Clinical Genetics 1990: 38: 401-409
Association of DNA-haplotypes in the human LDL-receptor gene with normal serum cholesterol levels H. SCHUSTER', s. HUMPHRIES2,G.RAUH', c . HELD', CH. KELLER', G.WOLFRAM' AND N. ZOLLNER' 'Medizinische Poliklinik der Universitat Munich, Germany and *Charing Cross Sunley Research Centre, Hammersmith, London, UK For the low density lipoprotein receptor (LDLR), many mutations have been characterized which identify this gene as one with an important role in lipid metabolism in patients with familial hypercholesterolemia (FH). Genetic heterogeneity at this locus raises the possibility that the LDLR may also contribute to variation in cholesterol levels in the normocholesterolemic population. We have determined genotypes at the LDLR locus using restriction fragment length polymorphisms (RFLPs) detected with the enzymes StuI, ApalI, PvuII and NcoI in 324 normocholesterolemic individuals from Germany. A significant association (p 0.01) was detected between the cutting site for the PvuII RFLP and lower cholesterol levels, and variation associated with this polymorphism explains 3% of the sample variance in cholesterol levels. In family studies we have determined four-BFLP haplotypes of 148 independent LDLR genes and have observed 9 haplotypes in the population. Three of these haplotypes containing the cutting site for PvuII are associated with a reduction in plasma LDL-cholesterol levels. Phylogenetic analysis indicates that these three haplotypes are related by evolutionary history, and this suggests that a single functionally important sequence change in the LDLR explains our observations. Our data confirm other reports and strongly suggest that the LDLR locus may be one of those genes involved in determining serum cholesterol levels in the normal population. Received 10 October 1989. revised 14 June, accepted for publication 21 June 1990 Key words: LDL receptor gene; normal cholesterol; RFLP
It is well known that high levels of serum cholesterol are associated with an increased risk of coronary artery disease (CAD) (Goldstein et al. 1973). Most of the cholesterol in the blood is carried in low density lipoprotein (LDL) particles and the major protein in these particles is the apolipoprotein B (apoB) (Kane & Have1 1989). Levels of LDL in the blood are determined both by the rate of production of LDL, from VLDL and lipolysis, and the rate of removal (Sparks & Sparks 1985). LDL is removed both by non-receptor mediated pathways
and by interaction between the apoB protein and a specific receptor on the surface of all cells called the LDL-receptor. Major defects in this receptor are found in about 1 in 500 individuals and cause elevated levels of serum LDL-cholesterol and a disorder called famiiial hypercholesterolemia (Goldstein & Brown 1984). The gene for the LDLR has been cloned and sequenced, and used in studies of patients with FH (Sudhof et al. 1985). Many mutations in the receptor have been characterized at the molecular level (Hobbs et al. 1988), and have greatly
402
SCHUSTER ET A L .
