Introduction

Table des matières

In 1988, Reaven noted that several risk fators commonly cluster together under one disorder entity that he originally described as syndrome X1. This syndrome has been characterized by the co-occurrence of hypertension, some degree of glucose intolerance, high triglyceride levels and low high-density lipoprotein (HDL) concentrations. The basic abnormalities underlying syndrome X have been presented as the resistance of insulin to mediate glucose disposal. Due to this underlying pathophysiology, many authors have also used the term insulin resistance syndrome to define this aggregation of risk factors. The syndrome has also been given other names, including the metabolic syndrome, the plurimetabolic syndrome and the deadly quartet. More recently, the National Cholesterol Education Program’s Adult Treatment Panel III report (NCEP ATP III) has recognized the importance of this syndrome in the prevention of cardiovascular disease (CVD)2. However, NCEP ATP III, used the term metabolic syndrome for this clustering of metabolic risk factors. This term avoids the implication that insulin resistance is the primary or the only cause of associated risk factors. Since the description of the syndrome by Reaven1, several other metabolic abnormalities have been associated with it, including obesity, particularly abdominal obesity, high apolipoprotein (apo) B levels, small dense low-density lipoprotein (LDL), and abnormalities in fibrinolysis and coagulation3.

At least three health authorities have provided practical tools to identify patients with the metabolic syndrome. However, the clinical criteria differ somewhat between organizations2,4,5. Tables 1, 2 and 3 summarize the criteria used by the NCEP ATP III, the World Health Organization (WHO) and the American Association of Clinical Endocrinologists (AACE) to clinically identify the metabolic syndrome. For NCEP ATP III, when 3 out of the 5 characteristics listed in Table 1 are present, a diagnosis of metabolic syndrome can be made. It is the only guideline that considers waist circumference to express the level of adiposity. In the NCEP ATP III an explicit demonstration of insulin resistance is not required. In contrast, the WHO guidelines view insulin resistance as a required component for diagnosis. In addition to insulin resistance, two other risk factors are required for the diagnosis of the metabolic syndrome. Microalbumineria has been also added to the list as a criterion. The AACE criteria seem to be a compromise between the NCEP ATP III and the WHO guidelines. Although the clinical criteria are listed, the number of risk factors required to claim the existence of the metabolic syndrome is not specified and left to clinical judgment. Both AACE and WHO guidelines included insulin resistance measurements that are beyond routine clinical assessment. Indeed, for these two authorities, values derived from an oral glucose tolerance test are among the risk factors for the metabolic syndrome. Although these measurements give additional information, they add time and cost to clinical practice. Therefore, the NCEP ATP III guidelines may be more suitable on a clinical basis.

Derived from NCEP ATP III2

Derived from WHO4

Derived from Einhorn et al.5

These working definitions of the metabolic syndrome have allowed the prevalence of this condition to be estimated in the population. Based on the NCEP ATP III definition, the prevalence of the metabolic syndrome has been established in the Third National Health and Nutrition Examination Survey (NHANES III). Overall, the unadjusted prevalence of the metabolic syndrome was approximately 22% in this US adult population. However, the prevalence of this condition increases with age both in men and women to reach almost 45% for subjects aged above 60 years (Figure 1). Considering the epidemic of obesity6, these numbers are very likely to be higher today compared to the estimates derived from 1988-1994 NHANES III survey. Accordingly, a large proportion of adults are affected by the metabolic syndrome. This condition is thus a rapidly growing threat to public health and a major challenge that physicians and public health agencies must face.

Figure 1. Age-specific prevalence of the metabolic syndrome among 8814 US adults aged at least 20 years, by sex, National Health and Nutrition Examination Survey III, 1988-1994. Taken from Ford et al.7

The pathogenesis of the metabolic syndrome is complex and is a direct consequence of the interactions between the effects of many susceptibility genes and many environmental exposures. The complexity can be even more appreciated by considering that each component feature of the metabolic syndrome is subjected to its own regulation through both genetic and environmental factors. Changes in lifestyle observed during the last decades are responsible for the rising prevalence of the metabolic syndrome. Indeed, physical inactivity combined with an atherogenic diet (rich in saturated fat, trans fatty acids and refined sugars) constitute the perfect combination giving rise to the metabolic syndrome. These lifestyle variables, acting either through obesity or insulin sensitivity, disturb the metabolic homeostasis and lead to the multiplex risk factors.

The National Heart, Lung, and Blood Institute (NHLBI), in collaboration with the American Heart Association (AHA) identified three potential etiologic categories of the metabolic sydrome: 1- obesity and disorders of adipose tissue, 2- insulin resistance, and 3- a constellation of independent factors (eg, molecules of hepatic, vascular, and immunologic origin) that mediate specific components of the metabolic syndrome8. Some investigators place greater priority to obesity, more specifically visceral obesity, to explain the clustering of risk factors. This argument is supported by the fact that obesity is strongly associated with all cardiovascular risk factors9,10. In fact, with the strong connection between visceral obesity and risk factors, it is possible to define the metabolic syndrome as a cluster of the metabolic complications associated with obesity. Adipose tissue is now recognized as an endocrine organ that secretes numerous proteins that exert various effects11. Indeed, hyperplasia and hypertrophy of adipocytes as seen in obesity leads to an increased production of leptin, tumour necrosis factor α, interleukin-6, resistin, acylation-stimulating protein and many other proteins, and a decreased production of adiponectin. Visceral adipose tissue may be particularly active in producing several of these factors. Through these mechanisms, it is clear that obesity play a central role in the pathogenesis of the metabolic syndrome. The role of obesity in the metabolic syndrome is also accentuated by the benefits observed on all its components following weight loss12 (Figure 2). Indeed, there is substantial evidence that weight loss, particularly the mobilisation of visceral adipose tissue, leads to simultaneous improvements of the metabolic profile. Taken together, these arguments place obesity, more specifically visceral obesity, at the heart of the metabolic syndrome.

Figure 2. Potential benefits of moderate (5-10%) weight loss in high risk patients with a cluster of atherothrombotic, pro-inflammatory metabolic abnormalities associated with hypertriglyceridaemic waist. Taken from Després et al.12.

Despite these facts, there are still disagreements as to whether insulin resistance or abdominal obesity is the primary contributor to the metabolic syndrome. It is true that there is a broad range of insulin sensitivity at any given level of body fat and a large spectrum of obesity at any given level of insulin sensitivity13,14. This also means that not all insulin-resistant individuals are overweight nor all overweight individuals are insulin resistant. Some investigators place a greater priority on insulin resistance by arguing that insulin resistance/hyperinsulinemic individuals, with or without obesity, are more likely to display the abnormalities of the metabolic syndrome15. They believe that insulin resistance or hyperinsulinemia directly causes other metabolic risk factors. One point of agreement is that insulin resistance generally increases with body fat content14. On this point, it was proposed that obesity should be viewed as a lifestyle variable that has an adverse effect on insulin-mediated glucose disposal thereby increasing the chance that abnormalities associated with the metabolic syndrome will develop15. However, in contrast to weight loss, there is yet little evidence that reduction in insulin resistance will improve components of the metabolic syndrome other than glucose intolerance. Clearly, the dissociation between obesity and insulin resistance is difficult to make because both are associated with the metabolic abnormalities. It is also obvious that both factors can play an independent role in the syndrome giving their independent effects on cardiovascular risk factors16 and CVD13. One thing is for sure: the rising prevalence of the metabolic syndrome is certainly propeled by the epidemic of obesity, which is driven by changeable factors such as high caloric diets and sedentary lifestyle.

The primary clinical outcome of the metabolic syndrome is CVD. However, it is also worth mentioning that individuals with this syndrome have an increased risk for type 2 diabetes17, which is also a high-risk condition for CVD. It seems obvious that a condition characterized by multiple risk factors would increase the risk of CVD. However, few studies have examined the relation between the metabolic syndrome and future development of CVD. In the Kuopio Ischemic Heart Disease Risk Factor Study, a population-based prospective study of 1209 Finnish men, it has been demonstrated that men with the metabolic syndrome are at increased risk of CVD and all-cause mortality18. Similar results were obtained in the Botnia study including 4483 subjects participating in a family study of type 2 diabetes in Finland and Sweden19. In the Framingham study, the metabolic syndrome alone, without diabetes, has been associated with a ten-year risk of CVD that ranged from 10% to 20% in men and that did not exceed 10% in women8. In addition, no gain in CVD risk assessment is obtained when the NCEP ATP III metabolic syndrome is added to the usual Framingham risk assessment algorithm20. This suggests that the risk associated with the metabolic syndrome is captured by the traditional risk factors including age, blood pressure, total cholesterol, diabetes and HDL cholesterol. However, the metabolic syndrome was highly predictive of new-onset diabetes in the Framingham cohort.

Large prospective studies such as Framingham conducted in the United States20 and the Prospective Cardiovascular Münster (PROCAM) study conducted in Germany21 have provided simple algorithms to predict the risk of CVD in originally asymptomatic individuals. These studies have greatly contributed to the identification of the major risk factors of CVD such as age, smoking, diabetes, hypertension and plasma LDL and HDL cholesterol concentrations. In addition, several lipid-lowering trials have clearly demonstrated the importance of targeting LDL cholesterol in order to reduce the risk of CVD22,23. For this reason, LDL cholesterol has become the primary target of therapy to reduce the risk of CVD2. However, a considerable proportion of patients with CVD have cholesterol levels in the normal range24,25 and a notable proportion of patients achieving significant LDL reduction with lipid lowering therapy still develop CVD26. These results suggest that there is a need to go beyond LDL reduction and traditional risk factors in order to properly identify high risk individuals.

