In the present study, we have determined the relative frequency of the R46L, I474V and E670G variants in the PCSK9 (protein convertase subtilisin/kexin type 9) gene and its association with plasma lipid levels and CHD (coronary heart disease) in healthy U.K. men and patients with clinically defined definite FH (familial hypercholesterolaemia). Genotypes were determined using PCR and restriction enzyme digestion in 2444 healthy middle-aged (50–61 years) men from the prospective NPHSII (Second Northwick Park Heart Study), with 275 CHD events (15 years of follow-up), and in 597 U.K. FH patients from the Simon Broome Register. In the NPHSII healthy men, the R46L genotype distribution was in Hardy–Weinberg equilibrium and the frequency of 46L was 0.010 [95% CI (confidence interval), 0.007–0.013], with one man homozygous for the 46L allele. There was significant association of the 46L allele with lower mean (S.D.) total cholesterol [5.74 (1.01) mmol/l for RR compared with 5.26±1.03 mmol/l for RL; P=0.001], apolipoprotein B [0.87 (0.24) g/l for RR compared with 0.75 (0.26) g/l for RL; P<0.0001] and low-density lipoprotein cholesterol [4.01 (0.95) mmol/l for RR compared with 3.62 (0.97) mmol/l for RL; P=0.02]) levels, after adjustment for age, general medical practice, smoking, body mass index and systolic blood pressure. As expected, 46L carriers had a low risk of definite or possible CHD [hazard ratio, 0.46 (95% CI, 0.11–1.84)], but this was not statistically significant (P=0.27). Two other common PCSK9 variants I474V [V allele frequency, 0.179 (95% CI, 0.17–0.19)] and E670G [G allele frequency, 0.034 (CI, 0.03–0.04)] were not associated with any significant effects on lipid levels or CHD risk. In FH patients, the frequency of 46L was 0.003 (95% CI, 0.00–0.01), which was significantly lower (P=0.037) than the healthy subjects. In the four FH patients carrying 46L, mean untreated total cholesterol levels were not different (P=0.91) in carriers and non-carriers (median, 10.3 mmol/l compared with 10.2 mmol/l respectively, after adjustment for age, gender and mutation type). In conclusion, the PCSK9 46L allele is more frequent in healthy U.K. men than in FH patients and is strongly associated with a protective plasma lipid profile risk for CHD. Its low frequency (approx. 2% carriers) means that it does not make a major contribution to determining population CHD risk in the U.K.
- apolipoprotein B (ApoB)
- coronary heart disease
- cardiovascular risk
- familial hypercholesterolaemia
- low-density lipoprotein cholesterol (LDL-C)
- protein convertase subtilisin/kexin type 9 gene (PCSK9)
- total cholesterol
The PCSK9 (protein convertase subtilisin/kexin type 9) gene, codes for an enzyme which has been shown to be involved in degrading the LDL-R [LDL (low-density lipoprotein)-receptor] protein in the lysosome of the cell and prevents it recycling to the cell surface [1,2]. Gain-of-function mutations in PCSK9 that increase activity therefore cause increased degradation of LDL-Rs, reduce the numbers of receptors on the surface of the cell [1,3] and hence monogenic FH (familial hypercholesterolaemia) . However, loss-of-function mutations that inactivate PCSK9 result in less degradation of the LDL-R and thus higher levels on the cell surface . This results in faster clearance of LDL from the plasma and has been reported to be associated with lower LDL-C (LDL cholesterol) levels and lower risk of CHD (coronary heart disease) [6–8].
PCSK9 is a highly polymorphic gene [9,10], and one of the DNA changes, 137G>T in exon 1, results in the replacement of Arg46 by leucine (R46L). The reported prevalence for 46L heterozygosity was 0.7% in black subjects (n=3363) and 3.2% in European Americans (n=9524), eight of whom were homozygous for the 46L allele . Carriage of the 46L allele was associated with significantly lower plasma levels of total cholesterol and LDL-C (9 and 15% respectively), and a 47% reduction in the rate of coronary events in those carrying the 46L allele. These results have been confirmed by others . A recent study from Norway has confirmed the association of PCSK9 R46L with hypocholesterolaemia and suggested that, in FH subjects, 46L carriers also had a better lipid-lowering response to statin therapy .
The aims of the present study were to determine the relative frequency of the PCSK9 46L allele and its association with plasma lipid levels and CHD in healthy U.K. men [12,13] and patients with clinically defined definite FH, of whom 35% had evidence of premature CHD . We also determined the frequency and examined the impact on lipid levels of two other reported common PCSK9 variants, I474V and E670G. Previous studies from Japan  and the U.S.A.  have also suggested that carriers of the rare alleles of these variants and haplotypes containing them have lower and higher plasma lipid levels respectively.