helped in our understanding of the structure-function relationships of this protein. Patients with FH and a defect in the LDLR thus identify this gene as one with an important role in lipid metabolism. It is also possible that common sequence changes in the gene may exist which have only a small effect on function of the receptor. If such sequence changes are common, they may make an important contribution to determining lipid levels within the normal population. Data to support this hypothesis come from two types of studies. Firstly cholesterol levels within normal individuals overlap with cholesterol levels in patients with FH (Keller et al. 1981)suggesting there may be a gradation in receptor function within the normal population. Even in the families of patients with FH, there is significant overlap in the distribution of LDL cholesterol between those children who have inherited the defective gene and those who have not (Kwiterovich et al. 1974). Secondly, several studies that have measured LDLR activity in human fibroblasts have shown a considerable range in such activity even in fibroblasts from normal individuals (Maartmann-Moe & Berg-Johnsen 1981a). Family studies on normal individuals suggested evidence for four different LDL-receptor variants each coded for by different alleles at the LDL-receptor locus (Maartmann-Moe et al. 1981b). In a twin study, the intra-pair difference in the ability of fibroblasts to bind LDL was significantly smaller in monozygotic compared to dizygotic twins (Magnus et al. 1981). Taken together these data strongly suggest that variation at the LDLR locus contributes to between-individual difference in LDLcholesterol levels in the normal population. Recently association has been shown between cholesterol levels and the PvuII RFLP in the LDLR gene in the Norwegian population (Pedersen & Berg 1988). Additional studies of the same sample have
revealed an even stronger association which is mediated through isoforms of the apolipoprotein E (apE) (Pedersen & Berg 1989). There have now been over 12 RFLPs within or around the LDLR gene reported (summarized in Leitersdorf et al. 1989). Haplotype analysis has demonstrated a high degree of allelic heterogeneity at this locus. A sample of 20 Caucasian American pedigrees contained 3 1 different 10RFLP haplotypes (Leitersdorf et al. 1989). In the Afrikaner population, pedigree analysis in 50 families has shown segregation in the normal population of at least 17 haplotypes characterized by 10 RFLPs (Kotze et al. 1989). In a German sample we have previously shown that 38 FH families contained 10 different 4-RFLP haplotypes (Schuster et al. 1989). We have now examined these RFLPs in a group of 324 normocholesterolemic individuals from Germany and report results of association between genetic variation in the LDLR gene and serum cholesterol and LDL-cholesterol levels.
Subjects and Methods
Subjects Three hundred and twenty-four normocholesterolemic unrelated individuals with cholesterol levels within the 95% percentile of the normal distribution as described before (Heme et al. 1981) have been genotyped for four RFLPs in the LDL receptor gene. Blood samples were taken 12 h after the last meal. HDL-cholesterol was determined by' the precipitation method and LDL-cholesterol was estimated by the Friedewald formula (Friedewald et al. 1972), and was confirmed in 50 of the samples by preparative ultracentrifugation according to Have1 et al. (1955). Blood cells were stored at -30°C for later DNA analysis.
LDLR GENE AND SERUM CHOLESTEROL LEVELS
DNA Analysis The gene probe used was pLDLR-ZHHI, a kind gift of Dr. D. W. Russell, Dallas, USA. The plasmid contains a 1.9 kb fragment of the full length cDNA which hybridizes with exon 11 to exon 18 of the gene. The fragment was excised from a IYo low-gelling temperature agarose gel and labelled by means of a random oligonucleotide primer with a3'P-dCTP. The StuI, ApalI, PvuII and NcoI RFLPs were used to determine genotypes and haplotypes of the LDLR gene. Total genomic DNA was extracted from white blood cells using a Triton X-100 lysis method (Kunkel et al. 1977). Five micro-. grams of high molecular weight genomic DNA was digested with the restriction enzymes (Amersham International) under conditions recommended by the manufacturer. Digested and electrophoresed DNA was transferred to nylon membranes (Ameisham International) and hybridized with the labelled probe in 7% sodium dodecyl sulphate (SDS) and 0.5 M phosphate buffer (PH 7.0) at 65 "C. Filters were washed finally in 1 x SSC (150 mM NaCl, 15 mM sodium citrate), 0.1% SDS and exposed to X-ray films (Kodak X-Omat AR) for 1-3 days.