The Quebec Cardiovascular Study has contributed substantially in finding new markers of risk that allow a more refined identification of individuals at high risk for CVD. In the Quebec Cardiovascular Study, more than 2000 men initially free of ischemic heart disease (IHD) have been followed for a period of five years. During that period, 114 of them developed IHD. During the years, investigators of this population-based cohort have taken advantage of this setting to identify non-traditional risk factors involved in IHD. In 1997, Lamarche et al.27 measured LDL peak particle diameter (LDL-PPD) in 103 case-control pairs to determine whether the LDL size can predict the risk of IHD. Despite the lack of difference in the mean LDL-PPD between case patients and control subjects, a clear shift in the distribution of LDL-PPD was observed between the two groups (Figure 3). In fact, the distribution of LDL-PPD in men who developed IHD during the follow-up tended to be shifted toward lower values compared with the control subjects. It has also been demonstrated that men in the first tertile of LDL-PPD distribution had a 3.6 fold increase in the risk of IHD compared with those in the third tertile. This effect was also independent of lipid variables including LDL cholesterol, triglyceride, HDL cholesterol, and apolipoprotein B concentrations. This important work clearly demonstrated that new cardiovascular risk factors can improve risk evaluation.

Figure 3. Frequency distributions of LDL-PPD in 103 pairs of case patients (solid bars) and control subjects (open bars). Also presented is the proportion of case patients in each tertile of the LDL-PPD distribution of control subjects and their corresponding mean LDL-PPD (± SD). Arrows on the x axis identify the tertile values of control subjects (25.64 and 26.05 nm). Taken from Lamarche et al.27.

In the same cohort, it has also been demonstrated that hyperinsulinemia28 and apoB levels29 are independent predictors of IHD. Taken together, it suggests that small, dense LDL, insulin and apoB levels can provide substantial information for assessment of IHD. In 1998, Lamarche et al.30 published another important work demonstrating that these three variables, the so-called metabolic triad of non-traditional risk factors, give substantially more information on the risk of IHD compared with the information provided by conventional lipid variables (Figure 4). Indeed, the risk of IHD was significantly increased in men who had elevated fasting plasma insulin and apoB levels and small, dense LDL particles, compared with men who had normal levels for these three risk factors. It is also worth mentioning that adjustment for traditional lipid variables did not attenuate this relationship. Figure 4 also highlights the superiority of non-traditional risk variables over traditional variables to discriminate individuals at high risk for IHD.

Figure 4. Risk of ischemic heart disease (IHD) according to the cumulative number of traditional and non-traditional risk factors. Traditional factors are LDL cholesterol, triglycerides and HDL cholesterol. Non-traditional factors are insulin, apoB and small, dense LDL particles. Odds ratios are adjusted for systolic blood pressure, family history of IHD and medication use. Taken from Lamarche et al.30.

Despite being a powerful tool to predict the risk of IHD, the utility of the metabolic triad is somewhat limited due to the costs associated with the measurements of insulin, apoB, and LDL size. Indeed, these new metabolic risk markers are not evaluated in a standard clinical visit and require additional costs. In addition, not all laboratories can perform such measurements. Unfortunately, these factors constitute major barriers for the use of these non-traditional risk factors in a clinical setting. Therefore, Lemieux et al.31 have developed a simple screening tool for the identification of men characterized by the metabolic triad (Figure 5). By using simple measurements such as waist circumference and triglyceride levels they have been able to identify the majority of subjects with the metabolic triad. This screening tool has been called the hypertriglyceridemic waist and provides a simple and inexpensive approach to identify high-risk patients.

Figure 5. Working hypothetical model providing rationale for use of waist circumference and triglyceride levels as screening tools for the atherogenic triad of new metabolic risk factors. Waist circumference is used as discriminant of elevated fasting insulin and apo B levels, whereas triglyceride concentrations is used as discriminant of small, dense LDL phenotype. Taken from Lemieux et al.31.

This section provides a brief overview of the methods and strategies for the genetic dissection of complex human traits such as the metabolic syndrome and its individual components. Globally the genetic dissection of complex traits integrates methods from the fields of genetic epidemiology and molecular biology. The methods and strategies employed to identify the genes have greatly changed throughout the years owing to the progress made in statistical analysis and the influence of the Human Genome Project. Heritability studies, complex segregation analyses, candidate gene linkage and association studies, genome-wide linkage scans and animal models are now all part of the arsenal to hunt-down the susceptibility genes.

By definition, complex traits are determined by the joint action of multiple genes and environmental factors32 (Figure 6). For a given phenotype, multiple causes, both genetic and nongenetic, and interactions among them, contribute to its variation. While these factors may make the task of finding genes difficult, their recognition is important and gives a broader perspective of the underlying complexity. The multiple etiologic factors also imply that the effect size of most of these factors is rather modest. In fact, in complex traits, genes with large individual effects are likely to be rare. A more realistic genetic component of many complex traits is oligogenic (a few genes each with a moderate effect) or even polygenic (many genes, each with a small effect). In that spirit, searching for a small effect size calls for a large sample size, which is often not the case in many genetic studies. Indeed multiple studies have generated conflicting findings due to insufficient power. It should also be accepted that a gene with small individual effect may have a substantial contribution to the manifestation of the trait by interacting with a second gene or an environmental factor. Gene-gene and gene-environment interactions are difficult to test with current technologies but are likely to be important in the context of complex traits.

Figure 6. Varying causes of phenotypic variation. Derived from Rao.32.

Genetic and etiological heterogeneity are also expected to be a major challenge in the genetic dissection of complex traits. The multifactorial aetiologies imply that similar phenotypes can be the result of different factors. Accordingly, the phenotype-genotype relationship is not exclusive. On a purely genetic sense, different genes may act for different populations having a distinct genetic background or a manifest characteristic (obesity, type 2 diabetes, etc.). Furthermore, the phenotype is influenced by an unknown number of polygenes and each polygenic effect depends on an unknown number of alleles33. With a more global perspective, the connections between the genome and the phenotypes cannot be viewed as a one-way flow of information. In fact, the multigene genotype acts through the primary biochemical and physiological subsystems which subsequently affect the phenotypic measures of health34. Between the genotype and the phenotype there is a dynamic and epigenetic network of cellular and organismal dimensions that constantly shapes the phenotype and produces feedback information to influence the expression of participating genes. These complicated networks act throughout the lifetime of an individual and at any moment are influenced by the previous and contemporary environmental exposures. Taken all together it is not surprising that the dissection of complex human traits is the greatest challenge that geneticists have ever faced.

Despite the complexity enumerated in the previous paragraph, the genetic basis of complex human phenotypes can be divided into two basic approaches: the unmeasured genotype and the measured genotype approaches35 (Figure 7). The unmeasured genotype approach is a purely statistical strategy that makes inference about the influence of genes based on the distribution of the phenotype. Because the inference is made from the phenotype to the genes, the approach is also called the top-down approach. A major advantage of the top-down approach is that no prior knowledge about the biology of the phenotypes is required to perform the analysis. Most of the genetic analysis using this approach used relatedness among family members to quantify the contribution of genetic factors and test hypotheses regarding a variety of general and specific models of inheritance. However, a major drawback of the unmeasured genotype is that specific gene(s) involved cannot be identified. Heritability studies and complex segregation analyses fall within this category and have been used in chapters 7 and 9, respectively.

Figure 7. Bottom-up and top-down approaches in the study of continuously distributed phenotypes. Taken from Bouchard et al.35.

In the measured genotype approach, influences about the roles of genes are made from DNA variations to the phenotype. For this reason, the approach is also referred to as the bottom-up approach. With this approach, genetic variations at the DNA level are genotyped and then tested for association and linkage with the phenotypes. The approach holds for evaluating the impact of a specific gene. DNA variations can be located within a gene believed, based on biological observations, to be involved in the trait (candidate gene approach) or can be random markers distributed across the genome (genome scan approach). Association and linkage studies fall within this category and have been used extensively within this project.

A combination of both bottom-up and top-down approaches is required to have a better understanding of the genetics underlying a complex trait. Throughout the years, the priority has switched from the top-down to the bottom-up approach due to the new technologies and the completion of the Human Genome Project. In the past, the top-down approach was the only option since genotype measurement was not possible. However, today with the high throughput technologies, most geneticists have favoured the bottom-up approach. Nevertheless, both approaches are required if one wants to appreciate the genetics of a phenotype from its genetic contribution to the specific genes implicated.

The traditional steps to achieve a global appreciation of the genetic basis of a quantitative phenotype are illustrated in Figure 8. These steps are successive and the necessity of moving to the following one is greatly motivated by positive results obtained in the preceding one. These steps also help to understand how specific genetic techniques fit into the larger arsenal of genetic epidemiological methods. The first step is to determine whether or not the phenotype of interest aggregates within families. In the context of family study, familial aggregation can be evaluated using a simple analysis of variance to compare the variance between families to the variance within families. Familial resemblance is claimed if the variance between families is significantly higher than the variance within families. If there is evidence of familial aggregation, the next step consists of evaluating heritability to verify whether or not the familial resemblance is partly or totally attributable to genetic factors. In a family study, heritability can be obtained by estimation of familial correlations36. The strength of the correlations between family members (i.e., spouse, offspring, etc) gives an appreciation of the familial resemblance and provides insights about the relative importance of genetic versus nongenetic factors. If there is evidence for a genetic effect, the next step is to determine whether a major gene effect can be detected in the phenotype. The most popular method to detect a major gene effect is segregation analysis. The analytical strategy of segregation analysis relies on fitting a variety of general and specific models of inheritance testing the existence of a major gene effect, the mode of transmission of this major gene effect and its allele frequency. The model providing the best fit to the data is chosen based on specific statistical criteria and inference is made based on the hypothesis tested by this chosen model. It should be noted that up to this step only the unmeasured genotype approach is used. Indeed, familial resemblance, heritability and segregation analysis provide only statistical evidence about the contribution of genes involved in the phenotype of interest but indicate nothing about the specific genes. Thus, the last step consists of performing association and linkage studies to identify the genes or the genomic regions underlying the observed genetic effect. These studies can be performed with DNA variants located in candidate genes or with random genetic markers evenly spaced throughout the genome (see next section).

