点击显示 收起
1 From the Departments of Internal Medicine and Physical Sciences and Biocenter Oulu, University of Oulu, Oulu, Finland, and Hebrew UniversityHadassah School of Public Health and Community Medicine, Jerusalem.
2 Supported by grants from the Finnish Foundation for Nutrition Research, the Sigrid Jusélius Foundation, the Medical Council of the Academy of Finland, and the Finnish Heart Association.
3 Address reprint requests to M Rantala, Department of Internal Medicine, University of Oulu, Kajaanintie 50, FIN-90220 Oulu, Finland. E-mail: Maire.Rantala{at}oulu.fi.
See corresponding editorial on page 668.
ABSTRACT
Background: The genetic variance determining plasma lipid and lipoprotein concentrations may modify individual responsiveness to alterations in dietary fat and cholesterol content.
Objective: The aim was to examine the role of apolipoprotein (apo) B DNA polymorphisms in responsiveness of plasma lipids and lipoproteins to diet.
Design: A controlled dietary intervention study was conducted in 44 healthy, middle-aged subjects with a 3-mo baseline, a 1-mo fat-controlled, a 1-mo high-fat, and a 1-mo habitual diet period. We also conducted a meta-analysis of all published dietary trials, including our own.
Results: In our own dietary study, the apo B XbaI restriction-site polymorphism affected the responsiveness to diet of the plasma LDL-cholesterol concentration (P < 0.05, repeated-measures analysis of variance). Especially during the high-fat diet, homozygous absence of the XbaI restriction site (X-/X-) was associated with a greater increase in LDL cholesterol (44 ± 5%) than was X+/X+ (27 ± 7%) or X+/X- (40 ± 5%). The high-fat diet also induced a larger increase in plasma LDL cholesterol in subjects with the R-/R- genotype (homozygous absence of the EcoRI restriction site) (59 ± 10%) than in those with the R+/R- (39 ± 6%) or R+/R+ (36 ± 4%) genotype. The M+/M+ genotype (homozygous presence of the MspI restriction site) was also more responsive (41 ± 3% increase in LDL cholesterol) than the M+/M- genotype (27 ± 10% increase). The meta-analysis supported the finding of the significant role of the EcoRI and MspI polymorphisms, but not that of the XbaI polymorphism.
Conclusions: The present study indicated that the apo B EcoRI and MspI polymorphisms are associated with responsiveness to diet.
Key Words: Apolipoprotein B cholesterol diet meta-analysis human polymorphism restriction endonuclease restriction-fragment-length polymorphism restriction-site polymorphism plasma lipids
INTRODUCTION
Epidemiologic studies have shown an increase in the risk of atherosclerosis and coronary artery disease with increasing serum total and LDL-cholesterol concentrations (13). The primary treatment for hypercholesterolemia is a reduction of dietary fat and cholesterol intake and a replacement of dietary saturated fats by unsaturated fats. However, the responsiveness of plasma lipids to dietary changes seems to vary notably from one individual to another (4). The metabolic basis for the variation is not well known, but a considerable part of it may be controlled genetically, potentially by the genes encoding apolipoproteins, lipid-processing enzymes, lipid transfer proteins, and receptors involved in the regulation of lipoproteins. Except for the mutations of the LDL receptor gene (ie, familial hypercholesterolemia), no other single genetic marker of insufficient responsiveness to dietary modification has been uniformly identified. For example, the studies on the role of apolipoprotein (apo) E polymorphism as a significant modifier of dietary responsiveness have yielded contradictory results (521).
The genetic variation in apo B, the almost exclusive apoprotein of LDL particles, may play a major role in modifying the response to diet. The variability in the response of serum lipid concentrations has been shown to relate directly to the responsiveness of the LDL apo B production rate to dietary cholesterol (22). Furthermore, the polymorphisms of the apo B gene are associated with plasma total and LDL-cholesterol concentrations (2326). Several studies elucidated the role of the apo B polymorphism in response to diet (7, 20, 2737) and showed an association in some (7, 20, 2730, 32, 35, 36) but not all cases (31, 33, 34, 37).