403
associated with RFLP alleles of haplotypes were calculated as described in Templeton et al. (1988). Results
Lipid and lipoprotein levels were obtained from 324 normolipidemic individuals. Sex subdivided mean (_+standarddeviation) age and lipid and lipoprotein estimates are summarized in Table 1. The difference in age between males and females was not significant; however, for triglycerides and HDLcholesterol the sex difference was highly significant (p < 0.0000 1). LDL-cholesterol levels were positively correlated with age in each sex group (r=0.51 for males, r=0.36 for females). Lipid and lipoprotein values were therefore adjusted for the age and sex effect by linear regression and the adjusted values used in subsequent analysis. Fig. la shows the map of the human LDLR gene and the location of the four variant restriction sites. For all four polymorphisms, genotype distribution and allele frequencies determined by gene counting, are shown in Table 2. The observed genotype distribution did not differ from that expected assuming Hardy-Weinberg equilibStatistical Methoh rium (data not shown). Differences between lipid and lipoprotein One-way analysis of variance was used to concentrations in males and females were compare the mean adjusted lipid and lipocompared using standard one-way analysis protein levels in groups of individuals with of variance. LDL-cholesterol levels were ad- different genotypes. Only for the PvuII justed for age and sex according to Siervogel polymorphism were statistically significant et al. (1980). Genotype distribution for each differences observed, in mean total cholesRFLP was determined by gene counting. terol levels (not shown). On further analysis, Chi-square analysis was used to test for differences in LDL-cholesterol were statistiHardy-Weinberg equilibrium. Genotype cally significant (Table 2), and differences in and haplotype association between LDL re- HDL-cholesterol were not. The PvuII rare ceptor genes and lipoprotein levels were allele characterized by the presence of the tested by variance analysis. Haplotypes variable cutting site was associated with sigwere deduced by inspection of the pedigrees nificantly lower levels of LDL-cholesterol and segregation of the RFLPs assuming no (Table 2). Subjects homozygous for the presrecombination. Significance was considered ence of the restriction site had a 12.5% to be at the 5% level. Average excess values lower mean cholesterol level than subjects
404
SCHUSTERETAL
homozygous for the absence of the restriction site (p < 0.01). Calculations show that the effect associated with the (+) allele (average excess) was to lower LDL-cholesterol by 3.0 mg/dl, and that associated with the (-) allele was to raise LDL-cholesterol by 1.0 mg/dl. The 4-RFLP haplotype at the LDL-receptor locus could be determined unambiguously in 74 unrelated normolipidemic subjects by following the inheritance of the LDLR gene in families. Of 16 possible haplotypes, only nine were observed, four of which were common (Table 3). Twenty different haplotype combinations were observed in this sample of 74 individuals, and the distribution of haplotypes and the mean adjusted LDL-cholesterol levels are presented in Table 4. Because many of the genotype classes are small, we have used the data to calculate the average excess associated with each haplotype (Table 3). By inspection, haplotypes 5 and 13, both of which are characterized by the presence of the PvuII site (+allele) are associated with a similar reduction in LDL-cholesterol level. Haplotype 6, which also carries the PvuII
cutting site, is associated with a smaller reduction in LDL-cholesterol. All other haplotypes are associated with a very small decrease or increase in LDL-cholesterol. Since haplotypes 5 and 13 are both associated with a reduction in LDL-cholesterol, this raises the possibility that there may have been sequence changes affecting LDLR function occurring independently on these two haplotypes. In order to investigate this we constructed a phylogenetic tree that reflects the evolutionary relationship between the haplotypes (Fig. lb). Such a phylogenetic tree of the LDL-receptor RFLPs has been presented recently (Leitersdorf et al. 1989). In the scheme presented in Fig. lb, all observed haplotypes can be derived uniquely from other haplotypes by a single mutational step, destroying or creating a restriction site. The exception to this is haplotype 7, which can be obtained by a single mutational step either from haplotype 3 by loss of the ApalI site, from haplotype 5 by loss of the PvuII site or from haplotype 8 by gain of the NcoI site. In addition, haplotype 1 was not observed in this sample and its existence has been inferred. Inspec-
Table 1 M e a n and standard deviation of lipid and lipoprotein levels in 324 normolipidemic subjects. LDLcholesterol levels have been adjusted for a g e and sex differences according to Siervogel Bt al.
