| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Article |
Correspondence to: Anne Tybjærg-Hansen
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Variation in apolipoprotein (apo)E genotypes predicts variation in plasma cholesterol and apoB; however, the context-dependent associations between high density lipoprotein (HDL) cholesterol, apoA-I, triglycerides, and lipoprotein[a] (Lp[a]) and this polymorphism remain unsettled. We genotyped 5,025 women and 4,035 men sampled to represent a white general population in the age range 20 to 80+ years (mean ages 58 and 57 years for women and men, respectively). The relative frequencies of the 
22, 32, 42, 33, 43, and 44 genotypes were 0.005, 0.127, 0.027, 0.564, 0.251, and 0.027, respectively. Variations in apoE genotype (in the order listed above) predicted stepwise increases in cholesterol and apoB in both genders (all ANOVAs: P < 0.001), and stepwise decreases in HDL cholesterol and apoA-I in women (both ANOVAs: P < 0.001), but not in men. In both genders 33 individuals had the lowest levels of nonfasting triglycerides, whereas the highest levels were found in individuals with 22 and 44 genotypes (both ANOVAs: P < 0.001). Finally, a stepwise increase in Lp[a] was seen in women (ANOVA: P < 0.001), but not in men. In women, the association between variation in nonfasting triglycerides and Lp[a], and variation in apoE genotypes was mainly seen in those with the highest alcohol consumption, similar to the consumption of most men. Variations in apoE genotype predicted 5% and 11% in women, and 2% and 6% in men, of the total variation in plasma cholesterol and apoB, respectively. Variation in levels of plasma lipoproteins is associated with variation in apoE genotypes in the population at large, with the most pronounced association in women, except for nonfasting triglycerides, for which the association is most pronounced in men.
Whereas the associations between variation in plasma cholesterol and apoB and the variation in apoE genotypes seem invariant, the associations with variation in plasma HDL cholesterol, apoA-I, nonfasting triglycerides, and Lp[a] seem context dependent. — Frikke-Schmidt, R., B. G. Nordestgaard, B. Agerholm-Larsen, P. Schnohr, and A. Tybjærg-Hansen. Context-dependent and invariant associations between lipids, lipoproteins, and apolipoproteins and apolipoprotein E genotype. J. Lipid Res. 2000. 41: 1812;–1822.
Supplementary key words: genes, atherosclerosis, interactions
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Three different apolipoprotein (apo)E alleles (2, 3, and 4) on the long arm of chromosome 19, each allele encoding one apoE isoform (apoE-2, apoE-3, apoE-4), result in six different genotypes (22, 32, 42, 33, 43, and 44) (1) (2). The apoE polymorphism explains about 20% of the variance in plasma levels of apoE (3). ApoE-2 differs from apoE-3 by a cysteine-for-arginine substitution at amino acid residue 158, whereas apoE-4 differs from apoE-3 by an arginine-for-cysteine substitution at residue 112 (4). ApoE mediates the catabolism of chylomicron and very low density lipoprotein (VLDL) remnants via a "remnant" receptor, and the binding of chylomicron remnants, VLDL, and intermediate density lipoproteins to the low density lipoprotein (LDL) receptor (4).
The apoE polymorphism is potentially one of the most important genetic predictors of plasma lipoprotein levels and thus of risk of ischemic heart disease (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14). It is well established that variation in plasma levels of cholesterol and apoB is associated with variation in apoE genotypes, and that these associations seem to exist in most contexts, that is, appear to be invariant. In contrast, the associations between variation in plasma levels of high density lipoprotein (HDL) cholesterol, apoA-I, triglycerides, and lipoprotein[a] (Lp[a]), and variation in apoE genotypes remain unsettled, possibly because these associations may be context dependent. Most previous studies (4) (5) (6) (7) (8) (9) have explored these associations in women and men combined or in men alone, and have not explored other contexts separately.