We wanted to examine the role of apo B DNA polymorphisms in the response to diet with a dietary trial design. Because dietary intervention studies in large groups of human subjects are laborious, each study permits only a limited number of comparisons and limited power to obtain conclusive results on genotype and environmental (ie, diet) interactions (38). An alternative to repeating a study with a large sample size is to combine evidence from other studies. Furthermore, systematic reviews help to sort out contradictory results. The technique of testing the statistical significances of combined data is commonly called meta-analysis (39). We carried out such an analysis of several dietary intervention studies by combining their effects and significances into overall estimates and then testing for combined significance.
SUBJECTS AND METHODS
Our dietary intervention study
Study subjects and study design
The subjects were selected from among the kitchen and technical staff of the Oulu University Central Hospital. The details of this study were reported previously (6, 40). Briefly, at the first stage, 200 employees were screened for their apo E phenotype. Of this cohort, 21 subjects with apo E phenotype 4 (genotypes E3/4 or E4/4) volunteered for the dietary trial, and these subjects were matched for age, sex, and body weight with 23 individuals having apo E phenotype 3 (genotype E3/3). All the subjects were healthy and were taking no medication regularly. Their body mass indexes (BMIs; in kg/m2) ranged from 19 to 29 and their ages from 21 to 54 y, the mean age being 38 y.
Baseline diet (the subjects' habitual diet) was analyzed by 7-d food consumption records. All study subjects were advised about how to prepare food diaries and all food diaries returned were also checked by an experienced dietitian. This food record information was used to calculate the nutrient content of the baseline diets. The intervention diets (low fat and high fat) were designed on the basis of the regular hospital meals. The meals were analyzed chemically for 7 d to confirm the calculated contents of nutrients in the hospital diets. All food during both intervention periods (including breakfast, lunch, dinner, and snacks, 7 d/wk, for 1 mo plus 1 mo in sequence) was delivered from the hospital kitchen to be consumed by the study subjects. The switch-back diet was again the subjects' habitual diet. The percentage distribution of energy during the baseline diet was 17% from protein, 46% from carbohydrate, and 37% from fat and the daily cholesterol intake was 480 mg on average. According to the food diaries, none of the study subjects were following any extreme baseline diets. The percentage distribution of energy during the low-fat diet was 20% from protein, 55% from carbohydrate, and 24% from fat and the daily cholesterol intake was 240 mg on average. During the high-fat diet, 18% of energy was derived from protein, 45% from carbohydrate, and 36% from fat, and daily cholesterol intake was 420 mg.
The study consisted of a 3-mo baseline period; a 1-mo intervention period with a fat-controlled, low-cholesterol diet; a 1-mo intervention period with a high-fat, high-cholesterol diet; and, finally, a 1-mo switch-back period. The most appropriate energy intake was chosen on the basis of the subjects' food consumption records and daily physical activity. The mean daily energy intakes were 12.6 MJ for men and 8.4 MJ for women. All study subjects volunteered for the study, which was approved by the Ethical Committee of the University of Oulu.
Laboratory methods
Analyses of the apo B restriction-fragment-length polymorphisms (RFLPs) were performed with DNA samples prepared by using the salting-out method described by Miller et al (41) involving the following sites: XbaI RFLP in exon 26 (the third base in codon 2488 causing a cytosine to thymidine change without changing the amino acid sequence), EcoRI RFLP in exon 29 (the first base in codon 4154 causing a guanosine to adenine change and changing glutamic acid to lysine), MspI RFLP in exon 26 (the second base in codon 3611 causing an adenine to guanosine change and changing arginine to glutamine), Bsp 1261I RFLP in exon 4 (the second base in codon 71 causing a cytosine to thymidine change and changing threonine to isoleucine), and a 3-codon insertion-deletion polymorphism (alanine-leucine-alanine) in the signal peptide region of the human apo B gene.
