Editorial Comment doi: 10.1111/joim.12207
Genetic determination of lipoprotein(a) and its association with cardiovascular disease: convenient does not always mean better The ups and downs in lipoprotein(a) research Lipoprotein(a) [Lp(a)] was discovered as early as 1963; however, it was not until the 1980s that the association between Lp(a) and cardiovascular disease (CVD) became evident. The first prospective studies of the relationship followed in the 1990s, but a few negative results at that time almost prevented further investigation of Lp(a). However, the following discoveries from initial genetic studies led to renewed interest (for review, see 1). 1 Copy number variation in the LPA gene with considerable size variability: the basis of this copy number variation is a so-called kringle-IV (KIV) repeat, varying in number from 11 to more than 50 [2–4]. 2 An inverse association between the apolipoprotein(a) [apo(a)] isoform size (number of KIV repeats) and serum Lp(a) concentration [2]. 3 An association between apo(a) isoform size and the prevalence or incidence of CVD events. In particular, this finding strongly supported the notion of causality between high Lp(a) concentrations and CVD events and was demonstrated by the first application of the now widely used Mendelian randomization study approach. In general, individuals with so-called small apo(a) isoforms that are determined at conception have on average markedly higher Lp(a) concentrations and a significantly higher risk of events [5]. Despite these convincing findings, there remained a lack of interest in Lp(a) and the apo(a) isoforms at least for some time for two main reasons. First, there were almost no drugs available that were known to decrease Lp(a) concentrations. Secondly, the apo(a) isoform size was difficult to measure in the laboratory, requiring gel electrophoresis followed by immunoblotting. These methods are complex and laborious and thus do not allow a
high-throughput approach as required for clinical routine. Meanwhile, evidence for the atherogeneity of Lp(a) became stronger with many prospective studies in the field of CVD measuring Lp(a) concentrations, which enabled the Emerging Risk Factor Collaboration to perform a meta-analysis with 126 634 study participants. The risk ratio for coronary heart disease (CHD) was reported to be 1.16 [95% confidence interval (CI) 1.11–1.22] per standard deviation increase in Lp(a) [6]. Another metaanalysis including studies of apo(a) size polymorphism revealed a combined relative risk of 2.08 (95% CI 1.67–2.58) for individuals with small versus large apo(a) isoforms (i.e. 11–22 vs. >22 KIV repeats) [7]. At the same time, Clarke et al. [8] published a study including CHD cases and controls, applying a gene chip with almost 50 000 SNPs in candidate genes and the subsequent replication of the most interesting SNPs in three independent populations. The authors identified the variant rs10455872 at the LPA locus with a minor allele frequency of 7% and an odds ratio for CHD of 1.70 (95% CI 1.49–1.95). Another SNP (rs3798220) showed an even higher odds ratio of 1.92 (95% CI 1.48–2.49), but had a lower minor allele frequency of 2%. Both variants were strongly associated with increased Lp(a) concentrations and small apo(a) isoform size [8]. These findings increased interest in using the SNPs for diagnostic purposes and risk prediction, and they are currently used as easy-to-measure surrogates for apo(a) isoforms in many studies. However, there are some concerns as I will discuss further below. Increasingly, the number of drugs that lower Lp(a) concentrations became a focus for interventional trials. For most of these agents, Lp(a) is not the main target but a ‘collateral win’ [9]. These drugs target classical lipid metabolism for the development of medications for patients with insufficient improvement in lipid profile on statin therapy. Therapeutic strategies include antisense oligonucleotides
ª 2014 The Association for the Publication of the Journal of Internal Medicine
1
F. Kronenberg