We conducted a study of 5,025 women and 4,035 men from the Danish general population, the Copenhagen City Heart Study (15), to assess whether variations in apoE genotypes predicted variation in plasma cholesterol, apoB, HDL cholesterol, apoA-I, nonfasting triglycerides, and Lp[a]. A priori we stratified our analysis by gender, but also determined whether these associations were context dependent within each gender. A particularly important aspect of our study is that we were able to explore whether apoE genotypes could predict variation in plasma nonfasting triglycerides. Because this is the natural state for most humans in the larger part of a 24-h cycle, it could be argued that this context is more important to study than that of the conventional fasting state, which occurs for only a few hours every morning. It should be emphasized, however, that conclusions about the associations between variation in nonfasting triglycerides and variation in apoE genotypes may not be comparable to conclusions based on fasting data.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Subjects
We studied individuals who participated in the third examination of the Copenhagen City Heart Study from 1991 through 1994. This prospective cardiovascular population study includes an almost equal number of women (55%) and men stratified into 10-year age groups from 20 to 80 years and above, drawn randomly from the general population of the city of Copenhagen around Copenhagen University Hospital, Rigshospitalet, using the Copenhagen Central Population Register. A detailed description of the first (1976 through 1978) and second (1981 through 1983) examination of this study has been published previously (15). The original cohort from 1976;–1978, supplemented with five hundred 20- to 24-year-olds at the second examination, and 500 individuals in each of the age groups 20 through 24, 25 through 29, 30 through 34, 35 through 39, 40 through 44, and 45 through 49 years at the third examination, were all invited to participate in the third examination at Rigshospitalet, Copenhagen. Of the 17,180 individuals invited, 10,049 participated, 9,259 gave blood, and 9,241 were genotyped. Fewer than 1% were nonwhites and 98.8% were Danish citizens, that is, for practical purposes were of Danish descent (16). All participants gave informed consent; 181 were excluded because of lipid-lowering medication. The study was approved by the Danish ethics committee for Copenhagen and Frederiksberg (No. 100.2039/91).
Questionnaire
A self-administered questionnaire was filled in before the examination, and validated by an investigator on the day of attendance. All subjects reported the weekly intake of alcoholic beverages in the form of number of beers, glasses of wine, or units of spirits [1 unit = 12 g of alcohol and corresponds to 1 beer (33 cl), 1 glass of wine, or 2 cl of spirits], and reported use of antihypertensive medication and diuretics, treatment for diabetes mellitus, smoking status, and physical activity, and women in addition reported menopausal status, and whether they used hormonal replacement therapy (HRT) (15).
Laboratory analyses
Total cholesterol, apoB, HDL cholesterol, triglycerides, apoA-I, glucose (Boehringer Mannheim, Mannheim, Germany), and Lp[a] total mass (Dako, Carpinteria, CA) levels were measured in nonfasting plasma (15). The polyclonal apo[a]-specific antiserum used in this assay was sensitive to apo[a] heterogeneity.
ApoE genotypes were identified by polymerase chain reaction (PCR) followed by digestion with HhaI as described (17), except that, because of the large sample size, we used a 5% agarose gel instead of a polyacrylamide gel. On the agarose gel we could not always detect the 48-bp band that distinguished between the 22 and 32 genotypes; we therefore retyped all 22 and 32 genotypes, using a second PCR (sense primer, 5'-ACATGGAGGAC GTGTGCGG-3'; antisense primer, 5'-ACGCGGCCCTGTTCC ACCA-3') followed by digestion of the PCR product (250 bp) with HaeII (22 homozygotes: 2 x 187 bp; 32 heterozygotes: 1 x 187 bp, 1 x 152 bp, 1 x 35 bp; and common bands of 32, 18, and 13 bp). The 187-bp band and 152-bp band were clearly distinguishable on an agarose gel. Seven of 48 22 individuals had originally been misclassified as 32.
Other measurements
Height, weight, waist/hip ratio, and systolic and diastolic blood pressure (BP) were measured as described (15). Body mass index (BMI) was determined as weight divided by height squared (kg/m2).
Statistical methods
Statistical analyses were performed for each gender separately, using the SPSS (Chicago, IL) program (18). A P value of <0.05 was considered significant. The distributions of plasma cholesterol, apoB, HDL cholesterol, apoA-I, nonfasting log10 triglycerides, and log10 Lp[a] appeared approximately normally distributed on inspection of bar graphs. The log10 transformation of Lp[a] was slightly overtransformed; however, for simplicity we still chose this transformation instead of a more unconventional transformation. When Kolmogorov-Smirnov normality tests were performed for distributions of these six traits as a whole as well as within each of the six genotypes separately, almost all 42 normality tests for each gender showed statistical evidence of non-normal distribution. Therefore, we have chosen mainly to show the results from the nonparametric analyses, except when nonparametric tests were not available: we used Levene's test for homogeneity of variance, a parametric test not dependent on the assumption of normality (18) (19), and analysis of covariance (ANCOVA) for multivariate analyses.