The methods for determining the XbaI and EcoRI DNA polymorphisms were published earlier (42). The MspI, Bsp 1261I, and signal peptide polymorphisms were measured by using specific oligonucleotide primers and polymerase chain reaction amplification of digested DNA fragments as reported previously (4345). The alleles were designated as + or -, ie, X+/X+ refers to the homozygous presence of the apo B XbaI restriction site, X+/X- refers to the heterozygous presence of the restriction site, and X-/X- refers to the homozygous absence of the restriction site. The EcoRI, MspI, and Bsp 1261I restriction sites were designated as alleles R, M, and B, respectively. The more frequent allele of the apo B signal peptide (insertion of alanine-leucine-alanine codons) was designated as I and the rare allele (deletion of alanine-leucine-alanine codons) as D.
For lipid analyses, all blood samples were collected after subjects fasted overnight (12 h). The baseline blood samples were drawn twice with a 2-mo interval, and the baseline values were means of these 2 measurements. The low-fat diet period lasted for 1 mo and the high-fat diet period lasted for 1 mo; blood samples were drawn weekly. Only the mean of the 2 values from the last 2 wk were used for analysis. The switch-back blood samples were drawn twice with a 1-mo interval and the mean of these 2 measurements was used for analysis.
The plasma cholesterol and triacylglycerol concentrations were analyzed enzymatically (46, 47) by using a Gilford Impact 400E Clinical Chemistry Analyser (Gilford Instruments Laboratories, Oberlin, OH). The plasma HDL-cholesterol concentration was determined after precipitation of the plasma sample with heparin-manganese (48). LDL-cholesterol concentration was calculated according to the Friedewald formula (49). The concentrations of apo B and apo A-I were determined with the liquid-precipitate technique by using the nephelometric method (Turbox-Kit; Orion Diagnostica, Espoo, Finland).
In our laboratory, each series of measurements contained an internal stardard to control the accuracy of the method, and the interassay CV for, eg, plasma total cholesterol concentration, was 4%. The intraassay variation was controlled by several parallel determinations of the same sample and the intraassay CVs for plasma total cholesterol, triacylglycerols, and HDL cholesterol were 2%, 5%, and 1%, respectively.
Statistical analyses
The study subjects in the different apo B genotypic subgroups (n = 44) did not differ significantly in their BMIs, whereas their sex and age distributions differed significantly (Table 1). Therefore, the lipid values (Table 2) were age- and sex-adjusted by linear regression as proposed by Siervogel et al (50). The results are expressed as means ± SEMs. The dietary interventioninduced changes in plasma lipids and lipoproteins and the effects of apo B genotypes especially were tested with respect to the intraindividual variation during the dietary intervention periods. This was done by using a layered design in the form of repeated measures across time [repeated-measures analysis of variance (ANOVA)]. The top layer in the model was the between-subject layer, in which the effect of having a certain apo B genotype (eg, X-/X-, X+/X-, or X+/X+) was tested with respect to the interindividual variation. The bottom layer was the within-subject layer, in which the repeated-measures factor for the diet periods (baseline, fat controlled, high fat, and switch back) was tested with respect to the variation from one dietary period to another. For the significant effects revealed by repeated-measures ANOVA, a further paired Student's t test (or signed-rank test when appropriate) and Tukey's test were performed to evaluate the respective P values. For most of the analyses, the JMP statistical software program (SAS Institute Inc, Cary, NC) was used.