MATERIALS AND METHODS
NPHSII (Second Northwick Park Heart Study) recruited 3012 healthy European Caucasian men aged 50–61 years from nine general medical practices in the U.K. and has followed them for the development of CHD from 1989 [12,13]. In brief, at baseline, all of the participants were free of a history of unstable angina, MI (myocardial infarction), coronary surgery, anticoagulant therapy (including aspirin), malignancy and cardiovascular disease or any other condition precluding informed consent. Ethical approval was granted by the Local Institutional Review Committee, and patients gave written informed consent. Participants attended for examination in a non-fasting state, having avoided smoking, vigorous exercise or heavy meals from midnight the day before. A resting 12-lead ECG was recorded at baseline and during the sixth year of surveillance, with results coded by Minnesota criteria . Weight was measured on a balance scale and height on a stadiometer to calculate BMI (body mass index; kg/m2). Smoking habits were assessed by questionnaires  completed at baseline and annually thereafter for 5 years. BP (blood pressure) was recorded with a random zero mercury sphygmomanometer after the subject had been seated for 5 min. Details of possible events were obtained through primary care medical practices, hospitals and coroners' offices. Clinical history, ECGs, cardiac enzymes and pathology were assessed by independent review to diagnose MI according to World Health Organization criteria . Normal limits for cardiac enzymes were those for the reporting laboratory. Other terminating events were a new major Q wave on the ECG after 5 years of follow-up (Minnesota codes 11, 12.1 to 12.7 and 12.8 plus 51 or 52)  and surgery for angina pectoris with CHD angiographically demonstrated. In the 2444 men with genotyping, there had been 275 terminating events comprising 161 acute CHD events, 67 coronary artery revascularization procedures, 22 silent MIs and 25 possible MIs.
Patients with clinically defined definite FH  were recruited for the Simon Broome British Heart Foundation study, which is a cross-sectional comparison of white Caucasian patients aged 18 years or more with treated heterozygous FH with and without clinically documented CHD [14,18]. Recruitment methods, inclusion and exclusion and diagnostic criteria were as defined previously [14,18]. Pretreatment cholesterol levels were obtained from patient records.
Molecular genetic analysis
Genomic DNA was isolated from whole blood samples using standard methods [19,20]. PCR primers were designed to introduce a recognition site for the restriction enzyme RsaI. Genotyping for the PCSK9 exon 1 c.137G>T (R46L) mutation was carried out using primers and conditions as follows. The primers used were 5′-CACGGCCTCTAGGTCTCCTC-3′ (forward) and 5′-AGGCCGTCCTCCTCGGTA-3′ (reverse), where the forced base is underlined. The 20 μl PCR reaction mixture contained 16.6 mmol/l ammonium sulfate, 67 mmol/l Tris/HCl (pH 8.3), 6.7 mmol/l MgCl2, 6.7 μmol/l EDTA, 10 mmol/l 2-mercaptoethanol, 0.17 mg/ml BSA, 10% (v/v) DMSO, 0.2 mmol/l dATP, dGTP, dTTP and dCTP, 8 pmol of each primer, 0.2 unit of Taq polymerase (Invitrogen) and 15 ng of genomic DNA. The PCR protocol consisted of 94 °C for 3 min, 94 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s for 34 cycles, 72 °C for 5 min and 16 °C for 3 min. Restriction digest was carried out with RsaI (New England Biolabs) for 4 h at 37 °C using 3 units of enzyme/digest. GG (wild type) homozygotes produced band sizes of 188 and 17 bp, whereas heterozygotes (GT) produced band sizes of 205, 188 and 17 bp. The TT (homozygote) produced one band size of 205 bp. Products were separated on a 2.0% (w/v) agarose gel and were visualized by staining with ethidium bromide. For R46L, all gels included a positive heterozygous control sample (confirmed by direct sequencing), and genotypes were read by two independent observers blinded to case/control status; any discrepancies were confirmed by PCR and digestion.