To test the first study hypothesis (which states that means and variances of plasma cholesterol, apoB, HDL cholesterol, apoA-I, nonfasting triglycerides, and Lp[a] do not differ as a function of apoE genotype in either gender), the Kruskal-Wallis analysis of variance (ANOVA) was used to evaluate heterogeneity of the means across genotypes. The Mann-Whitney U test was used as a post-hoc test for two-genotype comparisons. Levene's test examined homogeneity of variance. Average allele effects were calculated as described (13), and allele frequencies were estimated by the gene-counting method.
To test the second study hypothesis (which states that the apoE genotype does not interact with other lipid or nonlipid cardiovascular risk factors in the prediction of the six lipid, lipoprotein, and apolipoprotein traits), homogeneity of the association of genotype and each of the following cardiovascular risk factors on the six lipid, lipoprotein, and apolipoprotein traits were tested using bivariate interaction terms in an ANCOVA also including genotype and the risk factor in question: cholesterol, apoB, HDL cholesterol, apoA-I, nonfasting triglycerides, Lp[a], BMI, waist/hip ratio, glucose, alcohol consumption, systolic BP, diastolic BP, age in 10-year groups, hypertension, diuretics, diabetes mellitus, smoking, and physical activity for both women and men separately, and in addition menopausal status and use of HRT for women.
Prior to tests for interactions, all continuous covariates were evaluated with regard to normality and homogeneity of variances across genotypes. Distributions of apoB in both genders and HDL cholesterol and glucose levels in females appeared approximately normally distributed on inspection of bar graphs, whereas statistically there was evidence of heterogeneity of variance. Logarithmically transformed, the traits still appeared approximately normally distributed and the heterogeneity of variance disappeared. Distributions of nonfasting triglycerides and Lp[a] were skewed and showed evidence of heterogeneity of variance across genotypes. Logarithmically transformed, these traits appeared approximately normally distributed on inspection of bar graphs in both genders, and showed no heterogeneity of variance across genotypes, except for Lp[a] in men; when 22 males were excluded from the analysis this heterogeneity almost disappeared (before exclusion, P = 0.001; after exclusion, P = 0.039). We have therefore chosen to ignore this heterogeneity when testing for interactions between apoE genotype and covariates in predicting levels of Lp[a] in men, which represents a limitation of that analysis. A further limitation is that alcohol intake was not normally distributed. All other continuous variables appeared normally distributed and showed no heterogeneity of variance.
All significant bivariate interactions apparent from the 216 ANCOVAs performed were further explored by dividing the interacting covariates into categories, tertiles, or quintiles, followed by tests of heterogeneity of means across the six genotypes. Bonferroni corrections for multiple comparisons (19 for women, 17 for men) after interaction tests were performed for each lipid trait in both genders (20).
The residual marginal "effect" of apoE genotype on the total variation in the six lipid, lipoprotein, and apolipoprotein traits was determined. First, the six lipid traits were adjusted by ANCOVA for age in 10-year groups, diabetes mellitus, physical activity at leisure, antihypertensive drugs, diuretics, BMI, and alcohol consumption in both women and men, and in addition for menopausal status and use of HRT in women. Second, an ANOVA on these adjusted lipid traits was used to calculate the genetic variance due to apoE (sG2) and the within genotype group variance (sW2) from sum of squares and mean square error from the ANOVA table. Finally, the proportion of the total phenotypic variability attributable to apoE genotype was estimated by sG2/(sG2+sW2) (21). If there was interaction between genotype and the covariate, the residual marginal effect of apoE genotype was calculated in tertiles of the interacting covariate.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The relative apoE genotype and allele frequencies in these white Danish individuals sampled from the general population were as follows: 22, 0.005; 32, 0.127; 42, 0.027; 33, 0.564; 43, 0.251; and 44, 0.027; also 2, 0.082; 3, 0.753; and 4, 0.165. Genotype frequencies were similar for women and men (
2: P = 0.4), and did not differ significantly from those predicted by the Hardy-Weinberg equilibrium (2: 0.2 < P < 0.3). However, as shown in previous studies (22) (23) (24) (25) the frequency of the 4 allele decreased significantly with age from 0.156 in the 20- to 30-year-olds to 0.137 in individuals above 80 years of age (P = 0.05). Table 1 shows basic characteristics of the 5,025 women and 4,035 men sampled from the general population.
