Clinical Genetics 1990: 38: 281-294

Gene-gene interaction between the low density lipoprotein receptor and apolipoprotein E loci affects lipid levels JANCHR. PEDERSEN'"

AND K i R E

BERG'"

'Institute of Medical Genetics, University of Oslo, and *Department of Medical Genetics City of Oslo, Oslo, Norway A restriction fragment length polymorphism (RFLP) at the low density lipoprotein receptor (LDLR) locus, detectable with the restriction enzyme PvuII has been studied in a second series of Norwegian subjects, believed to be representative of the general population. The results confirm the association of normal alleles at the LDLR locus with differences in age- and sexcorrected total and LDL cholesterol. A gene identified as one of the alleles in this RFLP appears to eliminate the effect of the apolipoprotein E4 (apoE4) gene on cholesterol (or its allele must be present for the apoE4 effect on lipid level to manifest itself). The findings in this series substantiate previous indications that normal alleles are of importance in the control of LDLR activity and that normal LDLR alleles contribute to the population variation in cholesterol. Finally, they confirm that an interaction between LDLR and apoE genes contributes to the population variation in total and LDL cholesterol.

Received 16 March, accepted f o r publication 21 March 1990 Key words: apolipoprotein E (apo E); DNA; gene-gene interaction; genetics; LDL cholesterol; LDL receptor (LDLR); low density lipoprotein (LDL); restriction fragment length polymorphism (RFLP); normal alleles; total cholesterol.

The function of the low density lipoprotein receptor (LDLR) is to mediate the cellular uptake of apolipoprotein particles. The LDLR receptor pathway has been well studied in familial hypercholesterolemia (FH) (Brown & Goldstein 1986). As part of studies on the genetics of common diseases and risk factors, we are studying lipid, apolipoprotein and LDLR variation in the population at large, as well as in families with monogenic lipid disorders (Berg 1987, 1989). Analysis of within-pair difference in LDLR activity parameters in dizygotic (DZ) and monozygotic (MZ) twins (Magnus et al. 1981) uncovered significantly

higher within-pair difference in DZ than in MZ pairs. This pointed to an important influence of genes on the normal variation in LDLR activity. This finding was confirmed in a twin study of LDLR in mononuclear cells (Weight et al. 1981). In a study of LDLR activity in fibroblasts from heterozygous FH patients (who have only one normal LDLR allele) and normal subjects, the distributions of LDLR parameters suggested control of normal LDLR activity by a small number of normal alleles (Maartmann-Moe et al. 1981a). The possibility that normal, genetic LDLR variation could contribute to the population variation in cholesterol was further suggested by a negative

288

PEDERSENANDBERG

correlation between LDLR activity and ageand sex-adjusted serum cholesterol (Maartmann-Moe et al. 1981b). Several RFLPs have been detected at the LDLR locus, including an RFLP detectable with the restriction enzyme PvuII located in intron 15 (Hobbs et al. 1985, Humphries et al. 1985). Extensive linkage disequilibrium between RFLP's within the LDLR locus has been found (Leitersdorf et al. 1989 and unpublished data of our own). A cDNA probe (pLDLR-ZHHl) (Yamamot0 et al. 1984) encoding the region from exon 10 to exon 18 was kindly given to us by Drs. Russell, Brown and Goldstein. We have previously reported an association between the PvuII RFLP and lipid levels (Pedersen & Berg 1988). In the same series we found evidence of interaction between the apolipoprotein E4 (apoE4) isoform and LDLR alleles (Pedersen & Berg 1989) in determining total and LDL cholesterol variation. This suggested that functional LDLR variants identified as alleles in this RFLP eliminate or are necessary for the lipid effect of the apoE4 isoform. Since this could be a chance occurrence, we have now examined a second series of subjects in an effort to confirm or reject the hypothesis that normal alleles at the LDLR locus contribute to the population variation in serum total and LDL cholesterol, alone and/or in interaction with apoE alleles. In the present paper we report the results of this second study.

Material and Methods

Two hundred and thirty-nine ostensibly healthy, unrelated subjects, 117 females and 122 males (mean age 42 yrs; range 31-66 yrs) who were spouses of M Z twins drawn from the population-based Norwegian Twin Panel (Berg 1984) were analyzed with respect to apoE isoforms and the PvuII RFLP at the LDLR locus.