Genotyping for PCSK9 exon 9 c.1420A>G (I474V) and exon 12 c.2009A>G (E670G) variants (rs562556 and rs505151 respectively) was carried out using PCR and RFLP (restriction-fragment-length polymorphism) conditions as follows. The primers used for amplifying exon 9 were 5′-TCCCTTCTCCCTTGTCTGTG-3′ (forward) and 5′-CTGTGGCTCTCTCCAGCAG-3′ (reverse) and for exon 12 were 5′-GATGTCGGAGGGAGAAATGA-3′ (forward) and 5′-GGCACCCAGAGTGAGTGAGT-3′ (reverse). The 20 μl PCR reaction mixture contained 16 mmol/l ammonium sulfate, 67 mmol/l Tris/HCl (pH 8.3), 1.5 mmol/l MgCl2, 0.01% Tween 20, 0.2 mmol/l dATP, dGTP, dTTP and dCTP, 0.032 pmol of each primer (Invitrogen), 0.3 unit of Taq DNA Polymerase (Bioline) and 15 ng of genomic DNA. The PCR cycling conditions consisted of 95 °C for 5 min, 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s for 35 cycles, and 72 °C for 5 min. Restriction digest for the I474V variant was carried out using 4 units of BccI (New England Biolabs)/sample, and restriction digest for the E670G variant was carried out using 2 units of Sau96I (New England Biolabs)/sample, both overnight at 37 °C. The A allele of the BccI digest produced band sizes of 267, 193, 96 and 25 bp, and the G allele produced band sizes of 267, 218 and 96 bp. The A allele of the Sau96I digest produced band sizes of 287, 69, 51, 33, 30, 18 and 6 bp, and the G allele produced band sizes of 215, 72, 69, 51, 33, 30, 18 and 6 bp. Products were separated on a 7.5% (w/v) polyacrylamide gel stained with ethidium bromide using the MADGE (microarray diagonal gel electrophoresis) system .
All statistical analyses were carried out using STATA (Intercooled Stata 9.2). Differences in proportions were determined by a χ2 test or Fisher's exact test. Lipid levels were compared by genotype after adjustment using analysis of covariance. R2 values were used to estimate the percentage of the sample variance explained by each genotype. CHD risk was assessed by Cox proportional hazards models [which allowed for varying lengths of follow-up, producing HRs (hazard ratios) with 95% CIs (confidence intervals)]. For each variant, the common homozygotes served as the reference category. Adjustments for possible confounding factors [age, BMI, smoking, cholesterol, triacylglycerol and SBP (systolic BP)] were made by including them as covariates in the model, and differences in the baseline hazard by general practice (recruitment site) were permitted (using the strata option in STATA). Haplotype analysis was conducted using the THESIAS (Testing Haplotype Effects In Association Studies) program. Pretreatment cholesterol levels were not available for all FH patients. Throughout, a P value of <0.05 was taken as statistically significant.
Healthy NPHSII men
The R46L genotype distribution was in Hardy–Weinberg equilibrium, and there was one healthy man homozygous for the 46L allele, the rare allele frequency was 0.010 (95% CI, 0.007–0.013) (Table 1). Table 2(a) shows the baseline characteristics of lipids levels of NPHSII men by PCSK9 R46L genotype. There was significant association of the 46L allele with lower mean (S.D.) total cholesterol [5.74 (1.01) mmol/l for RR compared with 5.26 (1.03) mmol/l for RL; P=0.001], ApoB (apolipoprotein B) [0.87 (0.24) g/l for RR compared with 0.75 (0.26) g/l for RL; P<0.0001] and LDL-C [4.01 (0.95) mmol/l for RR compared with 3.62 (0.97) mmol/l for RL; P=0.02] levels in NPHSII men, after adjustment for age, general medical practice, smoking, BMI and SBP. The 46L carriers had a low risk of definite or possible CHD, with a HR of 0.46 (95% CI, 0.11–1.84), but this was not statistically significant (P=0.27).
The majority of the men had lipid levels available at baseline and for five annual visits, and these are shown in Figure 1. The mean levels of the 46L carriers were approx. 0.5 mmol/l lower at baseline and in the one LL man this was 2.3 mmol/l lower. These levels were maintained over the next 5 years. The 46L carriers were at the 33rd percentile of the total cholesterol distribution of the whole sample, whereas the LL man was at the 0.4th percentile. Over the 5 year period, there was a progressive drop-out of subjects with recorded lipid levels (ten 46L carriers and 659 non-carriers). During this time, there was a small, but non-significant, 3.8% increase in the mean levels in 46L carriers and a statistically significant 3% decrease in levels in non-carriers, the result of which was that the difference between the two groups lost statistical significance by year 4.
To put the impact of this variant in context, the proportion of sample variance for each of the lipid traits was determined and compared with the impact of the well known common APOE (apolipoprotein E) gene e2, e3 and e4 alleles. Although APOE explained 5.1% of the variance in LDL-C, 3.5% for ApoB and 2.9% for total cholesterol in this sample, the PCSK9 R46L genotype explained 0.6, 0.9 and 0.6% respectively. There was no evidence for an interaction with APOE genotype in NPHSII men, with effects on cholesterol levels of variation in the two genes not being different from additive (P interaction=0.39).
