Introduction

Clinical familial hypercholesterolemia (FH) is characterized by LDL cholesterol levels of at least twice normal, with a bimodal inheritance pattern within affected families. In addition to dyslipidemia, affected individuals often present with tendon xanthomas and premature coronary disease. Mutations in either the LDL receptor gene (familial hypercholesterolemia, or FH) or the apolipoprotein B gene (familial defective apo B, or FDB) have been found to cause this disorder; however, these mutations explain only about 70% of all cases of the disease. A third locus (FH3) with a LOD score of 2.88 on chromosome 1p32 was initially reported by Varret et al. (1999), based on analysis of several French and Spanish FH families. Hunt et al. (2000), whose analysis of a large Utah FH pedigree identified a linkage at the same approximate location with a LOD score of 6.8, subsequently confirmed this linkage.

Recently, Abifadel et al. (2003) reported the results of mutation screening in the FH3 region. They identified two mutations in the gene PCSK9 (F216L and S127R), observed in three families with autosomal-dominant hypercholesterolemia. These mutations explained 12.5% of the hypercholesterolemia families examined in their study. Here we describe detection of another mutation in PCSK9 and propose that it is the causal mutation underlying our FH3 linkage.

Materials and methods

Pedigree ascertainment and description

Ascertainment of kindred 1173 (K1173) has been previously described by Haddad et al. (1999) and Hunt et al. (2000). The kindred was expanded to 95 individuals during the course of this study.

Genotyping

A genome search was performed on 31 members of K1173 as described by Hunt et al. (2000). Additional markers were genotyped on the expanded pedigree within a 17-cM region centered on the peak LOD score. The majority of markers are tri- or tetranucleotide repeats derived from public databases. Additional markers were developed by searching the public sequence databases for short tandem repeats within this region; these markers are indicated by the prefix MYR (sequence available upon request). The relative order of the markers within this interval was determined by use of the UCSC Human Genome Project Working Draft (http://genome.ucsc.edu; International Human Genome Sequencing Consortium 2001).

Linkage analysis

Linkage analysis was performed as described by Hunt et al. (2000).

Gene identification and assembly

Genes and EST clusters were identified using the UCSC Human Genome Project Working Draft (http://genome.ucsc.edu). Copy DNA sequences of incomplete genes were extended into their respective 5′UTRs by biotin capture 5′RACE (Tavtigian et al. 1996).

Mutation screening

Mutation screening was performed as described by Tavtigian et al. (1996).

Results

K1173 has been previously described in Haddad et al. (1999) and Hunt et al. (2000). During the course of this study K1173 has been expanded to a total of 95 individuals, including 32 affected individuals. All affected individuals had total cholesterol values at or above the 95th percentile for their age (Fig. 1a).

Fig. 1 a
figure 1

Pedigree drawing of K1173 showing phenotypes of family members. Only haplotype carriers and their spouses are shown. The shading of the symbols represent the following: black affected individuals, grey individuals of unknown phenotype, white unaffected individuals; * individual on statin treatment at the time of cholesterol measurement, and with historical total cholesterol levels >11.6 mmol/l, ** individuals with tendon xanthomas, statin treatment and historically high total cholesterol levels. Numbers under the symbols represent: top line individual identifier; middle line current age; bottom line highest recorded total cholesterol value (mmol/l). b Multipoint LOD scores for the linked region using a dominant model. Markers are listed in Table 1. Recombinant individuals in K1173 are indicated by the arrows below the LOD plot labeled with generation and individual number. Solid arrows depict haplotype sharing by affected individuals, dashed arrows indicate exclusion of the same region by unaffected family members. The dashed box outlines the minimal linked region as defined by recombinants from K1173. The location of PCSK9 is indicated by a vertical arrow

Full size image

Hunt et al. (2000) reported a genome search for K1173 performed using 583 genetic markers with an average spacing of 5.7 cM. Multipoint linkage analysis using a dominant model identified a highly significant LOD score (6.8) on chromosome 1p32. In this study we have typed a total of 52 markers spanning this 17-cM linked region on all sampled individuals from K1173. Results of multipoint linkage analysis performed using MCLINK (Thomas et al. 2000) with a dominant model are shown in Fig. 1b. This analysis produced an extremely significant LOD score of 9.6 with very well defined recombinant boundaries.

All 32 affected individuals in the pedigree shared a single STR-based haplotype from marker 1-MYR0194 to marker 1-MYR0213 (Fig. 1 and Table 1). No unaffected pedigree members inherited this portion of the segregating haplotype, suggesting that the mutation in this pedigree is 100% penetrant. A total of 13 recombinant individuals were identified, ten affected and three unaffected. All of the recombinants were internally consistent and were used to define a 7.5-Mbp region that was likely to contain the disease-causing gene (Fig. 1b). Table 1 lists the markers, the chromosomal location and the multipoint LOD score at each marker in K1173.