View this table:
TABLE 1.. Demographic data of study subjects in dietary intervention trials1
View this table:
TABLE 2.. Plasma lipid and lipoprotein values during the dietary trial according to apolipoprotein (apo) B genotype1
Meta-analysis
All published reports of dietary interventions conducted to investigate the effect of apo B DNA polymorphisms on plasma lipid responsiveness among healthy adults between 1980 and 1998 were identified by literature searches (National Center for Biotechnology Information at the National Library of Medicine, database coverage being Medline and Premedline; 51). Including our present study (6), 14 eligible reports were found (7, 20, 2737). Five reports turned out to be publications of the North Karelia Study (7, 2730) and were analyzed as one study in our meta-analysis. All the dietary interventions were conducted with use of solid food diets and the details of the dietary interventions were available. These details included the dietary fat content; the amounts of saturated, polyunsaturated, and monounsaturated fatty acids and dietary cholesterol; the duration of the trials; and the respective plasma lipid and lipoprotein values. Data on the apo B XbaI, EcoRI, MspI, signal peptide, and Bsp 1261I polymorphisms were available in most studies. The effects of dietary modification, ie, the change from a low-fat or low-cholesterol diet or a high-polyunsaturated, low-saturated fatty acid diet to a high-fat or high-cholesterol or low-polyunsaturated, high-saturated fatty acid diet, on plasma total, LDL-cholesterol, HDL-cholesterol, triacylglycerols, apo B, and apo A-I concentrations were investigated.
The 3 genotypes, -/-, -/+, and +/+, of each apo B polymorphism are indicated by the number of restriction sites P (P = 0, 1, 2). For each genotype, we were provided with the response to diet, sample size, variable mean, and SD (or variance) for both the low-fat and high-fat diets and for the male (M) and female (F) groups. We set up a new group called "all" (A) in each genotype group by combining the data for males and females and by computing the averages of variables and variances weighted by sample sizes and degrees of freedom, respectively.
From the data of each diet study (indexed by i where necessary), we first computed the effect size and the pooled SD for the 9 P by X subgroups (P = 0, 1, 2; X = M, F, A). We determined the response to diet, PX, as a difference between the high-fat (H) and low-fat (L) values of variables and the corresponding variances s2 for the subgroups from the following:
RESULTS
The fat content of the diet consumed in our own study was determined first by analyzing the regular hospital meals chemically for 1 wk (courtesy of the Agricultural Research Centre, Jokioinen, Finland). According to results of calculations made from the daily records and the chemical analysis, the fat content of the diet determined by these 2 methods did not differ significantly. The intervention diets were based on the regular hospital meals; the low-fat diet was prepared by limiting the amount of dairy products and adding margarine, polyunsaturated salad dressings, and vegetables; and the high-fat diet was prepared by increasing the amount of fatty dairy products, adding cold meats, and limiting vegetables. Thus, the fat contents of the intervention diets were easy to determine reliably by calculations made according to the daily records. The fat contents of diets in all other studies included in the meta-analysis were determined as indicated in the original reports.
Our dietary intervention study
Distribution of apo B polymorphisms
The allele frequencies for the rare alleles were 39% for X+, 24% for R-, 6% for M-, 26% for B-, and 25% for D. The genotype variants are presented in Table 1. For all the polymorphisms studied, genotype distributions were not significantly different from the Hardy-Weinberg prediction.