|
Means and variances of lipids, lipoproteins, and apolipoproteins as a function of apoE genotype
There was a stepwise increase as a function of genotype (22, to 32, to 42, to 33, to 43, to 44) and alleles (2, to 3, to 4) in cholesterol and apoB, in both women and men (all ANOVAs: P < 0.001) ( Fig 1 and Fig 2); for cholesterol this pattern was confirmed in the same individuals on levels measured 10 and 15 years earlier (15) (all ANOVAs: P < 0.001; data not shown). On post-hoc Mann-Whitney U tests, most two-genotype comparisons ( Table 2) of mean values (Fig 1) were statistically significant for both traits in either gender. The absolute increase from 22 to 44 of age-adjusted levels of cholesterol and apoB was 1.1 mmol/l and 46 mg/dl in women, and 0.8 mmol/l and 43 mg/dl in men, respectively.
|
|
|
A significant stepwise decrease was observed for HDL cholesterol and apoA-I as a function of apoE genotype in women (P < 0.001 for both traits), but not in men (Fig 1 and Fig 2); for HDL cholesterol this pattern was confirmed on levels measured 10 years earlier in the same individuals (15) (women, P < 0.001; men, P = 0.17; data not shown). For apoE alleles from 2 to 3 to 4 there were stepwise decreases in HDL cholesterol and apoA-I in women, with a weak similar trend in men (Fig 2). On post-hoc Mann-Whitney U tests, 32 versus 33, 32 versus 43, 32 versus 44, and 33 versus 43 differed for both HDL cholesterol and apoA-I in women (Table 2, Fig 1). The absolute decrease from 22 to 44 of age-adjusted levels of HDL cholesterol and apoA-I in women was -0.16 mmol/l and -10 mg/dl, respectively (Fig 1).
In both genders 22 and 44 individuals had the highest nonfasting triglyceride levels whereas 33 individuals had the lowest levels (Fig 1, Table 2). This nonfasting triglyceride pattern was confirmed in the same individuals for values measured 15 years earlier (15) (women, P = 0.02; men, P = 0.001; data not shown). Most heterozygotes appeared to have intermediate levels of plasma triglycerides (Fig 1). Accordingly, both the 2 and 4 alleles predicted higher levels of nonfasting triglycerides than the 3 allele (Fig 2). The absolute increases in plasma triglycerides from 33 to 22 in women and men were +0.4 and +0.7 mmol/l, and from 33 to 44 +0.2 and +0.7 mmol/l, respectively (Fig 1).
Levels of Lp[a] increased from 22 to 33 in women, but not in men (Fig 1). On post-hoc Mann-Whitney U tests, 22 versus 32, 22 versus 42, 22 versus 33, 22 versus 43, 32 versus 33, and 32 versus 43 differed significantly in women (Table 2, Fig 1). Accordingly, the 2 allele predicted lower Lp[a] levels than the two other alleles in women, but not in men (Fig 2). The absolute increase in Lp[a] levels from 22 to 33 in women was 13 mg/dl (Fig 1).
There was evidence of heterogeneity of variance between apoE genotypes for cholesterol, not present in HDL or Lp[a] in both genders, and for log10 Lp[a] in men (data not shown): this appeared to be explained by a larger variance for 22 individuals than for individuals with other genotypes. A likely explanation is that some but not all 22 individuals develop type III hyperlipoproteinemia (1). The heterogeneity of variance seen for apoB in both genders across genotypes and for HDL cholesterol in women mainly reflected that the variance increased with increasing mean levels of apoB and HDL cholesterol.