(1980) Sex Males Females
n
Age
Trgl'
Chol
LDL-C
ad-LDL-C
HDL-CC
144 180
38f21 41f18
110245 89 f 40
201 f 31 208 34
132f23 134f24
134il9 133dz22
49f 12 59f13
~
* F = 47.60;p < 0.00001. F= 19.83;p
Association of DNA-haplotypes in the human LDL-receptor gene with normal serum cholesterol levels H. SCHUSTER', s. HUMPHRIES2,G.RAUH', c . HELD', CH. KELLER', G.WOLFRAM' AND N. ZOLLNER' 'Medizinische Poliklinik der Universitat Munich, Germany and *Charing Cross Sunley Research Centre, Hammersmith, London, UK For the low density lipoprotein receptor (LDLR), many mutations have been characterized which identify this gene as one with an important role in lipid metabolism in patients with familial hypercholesterolemia (FH). Genetic heterogeneity at this locus raises the possibility that the LDLR may also contribute to variation in cholesterol levels in the normocholesterolemic population. We have determined genotypes at the LDLR locus using restriction fragment length polymorphisms (RFLPs) detected with the enzymes StuI, ApalI, PvuII and NcoI in 324 normocholesterolemic individuals from Germany. A significant association (p 0.01) was detected between the cutting site for the PvuII RFLP and lower cholesterol levels, and variation associated with this polymorphism explains 3% of the sample variance in cholesterol levels. In family studies we have determined four-BFLP haplotypes of 148 independent LDLR genes and have observed 9 haplotypes in the population. Three of these haplotypes containing the cutting site for PvuII are associated with a reduction in plasma LDL-cholesterol levels. Phylogenetic analysis indicates that these three haplotypes are related by evolutionary history, and this suggests that a single functionally important sequence change in the LDLR explains our observations. Our data confirm other reports and strongly suggest that the LDLR locus may be one of those genes involved in determining serum cholesterol levels in the normal population. Received 10 October 1989. revised 14 June, accepted for publication 21 June 1990 Key words: LDL receptor gene; normal cholesterol; RFLP
It is well known that high levels of serum cholesterol are associated with an increased risk of coronary artery disease (CAD) (Goldstein et al. 1973). Most of the cholesterol in the blood is carried in low density lipoprotein (LDL) particles and the major protein in these particles is the apolipoprotein B (apoB) (Kane & Have1 1989). Levels of LDL in the blood are determined both by the rate of production of LDL, from VLDL and lipolysis, and the rate of removal (Sparks & Sparks 1985). LDL is removed both by non-receptor mediated pathways
and by interaction between the apoB protein and a specific receptor on the surface of all cells called the LDL-receptor. Major defects in this receptor are found in about 1 in 500 individuals and cause elevated levels of serum LDL-cholesterol and a disorder called famiiial hypercholesterolemia (Goldstein & Brown 1984). The gene for the LDLR has been cloned and sequenced, and used in studies of patients with FH (Sudhof et al. 1985). Many mutations in the receptor have been characterized at the molecular level (Hobbs et al. 1988), and have greatly
402
SCHUSTER ET A L .