Baseline lipid values
Of the subjects screened for the dietary intervention study (n = 200), lipid values were available for 187 subjects. Among them, the X+ allele tended to associate with higher plasma total cholesterol concentrations, the age- and sex-adjusted mean plasma total cholesterol concentration being 5.79 ± 0.20, 5.40 ± 0.11, and 5.25 ± 0.12 mmol/L for the X+/X+, X+/X-, and X-/X- subjects, respectively. The M- and B- alleles were associated with both high plasma total and LDL-cholesterol concentrations. The M+/M- subjects had plasma total and LDL-cholesterol concentrations of 5.79 ± 0.19 and 3.73 ± 0.22 mmol/L and the M+/M+ subjects had concentrations of 5.32 ± 0.08 and 3.14 ± 0.09 mmol/L, respectively (P < 0.05 for both, Tukey's test). The B-/B- subjects had plasma total and LDL-cholesterol concentrations of 6.36 ± 0.32 and 4.09 ± 0.33 mmol/L, the B+/B- subjects had concentrations of 5.41 ± 0.12 and 3.25 ± 0.13 mmol/L, and the B+/B+ subjects had concentrations of 5.30 ± 0.10 and 3.12 ± 0.11 mmol/L, respectively (P < 0.01 and P < 0.05). In addition, the apo B signal peptide polymorphism was associated with the plasma LDL-cholesterol concentration: the D/D subjects had the highest LDL-cholesterol concentration (3.98 ± 0.32 mmol/L), the I/D subjects had intermediate values (3.29 ± 0.13 mmol/L), and the I/I subjects had the lowest values (3.09 ± 0.11 mmol/L; P < 0.05).
The demographic data of the 44 study subjects are shown in Table 1. The baseline plasma total cholesterol values of the study subjects were higher among those with the X+/X+ genotype than in those with the X-/X- and X+/X- genotypes (Table 2). No other significant differences between the genotypes were found in the baseline lipid values.
Effect of apo B polymorphisms on response to diet
The plasma total, HDL-cholesterol, LDL-cholesterol, apo B, and apo A-I concentrations decreased significantly during the low-fat diet and increased during the high-fat diet. In addition, the plasma triacylglycerol concentrations increased significantly during the low-fat diet and decreased during the high-fat diet (Table 2).
The XbaI polymorphism significantly affected the diet-induced responses of plasma total and LDL-cholesterol concentrations when all the dietary periods (baseline, low fat, high fat, and switch back) were analyzed together (P < 0.05, repeated-measures ANOVA). The subjects with the X-/X- genotype had the greatest increase in their plasma total and LDL cholesterol during the high-fat diet and the greatest decrease during the low-fat diet (Table 2 and Figure 1). The effects of the EcoRI, Bsp 1261I, MspI, and signal peptide polymorphisms on the response to diet were not significant when all the diet periods were analyzed together. The percentage increase in plasma total cholesterol during the high-fat diet was significantly greater in the subjects with the M+/M+ genotype than in those with the M+/M- genotype (29 ± 2% compared with 16 ± 5%; P = 0.02) and greater in the R-/R- subjects (35 ± 6%) than in the R+/R- (28 ± 3%) and R+/R+ subjects (26 ± 2%; P = 0.21). The percentage changes in plasma LDL-cholesterol concentrations in the different genotype groups are shown in Figure 1.
FIGURE 1. . Percentage changes in plasma LDL-cholesterol concentrations in the different apolipoprotein B genotype groups (n = 44). X refers to the XbaI polymorphism, R refers to the EcoRI polymorphism, M refers to the MspI polymorphism, I and D refer to a signal peptide insertion or deletion, and B refers to the BspI polymorphism; the genotypes are explained in more detail in Methods. Open bars indicate the percentage increase in the plasma LDL-cholesterol concentration (change from the low-fat diet to the high-fat diet). Solid bars indicate the percentage decrease in the plasma LDL-cholesterol concentration (change from the baseline diet to the low-fat diet). *Significantly larger change in X-/X- subjects than in X+/X- and X+/X+ subjects (P < 0.05); significantly larger change in R-/R- subjects than in R+/R+ and R+/R- subjects (P < 0.05). Number of study subjects in brackets.