Interaction of apoE genotype with other cardiovascular risk factors in the prediction of lipid, lipoprotein, and apolipoprotein traits
Of 216 interactions tested ( Table 3), 25 were significant (P < 0.05): in 16 interactions, the association between variation in the trait examined and variation in apoE genotypes showed irregular patterns rather than monotonic and consistent associations in different strata of the interacting covariate, and thus suggested chance observations, whereas the 9 interactions described below appeared with some consistency between related traits, and/or appeared biologically plausible, that is, were consistent with knowledge that the interacting covariate and the dependent lipid trait might be associated, and that the same or a similar association was found with related traits.
|
Three of four interactions involving apoE genotype and cholesterol predicting variation in apoB, or apoE genotype and apoB predicting variation in cholesterol, were highly significant (P < 0.001). The cause of these interactions may be that individuals with the 22 genotype (28 women and 20 men) were not equally distributed in the apoB quintiles (data not shown): there were 22 women in the first, 5 in the second, and 1 woman in the third quintile; for men there were 18 in the first, 1 in the second, and 1 in the fifth quintile. The stepwise increase in apoB by apoE genotypes remained significant in all cholesterol quintiles; however, the stepwise increase in cholesterol disappeared when data were stratified in apoB quintiles (data not shown).
Four significant interactions (and one borderline) between apoE genotype and glucose in predicting variation in HDL cholesterol and apoA-I in both genders and nonfasting triglycerides in men, were present: in women, the stepwise decrease in HDL cholesterol and apoA-I as a function of genotype was present only in the highest glucose quintile, and this was most likely the cause of the interaction, whereas for the remaining three interactions the cause was not evident ( Fig 3).
|
Alcohol consumption interacted with apoE genotype in women in predicting variation in nonfasting triglycerides and Lp[a]: variations in apoE genotypes predicted variation in nonfasting triglycerides and Lp[a] mainly in the highest tertile of alcohol consumption ( Fig 4). Although these interactions were not found in men, the stratified results are still shown for men for comparison.
|
Contribution of apoE genotype to variability in lipids, lipoproteins, and apolipoproteins
In women, variation in apoE genotypes predicted 4.5 and 10.7% of the variation in cholesterol and apoB after adjustment for age in 10-year groups, diabetes mellitus, physical activity at leisure, antihypertensive drugs, diuretics, BMI, alcohol consumption, menopausal status, and use of HRT ( Table 4). Because of significant interactions between apoE genotype and alcohol consumption in predicting variation in nonfasting triglycerides, and Lp[a] in women, and between apoE genotype and glucose levels in predicting variation in HDL cholesterol and apoA-I in women and HDL cholesterol, apoA-I, and nonfasting triglycerides in men, the marginal genotype "effects" on these lipid traits were estimated in tertiles or quintiles of the interacting covariates ( Table 5). In men, variation in apoE genotypes predicted 2.3, 6.4, and 0.4% of the variation in cholesterol, apoB, and Lp[a], respectively (Table 4).
|
|
For comparison, age in 10-year groups alone predicted 23.5, 20.2, 0.2, 2.7, 11.2, and 0.8% in women and 7.3, 7.6, 0.8, 2.4, 2.6, and 0.1% in men of the total variation in cholesterol, apoB, HDL cholesterol, apoA-I, nonfasting triglycerides, and Lp[a], respectively (data not shown).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The main novel observations in the present study are as follows: General: i) comprehensive two-genotype comparisons of all possible apoE genotype combinations for six lipid, lipoprotein, and apolipoprotein traits in both genders (Fig 1 and Table 2); ii) comprehensive tests of interaction between apoE genotype and 20 other covariates in this relatively large sample (Table 3). Specific: iii) the contribution of apoE genotype to the total variance in cholesterol and apoB is considerably greater in women than in men; iv) 42 individuals have cholesterol and apoB levels between 32 and 33 individuals; v) variations in apoE genotypes predict variation in plasma HDL cholesterol, apoA-I, and Lp[a] in a gender-specific manner; vi) in women, variations in apoE genotypes predict variation in nonfasting triglycerides and Lp[a] dependent on alcohol consumption, and variation in HDL cholesterol and apoA-I dependent on glucose levels. In men, variations in apoE genotypes predict variation in HDL cholesterol, apoA-I, and nonfasting triglycerides dependent on glucose levels.
The demonstrated cholesterol and apoB lowering associated with the 2 allele, and the cholesterol and apoB increase associated with the 4 allele, are well established (4) (6) (7) (8) (9) (10); however, the reported order of the six genotypes associated with stepwise increases of mean levels of cholesterol and apoB varies somewhat (4) (13). We and others (14) found that apoB increases from 22 to 32 to 42 to 33 to 43 to 44 and that cholesterol increases from 22 
32 to 42 to 33 to 43 to 44 in both genders.
