the serum lipid levels, however, the relative contributions of genetic and lifestyle factors are not clear ; it is very difficult to arrive at an assignment of weights.
The aim of the present study is to quantitatively estimate the relative contributions of selected genetic and lifestyle factors to the inter-individual variations in serum lipid and apolipoprotein levels in a rural Japanese population, using the multiple regression model. Serum total cholesterol (TC), low density lipoprotein cholesterol (LDL-C), triglycerides (TG), apolipoprotein B (apo B), high density lipoprotein cholesterol (HDL-C), apo AI, apo AII and apo CIII levels were determined as the dependent variables of the function. On the basis of established or putative roles of the candidate gene products in lipoprotein metabolism, four restriction fragment length polymorphisms (RFLPs) were selected with Xba I and EcoR I at the apo B, and Msp I and Sac I at the apo AI-CIII gene loci as the genetic independent variables. As for lifestyle factors, dietary intake, cigarette smoking, alcohol consumption and physical activity were also entered into the model for the current study.

Study Area
The study area is H-Y district, Shiso County, Hyogo Prefecture, Japan. This district is an agricultural area where rice and vegetables are produced on a relatively small scale. However, majority of the residents are not only farmers but also regular employees of offices or factories. They are engaged in farming only during the rice-planting and harvesting seasons. Associations between lifestyle factors and serum lipids or apolipoproteins The same ANOVAs were done to estimate the associations between lifestyle factors and serum biochemical traits (Appendix 1). Our findings were almost consistent with the existing knowledge6-11).As for dietary intakes, significant associations were observed between carbohydrate intake and serum apo B and HDL-C ; dietary fiber and serum LDL-C and HDL-C ; n-3 fatty acids and serum LDL-C, apo B and HDL-C ; and Keys dietary score and serum LDL-C. Smoking was significantly associated with serum LDL-C, log TG and HDL-C. The positive association of drinking or physical activity with serum HDL-C, apo AI and apo AII was statistically significant and relatively strong.

Multiple regression analysis
Multiple regression analysis was performed to estimate the relative contributions of selected genetic and lifestyle factors to the inter-individual variations in serum lipid or apolipoprotein levels. These results are shown in Table 2 and Figure 1. The model included sex, age, four genetic polymorphisms, four dietary factors, smoking, drinking and physical activity (13 variables in total) as the independent variables and accounted for 6.98%, 13.17%, 6.99%, 13.43%, 22.54%, 20.17%, 16.55%, and 8.75% (R2) of the total variation in serum TC, LDL-C, log TG, apo B, HDL-C, apo AI, apo All and log apoCIII, respectively ( Figure 1). The contribution, Pi, of individual RFLP to interindividual variation in serum TC was 0.08% to 1.60%, 0.06% to 1.69% in LDL-C, 0.04% to 0.89% in log TG, 0.06% to 1.99% in apo B, 0.05% to 2.59% in HDL-C, 0.54% to 2.73% in apo AI, 0.12% to 1.96% in apo AII and 0.09% to 0.48% in log apoCIII. In general, Pi for a RFLP did not appear to be larger than that for a lifestyle factor within the model : Allele frequencies and serum lipid levels The allele frequencies of four RFLPs for apo B and apo AI-CIII genes are shown in Appendix 2. The minor allele frequencies of Xba I and EcoR I RFLPs in H-Y district were much lower than those in Caucasian populations12,13), although they were almost the same as those from other Japanese14) and Chinese populations15).
On the other hand, the minor allele frequencies of Sac I and Msp I RFLPs were four times or more than those in Caucasian populations13), but they were not different from those in other Japanese populations16-18).
There was no significant deviation of observed genotype frequencies from those predicted by the Hardy-Weinberg law in H-Y district (data not shown).
The means and standard deviations of serum lipids and apolipoproteins are presented in Appendix 3. The levels of serum TC in H-Y district were similar to those of the whole Japan19), although the average values of serum HDL-C were a little lower. Therefore, the subjects of the present study did not appear to differ markedly from average Japanese.

Associations between RFLPs and serum lipids or apolipoproteins
Genotypes of Xba I RFLP at the apo B gene locus were not associated with any apo B-related traits in H-Y district ( Table 1). The lack of the significant association was also observed in another Japanese population14), while the significant associations were observed in several Caucasian populations20-24). Xba I polymorphism results from a silent C to T transversion in the 26th exon of the apo B gene and does not change the apo B amino acid sequence25). Therefore, this polymorphism might work as a linkage marker for the particular traits in some populations but not in others. On the other hand, Xba I genotypes were related to serum HDL-C and apo AI. Our findings agreed with those from Asian studies15,26). Although the basis for these associations remains unknown, the allelic variant may be in linkage disequilibrium with another functional mutation within the candidate gene or even at another gene that is physically or genetically linked to the gene tested26). In this case, the variant serves as a marker for the actual causative mutation.
Genotypes of EcoR I at the apo B locus were significantly associated with serum apo B. This association may suggest that the genetic variants which encode structural changes in the protein might directly or indirectly affect the lipoprotein metabolism and serum apo B concentration 27). However, the positive, negative and non-associations took nearly equal share in the number of the literatures cited15,20,22,28-30) Hegele et al. reported that the lack of a significant genotype-phenotype association for EcoR I was not surprising, because functional studies failed to show that this protein polymorphism had an efect31). On the other hand, a significant association of EcoR I with serum apo AII might be explained by the linkage disequilibrium hypothesis.
Genotypes of Sac I and Msp I RFLPs at the apo AI-CIII locus were shown to be associated with variations in serum HDL-C, apo AI and apo AII. Particularly, the significant association of Sac I polymorphism with serum HDL-C is consistent with many other studies31-33). This association could be due to linkage disequilibrium between this site polymorphism and functional variants in the promoter region of apo CIII gene34). However, some other studies showed no association of Sac I polymorphism with HDL-C35,36). The persons with M+M+ genotype of Msp I RFLP tended to have higher level of serum HDL-C37) or apo AI32) than those with M-M-genotype.