targeting apolipoprotein B mRNA such as mipomersen [10], inhibitors of proprotein convertase subtilisin/kexin type 9 (PSCK9) [11] or inhibitors of the cholesteryl ester transfer protein CETP such as anacetrapib [12]. These drugs were all shown to decrease Lp(a) levels by up to 30–40% in addition to the changes in their main targets of LDL or HDL cholesterol. This effect on several targets makes it difficult, if not impossible, to determine whether a lowering of Lp(a) indeed contributes to a lowering of CHD risk. The same holds true for lipid apheresis, which, in addition to LDL cholesterol, lowers Lp(a) concentrations and other factors associated with CVD [13, 14]. An exception might be a recent report on a very specific Lp(a)-lowering apheresis system that resulted also in promising changes in the coronary arteries [15]. Thus, as a result of genetic and interventional studies, Lp(a) became more attractive than ever not only for risk prediction but also as a target for intervention. The study by Hopewell et al In this issue of the Journal of Internal Medicine, Hopewell et al. [16] investigated plasma Lp(a) levels and apo(a) isoform size in 995 CHD cases and 998 controls. The authors used the laborious method of immunoblotting to determine apo(a) isoform size. Individuals in the top fifth of Lp(a) levels had more than a twofold higher risk of CHD compared with those in the bottom fifth, and this association was unaltered after adjustment for apo(a) KIV repeats. Individuals in the bottom fifth of KIV repeats (corresponding to approximately ≤22 KIV repeats) had about a twofold higher risk of CHD compared with those in the top fifth; however, this association was no longer significant after adjustment for Lp(a) levels. The authors concluded that the effect of KIV repeats on CHD risk is mediated through their impact on Lp(a) levels, suggesting that absolute level of Lp(a), rather than apo(a) isoform size, is the chief determinant of CHD risk. The main conclusion of the study by Hopewell et al. is not new. However, the study includes a considerable number of patients and controls. It is a single study in which all measurements were taken in one laboratory, thus avoiding the many disadvantages of meta-analyses and so probably providing more accurate results. Studies performed in a similar way included the cross-sectional study by Sandholzer et al. [5] in six ethnic populations and 2
ª 2014 The Association for the Publication of the Journal of Internal Medicine Journal of Internal Medicine
Editorial: Apo(a) isoforms and CVD
the prospective Bruneck study in a general population sample [17]. What is the advantage of apo(a) isoform measurement compared with SNP analysis? The advantages of SNP analysis clearly include ease of use in the routine laboratory setting, suitability as a high-throughput method and simple interpretation of the data. Ideally, the SNP or a panel of SNPs should provide at least the same information as the alternative methods. However, this is not the case for the two most commonly used SNPs that show the strongest association between Lp(a) concentrations and CVD events (rs10455872 and rs3798220). These two SNPs have a minor allele frequency of 7% and 2%, respectively, and about 15% of the population carry at least one minor allele of the two SNPs. It was claimed that they tag carriers of small apo(a) isoforms. However, Fig. 1 shows the results from almost 3000 individuals in a general Caucasian population who were genotyped for these two SNPs and phenotyped for the apo(a) isoforms. These data clearly show that 47% of all subjects expressing a small apo(a) isoform were not tagged by one of these two highrisk SNPs. Furthermore, about 11% of all subjects carrying at least one minor allele of these two SNPs did not actually have a small apo(a) isoform. This means that these two SNPs are far from being a
rs10455872 or rs3798220 Non-carrier 98%
rs10455872 or rs3798220 Apo(a) phenotypes
2% Carrier
Non-carrier 47%
53% Carrier Corresponds to 11% of all carriers of rs10455872 or rs3798220
Fig. 1 Apolipoprotein(a) [apo(a)] phenotypes and the carrier and noncarrier status of the two single nucleotide polymorphisms rs10455872 and rs3798220 derived from a large Caucasian general population sample of almost 3000 individuals.