Figure 8. Flow chart describing the different steps in the investigation of the genetic basis of a quantitative phenotype. Derived from Bouchard et al.35.

The development of rapid throughput genotyping assays and the widespread availability of DNA from large population studies have promoted the use of association studies. The number of association studies published in the literature has steadily increased throughout the years and is now reaching a cadence of 75 to 100 per week37. The phenotype under study may be the presence or absence of disease (discrete phenotype) or a quantitative measure38. Slightly different analytical techniques are used for discrete and quantitative phenotypes. For a discrete trait, the allele frequency at the polymorphic marker is compared between a case and control groups of unrelated individuals. In the presence of association, allele frequencies differ between cases and controls. For a quantitative phenotype, the test compares mean phenotypic values among individuals of different genotypes at the marker locus. For a rare variant, the analysis may also be done by grouping subjects based on the presence or absence of a particular allele. The mean phenotypic value can then be compared between carriers and noncarriers of that specific allele.

There are three reasons why an association between a marker locus and a trait may be observed39. First, it is possible that the relationship is the causal one and the genotyped marker is itself functional. This argument is particularly valid if the different alleles at the genotyped marker result in changes in the amino acid structure of the protein and functional studies have confirmed the effect on the gene product. A second option is that the genotyped marker is not itself functional, but is in linkage disequilibrium with other polymorphisms that are functional. Linkage disequilibrium is by definition a non-random association of alleles at adjacent loci40. If two genetic markers are in close proximity on the same chromosome they tend to be co-inherited. The loci are said to be in disequilibrium when a particular allele at the first locus is found together with a specific allele at a second locus more often than would be expected by chance. Accordingly, a significant association can be the result of a functional variant in linkage disequilibrium with the tested marker. This existing relationship between neighbouring loci has recently justified the launch of the International HapMap Project met to facilitate the discovery of sequence variants that affect common diseases41. Finally, the observed association can be spurious and result from population stratification42. These spurious associations occur due to ethnic admixture and unintentionally drawn from two or more ethnic groups present in the studied population. Any trait having a higher frequency in a particular ethnic group will have a positive association with any genetic markers having a higher allele frequency in that same ethic group. Population stratification has clearly impaired the credibility of association studies. To prevent these false positive associations, two solutions have been proposed. First, by typing several dozen random markers, one can empirically detect and correct for stratification43-45. Secondly, one can use family-based studies such as transmission disequilibrium test (TDT)46,47. TDT and other family-based tests of associations are immune to false-positives caused by ethnic admixture and their use has been substantially propelled by the concern driven by population stratification47.

Other than population stratification, association studies received criticisms due to the lack of replication48,49. A recent meta-analysis has demonstrated that among 166 associations (between gene variants and disease) investigated more than three times, only 6 have been consistently replicated42. The possible reasons for the inconsistency include false positives due to type 1 error (driven by publication bias for positive associations), false positives due to population stratification, false negatives due to lack of power in potential replication studies, and true differences between study population. Considering these factors, caution is emphasized before drawing conclusions from a single report of an association between a genetic variant and a particular phenotype or disease42. However, association studies are widely anticipated to contribute to the understanding of complex traits50-52. To achieve this expectation, the association study needs to be well conducted and some journals have now proposed criteria for acceptance and publication of genetic association studies53,54. Ideally, association studies must have a large sample size, small P values, report associations that make biological sense and alleles that affect the gene product in a physiologically meaningful way. In addition, the association gains credibility when the finding is replicated in an independent cohort, the association is observed both in family based and population-based studies, and the odds ratio and/or attributable risk is high. Obviously, very few studies will meet these criteria but they can help judging the credibility of the finding. It is also important to prioritize polymorphisms50. We now know that the 3 billion base pairs of the human genome contain more than 10 millions genetic variations and approximately 30 thousand genes41,55-57. Accordingly, the prior probabilities that given polymorphisms (located or not in a candidate gene) play a role in disease or any health related phenotype is very low. To increase this likelihood, one should follow genes located in chromosomal regions that co-segregate with the phenotype and/or selected genetic variants that have a demonstrated functional consequence. To take full advantage of association studies, it is important to fit this technique into a larger arsenal of genetic epidemiological methods and molecular studies.

Linkage study is another important method to identify genes contributing to diseases or health-related phenotypes. In contrast to association studies, which seek to identify particular variants that are associated with the phenotype at the population level, linkage tests for evidence of cosegregation between a marker locus and a trait within families. Indeed, relatedness among subjects is required in order to perform linkage analyses. Linkage analyses have greatly evolved during the years to exploit the genetic information present in kindreds consisting of sibling pairs to extended pedigrees. Linkage analyses fall into two main categories: parametric (model-based) that strongly models the genetic effect and nonparametric (model-free) that makes few assumptions regarding the etiologic model underlying the phenotypic distribution.

The first linkage method in humans was proposed by Morton in 195558. In this seminal paper, he introduced the lod score method (and simultaneously the concept of lod score) which has been recognized as a major milestone in the genetic dissection of human traits32. This method is still in use today and has been remarkably successful in identifying disease genes for Mendelian disorders59. However, before performing the lod score method, prior knowledge is required about the mode of inheritance of the trait. Unfortunately, this information is unknown for complex traits and misspecification of the required parameters may substantially reduce the power60. Accordingly, robust nonparametric linkage methods have been developed for complex traits, which rely on patterns of allele sharing between related subjects to infer linkage.

The Haseman and Elston sib-pair linkage method61 is a classic example and a widely used nonparametric linkage analysis. This method simply regresses the squared phenotypic difference in sibs on the number of alleles shared identical by descent—that is, that are direct copies of the same ancestral alleles. The idea is that sibs that share a greater proportion of alleles identical by descent have a more similar phenotype. In contrast, under the null hypothesis, no relation is observed between the sibs’ phenotypes and the degree of allele sharing. Instead of using the squared sib-pairs trait difference, current methods look at the phenotypic covariance or correlation as a function of the identical by descent sharing using the variance components approach62,63. Variance component-based linkage analysis has become a widespread statistical tool to identify quantitative trait locus (QTL) involved in complex traits. This increased popularity is mainly explained by its ability to accommodate large pedigrees and to test important biological phenomenon such as epistasis, gene-environment interaction and oligogenic model64.

Linkage and association studies should not be viewed as distinct genetic statistical tool used independently but more as complementary approaches to find the genes. It has been proposed that genes with small or subtle effects may not be detectable by linkage65. However, linkage studies can succeed where association fails and vice-versa (Figure 9). The power to detect the effect of genes depends on the effect size, the allele frequency and the sample size. Obviously, the ability of both association and linkage studies to detect the genes increases when the sample size and the effect size increase. On the other hand, the allele frequency will really dictate whether linkage or association is more powerful. For a more frequent allele, association is favoured and for a rare allele, linkage is prefered. However, these theoretical concepts may be useless since the knowledge about the allele frequency is rarely known in advance. Usually, more practical reasons force the utilities of one to another depending whether a candidate gene or a genome scan strategy is adopted.

Figure 9. Optimal mapping strategies for different types of loci. Taken from Ardlie et al.40.

Genome scans are simply large-scale applications of linkage and association methods. The objective of the approach is to identify the chromosomal regions within which one or more disease-predisposing genes lie. Genes contained within such linked regions become positional candidates and are next examined for mutations potentially causing the signal. Conceptually, genome scan studies can be divided into three steps: 1-scan the entire genome with a dense collection of genetic markers; 2-calculate an appropriate linkage or association statistic at each markers along the genome; and 3-identify the regions in which statistics show a significant deviation from what would be expected by chance66. One of the great advantage of the genome scan approach is that it can be applied without prior knowledge of the biological basis of the disease or the phenotypes under study.

For genome-wide linkage scan the usual practice is to test about 300-400 highly polymorphic markers, usually microsatellites, distributed approximately evenly across the genome with an average spacing between markers in the order of 10 cM. Localization of the locus by linkage analysis can achieve only a certain level of precision. Usually the minimal interval of a QTL in humans range from 10 to 30 Mb which contains approximately 100 to 300 genes67. It should also be emphasized that the estimated peak locations are generally not very precise68. Accordingly, there is a large gap between QTL and gene finding with genome-wide linkage scan69. Nevertheless, this strategy has been highly successful in the identification of genes responsible for simple Mendelian traits70,71. However, the general picture derived from genome-wide linkage scan in complex traits is one of the difficulty in locating genes and replicating previous reported linkage signals. This difficult picture is not related to the approach but more to the nature of complex traits characterized by locus heterogeneity, epistatis, low penetrance, variable expressivity, pleiotropy and limited statistical power39,70.

