Multiple Rare Alleles Contribute to Low Plasma Levels of HDL Cholesterol

See allHide authors and affiliations

Science  06 Aug 2004:
Vol. 305, Issue 5685, pp. 869-872
DOI: 10.1126/science.1099870


Heritable variation in complex traits is generally considered to be conferred by common DNA sequence polymorphisms. We tested whether rare DNA sequence variants collectively contribute to variation in plasma levels of highdensity lipoprotein cholesterol (HDL-C). We sequenced three candidate genes (ABCA1, APOA1, and LCAT) that cause Mendelian forms of low HDL-C levels in individuals from a population-based study. Nonsynonymous sequence variants were significantly more common (16% versus 2%) in individuals with low HDL-C (<fifth percentile) than in those with high HDL-C (>95th percentile). Similar findings were obtained in an independent population, and biochemical studies indicated that most sequence variants in the low HDL-C group were functionally important. Thus, rare alleles with major phenotypic effects contribute significantly to low plasma HDL-C levels in the general population.

Many clinically important quantitative traits are highly heritable, but progress in the elucidation of their genetic architecture has been limited. Because quantitative traits do not segregate in Mendelian fashion in most families, their distribution in the population is presumed to reflect the cumulative contribution of multiple common DNA sequence variants that each has a small effect (1). Sequence variants with strong phenotypic effects may also contribute to variation in complex traits (2). Although these variants are likely to be rare individually, they may be sufficiently common in aggregate to contribute to variation in common traits in the population. Whereas most Mendelian disorders are caused by a spectrum of different mutations in a gene (or genes) (3), the contribution of rare alleles to more common, quantitative traits has not been systematically evaluated.

In this study, we evaluated the hypothesis that rare sequence variations contribute significantly to low plasma levels of high-density lipoprotein cholesterol (HDL-C), a major risk factor for coronary atherosclerosis. If this hypothesis is correct, then mutations that impair HDL production or enhance HDL catabolism should be significantly more common among individuals with low plasma levels of HDL-C than among those with high plasma levels of HDL-C. Furthermore, sequence variants with major phenotypic effects are likely to be found exclusively at one extreme or the other, whereas alleles found in both the high HDL-C and the low HDL-C groups are likely to be neutral with respect to plasma HDL-C levels. The prevalence of mutations with major effects on plasma HDL-C levels is not known. Molecular defects causing rare genetic forms of HDL deficiency have been identified in the genes encoding apolipoprotein AI (APOA1), the major protein component of HDL (4); the adenosine triphosphate binding cassette (ABC) transporter A1 (ABCA1), required for the efflux of cholesterol from cells to HDL particles (5); and lecithin cholesterol acyltransferase (LCAT), the enzyme that catalyzes the formation of cholesteryl esters in HDL (6). Homozygotes for mutations in these genes have virtually no circulating HDL-C, whereas heterozygotes have about half the normal plasma level of HDL-C. Mutations in ABCA1 have also been associated with a less severe form of familial hypoalphalipoproteinemia (7, 8), but systematic screening of these genes in population-based samples has not been reported.

To determine whether sequence variations in these three candidate genes contribute to low plasma levels of HDL-C in the general population, we sequenced the coding regions and consensus splice sites of each gene in 32 individuals from each of four gender and race groups [black men, white men, black women, and white women (table S1)] constituting the upper and lower 5% of the distribution of plasma HDL-C levels in a population-based study of Dallas County residents (9). Nonsynonymous sequence variations were significantly more common in the low HDL-C group than in the high HDL-C group (Table 1). Of the 128 individuals with low plasma levels of HDL-C, 21 (16%) had sequence variants not present in the high HDL-C group (Table 2). In contrast, only 3 (2%) of the individuals in the high HDL-C group had sequence variants not found in the low HDL-C group (P < 0.0001, Fisher's exact test). A total of 20 of the 21 mutant alleles found only in the low HDL-C group were in ABCA1 (fig. S1) and two of these (W590S and N1800H) were previously identified in individuals with hypoalphalipoproteinemia (6).

Table 1.

Sequence variations in the coding regions of ABCA1, APOA1, and LCAT. Values represent the numbers of sequence variants identified in 256 individuals from the Dallas Heart Study (DHS) (128 with low HDL-C and 128 with high HDL-C) and 263 Canadians (155 with low HDL-C and 108 with high HDL-C) (17). NS, nonsynonymous (nucleotide substitutions resulting in an amino acid change); S, synonymous (coding sequence substitutions that do not result in an amino acid change). GenBank accession numbers for DHS ABCA1, APOA1, and LCAT sequences are NM_005502, NM_000039, and NM_000229, respectively.

Sequence variants unique to one group Sequence variants common to both groups
Low HDL-C High HDL-C
ABCA1 14 6 2 5 10 19
APOA1 1 0 0 1 0 1
LCAT 0 1 1 0 1 1
ABCA1 14 2 2 3 7 5
APOA1 0 1 0 0 2 0
LCAT 6 1 0 0 0 0
Table 2.