Meta-analysis
The dietary intervention studies included in the meta-analysis are listed in Table 3. The data from the study by Friedlander et al (35) were divided into 2 separate data sets (a and b), because there were data from 2 separate dietary trials. All the studies had been conducted on healthy subjects, and the average duration of the dietary intervention was 6 wk (from 3 to 12 wk). The distribution of apo E phenotypes in the present study and that of Gylling et al (20) differed from that of the other studies, because these studies had been designed to include equal numbers of subjects with apo E3 (E3/3) and apo E4 (E3/4 or E4/4) phenotypes. Most of the studies had been designed to alter serum lipids, with several dietary factors being modified simultaneously, ie, the intake of total fat, saturated fatty acids, polyunsaturated fatty acids, or a combination of these, and the intake of dietary cholesterol (6, 7, 20, 2730, 3237), whereas one study was designed to alter serum lipids with one single dietary modification (31) (Table 4). Because all of the dietary modifications resulted in altered plasma total and LDL-cholesterol and apo B concentrations, and because it has been shown that persons responding to dietary cholesterol also tend to be more sensitive to changes in the quality of dietary fat (53), the different dietary studies were not considered to be distinct from each other, and thus suitable for meta-analysis. The X+ allele was significantly less common in one study (33) and the R- allele more frequent in one study (34) than in the other studies (Table 5). The main results of the meta-analysis are presented in Table 6. The comparison of the diet-induced changes (z scores) in plasma LDL-cholesterol concentrations between genotype groups +/- and +/+ are presented in Figure 2.
View this table:
TABLE 3.. Baseline characteristics and respective percentage changes in plasma LDL cholesterol concentration in relevant dietary intervention studies included in the meta-analysis1
View this table:
TABLE 4.. Comparison of dietary variables in the relevant intervention studies included in the meta-analysis1
View this table:
TABLE 5.. Allele frequencies of the rare alleles of apolipoprotein B restriction-fragment-length polymorphisms in relevant dietary intervention studies1
View this table:
TABLE 6.. Overall scaled effect sizes (z scores) and 95% CIs of apolipoprotein (apo) B genotypes on diet-induced responsiveness of plasma lipids and lipoproteins1
FIGURE 2. . Comparison of diet-induced changes (z scores) in plasma LDL-cholesterol concentrations between 2 genotype groups (+/- and +/+). The apolipoprotein B genotypes are explained in more detail in Methods. The horizontal bars represent 95% CIs. Overall: meta-analysis of the mean weighted z scores over separate studies (*P < 0.05, **P < 0.01). The numbers in parentheses correspond to the reference number of the relevant dietary intervention study.
Effect of apo B XbaI polymorphism on response to diet
The average diet-induced alterations in the plasma total cholesterol, HDL-cholesterol, and triacylglycerol concentrations did not differ between the subjects with the genotypes X+/X+, X+/X-, or X-/X-. The heterozygous presence of the XbaI restriction site was associated with a greater increase in the plasma LDL-cholesterol and apo A-I concentrations than was the homozygous presence of the restriction site (Figure 2 and Table 6). The plasma apo B concentration increased more in the X+/X+ subjects than in the X-/X- subjects (Table 6).
Effect of apo B EcoRI polymorphism on response to diet
The distribution of EcoRI polymorphisms in different dietary studies was skewed, and most studies included only one female or one male R-/R- subject. The overall effect analysis showed that the increase in plasma total cholesterol, LDL-cholesterol, and apo B concentrations was significantly greater in R-/R- subjects than in R+/R+ subjects (Table 6).
Effect of apo B MspI polymorphism on response to diet
The MspI polymorphism was analyzed in 5 dietary trials (6, 20, 29, 33, 35) (Table 4), and the distribution of this polymorphism was also skewed, M-/M- homozygotes being very rare. The M+/M+ genotype was associated with a greater alteration in plasma LDL-cholesterol, plasma triacylglycerol, and apo A-I concentrations than the M+/M- genotype. There was no significant effect of the apo B MspI polymorphism on the plasma total cholesterol, HDL- cholesterol, or apo B response (Table 6).