Genome-wide association studies has been proposed as an alternative to facilitate the mapping of complex disease loci72. Similar to genome-wide linkage scan, this approach does not require prior knowledge about the molecular basis of the disease/phenotypes. As it is the case for linkage scan, the objective of association scan is to identify chromosomal regions harbouring susceptibility genes. However, with this new strategy, the resolution attain will be much higher (map the gene to smaller genetic interval), in the order of kilobases. Genome-wide association scan has been driven by empirical studies showing strong associations between nearby SNPs73-76. This strong allelic association between variants is known as linkage disequilibrium. The practical implication of linkage disequilibrium is that a few carefully chosen SNPs (tag SNPs) can be genotyped to capture much of the information in a chromosomal region. Therefore the tag SNPs could serve as genetic markers to detect association between a particular chromosomal region and the disease/phenotype, whether or not the tag SNPs themselves have functional effects. The search for causal genetic variants can then be limited to the regions showing association. However, because of the central role of linkage disequilibrium in the concept of genome-wide scan association study and the variable nature of disequilibrium in the human genome73,74,76,77, the utility of such approach remains to be seen.

Almost a decade ago, Bouchard78 proposed a multi-layered model of the metabolic syndrome where genes actions, including their interaction with each other and with the environment, can be operative at all levels. This structure recognized that genes can act on the causes of the metabolic syndrome (visceral obesity, insulin resistance and even behaviours affecting healthy lifestyle choices) and on its individual components. In this complicated context, it is essential to properly define the phenotype under study. It is recognized that affection status of clinical diseases based on a discrete binary scale (affected or unaffected) contains considerably less genetic information in comparison to a quantitative disease-related phenotypes79,80. Accordingly, in this project we relied mainly on quantitative phenotypes to identify the susceptibility genes of the metabolic syndrome. In the last chapter, we tested the genetics of a quantitative metabolic syndrome variable created by factor analysis. This was an attempt to identify the genes contributing to the underlying cluster of risk factors defining the metabolic syndrome. However, for most of the project, we used quantitative individual components of the metabolic syndrome to identify the metabolic syndrome genes. Many phenotypes were investigated including obesity, lipoprotein/lipid and glucose/insulin variables. However, a major part of the work has focused on the genetics of LDL particle size. Understanding the genetics of LDL size is not only important because it is a component of the metabolic syndrome, but also because it was recently recognized as an independent marker of cardiovascular risk81.

The following is an attempt to summarize the growing evidence of genetic control on LDL particle heterogeneity.

LDL cholesterol is a well-known risk factor for coronary heart disease (CHD) and is now recognized as the primary target of lipid lowering therapy2. However, it is known that LDL particles are heterogeneous in terms of size, density, chemical composition and electric charge82-84. Data from case-control85 and prospective27,81,86,87 studies suggested that small, dense LDL particles are associated with increased risk of CHD. The atherogenicity of these particles is attributed to several possible biological mechanisms including greater susceptibility for oxidation88-92, decreased affinity for the LDL receptor93-97, increased binding to the arterial wall98-101 and greater facility to cross the arterial wall102,103 as well as having negative effects on the endothelium function104. Additional evidence for the relevance of the small, dense LDL on atherosclerotic lesions development and CHD progression are derived from an animal model105 and lipid lowering trials in human106,107. On the other hand, recent findings from the Cholesterol and Recurrent Events (CARE) trial108 support earlier case-control109-112 and prospective113 studies showing that small, dense LDLs are not risk factors for CHD. In fact, some of these studies have shown that larger LDL particles are associated with CHD. While these studies disagree as to which LDL particle size (small or large) is related to CHD risk114, defining the genetic and environmental factors modulating LDL particle properties may be helpful in understanding its relationship with CHD.

Multiple approaches have been used to hunt down the genes involved in complex human diseases and diseases-related risk factors. Through the years, methods and strategies have evolved following the progress made in genetic epidemiology and the completion of the Human Genome Project. Genetic studies on LDL particles represent a perfect example of this phenomenon. Several studies have investigated the genetics of LDL particle heterogeneity. Heritability studies, complex segregation analyses, linkage and association studies with candidate genes, and genome-wide linkage scans are all part of the arsenal used for dissecting the genetic architecture of this trait. The following is an attempt to summarize the growing evidence of genetic control on LDL particle heterogeneity.

Several studies have shown that small, dense LDL are associated with a constellation of other well-recognized lipoprotein-related risk factors, including increased plasma triglyceride and apoB levels as well as decreased HDL cholesterol concentrations. Furthermore, small, dense LDL particles coexist in the same subjects as part of multifaceted phenotypes including the metabolic syndrome, the atherogenic lipoprotein phenotype (LDL subclass pattern B) and familial combined hyperlipidemia (FCHL)115. Thus, small, dense LDL may be a qualitative trait representing a common atherogenic lipoprotein/metabolic profile and the proposed genetic loci responsible for small, dense LDL may in fact be responsible for a more extensive syndrome. However, throughout the following section a more narrow view of the phenotypes that characterize LDL particle heterogeneity is taken.

A number of analytical techniques are available for characterizing LDL heterogeneity some technicality must be addressed before going through genetic ground. LDL heterogeneity was first described using analytical ultracentrifugation (AnUC)116. Over the years, this technique was replaced by others including density gradient ultracentrifugation (DGU), gradient gel electrophoresis (GGE), and more recently by nuclear magnetic resonance (NMR) spectroscopy. The phenotypes derived from these techniques are those used in the genetics studies performed so far. Based on GGE, a continuous variable can be defined as LDL peak particle diameter (LDL-PPD), reflecting the size of the major LDL subclass in an individual subject. A dichotomous classification can also be defined based on GGE and referred to as LDL subclass patterns, or phenotypes, A and B . LDL subclass phenotype A is characterized by a predominance of large LDL particles and skewing of the densitometric scan toward small particles, while LDL subclass phenotype B is characterized by a predominance of small LDL particles and skewing of the curve toward large particles117. Other phenotypes can be constructed using GGE, including LDL score which is calculated using the migration distance (mm) of each peak multiplied by its respective relative area118 and LDL type which is a weighed average of seven possible categories of LDL, resulting in a variable ranging from 1 (largest) to 7 (smallest)119. A more detailed description of these techniques are found in the following published reviews85,120,121.

The first evidence for a genetic determination of LDL properties was reported by Fisher et al. in 1975122. Five families, including 11 mating and 16 offspring, were examined for their LDL molecular weight. Only subjects having monodisperse LDL, that is, LDL that is found to be present as a single, essentially homogeneous population of macromolecules, were included in the study. Correlation coefficients between pairs of relatives revealed a significant parent-offspring correlation (0.82, p < 0.01) but absence of correlation between fathers’ and mothers’ (0.32, p = NS). These results provided evidence for the genetic contribution to LDL molecular weight. To further determine the degree of resemblance of the offsprings to their parents, a regression coefficient of the mean molecular weight of the offsprings on the mean parental molecular weight was calculated. The regression coefficient was 0.30 (p < 0.01), which made the authors conclude that approximately 30% of the observed LDL molecular weight variance is due to additive gene action. In addition, based on the five families, the authors postulated a model consistent with a single gene (two alleles) locus genetic mode of inheritance without dominance. Although the sample size used in this study was relatively small, it demonstrated for the first time that LDL characteristics segregate within families.

Since this earlier report, accumulating evidence of familial and ethnic aggregation of LDL subclasses have emerged in the literature. Haffner et al.123,124 demonstrated a significant difference between ethnic groups in LDL size among 1571 subjects from the Insulin Resistance Atherosclerosis Study and 466 subjects from the San Antonio Family Heart Study. These studies cannot disentangle the effect of the genetic background from the effect mediated by the difference in lifestyles between ethnicity, but clearly motivated genetic studies in the field.

Studies using identical (monozygotic, MZ) and fraternal (dizygotic, DZ) twins have been used to assess the heritability of LDL size. The first study on this issue was based on 119 MZ and 113 DZ twin pairs participating in the third examination of the National Heart, Lung, and Blood Institute Twin Study125. In this study, the LDL subfractions were separated by GGE and the heritability analysis used LDL type. The LDL type intraclass correlation coefficient in MZ twins was significantly higher than the correlation coefficient in DZ twins (0.58 vs 0.32, p < 0.005), with an heritability of 0.52 prior to controlling for covariate effects. After adjustment for BMI, alcohol consumption, cigarette smoking, and physical activity, the heritability decreased to 0.39. Despite their great magnitude, these estimates were not statistically significant suggesting the lack of heritability for LDL type. Similar results were obtained when only the major LDL band (LDL-PPD) was used as a variable. Thus, the authors concluded that LDL particle size is not greatly influenced by genetic factors within this population. It is noteworthy that the authors used the more conservative among component126 estimate of heritability due to some indication of unequal total variance between zygosities. Although, this procedure is considered more suitable in such case, the power to detect significant heritability is substantially reduced.

The heritability estimates were also analyzed based on 203 monozygotic and 145 dizygotic pairs of adult female twins who participated in the second examination of the Kaiser Permanente Women Twins Study127. The classical heritability estimate for LDL-PPD was 0.54, and the within-pair estimate was 0.48. These estimates were not changed substantially when the analyses were restricted to postmenopausal, nondiabetic, non β-blocker users or Caucasian pairs with heritability ranging from 0.34 to 0.5. Thus the authors suggested that between one third and one half of the variability in LDL size appears to be attributable to genetic influences in this sample of women twins.

Heritability estimates of LDL-PPD was also evaluated using family data. The first family study on this issue was based on 780 individuals members of 85 families participating in the Genetic Epidemiology of Hypertriglyceridemia (GET) study128. The GET study is based on 2 family studies one ascertained through hyperlipidemic probands surviving a myocardial infarction and the second through hypertriglyceridemic probands without CHD. After accounting for age and sex effects, results suggested that approximately one third of the residual variance in LDL-PPD (h2 = 0.34) was attributable to additive genetic effects. Higher heritability coefficients were observed for LDL size in the Ashkenazi Jewish families ascertained for exceptional longevity129. In this study, LDL size was characterized by NMR and heritability was estimated at 0.46 in women and at 0.60 in men. These results demonstrated that LDL size is highly heritable irrespective of the analytical methods used to characterize the particles and suggested that the measurement error inherent to each technique does not mask the genetic signal.