F. Kronenberg
good surrogate for risk evaluation, instead of measurement of small apo(a) isoforms, because approximately half of the small apo(a) isoforms will remain undetected when only these two SNPs are genotyped. From a population perspective, we might see the iceberg, but we might underestimate the problem. From an individual’s perspective, the false-negative rate might simply be too high when these two SNPs are used for risk prediction. Whether an extended panel of SNPs is able to tag a markedly larger fraction of small apo(a) isoforms remains unclear at present. If Lp(a) from small apo(a) isoforms has indeed a higher atherogenic potential as suggested earlier [17], it might be hypothesized that the different molecular weight of small and large apo(a) proteins is more likely to be responsible for this increased risk than a simple SNP that does not necessarily result in a structural change of the apo(a) protein. It can be argued that the apo(a) isoform should be a better predictor than the SNPs rs10455872 and rs3798220. The two SNPs demonstrated odds ratios for CHD of 1.70 and 1.92, respectively, which are not much lower than the odds ratio for small apo(a) isoforms of about 2.0. However, it is noteworthy that only about 15% of the population carry at least one minor allele of the two SNPs, whereas 25–35% of the population carry small apo(a) isoforms. Therefore, only a fraction of the high-risk carriers might be correctly classified with the two SNPs, compared to use of the apo(a) isoforms. An alternative approach is the estimation of KIV repeats by quantitative PCR, which measures the total number of KIV repeats of the two apo(a) alleles [18]. The consequence is that individuals with one very short KIV repeat allele and one very large allele would be placed in the same category as individuals with two intermediate copy number alleles. These two situations are, however, associated with very distinct Lp(a) concentrations. Despite this limitation, previous studies have shown a significant association between the sum of KIV repeats and CVD risk [18], which might have underestimated the strength of the association. Finally, and to make the situation even more complex, not all apo(a) isoforms coded in the DNA are indeed expressed in plasma. In about half to one-third of individuals, only one isoform is expressed in plasma although two different alleles are present at the DNA level [19]. With regard to risk prediction, it might be questioned why one
Editorial: Apo(a) isoforms and CVD
should be interested in an allele that is not expressed at all in plasma when the site of pathogenic action of Lp(a) is expected to be in the vascular system. Situations in which the apo(a) isoform might be more useful than currently known SNPs Usually, Lp(a) concentrations seem to be quite stable throughout life because of fairly strong genetic control [1]. However, certain diseases result in pronounced changes in the concentration of Lp(a). For example, chronic kidney disease results in an increase in Lp(a) that depends on the type of disease (whether or not there is a nephrotic component) and on the type of treatment (peritoneal dialysis is associated with higher concentrations than haemodialysis, with normalized concentrations after successful kidney transplantation). It is interesting that in patients with end-stage renal disease treated with haemodialysis, Lp(a) is increased only in those with large apo(a) isoforms compared with isoform-specific control subjects. This might explain why the apo(a) isoform is more predictive of CVD risk than Lp(a) concentrations in these patients (for review see 1, 20). Another example is type 2 diabetes mellitus (T2DM). Recent studies have described an unexpected association between very low Lp(a) concentrations (
Genetic determination of lipoprotein(a) and its association with cardiovascular disease: convenient does not always mean better The ups and downs in lipoprotein(a) research Lipoprotein(a) [Lp(a)] was discovered as early as 1963; however, it was not until the 1980s that the association between Lp(a) and cardiovascular disease (CVD) became evident. The first prospective studies of the relationship followed in the 1990s, but a few negative results at that time almost prevented further investigation of Lp(a). However, the following discoveries from initial genetic studies led to renewed interest (for review, see 1). 1 Copy number variation in the LPA gene with considerable size variability: the basis of this copy number variation is a so-called kringle-IV (KIV) repeat, varying in number from 11 to more than 50 [2–4]. 2 An inverse association between the apolipoprotein(a) [apo(a)] isoform size (number of KIV repeats) and serum Lp(a) concentration [2]. 3 An association between apo(a) isoform size and the prevalence or incidence of CVD events. In particular, this finding strongly supported the notion of causality between high Lp(a) concentrations and CVD events and was demonstrated by the first application of the now widely used Mendelian randomization study approach. In general, individuals with so-called small apo(a) isoforms that are determined at conception have on average markedly higher Lp(a) concentrations and a significantly higher risk of events [5]. Despite these convincing findings, there remained a lack of interest in Lp(a) and the apo(a) isoforms at least for some time for two main reasons. First, there were almost no drugs available that were known to decrease Lp(a) concentrations. Secondly, the apo(a) isoform size was difficult to measure in the laboratory, requiring gel electrophoresis followed by immunoblotting. These methods are complex and laborious and thus do not allow a