helped in our understanding of the structure-function relationships of this protein. Patients with FH and a defect in the LDLR thus identify this gene as one with an important role in lipid metabolism. It is also possible that common sequence changes in the gene may exist which have only a small effect on function of the receptor. If such sequence changes are common, they may make an important contribution to determining lipid levels within the normal population. Data to support this hypothesis come from two types of studies. Firstly cholesterol levels within normal individuals overlap with cholesterol levels in patients with FH (Keller et al. 1981)suggesting there may be a gradation in receptor function within the normal population. Even in the families of patients with FH, there is significant overlap in the distribution of LDL cholesterol between those children who have inherited the defective gene and those who have not (Kwiterovich et al. 1974). Secondly, several studies that have measured LDLR activity in human fibroblasts have shown a considerable range in such activity even in fibroblasts from normal individuals (Maartmann-Moe & Berg-Johnsen 1981a). Family studies on normal individuals suggested evidence for four different LDL-receptor variants each coded for by different alleles at the LDL-receptor locus (Maartmann-Moe et al. 1981b). In a twin study, the intra-pair difference in the ability of fibroblasts to bind LDL was significantly smaller in monozygotic compared to dizygotic twins (Magnus et al. 1981). Taken together these data strongly suggest that variation at the LDLR locus contributes to between-individual difference in LDLcholesterol levels in the normal population. Recently association has been shown between cholesterol levels and the PvuII RFLP in the LDLR gene in the Norwegian population (Pedersen & Berg 1988). Additional studies of the same sample have
revealed an even stronger association which is mediated through isoforms of the apolipoprotein E (apE) (Pedersen & Berg 1989). There have now been over 12 RFLPs within or around the LDLR gene reported (summarized in Leitersdorf et al. 1989). Haplotype analysis has demonstrated a high degree of allelic heterogeneity at this locus. A sample of 20 Caucasian American pedigrees contained 3 1 different 10RFLP haplotypes (Leitersdorf et al. 1989). In the Afrikaner population, pedigree analysis in 50 families has shown segregation in the normal population of at least 17 haplotypes characterized by 10 RFLPs (Kotze et al. 1989). In a German sample we have previously shown that 38 FH families contained 10 different 4-RFLP haplotypes (Schuster et al. 1989). We have now examined these RFLPs in a group of 324 normocholesterolemic individuals from Germany and report results of association between genetic variation in the LDLR gene and serum cholesterol and LDL-cholesterol levels.
Subjects and Methods
Subjects Three hundred and twenty-four normocholesterolemic unrelated individuals with cholesterol levels within the 95% percentile of the normal distribution as described before (Heme et al. 1981) have been genotyped for four RFLPs in the LDL receptor gene. Blood samples were taken 12 h after the last meal. HDL-cholesterol was determined by' the precipitation method and LDL-cholesterol was estimated by the Friedewald formula (Friedewald et al. 1972), and was confirmed in 50 of the samples by preparative ultracentrifugation according to Have1 et al. (1955). Blood cells were stored at -30°C for later DNA analysis.
LDLR GENE AND SERUM CHOLESTEROL LEVELS
DNA Analysis The gene probe used was pLDLR-ZHHI, a kind gift of Dr. D. W. Russell, Dallas, USA. The plasmid contains a 1.9 kb fragment of the full length cDNA which hybridizes with exon 11 to exon 18 of the gene. The fragment was excised from a IYo low-gelling temperature agarose gel and labelled by means of a random oligonucleotide primer with a3'P-dCTP. The StuI, ApalI, PvuII and NcoI RFLPs were used to determine genotypes and haplotypes of the LDLR gene. Total genomic DNA was extracted from white blood cells using a Triton X-100 lysis method (Kunkel et al. 1977). Five micro-. grams of high molecular weight genomic DNA was digested with the restriction enzymes (Amersham International) under conditions recommended by the manufacturer. Digested and electrophoresed DNA was transferred to nylon membranes (Ameisham International) and hybridized with the labelled probe in 7% sodium dodecyl sulphate (SDS) and 0.5 M phosphate buffer (PH 7.0) at 65 "C. Filters were washed finally in 1 x SSC (150 mM NaCl, 15 mM sodium citrate), 0.1% SDS and exposed to X-ray films (Kodak X-Omat AR) for 1-3 days.