Blood Sampliitg The subjects were bled at the Institute of Medical Genetics, University of Oslo after an overnight fast. Biochemical tests and marker analyses were conducted without unnecessary delay. Whenever samples had to be stored, this was done at -96°C. Liyid Analyses Cholesterol was measured by the enzymatic method of Roschlau et al. (1974) and triglycerides by the enzymatic method of Wahlefeld (1974), employing a Multistat microcentrifugal fast analyzer (Instrumentation Laboratory, Lexington, Mass.). All results were recorded in mmol/l. LDL cholesterol was calculated according to Friedewald et al. (1972). Apolipoprotein E phenotyping ApoE phenotypes were determined by high resolution, two-dimensional polyacrylamide gel electrophoresis (Anderson & Anderson 1977), employing the modifications of Berrresen & Berg (1981). DNA Analyses DNA was prepared from buffy coat, essentially as described by Kunkel et al. (1977), and digested with the restriction enzyme PvuII, using conditions recommended by the manufacturer (Amersham International, Amersham, England). DNA fragments were separated electrophoretically according to size in 0.7% asarose gels. Transfer to Zeta Probe nylon membranes (Biorad, Richmond, Ca.) was done with the alkaline variant of the Southern blotting technique (Southern 1975). Prehybridization for 5 min was done in 0.5 mmol/l sodium phosphate buffer (Church & Gilbert 1984) with 10% SDS and 1 mmol/l EDTA. The probe pLDLR-2HH 1 was labelled with '*dCTP with a commercial nick translation kit (Amersham International, Amersham, England). Hybridization was performed at 65 "C

289

GENE-GENE INTERACTION

for 16 h in 500 mmol/l sodium phosphate buffer with 10% SDS and 1 mmol/l EDTA. Post-hybridization rinsing was done for 5 min in 133 mmol sodium phosphate buffer with 1YOSDS and 1 mmol/l EDTA followed by a 30-min wash with the same solution preheated to 65°C. A final high stringency 30-min wash was done in 5 mmol/l sodium phosphate buffer containing 0.1 % SDS. Blots were submitted to autoradiography for 2-3 days at - 70°C with Kodak XAR 5 films and a Kodak Super Rapid intensifying screen (Eastman Kodak Co., Rochester, N.Y.). Homozygotes for absence of the PvuII restriction site (genotype A2A2), exhibit a 16.5 kilobase (kb) band but no 14.0 kb or 2.5 kb band; homozygotes for presence of the site (genotype AlAl), lack the 16.5 kb band but have a 14.0 kb and a 2.5 kb band; and a heterozygote for the site (genotype AlA2) exhbits all three fragments (Fig. 1). In addition to the variable fragments, a 3.6 kb fragment was always observed.

Statistical Analyses Gene frequencies were determined by gene counting. 2 analyses were used to compare the population distribution with respect to LDLR and apoE genotypes, to test for Hardy-Weinberg equilibrium, and to analyze the distribution of subjects with different genotypes according to quartiles of lipid levels. Lipid and apolipoprotein concentrations were adjusted for age and sex according to Siervogel et al. (1980). Student’s t-test and analysis of variance were used to compare differences between means. The significance of outcome of analyses of variance were by statistics. less than 0.05 were considered significant. The size of the effect Of individualgenes on total and LDL cholesterol levels was measured as advised by Sing & Davignon (1985).

Results

Population Distribution of LDLR and ApoE Genotypes Table 1 shows the distribution of subjects according to genotypes in the LDLR and apoE polymorphisms. The distribution for each polymorphism was in excellent agreement with the expectations assuming Hardy-Weinberg equilibrium. No association was detected between the two polymorphisms, using x2 analysis. The frequencies of the LDLR (A) alleles and apoE (E) alleles were similar to those found in our previously published series (Pedersen & Berg 1989) (figures in brackets are for the previous series): A1 =0.18 (0.19), A2=0.82 (0.81); E2=0.09 (0.08), E3 =0.80 (0.76), and E4=0.11 (0.16).

416.5 414.0

4 3 6

+ 2.5

Fig. 1. DNA fragments observed by Southern Blot hybridization using an LDLR probe following Puvll digestion of individuat samples. The samples were from subjects of the following genotypes: A l A l (homozygous presence of site), A1A2 (heterozygotes) and A2A2 (homozyaous absence of site). The 2.5 kb band is frequently yr;; weak in the heterozygotes (AlA2).

290

PEDERSENANDBERG Table 1

Distribution of 239 unrelated Norwegians with respect to genotype determined by a polymorphic Pvull restriction site at the LDLR locus and apoE genotype No. with LDLR genotype

ApoE genotype

E2E2 E2E3 E2E4 E3E3 E3E4 E4E4

AlAl

A1A2

3 1 6 2

11 1 41 10

12

63

no,

A2A2

observed

4 19 102 37 2

4 33 2 149 49 2

164

239

It was concluded that the series analyzed was in Hardy-Weinberg equilibrium and probably representative of the population, and that there is no association belween the two polymorphisms. Total and LDL Cholesterol in Subjects With Different LDLR Genotypes Table 2 shows mean age- and sex-adjusted serum and LDL cholesterol in the 239 subjects according to genotype in the PvuII RFLP at the LDLR locus. Homozygotes for A1 had significantly lower levels of both total and LDL cholesterol than subjects with other genotypes. No effect of LDLR