Another novel finding is that the stepwise increase in cholesterol associated with the six apoE genotypes disappeared in quintiles of apoB, whereas in contrast quintiles of cholesterol did not abolish the stepwise increase in apoB. This suggests that apoB is the factor primarily associated with apoE genotype. These and previous studies (4) (26) support the hypothesis that the affinity of the apoE isoforms to the LDL receptor are inversely correlated with the ability of the LDL receptor to remove apoB-containing lipoproteins from plasma, and thereby positively correlated with plasma cholesterol levels.
It has previously been demonstrated that the associations between variation in lipid and lipoprotein levels with variation in apoE genotypes are context dependent, with many different environmental and biological factors as proposed interactors (27). Conflicting results concerning the association between variation in levels of HDL cholesterol and apoA-I with variation in apoE genotypes have been reported (4); however, Reilly et al. (10) (11) (12) demonstrated that these associations were influenced by gender, which could explain the controversy. In agreement with this, we demonstrated a stepwise decrease in HDL cholesterol and apoA-I levels from 22 to 32 to 42 to 33 to 43 to 44 in women, but not in men. Kaprio et al. (3) estimated that the 4 allele predicted an increase in triglycerides in women, and a decrease in HDL cholesterol in men. We find, however, that the 4 allele predicted an increase in nonfasting triglycerides and a decrease in HDL cholesterol in both women and men. Furthermore, it has been suggested that variation in HDL cholesterol associated with variation in apoE genotypes in women is dependent on the use of HRT (3) (28) in agreement with the present findings of statistically significant borderline interactions (HDL cholesterol, P = 0.15; apoA-I, P = 0.09) between apoE genotype and HRT in postmenopausal women on levels of HDL-cholesterol and apoA-I. The stepwise decreasing pattern in HDL cholesterol and apoA-I from 22 through on to 44 was present in premenopausal (n = 1,437) and untreated postmenopausal women (n = 2,859), but not in postmenopausal women treated with HRT (n = 683), where neither HDL cholesterol nor apoA-I showed a decreasing pattern in mean levels from 22 to 44 (Kruskal-Wallis ANOVA P = 0.26, and P = 0.27; data not shown). The variation in cholesterol and apoB associated with variation in apoE genotypes was not abolished by HRT; the stepwise increase from 22 through on to 44 was present in all three female groups.
Why the variation in HDL cholesterol, apoA-I, and nonfasting triglycerides associated with variation in apoE genotypes is most obvious in the quintile of highest plasma glucose is not completely clear. However, it is well known that patients with elevated glucose and insulin resistance, and thus non-insulin-dependent diabetes mellitus, have hypertriglyceridemia as well as low HDL levels due to abnormalities in both production and clearance of VLDL triglycerides and HDL cholesterol/apoA-I (29); HDL particles containing both cholesterol and apoA-I are produced as excess surface material when triglycerides in VLDL are hydrolyzed. In accordance with this in the present study, levels of nonfasting triglycerides increased and levels of HDL cholesterol and apoA-I decreased as a function of increasing glucose levels in both genders (data not shown). It therefore could be speculated that the rather discrete variation in HDL cholesterol and apoA-I associated with variation in apoE genotypes is more pronounced in the highest quintile of glucose because this is a substrate-rich environment with plenty of triglycerides, and thus a great potential for production of HDL particles. In such a context, variations in apoE genotypes may in particular predict variation in HDL cholesterol and nonfasting triglyceride levels.
It is interesting that our results on variation in apoE genotypes as predictors of variation in nonfasting triglycerides are in agreement with previous results in studies of fasting triglycerides (6) (7) (9): individuals in the present study with the 2 or 4 allele when compared with individuals with the most common genotype (33, "wild-type") had the highest nonfasting triglyceride levels. However, in contrast to the present findings some studies did not find an association between the 44 genotype and higher triglyceride levels (30) (31). In the present study alcohol consumption interacted with apoE genotype on nonfasting plasma triglycerides in women, but not in men. In women, this could be explained by an association between the 4 allele and raised nonfasting triglyceride levels mainly in the 30% of women with a weekly alcohol intake above 6 units. In men, the mean weekly consumption of alcohol was almost 2H times higher than in women (14 units) and 60% of men drank more than 6 units of alcohol per week, possibly explaining the lack of a statistically significant apoE-alcohol interaction in men. This may explain why levels of nonfasting triglycerides associated with 4 were higher in men than in women in the Copenhagen City Heart Study and why high triglyceride levels were not associated with the 44 genotype in a large Turkish study (31), where members of the largest part of the population are Muslims with restricted alcohol intake.