By means of a new metric representing coordinated size variation between HDL and LDL size particles, Rainwater et al.130 conducted an original study to test the hypothesis that there are “lipoprotein size genes”. The new metric, named ΔLDL, is a metric for LDL particle size phenotype that optimally reflects the size correlation between LDL and HDL particles. ΔLDL was subjected to quantitative genetic analyses using 1157 Mexican Americans participating in the San Antonio Family Heart Study. Heritability of ΔLDL was highly significant and indicated that nearly half (44%) of the residual variance (after adjustment for sex, age, diabetes status, contraceptive use, and hypertension medications) in ΔLDL was explained by additive gene effects. After including triglyceride levels in the model as a covariate, the heritability estimate decreased from 0.44 to 0.30, indicating gene(s) common for both traits. These data indicate that particle size phenotypes are under substantial genetic control.

Taken together, the above studies suggested that 30% to 60% of the variance in LDL particle size is attributable to genetic factors, with the remainder due to nongenetic influences. Accordingly, these genetic studies also pointed out the importance of nongenetic factors on LDL subclasses since approximately 50% of the variance is attributable to nongenetic factors. A number of environmental influences have been identified, including, among others, dietary factors132, physical activity133, abdominal obesity134, insulin resistance and hyperinsulinemia135. The combination of genetic and environmental influences provides opportunities to develop prevention strategies to reduce CHD risk among genetically susceptible individuals136.

Heritability estimates obtained from twin and family studies reinforced the interest in finding gene(s) underlying that genetic effect. The following are the different lines of evidence that provided the existence of a single gene with major effect on the phenotype, including commingling analyses, segregation analyses and complex segregation analyses.

The first studies that investigated the inheritance of LDL heterogeneity were derived from fitting the data into pedigrees under an hypothetical genetic model. Fisher et al.122 were the first to provide evidence for a single gene-two allelic system locus affecting LDL heterogeneity. Using pedigrees from five families, they proposed a model of two alleles, one a determinant for high, the other for low LDL molecular weight. A decade later, Austin et al.140 evaluated the lipoprotein subclasses (pattern A/B) by GGE in 79 healthy members of sixteen nuclear families living in a local Mormon community. Their data proposed a genetic model consistent with a single-locus, bi-allelic system as well. The estimated frequency of the allele leading to the phenotype characterized by a predominance of small, dense LDL subclasses (pattern B) was approximately 15% under a dominant mode of inheritance. However, in contrast to the observation of Fisher et al122, expression of the phenotype appears to be age dependent, in that most affected subjects in the population were older than 40 years. Although, different techniques were used to detect LDL properties between these two studies, it is possible that the LDL pattern reported in the later and the molecular weight reported in the former represent the same trait. Although these studies were limited by their sample size, they provided additional evidence in favor of a single gene affecting LDL density and size.

The presence of a major gene effect in addition to its mode of inheritance has been also investigated using complex segregation analyses. The results of these studies are summarized in Table 5. Two years after having proposed a single gene-two allelic system locus affecting LDL patterns, Austin et al.141 have confirmed their results on an enlarged sample of the same Mormon community containing 61 healthy families including 301 members. The model providing the best fit to the data included a single gene with a dominant mode of inheritance and a frequency of 25% and reduced penetrance for men under age 20 and for premenopausal women. It should be noted, however, that both recessive and additive modes of inheritance could not be rejected. Similar results were observed for 234 individuals of 78 nuclear families with FCHL142. In this sample, complex segregation analyses suggested that LDL subclass pattern B is controlled by a single major genetic locus (with either a dominant or an additive mode of inheritance) and a small, but significant, multifactorial inheritance component. The prevalence of LDL subclass pattern B allele was also common in these families (≈ 0.30), suggesting that the proposed allele for pattern B is just as likely to occur in families with FCHL as in healthy families. Again, reduced penetrance for pattern B allele in FCHL families was observed for men under age 20 and for women under age 50.

The two later complex segregation studies were based on the dichotomization of the LDL subfraction into two discrete phenotypes. It is possible that this dichotomous definition oversimplifies the biochemical heterogeneity of LDL particles. de Graaf et al.144 were concerned by such procedure since much information is lost, i.e. we do not know whether an individual is close to or far from the LDL size threshold for the pattern A / pattern B classification which results in a lost of power79. Accordingly, they constructed a continuous variable, named parameter K, that reflect LDL subfraction profile and which is characterized by the relative contribution of the three major LDL subfractions, LDL1, LDL2, and LDL3, determined by DGU. Analysis for this quantitative trait was performed on 19 healthy Dutch families including 159 individuals. Results indicated that the LDL subfraction profile is controlled by a major autosomal, highly penetrant recessive allele with a population frequency of 19% and an additional multifactorial inheritance component. The penetrance of the more dense LDL allele increases with age, for both sexes, and was higher for men than women. Furthermore, it appeared that oral contraceptive use was associated with a high penetrance of the more dense LDL subfraction profile. Also concerned by the possibility that the dichotomous trait may not provide the best reflection of LDL size distribution, Austin et al.138 reanalyzed their healthy subjects living in a Mormon community but this time by using LDL-PPD instead of the dichotomous classification reported earlier141. The model providing the best fit to the data consisted of a single major gene effect with Mendelian inheritance, and with no additional multifactorial inheritance component. However, the available sample was not sufficient to distinguish dominant versus recessive mode of inheritance. Thus, analysis of the continuous LDL-PPD variable was not superior to the dichotomous LDL subclass pattern classification in determining the mode of inheritance of LDL subclasses in this healthy families sample. The mode of inheritance of parameter K was also investigated in a large sample of Dutch families with FCHL145. The genetic basis of LDL subfraction profile in this family was best described by a common, major autosomal gene effect with a population frequency of 42% and a recessive mode of inheritance with a polygenic heritability component of 25%. Subsequently, the mode of inheritance of LDL-PPD was investigated in 373 subjects from 80 kindreds residing in kibbutz settlements in Israel139. Complex segregation analyses on sex- and age-adjusted LDL-PPD were inconclusive in this study since both the mixed recessive genetic model and the mixed environmental model could not be rejected. However, when the regression model for sex and age allowed coefficients to be ousiotype (genotype class) specific, the mixed environmental model was rejected while a major Mendelian model was not. Indeed a major additive gene (codominant) model for LDL-PPD with an allele frequency of 24% for small LDL particles could not be rejected. In addition this model contains a large polygenic component (74%). The authors postulated that the ethnic homogeneity and the lifestyle similarity of the sample may explain the high contribution of polygenic factors to LDL-PPD. More recently, the genetic influence of LDL-PPD was modeled in 48 Finnish FCHL families143. Complex segregation analyses in these families suggested that the trait is the result of the additive effects of multiple genes where a recessive major gene effect of low frequency (6%) may contribute to large LDL-PPD in women. For men, they could not established that LDL-PPD follows a strictly polygenic model, but the results indicated that LDL size is unlikely to be influenced by a major gene effect in this population.

With the exception of the later study, results from complex segregation analyses support the concept of a major gene effect involved in LDL size and density. However, some dissimilarities were found between the studies in regard to the mode of inheritance, allele frequency and the presence or not of a multifactorial inheritance component. This discrepancy could be explained by differences in family structures, criteria for proband ascertainment, and the use of different techniques to characterize LDL heterogeneity. Nevertheless, these studies unanimously provided evidence about the contribution of a major gene effect and clearly motivate the race to hunt-it down.

Many investigators have used linkage analyses to identify the genes underlying the genetic contribution of LDL particle characteristics. The early studies have been performed using candidate gene strategies by studying genetic variations located within or in proximity of genes coding protein products known to be involved in lipoprotein/lipid metabolism. On the other hand, recent studies have used a genome-wide scan approach in order to identify chromosomal regions influencing LDL size-related phenotypes. Table 6 presents a summary of the loci and genes, ordered by chromosome number, that have provided evidence of linkage using these two strategies. It should be noted that only positive findings are provided in this table and careful examination of the literature might in fact show significant evidence against linkage for certain loci presented. It is also worth mentioning that opposite results for the same gene may not necessarily imply controversy giving the different study samples.