403
associated with RFLP alleles of haplotypes were calculated as described in Templeton et al. (1988). Results
Lipid and lipoprotein levels were obtained from 324 normolipidemic individuals. Sex subdivided mean (_+standarddeviation) age and lipid and lipoprotein estimates are summarized in Table 1. The difference in age between males and females was not significant; however, for triglycerides and HDLcholesterol the sex difference was highly significant (p < 0.0000 1). LDL-cholesterol levels were positively correlated with age in each sex group (r=0.51 for males, r=0.36 for females). Lipid and lipoprotein values were therefore adjusted for the age and sex effect by linear regression and the adjusted values used in subsequent analysis. Fig. la shows the map of the human LDLR gene and the location of the four variant restriction sites. For all four polymorphisms, genotype distribution and allele frequencies determined by gene counting, are shown in Table 2. The observed genotype distribution did not differ from that expected assuming Hardy-Weinberg equilibStatistical Methoh rium (data not shown). Differences between lipid and lipoprotein One-way analysis of variance was used to concentrations in males and females were compare the mean adjusted lipid and lipocompared using standard one-way analysis protein levels in groups of individuals with of variance. LDL-cholesterol levels were ad- different genotypes. Only for the PvuII justed for age and sex according to Siervogel polymorphism were statistically significant et al. (1980). Genotype distribution for each differences observed, in mean total cholesRFLP was determined by gene counting. terol levels (not shown). On further analysis, Chi-square analysis was used to test for differences in LDL-cholesterol were statistiHardy-Weinberg equilibrium. Genotype cally significant (Table 2), and differences in and haplotype association between LDL re- HDL-cholesterol were not. The PvuII rare ceptor genes and lipoprotein levels were allele characterized by the presence of the tested by variance analysis. Haplotypes variable cutting site was associated with sigwere deduced by inspection of the pedigrees nificantly lower levels of LDL-cholesterol and segregation of the RFLPs assuming no (Table 2). Subjects homozygous for the presrecombination. Significance was considered ence of the restriction site had a 12.5% to be at the 5% level. Average excess values lower mean cholesterol level than subjects
404
SCHUSTERETAL
homozygous for the absence of the restriction site (p < 0.01). Calculations show that the effect associated with the (+) allele (average excess) was to lower LDL-cholesterol by 3.0 mg/dl, and that associated with the (-) allele was to raise LDL-cholesterol by 1.0 mg/dl. The 4-RFLP haplotype at the LDL-receptor locus could be determined unambiguously in 74 unrelated normolipidemic subjects by following the inheritance of the LDLR gene in families. Of 16 possible haplotypes, only nine were observed, four of which were common (Table 3). Twenty different haplotype combinations were observed in this sample of 74 individuals, and the distribution of haplotypes and the mean adjusted LDL-cholesterol levels are presented in Table 4. Because many of the genotype classes are small, we have used the data to calculate the average excess associated with each haplotype (Table 3). By inspection, haplotypes 5 and 13, both of which are characterized by the presence of the PvuII site (+allele) are associated with a similar reduction in LDL-cholesterol level. Haplotype 6, which also carries the PvuII
cutting site, is associated with a smaller reduction in LDL-cholesterol. All other haplotypes are associated with a very small decrease or increase in LDL-cholesterol. Since haplotypes 5 and 13 are both associated with a reduction in LDL-cholesterol, this raises the possibility that there may have been sequence changes affecting LDLR function occurring independently on these two haplotypes. In order to investigate this we constructed a phylogenetic tree that reflects the evolutionary relationship between the haplotypes (Fig. lb). Such a phylogenetic tree of the LDL-receptor RFLPs has been presented recently (Leitersdorf et al. 1989). In the scheme presented in Fig. lb, all observed haplotypes can be derived uniquely from other haplotypes by a single mutational step, destroying or creating a restriction site. The exception to this is haplotype 7, which can be obtained by a single mutational step either from haplotype 3 by loss of the ApalI site, from haplotype 5 by loss of the PvuII site or from haplotype 8 by gain of the NcoI site. In addition, haplotype 1 was not observed in this sample and its existence has been inferred. Inspec-
Table 1 M e a n and standard deviation of lipid and lipoprotein levels in 324 normolipidemic subjects. LDLcholesterol levels have been adjusted for a g e and sex differences according to Siervogel Bt al.
(1980) Sex Males Females
n
Age
Trgl'
Chol
LDL-C
ad-LDL-C
HDL-CC
144 180
38f21 41f18
110245 89 f 40
201 f 31 208 34
132f23 134f24
134il9 133dz22
49f 12 59f13
~
* F = 47.60;p < 0.00001. F= 19.83;p