Our results on triglycerides in the nonfasting state may differ from those in the fasting state; however, studies of both fasting (6) (7) (9) and nonfasting triglycerides (30) have demonstrated that high triglyceride levels are associated with the 4 allele, in accordance with findings in the present study on nonfasting triglycerides. In spite of this, dietary status may still have some impact on the association between, particularly, the 44 genotype and triglyceride levels, as 44 individuals in the Turkish Heart Study (31) had fasting triglyceride levels similar to those of 33 individuals. Therefore, the fasting and nonfasting state may represent two different contexts with respect to apoE genotype and effect on triglycerides. Although it is conventional to measure triglycerides in the fasting state to minimize variability due to fat intake, it could be argued that studies of the nonfasting state may be more relevant as most humans spend more hours in the nonfasting state (up to 8 h after the last meal) than in the fasting state (more than 8 h after the last meal). It should be emphasized, however, that conclusions about genotypes as predictors of nonfasting variation may not be comparable to conclusions based on fasting data.
The association between variation in Lp[a] levels and variation in apoE genotypes is not quite clear (7) (32) (33) (34). One study suggests that variation in levels of Lp[a] is not directly associated with variation in apoE phenotypes: variations in apoE phenotypes predict variations in lipids and lipoproteins, which in turn seem to influence levels of Lp[a] (32). In support of this we found, at least in men, that apoE genotype interacted with levels of cholesterol and triglycerides in the prediction of Lp[a]. In women, apoE genotype interacted with alcohol consumption in predicting Lp[a], due to an effect in those who drank most. Potentially, the interaction between apoE genotype and alcohol consumption in predicting triglycerides could also explain the interaction between apoE genotype and alcohol consumption in predicting Lp[a] levels, as apo[a] associates with triglyceride-rich lipoproteins at least in the postprandial state (34). In the present study the association between variation in levels of Lp[a] and variation in apoE genotypes was strongest in women, whereas an overall association did not seem to exist in men. Another study found no overall association between variation in levels of Lp[a] and variation in apoE phenotypes in 466 white men, but suggested that the association with variation in apoE phenotypes was dependent on apo[a] isoforms (33). We suggest that variation in levels of Lp[a] associated with variation in apoE genotypes is gender specific and context dependent.
Potential limitations of the present study need to be discussed. Chance findings are always a crucial issue when multiple statistical tests are performed. We chose to perform 216 interaction tests between apoE genotype and several covariates in the prediction of the six lipid traits, because we did not want to miss any interaction that potentially could explain an important context-dependent effect. We did not trust solely in the statistical P value, but stratified all significant and borderline significant interactions in strata of the interacting covariate. On inspection of these bar graphs, we evaluated if the interaction was biologically plausible, that is, was consistent with knowledge that the interacting covariate and the dependent lipid trait might be associated, and that the same or similar associations were found with related traits, or if it showed irregular patterns most likely due to chance findings. Furthermore, misclassification of genotypes is unlikely as all photographs of gels were scrutinized by two different researchers unaware of the associations between variation in lipids, lipoproteins, and apolipoproteins and variation in apoE genotypes. In addition, the genotype distribution in the general population was similar to that predicted by the Hardy-Weinberg equilibrium, supporting that no major misclassification took place. We cannot, however, totally exclude misclassification of a few apoE genotypes, as we genotyped more than 9,000 individuals.
In conclusion, the present results suggest that in the population at large variation in plasma lipoprotein levels is associated with variation in apoE genotypes, especially in women. Whereas the associations between variation in levels of cholesterol and apoB and variation in apoE genotypes seem invariant between studies, the associations with variation in levels of HDL cholesterol, apoA-I, nonfasting triglycerides, and Lp[a] seem from our studies to be context dependent. Our data suggest that associations between cardiovascular disease and variation in apoE genotypes will most likely differ substantially by both genotype and gender due to variation in levels of lipids, lipoproteins, and apolipoproteins.
|
ACKNOWLEDGMENTS |
|---|
Pia T. Petersen, Mette Refstrup, and Hanne Damm provided assistance. This work was supported by the Danish Heart Foundation, The Danish Medical Research Council, Chief Physician Johan Boserup's and Lise Boserup's Fund, Sven Hansen's and Wife Ina Hansen's Fund, the Beckett Foundation, the European Organization for the Control of Circulatory Diseases, and the Danish Stroke Association.