Nonsynonymous sequence variations in ABCA1, APOA1, and LCAT in Dallas Heart Study participants with low (n = 128) or high (n = 128) plasma HDL-C levels. The effect of each amino acid (18) substitution on protein function was predicted with the use of PolyPhen (11, 19). Substitutions affecting known functional residues in the protein or that resulted in large changes in surface accessibility or side chain volume of the affected amino acid were scored as probably damaging. Substitutions that negatively affected predicted transmembrane regions or that were predicted to have smaller effects on surface accessibility or side chain volume of the affected amino acid on the basis of inferred protein structure were scored as possibly damaging. Conservative substitutions at poorly conserved residues were scored as benign. NA, not applicable.

Nucleotide change Amino acid change (View inline) n Race Predicted effect
c.593C>A S198X 1 Black NA
c.742G>A P248A 1 Black Benign
c.1201A>C K401Q 1 White Benign
c.1769G>C W590SView inline 1 Black Probably damaging
c.1913G>A R638Q 1 Black Possibly damaging
c.2320A>T T774SView inline 4 Black Benign
c.2320A>T E815G 1 White Probably damaging
c.2444A>G S1181F 1 White Possibly damaging
c.3542C>T R1341T 1 Black Possibly damaging
c.4022G>C S1376G 1 Black Benign
c.4126A>G R1615Q 1 Black Possibly damaging
c.4844G>A A1670T 1 Black Possibly damaging
c.5008G>A N1800HView inline 1 White Possibly damaging
c.5398A>C D2243E 4 Black Benign
c.152G>C R51T 1 Black Possibly damaging
c.1486C>T R496W 1 White Probably damaging
c.5039G>A R1680Q 1 White Possibly damaging
c.340G>A V114M 1 Black Benign
  • View inline* These sequence variations have been identified previously.

  • Thus, one of six individuals with HDL-C levels below the fifth percentile in the Dallas Heart Study had a rare mutation in ABCA1 or APOA1. Similar analyses were performed in a second sample comprising white Canadians who had either low (10 to 34 mg/dl) or high (58 to 116 mg/dl) levels of HDL-C (Table 3). A total of 21 (14%) of the 155 Canadians in the low HDL-C group had sequence variants that were not present in the 108 Canadians with a high plasma level of HDL-C. In contrast, only three individuals in the high HDL-C group (3%) had sequence variants not present in the low HDL-C group (P < 0.001, Fisher's exact test). Four sequence variants found in the Canadian low HDL-C group (ABCA1 P85L, N1800H, S1731C, and R1851X) had been previously identified in subjects with hypoalphalipoproteinemia (10). One mutation (ABCA1 N1800H) was found in both the Canadian and Dallas low HDL-C groups. In both Canadian and Dallas groups, almost all of the excess sequence variations in the low HDL-C group were in ABCA1. This may reflect the size of the coding sequence of ABCA1, which is three times larger than those of the other two genes; it is also possible that sequence changes occur more frequently at this locus or that mutations arising in ABCA1 are more likely to persist in the population.

    Table 3.

    Nonsynonymous sequence variations in ABCA1, APOA1, and LCAT in white Canadians with low (n = 155) or high (n = 108) plasma HDL-C levels. Cholesterol efflux data for the controls is the mean ±SD. ND, not done; NA, not applicable. Functional effects of amino acid substitutions were estimated as described in the legend to Table 2.

    Nucleotide change Amino acid change (View inline) n Predicted effect Cholesterol efflux (% of total cellular cholesterol/2 hours)
    Low HDL-C ABCA1
    c.254C>T P85LView inline 1 Probably damaging 0.8
    c.917G>A R306H 1 Benign ND
    c. 1375A>C T459P 1 Possibly damaging 0.28
    c. 1651C>G H551D 1 Probably damaging 0.32
    c. 1769G>T W590L 1 Probably damaging 0.31
    c.2893T>C R965C 1 Probably damaging 0.59
    c.3077T>C L1026P 1 Benign 0.25
    c.3966G>A W1322X 1 NA 0.38
    c.4156G>C E1386Q 1 Benign 0.51
    c.4430G>T C1477F 1 Probably damaging 0.34
    c.5116G>A D1706N 1 Possibly damaging 0.38
    c.5192C>G S1731CView inline 1 Possibly damaging 0.28
    c.5398A>C N1800HView inline 1 Possibly damaging 0.27
    c.5864C>T R1851XView inline 1 NA 0.26
    c.6217A>G T2073A 1 Possibly damaging 0.28
    Controls (n = 46) 0.52 ± 0.07
    Low HDL-C APOA1
    Low HDL-C LCAT
    c.167T>C L56P 1 Possibly damaging ND
    c.254G>A W85X 1 NA ND
    c.382G>A G128S 1 Probably damaging ND
    c.463A>G N155D 1 Possibly damaging ND
    c.901G>A D301N 1 Benign ND
    c. 1103G>T G396V 1 Probably damaging ND
    High HDL-C ABCA1
    c. 1664A>C K555T 1 Benign ND
    c.2320A>C T774PView inline 2 Benign 0.63/0.57
    High HDL-C APOA1
    High HDL-C LCAT
  • View inline* These sequence variations have been identified previously.