genotype was observed on triglycerides or HDL parameters. Employing the method of Sing & Davignon (1985), it was calculated that the average effect of the A1 allele at the LDLR locus is to lower cholesterol level by 0.20 mmol/l and LDL cholesterol level by 0.23 mmol/l. The effect of the A2 allele is to increase serum cholesterol by 0.1 mmol/l and LDL cholesterol by 0.05 mmol/l. Table 3 shows the number of individuals in different quartiles of the population distribution of total cholesterol according to genotype in the LDLR polymorphism. More than half of the A l A l homozygotes had a value in the lowest percentile as opposed to less than a quarter of the subjects with other genotypes. When the subjects with A1A2 and A2A2 genotypes as well as those in the three highest quartiles of the population distribution were amalgamated to form a 2 by 2 table, a x2 with Yates correction of 5.7 was obtained and this was highly significant (p < 0.0 17). The odds ratio was 4.6 with lower 95% limit 1.24 and the relative risk was 2.5 with 95% limits 1.29 and 4.84, respectively. The distributions were the same for total and LDL cholesterol. Total and LDL Cholesterol in Subjects With Different Apolipoprotein E Genotypes Table 4 shows mean age- and sex-adjusted serum and LDL cholesterol in the 239 sub-

Table 2 Sex- and age-adjusted total cholesterol (TC) and LDL cholesterol (LDLC) in 239 unrelated healthy Norwegians according to genotype in an RFLP at the LDLR locus detected with the restriction enzyme Pvull LDLR genotype

AlAl A1A2 A2A2

TC

N 12 63 164

(mmol/l) S.E.M. 5.10 5.95' 5.96b

0.234 0.172 0.098

Table 3 Distribution of 239 unrelated Norwegians according to genotype in an RFLP detectable with the restriction enzyme Pvull at the LDLR locus and quartile of LDL cholesterol distribution

LDLC (rnmolll) S.E.M.

3.20 3.99' 4.07"

0.178 0.162 0.096

t-test for differences from genotype AlA1: 't=1.74; p = 0.04, b1t2.39; p=0.009, Y=2.08; p=O.O2, 9'2.41; p = 0.009.

Percentiles Genotype

75

All

AlAl A1A2 A2A2

7 17 36

3 13 44

1 18 40

1 15 44

12 63 164

Total

60

60

59

60

239

G E N E - G E N E INTERACTION Table 4 Sex- and age-adjusted total cholesterol (TC) and LDL cholesterol (LDLC) in 239 unrelated healthy Norwegians according to genotype in the apoE polymorphism ApoE genotype E2E2 E2E3 E2E4 E3E3 E3E4 E4E4

N

TC (mmol/l)

S.E.M.

4 33 2 149 49 2

4.99 5.56 6.58 5.83 6.43 6.36

0.89 0.17 0.40 0.10 0.20 0.86

LDLC (mmol/l) S.E.M. 3.24 3.58 4.43 3.94 4.53 4.54

1.40 0.18 0.50 0.10 0.19 0.98

jects according to genotype in the apoE polymorphism. The well-known -effect of apoE genes on total and LDL cholesterol is obvious. It was calculated that in this series the mean effects of the apoE2 and apoE3 genes were to lower cholesterol by 0.51 and 0.04 mmol/l, respectively, and the mean effect of the apoE4 gene was to increase total cholesterol by 0.52 mmol/l. Total and LDL Cholesterol in Subjects With DifSerent Combinations of LDLR and ApoE Genes Table 5 shows mean age- and sex-adjusted total and LDL cholesterol according to presence or absence of the A1 allele at the

29 1

LDLR locus and the apoA4 isoform. The level of total as well as LDL cholesterol was significantly hgher in subjects possessing the apoE4 isoform and lacking the A1 gene in the LDLR polymorphism (p 0.0005 for both total and LDL cholesterol) than in the other three groups combined. In subjects possessing the A1 gene in the LDLR polymorphism, there was no significant difference in serum cholesterol or LDL cholesterol level between subjects having and those lacking the apoE4 isoform. In subjects lacking the A1 gene in the LDLR polymorphism, the mean serum cholesterol average effect was 0.9 mmol/l higher and LDL cholesterol was 0.88 mmol/l higher in those possessing than in those lacking the apoE4 isoform. Table 6 shows the result of two-way analysis of variance of mean age- and sexadjusted total and LDL cholesterol according to presence or absence of the A1 gene in the LDLR polymorphism and the apoE4 allele. In this analysis subjects possessing the apoE2 isoform were omitted since the apoE2 allele, which is associated with lower lipid levels as well as hyperlipidemia of type I11 (Utermann et al. 1975, 1979), could in theory influence the analysis. When subjects possessing the apoE2 isoform were included

-=

Table 5 Age- and sex-corrected total cholesterol (TC) and LDL cholesterol (LDLC) in 239 unrelated, healthy Norwegians according to presence or absence of the A1 allele in an RFLP at the LDLR locus, and the apoE4 isoform, respectively ~

LDLR gene Present Absent

ApoE4 absent TC (S.E.M.) LDLC (S.E.M.) (mrnolll) 5.81 (0.17) 3.87 (0.16) 5.74 (0.10) 3.85 (0.12)

ApoE4 present TC (S.E.M.) LDLC (S.E.M.) (rnmol/l) (n=61)

5.83 (0.36) 3.85 (0.29)

(n=14)

(n = 125)

6.65 (0.21) 4.73 (0.20)

(n = 39')

'=Comparison between this group and the other three combined in one group. t =4.041 p