The genotype was also determined for two other common variants in PCSK9, the I474V and E670G variants. For both, genotype distribution was in Hardy–Weinberg equilibrium and, as shown in Table 1, in healthy men the 474V frequency was 0.18 (95% CI, 0.17–0.19) and the 670G frequency was 0.03 (95% CI, 0.028–0.039). Tables 2(b) and 2(c) show that there was no significant association with any of the measured plasma lipid traits. There was also no significant association with CHD risk [for I474V V allele carriers compared with II men, HR=1.16 (95% CI, 0.89–1.50); for E670G G allele carriers compared with EE men, HR=1.20 (95% CI, 0.73–1.98)]. These estimates were not substantially altered by adjustment for age, general medical practice, smoking, BMI and SBP and remained non-significant (results not shown). When results from these two variants was combined into haplotypes, three common haplotypes were seen (V474/G670 frequency, 0.79; V474/E670 frequency, 0.03; V474/E670 frequency, 0.18) with significant evidence for linkage disequilibrium (D′=0.93). There were no significant associations between any haplotypes and plasma levels of total cholesterol or LDL-C, or with CHD risk (results not shown).
The genotype was only determined for the PCSK9 R46L variant, and was obtained in 597 FH patients. The genotype distribution was in Hardy–Weinberg equilibrium, and the 46L allele frequency was 0.03 (95% CI, 0.00–0.01), which was significantly lower than in the NPHSII healthy subjects (P=0.037). In the four FH patients carrying 46L, whose characteristics are shown in Table 3, there was no difference in mean untreated total cholesterol levels compared with the FH patient group overall [median, 10.3 compared with 10.2 mmol/l respectively (P=0.91); geometric mean, 10.4 (1.4) compared with 10.3 (1.8) mmol/l respectively (P=0.94), after adjustment for age, gender and mutation type]. None of the 46L carriers had evidence of premature CHD compared with the reported  prevalence in the sample as a whole (38.5%) or in those with a detected mutation causing FH  (38.6%), although these differences were not statistically significant (all P>0.2). Because of low numbers, no analysis of response to lipid-lowering therapy by R46L genotype was attempted.
The present study strongly confirms earlier reports  that, in healthy middle-aged men, carriers of the 46L allele in PCSK9 have lower plasma levels of total cholesterol, LDL-C and ApoB, and demonstrate that this effect is maintained over at least 5 years of follow-up. The one individual who was homozygous for this variant had LDL-C levels that put him at the 0.5th percentile of the sample of healthy men from the general U.K. population. In this group of healthy men, the 46L variant was not associated with a significant effect on any other measured lipid trait. As expected, carriers also had a 54% lower risk of early CHD, which, although not statistically significant in this sample, is of similar magnitude to the 47% lower risk reported previously . On the basis of results from the effect on CHD risk reduction seen with lipid-lowering therapy , the observed decrease in LDL-C would be predicted to result in only a 10% reduction in risk. This supports the view  that the life-long lower LDL-C experienced by 46L carriers has a greater than expected CHD benefit than lowering in middle-age. In support of this, none of the FH patients who were carrying the 46L variant had evidence of CHD. It is well known that the age of onset of CHD varies widely in FH patients [23,24] and this suggests that variants in the PCSK9 gene may be explaining in part this variability.
The possible mechanism of the LDL-C-lowering effect of 46L has been discussed elsewhere [1,5]. The enzyme encoded by PCSK9 is involved in degrading the LDL-R protein in the lysosome of the cell and preventing it recycling. Loss-of-function mutations in the PCSK9 gene result in less degradation of LDL-Rs, increased numbers of receptors on the surface of the cell and hypocholesterolaemia . There are two possible sites for this effect. The first is in a post-ER (endoplasmic reticulum) compartment, such as the Golgi apparatus, where PCSK9 binds to the LDL-R and targets it for degradation in the acidic lysosome. A second possibility is that secreted PCSK9 binds to surface-located LDL-R and, after internalization, prevents LDL-R recycling from the endosome to the surface and re-routes it to the lysosome. Currently, it is still unclear whether the LDL-R is itself cleaved directly by PCSK9. A definite proof of the effect of 46L in vitro has not yet been reported, but a direct effect of the substitution on an aspect of PCSK9 function seems highly likely. The R46L substitution has no measurable effect on the synthesis, processing or secretion of PCSK9 in transfected HEK-293 cells (human embryonic kidney cells) , but modelling indicates that Arg46 is in a N-terminal extension of the prodomain, which may make intramolecular contacts with the catalytical C-domain after autocatalytical cleavage. It has been suggested  that the substitution of a hydrophobic leucine residue for a polar arginine residue may alter this association and influence function in this way. The amino acid at position 46 is hydrophobic in most species .