Relative contributions of genetic and lifestyle factors to serum lipids and apolipoproteins
In the present study, the contribution, Pi, of individual RFLP to inter-individual variation in serum TC, LDL-C, log TG, apo B, HDL-C, apo AI, apo AII and log apoCIII was between 0.04% and 2.73% (Table 2). In a study in a genetically isolated population31), the multiple regression analysis was performed to estimate the percentage of phenotypic variation determined by genetic variation. The model included sex, age, log BMI, colony of origin and 4 to 8 genotypes as the independent variables. All of the candidate genes accounted for 7.8% of the total variation in plasma TC, 7.6% of LDL-C, 7.3% of non-HDL-C, 7.2% of apo B, 3.2% of log TG and 5.1% of HDL-C. The partial R2 (presented as Pi in our study) of individual RFLP to serum trait ranged from 0.35% to 2.96%. Peacock et al.22) reported that EcoR I genotypes explained 1.3%, 1.6%, 2.9%, 2.2%, 0.5% and 3.0%, respectively, of total variations in serum TC, LDL-C, HDL-C, TG, apoAI and apoB in Swedish healthy individuals. Kessling et al.38) reported that the contribution of Xba I RFLP to serum LDL-C was 1.39%; HDL-C, 0.40%; apo AI, 0.11%; and EcoR I to apo AI, 2.99% (all adjusted for 6 covariates).
In the present study, the contribution of a lifestyle factor tended to be a little larger than that of any RFLP in the corresponding model. Pi of individual dietary factor ranged from 0.03% to 2.67%, smoking from 0.09% to 2.60%, drinking from 0.06% to 5.99%, and physical activity from 0.07% to 4.48% (Table 2, Figure 1).
According to a study by Perusse et al.39), diet, physical activity, smoking and drinking accounted for 6% of the total variation in serum LDL-C and 13% in serum HDL-C. These percentages are close to our data : 8.96% in LDL-C and 11.81% in HDL-C. Freeman and colleagues40) observed that the contribution of drinking, smoking and age to the level of HDL-C was 10.2% (7.4% in our study). Adachi et al.41) reported that the proportion of drinking, smoking and dietary variables was 16.9% for apo B, 37.7% for apo AI and 35.0% for apo CIII. The corresponding figures in the present study were 8.0%, 6.0%, and 7.2%.
Several family and twin studies mainly in Caucasian populations reported the genetic proportions ranged from 40% to 68% for serum LDL-C39,42,43), 14-66% for apo B44,45), 36-62% for HDL-C39,46,47) and 0-53% for apo AI44, 45,48). These figures were much higher than our and other candidate gene studies, although the methods are not comparable with each other. These differences might be, at least in part, due to their overestimation of the genetic part of the inter-individual variations in serum lipids or apolipoproteins besides ethnical difference. First, the overestimation might come from "shared environment". Family members were living together at a home for a long time, and twins grew up together at least in their early life. Second, "family aggregation" of serum lipids and lipoproteins in families might result in the overestimation also49). Third, a study50) reported that the contribution of genetic factors to serum lipids was larger in the younger persons than in the older because of "survivor effect" and accumulated exposures to environment. As the subjects in many genetic studies were relatively young, the large genetic proportion might be partially explained by age.
It should be also stressed to add more genetic markers to the model and/or to combine them, e.g., to investigate the haplotypes, because the multiple common mutations have more important effect on the "normal" inter-individual variations in serum lipids.
In the present study, it is the limitation that 24-hour recall method was used for assessing the dietary intake, because this measure does not reflect long-term dietary habits retrospectively. To overcome this weak point more or less, the absolute value of dietary intake was not employed but categorized into four patterns by quartile.
Obesity-related indices were not included in the analyses, because they were related not only to the energy output and input but also to many host factors, i.e., because it is difficult to say clearly if they are lifestyle markers or genetic markers.
In conclusion, some polymorphisms in selected candidate genes appeared to be associated with some serum lipids or apolipoproteins, but the effect of individual RFLP on the inter-individual variations in serum lipids was relatively subtle. There was no marked difference between the contribution of individual RFLP and that of individual lifestyle factor to any serum trait in the present study. We did not find out any convincing evidence that the role of the genetic factors was more important than that of the environmental factors in lipoprotein metabolism in our study population, although many researchers reported that the environmental factors are less important than the genetic factors in Western populations12). Appendix 3. Means and standard deviations of serum lipids and apolipoproteins.