The APOB gene was of particular interest since it is the principal protein component of LDL particles. With the classic logarithm of the odds (LOD) score-linkage method the first two linkage studies rejected clearly the involvement of this locus with LDL subclass pattern B after obtaining LOD score of -13.3 and -7.5154,155. In addition, no evidence of linkage to the APOB locus was observed for LDL-PPD in families ascertained for coronary artery disease (CAD)148. However, a subsequent study performed in dizygotic twin pairs indicated, for the first time, positive linkage between LDL-PPD and the APOB locus146. Thus, it is possible that the APOB locus has an effect on LDL size in particular subgroups of the population, perhaps in women. Because low-density lipoprotein receptor (LDLR) is responsible for the clearance of apoB-containing lipoproteins, the LDLR locus on chromosome 19p was also a reasonable candidate gene for linkage analyses. Using parametric linkage analyses with reduced penetrance of pattern B, Nishina et al.153 obtained evidence of linkage to the LDLR locus (LOD = 4.27). This finding was confirmed by a subsequent study using quantitative sib-pair linkage analyses in CAD families148. Borderline significant evidence of linkage was also observed between the LDLR locus and LDL-PPD in dizygotic twin pairs from the Kaiser Permanente Women Twins Study (p = 0.082)146. On the other hand, results from the Dutch FCHL families149 and from families identified through hyperlipidemic probands156 showed no evidence of linkage between the LDLR locus and either the LDL-PPD and the dichotomized pattern A/B phenotype. It is also worth mentioning that a follow-up study of the original families in which linkage to this locus has been demonstrated153 found no mutation in the coding sequence of the LDLR gene, suggesting that a nearby gene was responsible for the linkage157. Using parametric linkage method and adjusting the phenotype for the within-genotype variance, Hokanson et al.150 found in heterozygous lipoprotein lipase (LPL)-deficient families a highly significant LOD score of 6.24 between LDL-PPD and the LPL gene, which encodes a rate-limiting enzyme in the formation of LDL particles. However, two other studies have been unable to confirm this linkage in different study samples148,156. To assess whether the hepatic lipase (HL) gene was linked to LDL size, Allayee et al.151 conducted sib-pairs analyses among the FCHL Dutch families using two microsatellite markers located near the HL gene (D15S643 and D15S148). In the quantitative analysis (LDL-PPD), both markers yielded evidence of linkage and in the qualitative analysis (pattern A/B) only marker D15S643 reached the level of significance. Finally, two other studies excluded the hypothesis of linkage with the HL locus146,156. The cholesteryl ester transfer protein (CETP) mediates the transfer of cholesterol ester from HDL to apoB-containing lipoproteins in exchange for triglyceride and thus constitutes an excellent candidate gene. Three independent studies using all sib-pairs linkage analysis have shown consistent evidence of linkage for LDL-PPD at this locus148,149,152. It should be noted however that the lecithin:cholesterol acyl transferase (LCAT) gene, which is responsible for the esterification of free cholesterol within HDL particles, is located nearby (≈ 10 Mb) the CETP locus and might be responsible for the signal. The APOAI-CIII-AIV gene cluster is also an interesting genetic locus potentially affecting LDL size. Rotter et al.148 originally suggested (p = 0.06) linkage to this locus with LDL-PPD. A subsequent study was unable to confirm this linkage with the quantitative phenotype, but did so with the qualitative phenotype149. However, two other studies rejected the hypothesis of linkage to the APOCIII locus146,156. Finally, the manganese superoxide dismutase (SOD2) gene was also linked to LDL size more than once. Although the influence of this candidate gene on plasma lipoproteins is less obvious, it was linked to LDL-PPD148 and the atherogenic lipoprotein pattern A/B149. However, a subsequent study provided significant evidence against linkage (LOD = -4.52) to the SOD2 locus with phenotype A/B156. No evidence of linkage was demonstrated for the other candidate genes tested with LDL particle phenotypes, including APOAII148,156, APOE-CII-CI gene cluster146,148,156, high-density lipoprotein binding protein (HDLBP)148, hormone sensitive lipase (HSL)146, insulin receptor (INSR)146,156, apo(a) (LPA)148 and microsomal triglyceride transfer protein (MTP)146,156.

Taken together, linkage studies based on the candidate gene approach have provided positive but mainly inconsistent results. Based on these observations, Austin et al.156 emphasized the necessity of finding new genetic loci, other than those harboring known candidate genes, to identify genes potentially involved in determining the small dense LDL phenotype. Genome-wide scans are particularly suited for this purpose. To date, two genome-wide linkage scans have been reported in the literature for LDL-PPD. Results of these genome-wide searches are indicated in bold in Table 6 and are illustrated in the Figure 10. The first whole-genome scan on LDL-PPD was performed on 240 individuals ascertained through 18 unrelated FCHL probands151. Results suggested a locus located approximately 12 Mb from the HL gene on chromosome 15 with a LOD score of 2.2. Suggestive linkage (LOD = 1.6) was also observed for a marker located on chromosome 19q13 which contains the APOE-CII-CI gene cluster. The second genome scan on LDL-PPD was based on 140 subjects from 26 familial hypertriglyceridemia families participating in the Genetic Epidemiology of Hypertriglyceridemia Study131. For the whole-genome scan, only one chromosomal region provided possible evidence of linkage on chromosome 6q (LOD = 2.1). When the LDL-PPD was adjusted for other lipoprotein covariates, the LOD score decreased slightly but the location of the peak remained unchanged, suggesting that the signal is independent of other lipoprotein levels. The SOD2 and LPA genes are located within the one-LOD score support interval. An additional genome scan on cholesterol concentrations within LDL size subfractions is also worth mentioning. Rainwater et al.147, found two QTLs on chromosome 3 and 4 with LOD score above 3 for LDL size 3 (LDL-3) a subfraction that contains small LDL particles. Suggestive linkage was also observed on 3p26-p25 and 6q24-q27 for LDL-3, 19p13-q12 for LDL-1 (a subfraction that contains large particles) and 19q13 for LDL-2 (a subfraction that contains particles with intermediate diameter). This study evaluated LDL size-related phenotypes, but QTLs identified are those affecting the cholesterol concentration within a particular subpopulation of LDL and do not correspond to QTLs affecting the size of the particles.

This genome scan and the two others on LDL-PPD have generated new leads in finding genes involved in LDL particle heterogeneity. Interestingly, these QTLs harbor a good number of candidate genes that have not been tested previously in linkage and association studies. Among these genome scans, only the two suggestive loci observed for LDL-PPD131 and LDL-3147 on chromosome 6q have shown replication (Figure 10). This locus contains the SOD2 gene which has been linked to the phenotype before148,149 (see Linkage studies). However, the number of loci identified by genome-wide scans clearly reveals the complex genetic architecture underlying LDL particle characteristics.

Figure 10. Ideogram of human karyotype showing chromosomal locations of genes and QTLs potentially involved in LDL size/density identified from various lines of evidence. Only positive findings are included in the figure (see text for the complete results). Red and purple lines indicate QTLs identified by genome-wide linkage scans in human and animal, respectively. Genes linked (green rectangle), associated (yellow rectangle) or both linked and associated (blue rectangle) to LDL particle characteristics are shown. Genes and QTLs are placed on the hybrid map showing the sequence and the cytogenetic locations. Information to construct the ideogram has been obtained from the UCSC Genome Browser (http://genome.ucsc.edu). The alternated black and white colors on the chromosomes have been used to distinguish a cytogenic band from the adjacent ones and do not correspond to the band colors observed on Giemsa-stained chromosomes. ABCA1, ATP-binding cassette, sub-family A, member 1; ADRB3, β3-adrenergic receptor; APO, apolipoprotein; CYBA, p22 phox; CETP, cholesteryl ester transfer protein; LDLR, low-density lipoprotein receptor; LIPC, hepatic lipase; LCAT, lecithin:cholesterol acyltransferase; LDLR, low-density lipoprotein receptor; LPL, lipoprotein lipase; MTP, microsomal triglyceride transfer protein; SOD2, manganese superoxide dismutase; SRB1, scavenger receptor class B type 1.

A considerable number of association studies have been conducted to identify the genes influencing LDL particle characteristics. Table 7 presents a summary of these studies organized by genes and ordered by chromosome number.

APOE.

The gene that encodes apoE lies on chromosome 19, and its three common alleles, ≥2, ≥3 and ≥4 code for the isoforms apoE2, apoE3, and apoE4, respectively. To the best of our knowledge, a total of nine studies have investigated the association between LDL size and apoE genotypes (Table 7). The largest among them was by far the one conducted by Schaefer et al.158 with 2258 men an women participating in the Framingham Offspring Study. In this study the age-, body mass index-, and plasma triglyceride-adjusted LDL particle type (a larger LDL type reflects smaller LDL particles) was significantly different in men with different apoE genotypes. However, the same trend was observed in men and women for higher LDL type from the ≥2 to the ≥4 subjects. The lowering effect of the ≥4 allele was confirmed in Japanese subjects159 and in men of North European descent160 showing that carriers of this allele had smaller LDL particle size than those without the ≥4 allele. Consistent with these observations, Haffner et al.161 demonstrated progressive decrease in LDL size in both men and women from apoE2/3, apoE3/3 and apoE3/4 genotypes. The same study also showed that the risk of having LDL subclass pattern B was higher for subjects carrying the apoE3/4 genotype compared to both apoE2/3 and apoE3/3 genotypes. Consistent with these observations Nikkilä et al.162 reported that LDL size was the lowest in E4/4 and increased in the order E3/4, E3/3 and E2/3. In contrast, an earlier study conducted in healthy middle aged men163 and a second one performed in children164 reported no difference in LDL particle size among the different apoE genotypes. To complicate even more the interpretation, two additional studies, one conducted with 132 subjects from a small 800 individuals island165, and the other performed in 212 subjects with or without recent onset of angina166, reported lower particle size among carriers of the ≥2 allele compared to noncarriers. In the former, the difference disappeared when data were adjusted for plasma triglyceride levels while the effect of the ≥2 allele in the later was still evident after such adjustment.

LIPC.

The human hepatic lipase (HL) gene is located on chromosome 15q and encodes for a protein that play an important role in lipoprotein metabolism. Two polymorphisms, namely -250G>A and -514C>T (also referred to as -480C>T), have been tested with LDL particle size/density. These two polymorphisms are in linkage disequilibrium196 and the rare allele is associated with lower HL activity185,186. Zambon et al.185 were the first to describe an association between the -250G>A polymorphism and LDL particle flotation rate (LDL-Rf) measured by DGU. They have shown that the less common A allele was associated with more buoyant LDL particles among normolipidemic subjects and men with CAD. This finding was then replicated in a group of premenopausal women showing more buoyant LDL particles among carriers of the T-514 allele186. However, a larger study, conducted in 2667 subjects participating in the Framingham Offspring Study, found no relationship between HL polymorphism at position -514 and the LDL particle size measured by GGE187. This lack of association between this variant and LDL size measured by GGE was also observed in a subgroup of unrelated subjects from FCHL Ducth families151 and in a cohort of healthy, middle-aged men160. The different methods used to characterize LDL particles might explain the inconsistency. However, an additional study rejected the hypothesis of association between -514C>T polymorphism and LDL particle size measured by NMR176.