Effect of the apo B signal peptide polymorphism on response to diet
The apo B signal peptide insertion-deletion polymorphism was determined in 7 dietary trials (6, 28, 31, 33, 3537); meta-analysis revealed no significant effect of this polymorphism on the plasma total, LDL-cholesterol, HDL-cholesterol, triacylglycerol, or apo A-I response. Plasma apo B concentration increased more with high-fat diet in I/D subjects than in I/I subjects (Table 6).
Effect of apo B Bsp 1261I polymorphism on response to diet
The Bsp 1261I polymorphism was determined in only 2 of the dietary studies available, both of them conducted in Finland (6, 29). The distributions of the allele frequencies did not differ significantly between these 2 studies. The apo B Bsp 1261I polymorphism had a significant effect on the plasma total and LDL-cholesterol, triacylglycerol, and apo A-I responses: the B-/B- genotype was associated with a greater increase with diet in the plasma total cholesterol, LDL-cholesterol, and apo A-I concentrations and a significantly smaller alteration in the plasma triacylglycerol concentration than the B+/B- genotype (Table 6).
DISCUSSION
Apo B mutations might affect the plasma lipid responses or the plasma apo B concentrations during dietary modifications by altering apo B secretion, structural stability, affinity for the LDL receptor, or interactions of apo Bcontaining lipoproteins with other lipoproteins, cells, or enzymes (eg, hepatic or lipoprotein lipase). The physiologic role of the apo B XbaI polymorphism in codon 2488 in exon 26 is still unclear. The polymorphism alters plasma lipid concentrations (2326, 54, 55) and LDL catabolism (5658) even though it does not alter the amino acid sequence. In our dietary intervention study, the genotype distribution of the subjects did not differ from that of other cohorts in Finland (20, 27, 36). In accordance with some previous studies (2326, 55), the X+ allele was associated with higher basal plasma total and LDL-cholesterol concentrations, although this association was not observed in all studies (5961). The greatest diet-induced response in plasma total and LDL cholesterol was related to the dose of the X- allele, which agrees with the results of some previous studies (35, 36). The X-/X- genotype has also been associated with a greater postprandial response than the X+/X+ genotype (62), although an opposite result was published for a Finnish population (27). Because the previous studies have given inconclusive or contradictory results, a meta-analysis could be expected to clarify the role of the apo B XbaI polymorphism as a modifier of diet-induced responsiveness. The meta-analysis did not support the results of our own dietary trial, however. Actually, the comparison of effect sizes between all genotype groups did not reveal any systematic effect of either XbaI allele; it can thus be concluded that the data available now do not strengthen the role of the apo B XbaI polymorphism in diet-induced plasma lipid changes.
The discrepancy between our results about the XbaI polymorphism and the meta-analysis may be due to several reasons. First, our finding could be due to chance alone (type 1 statistical error). In that case, the meta-analysis would increase the statistical power. Second, the comparability of the separate studies pooled in the meta-analysis may have varied according to the selection of study subjects, the dietary interventions carried out, or the compliance of the subjects, and the combination of data may hence have masked the true effect. However, all the studies included are considered reliable and, furthermore, it has been shown that the diet-induced responsiveness of plasma lipids does not differ between modifications of dietary cholesterol content or the quality of fat (53). Our study differed in some respects from the others. Here, the number of subjects with apo E3 and apo E4 phenotypes did not differ between the apo B XbaI genotype groups, even though the apo E phenotype itself did not explain the responsiveness to diet (6). In addition to the apo E phenotype, there may have been some other, so far unknown, confounding genetic factors. As long as the possible functionally effective mutation linked with the XbaI polymorphism is unknown, the role of the apo B XbaI polymorphism will remain unclear.
The apo B EcoRI polymorphism in exon 29 changes the amino acid sequence but its functional role is unclear. The R- allele has been associated with high plasma total cholesterol concentrations and coronary heart disease in some studies (5961, 63), but not in all (55, 64). The EcoRI polymorphism has been advocated as a "variability" gene affecting the plasma apo B concentration (65, 66). Moreover, the polymorphism is restrictive, ie, in identical twins it reduces the variability in the serum cholesterol concentration between co-twins (67). Because identical twins share the same genetic code, the reason for the difference in the cholesterol concentration between co-twins must lie in environmental factors such as diet.