  • To test whether differences in population substructure between the low HDL-C and high HDL-C groups contributed to the observed differences in the prevalence of rare sequence variants, we compared the frequency of synonymous substitutions in the high and low HDL-C groups. The frequency of synonymous substitutions was similar in the high HDL-C and low HDL-C groups in both the U.S. and Canadian populations (Table 1). Furthermore, the racial composition of the low HDL-C and high HDL-C groups was matched. Thus, it is unlikely that the excess sequence variations in the low HDL-C group is attributable to either population stratification or chance variations in allele frequencies.

    Analysis using a computer program [Poly-Phen (11)] that predicts effects of amino acid changes on protein function (12) indicated that 72% of the nonsynonymous sequence variants identified in the low HDL-C group were likely to adversely affect protein function. In a previous study (12), 69% of nonsynonymous mutations associated with a functional disorder were predicted to be damaging with the use of this algorithm, compared to 32% of nonsynonymous sequence variants identified in DNA samples from arbitrarily selected individuals. Thus, the proportion of nonsynonymous sequence variants predicted to be damaging in the low HDL-C groups with the use of PolyPhen was comparable to that obtained for disease-associated sequence variants (12, 13).

    Biochemical measurement of cholesterol efflux rates in cells from Canadian subjects with nonsynonymous sequence variants in ABCA1 provided further evidence that the sequence variants identified in this study were causally related to low plasma levels of HDL-C (Table 3). Eleven of 14 ABCA1 sequence variants represented in the low HDL-C group were associated with cholesterol efflux rates that were two standard deviations or more below the mean value obtained in subjects with normal HDL-C levels [i.e., ≤38% of total cellular cholesterol per 2 hours (Table 3)]. Conversely, the cholesterol efflux was within the normal range in cells from two individuals in the high HDL-C group who had a missense mutation (Thr774→Pro774) in ABCA1. These findings are consistent with the notion that the excess of sequence variants in subjects with low HDL-C reflects an accumulation of damaging alleles in this group.

    The results of this study provide direct evidence that rare variants contribute significantly to low plasma levels of HDL-C, a common quantitative trait. To determine whether any of the common sequence variants identified in this study influenced plasma levels of HDL-C, we tested for associations between sequence variations with frequencies > 10% in at least one racial group and plasma levels of HDL-C in the Dallas Heart Study (table S1). Previously we showed that functionally significant sequence variations in APOE (14) and APOA5 (15) are associated with plasma lipid levels in all four major gender and race groups in this population. In contrast to those results, we found no sequence variation in ABCA1 that was systematically associated with plasma levels of HDL-C in men and women of both race groups (table S2). Two sequence variants were associated with plasma HDL-C levels in two groups (white men and black men), but the effect of these variants on HDL-C levels was modest. For example, plasma HDL-C concentrations were 52 ± 17 mg/dl in black men homozygous for the Met883 allele of ABCA1 and 48 ± 15 mg/dl in homozygotes for the common allele (Ile883) at this locus. These data are consistent with haplotype analyses of the ABCA1 gene (16) and with quantitative trait linkage mapping studies, which have not provided consistent support for common variants at any locus contributing to variation in plasma HDL-C levels. Nonetheless, our results do not exclude a role for common sequence variations in hypoalphalipoproteinemia. Because we screened primarily the coding sequences and flanking intronic regions of the three candidate genes, it is possible that common sequence variations in intronic or regulatory regions not sequenced in this study confer susceptibility to low plasma levels of HDL-C. Alternatively, specific haplotypes or combinations of alleles at these loci or common sequence variants at other loci may contribute to low plasma HDL-C levels.

    The strategy used in this study can be generalized to analyze the relationships between sequence variations in candidate genes and other quantitative traits in humans. Screening individuals with low or high levels of a trait increases the likelihood of detecting functionally significant sequence variations. Comparison of the numbers of sequence variants found in individuals with trait levels at opposite ends of the distribution provides a method to assess the significance of the variants identified, including rare variants that are unlikely to be detected by haplotype-based association studies. The method is unbiased, because the selection of individuals has its basis solely in their phenotype and has the added advantage of not requiring ascertainment of families.

    Supporting Online Material

    Materials and Methods

    Fig. S1

    Tables S1 and S2

    References and Notes

    View Abstract

    Navigate This Article