Table 1 Multipoint LOD scores, chromosomal location of markers, and shared haplotype found in K1173, 3317, 1103 and 602. Markers are listed in physical order; novel markers developed at Myriad for this study are indicated by the prefix MYR (sequence available upon request). Chromosomal location given is according to the UCSC genomic assembly, July 2003 freeze. Multipoint LOD scores for the dominant model are listed for each marker. The region where the haplotype is shared between multiple kindreds is shaded grey. The dashed box represents the minimal linked region as defined by recombinant individuals from K1173, as in Fig. 1. Allele frequencies (obtained by genotyping a large set of Utah families) are shown for specific alleles on the K1173 shared haplotype
Full size table

We attempted to expand K1173 further by searching for the K1173 shared STR-based haplotype in other Utah dyslipidemia pedigrees that had been genotyped during our original genome search. Three such kindreds were identified; however, none of the individuals sharing the haplotype in these kindreds were affected. Table 1 shows the region shared in these kindreds (K1103, K602, and K3317) compared with the haplotype carried by affected individuals in K1173.

All genes identified within the minimal recombinant region were mutation screened on haplotype carriers from K1173 and K3317. A single nucleotide exchange (G→T) was found on the disease haplotype in K1173 that was not present on the same haplotype in K3317, this was the only observed difference between the shared haplotype carriers from the two pedigrees. The T allele was not detected in K1106 or K602, and 338 control chromosomes were all found to be homozygous for the wild-type G (data not shown). The mutation results in a missense change of D374Y in the gene PCSK9.

PCSK9 is a member of the peptidase S8 family of subtilases (serine proteases). The subtilase superfamily of serine proteases can be grouped into six families (A–F) based on sequence homology (Siezen and Leunissen 1997). PCSK9 shows greatest homology with the C family of serine proteases, which includes proteinase K. Comparison of PCSK9 to other family members indicates that while D374 is not a highly conserved residue it is located within a region of conservation in this family, and the most common residue at this location is D or G.

Discussion

Expansion of K1173, along with increased marker density, has narrowed the linked region to approximately 7.5 Mbp in size. We were also able to identify three distantly related pedigrees (K1103, K602 and K3317) with individuals that share the same STR-based haplotype but who are not affected. The identification of these pedigrees, in combination with the observation that the mutation in K1173 is 100% penetrant, led us to hypothesize that the mutation in K1173 is a recent event. Therefore, K1103, K602 and K3317 served as controls for mutation screening of the genes within the region. The causal mutation could be expected to reside on the haplotype in K1173 but be absent from the same haplotype in K1103, K602 and K3317.

Mutation screening of the coding exons within the region identified a single variant, a missense mutation in the gene PCSK9 (D374Y), located on the disease-associated haplotype in K1173 that was absent from the same haplotype in the other three kindreds. PCSK9 is a recently discovered member of a family of mammalian secretory kexin-like subtilases (Seidah et al. 2003). Members of this family are involved in protein activation, inactivation, and regulation of cellular location. Another family member, SKI-1 is involved in processing sterol regulatory element binding proteins (SREBPs) and hence plays an important role in cholesterol homeostasis. In a recent study, Maxwell et al. (2003) demonstrated that PCSK9 expression is downregulated in mice fed a high cholesterol diet, and that its expression is regulated by SREBPs. The catalytic domain of these enzymes, which is primarily responsible for substrate specificity, includes a triad of active residues comprised of Asp, His and Ser.

The Grantham chemical difference matrix (Grantham 1974) is a formula for calculating chemical difference values for amino acid substitutions that takes into account composition, polarity, and molecular volume. The chemical difference score correlates with the likelihood of an amino acid substitution affecting protein functioning. Recent studies propose that the average deleterious mutation has a chemical difference score of 93; in contrast, the average neutral substitution has a score of 63 (Miller and Kumar 2001; Abkevich V, Zharkikh A, Deffenbaugh AM, Frank D, Chen Y, Shattuck D, Gutin A, Tavtigian SV, manuscript submitted). The three mutations observed to date in PCSK9 have chemical difference scores of 22 (F216L), 110 (S127R), and 160 (D374Y). The mutation in K1173 shows 100% penetrance with affected individuals being diagnosed as young as one year old with a severe clinical phenotype (average untreated total cholesterol value in affected individuals of 9 mmol/l in K1173). The high chemical difference score for the D374Y mutation is consistent with the severity and penetrance of the mutation observed in K1173.

The identification of a causal mutation in PCSK9 resulting in hypercholesterolemia in a Utah pedigree substantiates the previously published work of Abifadel et al. (2003). This gene underlies a novel mechanism contributing to dyslipidemia, and it will be important to identify PCSK9’s substrate(s) and the impact of the mutations on its functioning. In addition, screening for variation in other hypercholesterolemia families will determine what fraction of disease families can be explained by mutations in the PCSK9 locus.