high-throughput approach as required for clinical routine. Meanwhile, evidence for the atherogeneity of Lp(a) became stronger with many prospective studies in the field of CVD measuring Lp(a) concentrations, which enabled the Emerging Risk Factor Collaboration to perform a meta-analysis with 126 634 study participants. The risk ratio for coronary heart disease (CHD) was reported to be 1.16 [95% confidence interval (CI) 1.11–1.22] per standard deviation increase in Lp(a) [6]. Another metaanalysis including studies of apo(a) size polymorphism revealed a combined relative risk of 2.08 (95% CI 1.67–2.58) for individuals with small versus large apo(a) isoforms (i.e. 11–22 vs. >22 KIV repeats) [7]. At the same time, Clarke et al. [8] published a study including CHD cases and controls, applying a gene chip with almost 50 000 SNPs in candidate genes and the subsequent replication of the most interesting SNPs in three independent populations. The authors identified the variant rs10455872 at the LPA locus with a minor allele frequency of 7% and an odds ratio for CHD of 1.70 (95% CI 1.49–1.95). Another SNP (rs3798220) showed an even higher odds ratio of 1.92 (95% CI 1.48–2.49), but had a lower minor allele frequency of 2%. Both variants were strongly associated with increased Lp(a) concentrations and small apo(a) isoform size [8]. These findings increased interest in using the SNPs for diagnostic purposes and risk prediction, and they are currently used as easy-to-measure surrogates for apo(a) isoforms in many studies. However, there are some concerns as I will discuss further below. Increasingly, the number of drugs that lower Lp(a) concentrations became a focus for interventional trials. For most of these agents, Lp(a) is not the main target but a ‘collateral win’ [9]. These drugs target classical lipid metabolism for the development of medications for patients with insufficient improvement in lipid profile on statin therapy. Therapeutic strategies include antisense oligonucleotides
ª 2014 The Association for the Publication of the Journal of Internal Medicine
1
F. Kronenberg
targeting apolipoprotein B mRNA such as mipomersen [10], inhibitors of proprotein convertase subtilisin/kexin type 9 (PSCK9) [11] or inhibitors of the cholesteryl ester transfer protein CETP such as anacetrapib [12]. These drugs were all shown to decrease Lp(a) levels by up to 30–40% in addition to the changes in their main targets of LDL or HDL cholesterol. This effect on several targets makes it difficult, if not impossible, to determine whether a lowering of Lp(a) indeed contributes to a lowering of CHD risk. The same holds true for lipid apheresis, which, in addition to LDL cholesterol, lowers Lp(a) concentrations and other factors associated with CVD [13, 14]. An exception might be a recent report on a very specific Lp(a)-lowering apheresis system that resulted also in promising changes in the coronary arteries [15]. Thus, as a result of genetic and interventional studies, Lp(a) became more attractive than ever not only for risk prediction but also as a target for intervention. The study by Hopewell et al In this issue of the Journal of Internal Medicine, Hopewell et al. [16] investigated plasma Lp(a) levels and apo(a) isoform size in 995 CHD cases and 998 controls. The authors used the laborious method of immunoblotting to determine apo(a) isoform size. Individuals in the top fifth of Lp(a) levels had more than a twofold higher risk of CHD compared with those in the bottom fifth, and this association was unaltered after adjustment for apo(a) KIV repeats. Individuals in the bottom fifth of KIV repeats (corresponding to approximately ≤22 KIV repeats) had about a twofold higher risk of CHD compared with those in the top fifth; however, this association was no longer significant after adjustment for Lp(a) levels. The authors concluded that the effect of KIV repeats on CHD risk is mediated through their impact on Lp(a) levels, suggesting that absolute level of Lp(a), rather than apo(a) isoform size, is the chief determinant of CHD risk. The main conclusion of the study by Hopewell et al. is not new. However, the study includes a considerable number of patients and controls. It is a single study in which all measurements were taken in one laboratory, thus avoiding the many disadvantages of meta-analyses and so probably providing more accurate results. Studies performed in a similar way included the cross-sectional study by Sandholzer et al. [5] in six ethnic populations and 2
ª 2014 The Association for the Publication of the Journal of Internal Medicine Journal of Internal Medicine
Editorial: Apo(a) isoforms and CVD
the prospective Bruneck study in a general population sample [17]. What is the advantage of apo(a) isoform measurement compared with SNP analysis? The advantages of SNP analysis clearly include ease of use in the routine laboratory setting, suitability as a high-throughput method and simple interpretation of the data. Ideally, the SNP or a panel of SNPs should provide at least the same information as the alternative methods. However, this is not the case for the two most commonly used SNPs that show the strongest association between Lp(a) concentrations and CVD events (rs10455872 and rs3798220). These two SNPs have a minor allele frequency of 7% and 2%, respectively, and about 15% of the population carry at least one minor allele of the two SNPs. It was claimed that they tag carriers of small apo(a) isoforms. However, Fig. 1 shows the results from almost 3000 individuals in a general Caucasian population who were genotyped for these two SNPs and phenotyped for the apo(a) isoforms. These data clearly show that 47% of all subjects expressing a small apo(a) isoform were not tagged by one of these two highrisk SNPs. Furthermore, about 11% of all subjects carrying at least one minor allele of these two SNPs did not actually have a small apo(a) isoform. This means that these two SNPs are far from being a
rs10455872 or rs3798220 Non-carrier 98%
rs10455872 or rs3798220 Apo(a) phenotypes
2% Carrier
Non-carrier 47%
53% Carrier Corresponds to 11% of all carriers of rs10455872 or rs3798220
Fig. 1 Apolipoprotein(a) [apo(a)] phenotypes and the carrier and noncarrier status of the two single nucleotide polymorphisms rs10455872 and rs3798220 derived from a large Caucasian general population sample of almost 3000 individuals.