In our dietary intervention study, plasma LDL-cholesterol concentrations increased during the high-fat diet by as much as 59 ± 10% in R-/R- subjects, whereas smaller responses were seen in R+/R+ (26 ± 2%) and R+/R- (39 ± 6%) subjects. In the present meta-analysis, the EcoRI polymorphism affected the response of the plasma total cholesterol, LDL-cholesterol, and apo B concentrations to diet. When switched from the low-fat to the high-fat diet, the R-/R- subjects had the greatest increase in plasma lipids. To confirm this result, a prospective study of a large number of R-/R- individuals should be carried out.
The apo B MspI restriction site in codon 3611 is located in the same exon as the XbaI restriction site. The exact role of this variation in the apo B gene, which causes an amino acid substitution (arginine to glutamine), is unclear. Interestingly, in the screening population of the present study, the rare allele (M-) was associated with a higher basal cholesterol concentration, whereas in our dietary intervention study, the common allele (M+) was associated with a greater response in the plasma LDL-cholesterol, triacylglycerol, and apo A-I concentrations. This finding agrees with that of Series et al (68), who showed that the M- allele is associated with diet-resistant hypercholesterolemia.
The apo B signal peptide insertion-deletion polymorphism alters the N-terminal signal sequence and may alter the apo B secretion and lipoprotein particle responses in the postprandial state (69, 70). The plasma lipid concentrations in subjects with the deletion allele may also be less responsive to an increased amount of dietary fat (69, 71). In the present meta-analysis, no significant effect was associated with either the insertion or the deletion allele in the plasma total, LDL-, or HDL-cholesterol response, but the apo B response was associated with the D allele. The role of the signal peptide polymorphism may be more important in the postprandial plasma lipid response than in the general dietary one.
The apo B BspI polymorphism in codon 71 causes a change in the amino acid sequence. Because this polymorphism has been determined in only 2 dietary intervention trials (6, 27), its role in dietary responsiveness remains unconfirmed.
The response in our own dietary trial was very good, the average percentage increase in plasma LDL cholesterol being 39% with the high-fat diet. The remarkable plasma lipid response in these free-living subjects was possible partly because all the food consumed was supplied by the hospital kitchen and because compliance was very good. As shown before, the apo E4 phenotype was not associated with dietary responsiveness in our trial (6) or in many others (714), although some reports showed greater responsiveness in subjects with the apo E4 allele (1521). Because our study population had more subjects with the apo E4 phenotype than did the other studies included in the meta-analysis, the apo E phenotype may have been a confounding factor in the meta-analysis.
In conclusion, a dietary intervention in 44 healthy subjects showed an association between the apo B XbaI, EcoRI, and MspI RFLP-determined genotypes and the diet-induced plasma lipid response. The results of the meta-analysis supported our finding of the association between the apo B EcoRI and MspI genotypes and responsiveness to diet. However, the determination of the apo B DNA polymorphisms does not now add much clinical value to dietary counseling. The heterogeneity of the study populations with respect to other possible genetic variations affecting plasma lipids, eg, the apo E polymorphism, may be a major confounder and should be taken into consideration in future studies.
ACKNOWLEDGMENTS
We gratefully acknowledge the skillful technical assistance of Saija Kortetjärvi, Marja-Leena Kytökangas, and Sari Pyrhönen; we thank Juha Tienari and Juhani Tuominen for guidance concerning meta-analysis, and we acknowledge Helena Gylling, Tatu Miettinen, Terhi Hakala, Matti Tikkanen, Kimmo Kontula, Mavis Abbey, Steve Humphries, and Jim Mann for providing the data for the meta-analysis.
REFERENCES