CETP.

The cholesteryl ester transfer protein (CETP) gene lies on chromosome 16q an encodes a protein that facilitates the exchange of triglyceride and cholesterol between lipoproteins. The LDL particles of CETP deficient patients comprised a group of abnormal heterogeneous particles which show polydispersity on GGE with a smaller mean particle size188. The most studied RFLP in this gene, Taq1B in intron 1, was associated with CETP activity and mass. In fact, the B2 allele (absence of the Taq1 restriction site) was associated with decreased CETP activity and mass, which mimics a mild form of CETP deficiency186,189. The Framingham Offspring Study has again provided the largest population-based cohort (n = 2916) investigating this polymorphic site with LDL size189. This study reported that the B2 allele in men was associated with increased levels of large LDL subfraction whereas B1B1 homozygous subjects had increased levels of small LDL subfraction. Therefore, the B2 allele was associated with increased LDL particle size, an effect seen only in men. The effect of the B2 allele was also observed in the Columbia University BioMarker Study, but this time only in women176. In this study, women homozygous for the B2 allele had larger particles compared to carriers of the B1 allele. This difference was also observed in men, but the small number of men tested might have lacked the power to detect a significant effect. In contrast, absence of association between the Taq1B polymorphism and LDL-Rf was reported in a cohort of normolipidemic premenopausal women186. A trend toward greater LDL size with increasing number of B2 allele was observed in the VA-HIT group, but the effect did not reach statistical significance190. This lack of association was also observed in a cohort of patients with type 2 diabetes191 and in Japanese subjects192. However, a second polymorphism in the CETP gene, namely I405V, showed a significant association in this Japanese study. They demonstrated lower LDL size in patients with the VV genotype compared to carriers of the I allele. A significant effect of this polymorphism was also observed in families ascertained for exceptional longevity129. However, in this cohort subjects carrying the VV genotype had larger LDL particles. The later study also reported no association between LDL size and a third gene variant in linkage disequilibrium with the Taq1B polymorphism named -629C>A. In contrast, carriers of the -629C>A polymorphism had increased LDL-PPD compared to CETP-629C homozygotes in a cohort of healthy, middle-aged men160. Finally, a less frequent missense mutation, D442G in exon 15, in the CETP gene was investigated in patients with CAD. The presence of this mutation was associated with lower CETP concentrations and larger LDL size193.

MTP.

The microsomal triglyceride transfer protein (MTP) gene, located on chromosome 4q, encodes for a protein essential in the assembly and secretion of apoB-containing lipoproteins in hepatocytes and enterocytes. A common functional polymorphism in the promotor of the MTP gene, -493G>T, was investigated in relation to LDL particle size. Couture et al.169, showed no association between the -493G>T genotype and LDL size in 2510 subjects participating in the Framingham Offspring Study. This absence of association was also observed in a group of viscerally obese men170. The only positive association between the -493G>T variant and LDL size was observed in a small sample of type 2 diabetes Chinese171. They reported smaller LDL particle size among TT homozygotes compared to the other genotypes.

LPL.

The lipoprotein lipase (LPL) gene is located on chromosome 8p and encodes a protein that is responsible for the hydrolysis of triglyceride within apoB-containing lipoproteins. Several mutations have been identified in the LPL gene and some of them have been associated with LDL particle size. First, a missense mutation at codon 188 cause a clear reduction in LPL activity. Carriers of this defective mutation had smaller LDL size compared with noncarrier family members173. Similarly, lower LDL size was observed among carriers of the LPL Tyr302-Ter mutation in an Italian family174. Hokanson et al.150 confirmed the LDL reducing size effect of LPL deficiency in five families with structural mutations in the LPL gene. Subsequently, the Ser447-Ter mutation in exon 9 was associated with larger LDL size160,175. This mutation cause a premature termination codon which surprisingly increases the enzymatic activity of LPL160,197. These observations suggested that the mutation that decreases LPL activity cause a reduction in LDL size and the mutations that increase LPL activity increase LDL size. This hypothesis was confirmed in a cohort of 206 heterozygote subjects carrying of either the null P207L or the defective D9N mutations177. In this study, LDL particle size was smaller in the P207L carriers than in the D9N subjects, suggesting that a greater reduction in LPL activity results in smaller LDL particle size. However, this relation between LPL genetic variants, LPL activity and LDL size is not without controversy given that the greater LDL particle size observed among carriers of the Ser447-Ter mutation was not confirmed in the Columbia University BioMarkers Study176.

APOA1/C3/A4/A5 cluster.

The APOA1/C3/A4/A5 cluster lies on chromosome 11q and encodes four proteins involved in lipoprotein metabolism. Genetic variations within individual gene have been associated with LDL particle characteristics. Russo et al.182 tested the SstI polymorphism on the 3’ untranslated region of the APOC3 gene and showed that male carriers of the S2 allele had significantly lower concentrations of large LDL and a significant reduction in LDL size. In women, there was no significant effect on LDL size. The other polymorphisms tested in the APOC3 gene (-455T>C, -625T>del and C3238G) showed no association with LDL particle size176,183. However, the SacI and the -625T>del polymorphisms located in the 3’ untranslated and the promotor regions of the APOC3 gene, respectively, were significantly associated with LDL charge168. In contrast, the same study reported no association between APOA1 restriction sites (PstI and MspI) and LDL charge. Recently, Austin et al.181 demonstrated, with several analytic approaches, that common SNP variants in the APOA5 gene are associated with LDL particle size in a community-based sample of Japanese American families. This study particularly pinpoints the -3A>G variant to decreased LDL size. However, considering the close proximity of the four genes in the cluster, it is difficult to infer with certainty that the effect observed with one of them is mediated by the gene tested. Indeed, a positive finding in one gene might be due to linkage disequilibrium of the tested marker with a polymorphism in a second gene within the cluster. Accordingly, further studies in the APOA1/C3/A4/A5 gene cluster will be required to identify the functional site.

Other candidate genes: APOB, CYP7, ACE, ADRB3, CYBA, FATP1, SR-BI, LDLR and ABCA1.

One study verified the effect of the APOB EcoRI polymorphism in a group of Caucasian men and showed no effect on either LDL-PPD or LDL score167. However, five polymorphisms in the APOB gene were found to influence LDL charge heterogeneity evaluated by relative electrophoretic mobility168. A common A to C substitution at position -204 in the promoter of the cholesterol 7α-hydrolase (CYP7) gene showed no association with LDL particle size178. The hypothesis that the angiotensin-converting enzyme (ACE) gene insertion/deletion polymorphism was associated with LDL size was also rejected in a small Japanese cohort159. However, a recent paper suggested a positive association between the Trp64Arg variant in the β3-adrenergic receptor (ADRB3) and LDL-PPD172. The authors reported that the LDL particle size was smaller in the subjects with the Arg64 allele than those without the Arg64 allele. The effect remained significant after triglyceride adjustment, but disappeared after adjustment for body mass index or parameters of insulin resistance. The phox 22 gene (CYBA), which is a small subunit of vascular NAD(P)H oxidase playing an important role in superoxide production, was also investigated in a group of healthy Japanese subjects194. A trend (p = 0.08) toward larger LDL-PPD was observed among carriers of the C242T polymorphism compared to noncarriers. In addition, the proportion of subjects with pattern B was significantly larger in the CC group than CT/TT group. No association was observed between a functional intronic variation in the fatty acid transport protein-1 (FATP1) gene and LDL-PPD in a cohort of healthy Swedish men195. However, the cholesterol concentration ratio of the largest and smallest LDL subfractions (LDL-I/LDL-III ratio) were found to be different between FATP1 intron 8 genotypes. No clear association was observed between three SNPs located within the scavenger receptor class B type 1 (SR-B1) gene and LDL size in the Framingham Study184. However, this study showed reduced LDL particle size in carriers of the A allele at the SR-BI exon 1 gene in the subgroup of diabetic subjects. Finally, no study investigated the effect of common polymorphisms in the LDLR gene on LDL particle characteristics. However, earlier studies showed that the LDL particles of patients with familial hypercholesterolemia are characterised by higher peak flotation rate and lower density198,199. Similarly, one patient with the Tangier disease was shown to have smaller particle size compared to control subjects highlighting the possible implication of the ATP-binding cassette 1 (ABCA1) gene179. However, the reducing LDL size effect of a defective mutation in the ABCA1 gene was not reported in a group of heterozygous subjects180.

The metabolic syndrome is characterized by a cluster of CAD risk factors including hypertension, upper-body obesity, glucose intolerance and the atherogenic lipoprotein phenotype which consists of elevated plasma triglyceride levels, low plasma levels of HDL-C and a predominance of small, dense LDL200. The strong association between the small, dense LDL phenotype and the atherogenic lipoprotein profile raises the question whether the gene proposed by complex segregation analyses is also responsible for the associated lipid and lipoprotein levels. Using factor analysis, Edwards et al.201,202 investigated the clustering of risk factors in the Kaiser Permanente Women Twins Study by examining the correlation structure among the components of the metabolic syndrome. Factor analysis reduced 10 correlated risk factors to 3 uncorrelated factors, each reflecting a different aspect of the metabolic syndrome. One of the factor was considered the lipid factor due to the strong factor loading for the lipid variables including triglyceride, HDL-C and LDL-PPD. Heritability estimates for the lipid factor was calculated using various approaches and ranges from 0.25 to 0.32. Thus the authors suggested that approximately a quarter to a third of the variance in this composite lipid factor may be attributable to genetic influences. Using a candidate gene strategy, the same research group subsequently found a strong evidence of linkage between the lipid factor and the CETP gene203. The authors proposed that the CETP gene variations influence the covariation in LDL size, triglyceride and HDL-C levels, and may account for a portion of the phenotypic correlation between these risk factors.