F. Kronenberg
good surrogate for risk evaluation, instead of measurement of small apo(a) isoforms, because approximately half of the small apo(a) isoforms will remain undetected when only these two SNPs are genotyped. From a population perspective, we might see the iceberg, but we might underestimate the problem. From an individual’s perspective, the false-negative rate might simply be too high when these two SNPs are used for risk prediction. Whether an extended panel of SNPs is able to tag a markedly larger fraction of small apo(a) isoforms remains unclear at present. If Lp(a) from small apo(a) isoforms has indeed a higher atherogenic potential as suggested earlier [17], it might be hypothesized that the different molecular weight of small and large apo(a) proteins is more likely to be responsible for this increased risk than a simple SNP that does not necessarily result in a structural change of the apo(a) protein. It can be argued that the apo(a) isoform should be a better predictor than the SNPs rs10455872 and rs3798220. The two SNPs demonstrated odds ratios for CHD of 1.70 and 1.92, respectively, which are not much lower than the odds ratio for small apo(a) isoforms of about 2.0. However, it is noteworthy that only about 15% of the population carry at least one minor allele of the two SNPs, whereas 25–35% of the population carry small apo(a) isoforms. Therefore, only a fraction of the high-risk carriers might be correctly classified with the two SNPs, compared to use of the apo(a) isoforms. An alternative approach is the estimation of KIV repeats by quantitative PCR, which measures the total number of KIV repeats of the two apo(a) alleles [18]. The consequence is that individuals with one very short KIV repeat allele and one very large allele would be placed in the same category as individuals with two intermediate copy number alleles. These two situations are, however, associated with very distinct Lp(a) concentrations. Despite this limitation, previous studies have shown a significant association between the sum of KIV repeats and CVD risk [18], which might have underestimated the strength of the association. Finally, and to make the situation even more complex, not all apo(a) isoforms coded in the DNA are indeed expressed in plasma. In about half to one-third of individuals, only one isoform is expressed in plasma although two different alleles are present at the DNA level [19]. With regard to risk prediction, it might be questioned why one
Editorial: Apo(a) isoforms and CVD
should be interested in an allele that is not expressed at all in plasma when the site of pathogenic action of Lp(a) is expected to be in the vascular system. Situations in which the apo(a) isoform might be more useful than currently known SNPs Usually, Lp(a) concentrations seem to be quite stable throughout life because of fairly strong genetic control [1]. However, certain diseases result in pronounced changes in the concentration of Lp(a). For example, chronic kidney disease results in an increase in Lp(a) that depends on the type of disease (whether or not there is a nephrotic component) and on the type of treatment (peritoneal dialysis is associated with higher concentrations than haemodialysis, with normalized concentrations after successful kidney transplantation). It is interesting that in patients with end-stage renal disease treated with haemodialysis, Lp(a) is increased only in those with large apo(a) isoforms compared with isoform-specific control subjects. This might explain why the apo(a) isoform is more predictive of CVD risk than Lp(a) concentrations in these patients (for review see 1, 20). Another example is type 2 diabetes mellitus (T2DM). Recent studies have described an unexpected association between very low Lp(a) concentrations (