However, although the effect seen in the NPHSII healthy men is statistically robust and is maintained over time, because of the relatively low frequency of 46L carriers the overall impact on LDL-C levels in the population is low, explaining approx. 12% of the effect of the well-known APOE gene common variants. Although the 46L-lowering effect on LDL-C of approx. 0.5 mmol/l is similar to that seen in APOE e2 carriers compared with APOE e3e3 subjects , the population level impact of APOE is greater because the variants are much more common (e2 carriers occur in approx. 12% of the U.K. population).
The two other PCSK9 variants examined in the present study, I474V and E670G, were not associated with any significant impact on lipid levels in the NPHSII healthy men. The 474V variant was first reported to be present at low frequency (0.05) and associated with approx. 7% lower LDL-C levels in carriers in a Japanese study , but, although the frequency was higher (0.18–0.20) in whites and blacks from the U.S.A. , no association with lipid levels were detected. The frequency in U.K. men was similar to that reported in the U.S.A., and also no effect on lipid levels was detected. For the E670G variant, in the original study in moderately hyperlipidaemic subjects from the U.S.A., homozygosity for the 670G allele was associated with approx. 20% higher LDL-C levels than EE subjects , an effect which again was not observed in either whites or blacks in the study by Kotowsky et al. . In the U.K. men examined in the present study, LDL-C levels in 670G homozygotes tended to be lower, although not statistically significant. The present study has the power (80%) to be able to detect a 0.14 mmol/l difference in LDL-C levels associated with carriage of the 474V variant and 0.6 mmol/l difference for homozygosity of the 670G variant. The original report suggested a larger effect with a PCSK9 haplotype , but no such effect was seen in the present study (results not shown). We were thus able to confirm the reports  of the negligible (or at the very best very modest) impact of these variants in healthy subjects and therefore did not examine their effect in FH patients.
For the FH patients examined, although the low frequency of 46L limits our ability to make strong interpretations, the lower frequency of the 46L variant and the possible protection from early CHD suggests that PCSK9 is one of the genes that is involved in modulating the expression of FH. Thus individuals with FH (commonly due to a mutation in LDLR or APOB) who also are carrying 46L may be less likely to be diagnosed with FH, resulting in the lower than expected allele frequency. Since all the patients examined had a diagnosis of definite FH (i.e. a personal or family history of tendon xanthoma ), it may be that the probable lower LDL-C levels in patients carrying 46L reduces the likelihood of xanthoma development. One of the patients (SB 2825, with an untreated total cholesterol level 8.6 mmol/l) has an affected father with the same LDLR mutation (age, 54 years) who did not carry the 46L variant , and his recorded untreated total cholesterol level was 13.4 mmol/l (results not shown), compatible with a lowering effect of the PCSK9 variant. Confirmation of these observation in larger patient cohorts, and particularly studies in the relatives of these patients, to determine their plasma lipid levels, will be informative.
We thank the members of the Simon Broome BHF (British Heart Foundation) study for access to patients samples, Dr Rossi Naoumova, Professor Gil Thompson, Dr Mary Seed, Professor Paul Durrington, Dr Paul Miller and Professor John Betteridge. The study was supported by a grant from the BHF (grant RG93008). S. E. H. and R. A. W. acknowledge BHF support (RG 2005/014), and a grant from the Department of Health to the London IDEAS Genetics Knowledge Park. C. H. is supported by the Pinto Foundation. M. S. is supported by a grant from The National Council for Scientific and Technological Development (CNPq) of the Ministry of Science and Technology (MCT), Brazil.
Abbreviations: ApoB, apolipoprotein B; ApoE, apolipoprotein E; BMI, body mass index; BP, blood pressure; CHD, coronary heart disease; CI, confidence interval; FH, familial hypercholesterolaemia; HR, hazard ratio; LDL, low-density lipoprotein; LDL-C, LDL cholesterol; LDL-R, LDL-receptor; MI, myocardial infarction; NPHSII, Second Northwick Park Heart Study; PCSK9, protein convertase subtilisin/kexin type 9; SBP, systolic BP
- © The Authors Journal compilation © 2007 Biochemical Society