To investigate the interrelationship between LDL particle size, triglyceride and HDL-C levels, Edwards et al.128 reported genetic correlations between pairs of traits. The genetic correlation between LDL-PPD and triglyceride was -0.87, suggesting that 76% [ρG 2 = (-0.87)2 = 0.76] of the additive genetic variance in LDL size is shared with triglyceride. The genetic correlation between LDL-PPD and HDL-C was more modest (0.65) but suggested that nearly 50% of the additive genetic variance in each of these traits is due to shared genes. However, based on the likelihood-ratio test, the hypothesis of complete pleiotropy was rejected for the two genetic correlations, suggesting the existence of unique genes for each trait. These results demonstrated that the observed phenotypic associations between these three traits are largely under genetic control and indicated that searching for genes implicated in LDL size may actually mean searching for genes also involved in triglyceride and HDL-C. A similar study conducted by Rainwater et al.130 reported a genetic correlation between lipoprotein size traits (ΔLDL and ΔHDL) and triglyceride. Triglyceride and ΔHDL were strongly correlated with ΔLDL, with genetic correlations of -0.76 and 0.56, respectively. Thus, shared genes accounted for 58% and 31% of the genetic variance in each pair of traits.

Small, dense LDL is also metabolically associated with elevated plasma apoB levels and both features are found in patients with FCHL115. Thus, some investigators searched for a common genetic mechanism between these two traits in families characterized by FCHL. Using bivariate segregation analysis, Juo et al.204 reported the evidence of a common genetic mechanism controlling both apoB levels and the distribution of LDL subfraction (parameter K) in FCHL families. The best-fitting model proposed a common gene with codominant allele for both traits, plus distinct polygenic component for each trait. This major gene explained 37% and 23% of the variance in parameter K and in apoB levels. On the other hand, Jarvik et al.205 have shown that LDL subclass phenotype B and apoB levels are two traits influenced by two mendelian locus independent of each other and modulating the risk of FCHL. This conclusion was drawn by showing: 1- that the major gene effect seen in segregation analysis for apoB levels remained after adjustment for LDL subclass phenotypes and 2- by showing lack of association between LDL subclass phenotype and the apoB level predicted genotypes in contingency analysis. Finally, using commingling analysis, Austin et al.206 reported bimodality of apoB levels in individuals with LDL subclass phenotype B. This finding suggested distinct genetic mechanisms for LDL subclass phenotype and apoB levels in FCHL families. The conflicting results between these studies may due to the different statistical strategies employed or may simply reflect the complexity of the genetic mechanisms for these traits.

Taken all together, it appears that distinct sets of genes influence LDL size: those that influence LDL size independent of triglyceride and other lipid parameters and those that affect several components of the lipid profile. Thus, in addition to the genes uniquely influencing LDL size, there appear to be genetic factors that are responsible for covariation in lipoprotein/lipid traits, which demonstrate the complexity of characterizing genetic influences on LDL size.

Few, but relevant studies on animal models have confirmed the presence of genetic factors influencing LDL size. First, LDL size vary substantially between different strains of mice, showing the effect of the genetic background. Jiao et al.207 characterized LDL size by liquid chromatography in 10 inbred strains and observed a LDL size range starting at 24.16 nm in BALB/c strain to 29.39 nm in SWR strain, with the whole spectrum of size within this interval for the other strains. In an attempt to test whether LDL size was an inherited trait in mice, three sets of recombinant inbred strains were produced by crossing strains with different LDL size. By this mean, authors have shown that LDL size of recombinant inbred strains segregated to one or another progenitor 88% of the time, implying that LDL size may be controlled by the product of a major gene. Attempts to identify the major LDL-size determining gene yielded only marginal significant results for a RFLP analysis in the APOB gene.

An attempt was also made to establish whether genes control variation in LDL size in baboons208. A 150 baboons members of 19 sire groups were investigated. Baboons were fed three diets contrasting in levels of fat and cholesterol. A multifactor ANOVA revealed that 18.3% of the variation in LDL size was explained by the sire groups. In addition, there was a significant sire×diet interaction on the phenotype, indicating that members of different sire groups responded differently to various dietary compositions. Taken together, these results suggested that genes influence LDL size and the patterns of LDL response to different diets in baboons. Recently, a genome-wide linkage scan was performed among an enlarged group of these baboons to localize the genes that control LDL size fractions209. Using GGE, four LDL size-related phenotypes were constructed based on fractional absorbance in four intervals of LDL (LDL4, 24-26 nm; LDL3, 26-27 nm; LDL2, 27-28 nm; and LDL1, 28-30 nm). The LDL median diameter was also estimated, which is a diameter where half the LDL absorbance is on larger and half is on smaller particles. Genome scans were performed on LDL size-related phenotypes taken from blood samples collected at the end of each experimental diets. On a high-cholesterol high-fat diet, a significant evidence of linkage (LOD = 4.22) for LDL2 was observed on the baboon homologue of human chromosome 20 and 22 (Figure). Two additional QTLs were suggested, one on the baboon homologue of human chromosome 16 for LDL3 when exposed to a low-cholesterol low-fat diet (LOD = 2.15), and one on the baboon homologue of human chromosome 5 for LDL3 when exposed to a low-cholesterol high-fat diet (LOD = 2.67). The later QTL is particularly relevant since the signal was also observed for the LDL median diameter (LOD = 2.21).

These results have clearly shown the usefulness of animal studies to identify the LDL size genes. Due to our ability of controlling tightly the animals environment, these studies might prove to be even more relevant in the future for testing gene-environment interactions.

Relatively little is known about gene-gene and gene-environment interactions in LDL particle characteristics, but it would be surprising if they were not important. A preliminary study has shown that the LDL bands of monozygotic twins were more concordant than dizygotic twins before but not after a 22 weeks exercise program, suggesting that the genetic contribution of LDL subfractions decreases with exercise210. It was also demonstrated that the LDL size response to a low-fat diet in children was predicted by the parental LDL subclass pattern211. Tentative evidence of interactions with LDL size phenotypes were also reported for specific loci. A significant interaction was observed between SR-BI exon 1 genotypes and type 2 diabetes on LDL size, indicating that diabetes status modifies the effect of this polymorphism on LDL particle size184. St-Pierre et al.170, for example, have shown an inverse effect of the MTP -493G>T genotypes according to visceral adipose tissue and fasting insulin. It is also apparent from association studies (Table 4) that the effect of some loci are sex-specific or reserved to subgroup of the population (diabetic for example). Zambon et al. 212 also reported an interesting pharmacogenetic interaction on LDL density. They showed that the -514C>T polymorphism in the HL gene promoter strongly influences the LDL flotation rate response in middle-aged men undergoing intensive lipid-lowering therapy. Although these studies are interesting examples, they demonstrate the high number of interactions that could be tested and the difficulty to do so in humans. Clearly, when the loci controlling small LDL will be mapped, there will be a greater potential for determining the gene-gene and gene-environment interaction effects.

As seen in the previous sections, the metabolic syndrome and its individual components are under genetic influences. However, the progress made in the search for single genes and QTLs associated with phenotypes related to the metabolic syndrome has been slow and difficult so far. Although a great deal of literature exists in the field, the overall picture is ambiguous and more research is clearly needed.

In the following chapters, we used a combination of measured (bottom-up) and unmeasured (top-down) genotype approaches to uncover the genetic architecture underlying the metabolic syndrome and its individual components. We believe that genetic dissection of complex traits requires multiple approaches in order to achieve our goals. In the first four chapters, we used a candidate gene approach. The genes were chosen based on their biological relevance with the metabolic syndrome. It is also well known that complex traits arise from interactions between multiple genes and environments, but not much has been done to date. Accordingly, in chapters 1 and 3 we also integrated the concept of pharmacogenetics and gene-gene interactions. In the next two chapters (5 and 6) we used a genome-wide search approach to identify novels or replicate previous QTLs acting on the variability of serum lipid, lipoprotein and apolipoprotein levels. This attempt is made to generate useful leads of positional candidate lipoprotein/lipid genes that will need to be tested in future studies. Chapters 7 to 9 focus on the genetic of LDL peak particle size. As shown in the previous sections, this component of the metabolic syndrome is an independent cardiovascular risk factor for which the genetic basis has just begun to be uncovered. From the preceding reports, its becoming clear that the small, dense LDL phenotype is under genetic influences. However, the specific genes remained to be identified. In this series of chapters, we used the traditional steps to understand the genetic basis of a quantitative phenotype (LDL-PPD) as presented in Figure 8. These steps include familial aggregation, heritability and segregation analyses as well as genome-wide linkage scan and association studies on positional candidate genes. Finally, the last chapter deals with the metabolic syndrome as a whole entity. The goal is to find the genetic loci contributing to the cluster of the metabolic syndrome-related phenotypes. The approach used might identify pleiotropic genes acting on several features of the metabolic syndrome or genes explaining the common variance of these clustering risk factors.

Genetics factors are involved in the development of the metabolic syndrome and its individual components.