Manuscript received October 14, 1999; and in revised form February 15, 2000; and in revised form June 19, 2000
Abbreviations: ANCOVA, analysis of covariance; ANOVA, analysis of variance; apo, apolipoprotein; BMI, body mass index; BP, blood pressure; HDL, high density lipoprotein; HRT, hormonal replacement therapy; LDL, low density lipoprotein; Lp[a], lipoprotein[a]; PCR, polymerase chain reaction; VLDL, very low density lipoprotein
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  Y.-C. Yen, G. W. Rebok, J. J. Gallo, M.-J. Yang, F.-W. Lung, and C.-H. Shih ApoE4 Allele Is Associated With Late-Life Depression: A Population-Based Study Am J Geriatr Psychiatry, October 1, 2007; 15(10): 858 - 868. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. M. Bennet, E. Di Angelantonio, Z. Ye, F. Wensley, A. Dahlin, A. Ahlbom, B. Keavney, R. Collins, B. Wiman, U. de Faire, et al. Association of Apolipoprotein E Genotypes With Lipid Levels and Coronary Risk JAMA, September 19, 2007; 298(11): 1300 - 1311. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Anuurad, G. Lu, J. Rubin, T. A. Pearson, and L. Berglund ApoE genotype affects allele-specific apo[a] levels for large apo[a] sizes in African Americans: the Harlem-Basset Study J. Lipid Res., March 1, 2007; 48(3): 693 - 698. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. H. Wittrup, R. V. Andersen, A. Tybjaerg-Hansen, G. B. Jensen, and B. G. Nordestgaard Combined Analysis of Six Lipoprotein Lipase Genetic Variants on Triglycerides, High-Density Lipoprotein, and Ischemic Heart Disease: Cross-Sectional, Prospective, and Case-Control Studies from the Copenhagen City Heart Study J. Clin. Endocrinol. Metab., April 1, 2006; 91(4): 1438 - 1445. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Stengard, S. L. R. Kardia, S. C. Hamon, R. Frikke-Schmidt, A. Tybjaerg-Hansen, V. Salomaa, E. Boerwinkle, and C. F. Sing Contribution of regulatory and structural variations in APOE to predicting dyslipidemia J. Lipid Res., February 1, 2006; 47(2): 318 - 328. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Frikke-Schmidt, B. G. Nordestgaard, P. Schnohr, R. Steffensen, and A. Tybjaerg-Hansen Mutation in ABCA1 Predicted Risk of Ischemic Heart Disease in the Copenhagen City Heart Study Population J. Am. Coll. Cardiol., October 18, 2005; 46(8): 1516 - 1520. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Benn, B. G. Nordestgaard, J. S. Jensen, P. Grande, H. Sillesen, and A. Tybjaerg-Hansen Polymorphism in APOB Associated with Increased Low-Density Lipoprotein Levels in Both Genders in the General Population J. Clin. Endocrinol. Metab., October 1, 2005; 90(10): 5797 - 5803. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Benn, B. G. Nordestgaard, J. S. Jensen, K. Nilausen, H. Meinertz, and A. Tybjaerg-Hansen Mutation in Apolipoprotein B Associated with Hypobetalipoproteinemia Despite Decreased Binding to the Low Density Lipoprotein Receptor J. Biol. Chem., June 3, 2005; 280(22): 21052 - 21060. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Ribalta, C. J.M. Halkes, J. Salazar, L. Masana, and M. C. Cabezas Additive Effects of the PPAR{gamma}, APOE, and FABP-2 Genes in Increasing Daylong Triglycerides of Normolipidemic Women to Concentrations Comparable to Those in Men Clin. Chem., May 1, 2005; 51(5): 864 - 871. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Song, M. J. Stampfer, and S. Liu Meta-Analysis: Apolipoprotein E Genotypes and Risk for Coronary Heart Disease Ann Intern Med, July 20, 2004; 141(2): 137 - 147. [Abstract] [Full Text] |
