Conclusion

The prevalence of the metabolic syndrome has risen tremendously over the past decades7. People with this syndrome are at increased risk to develop diabetes mellitus17 and CVD19 as well as at increased mortality from CVD and all causes213. Understanding the genetic contribution and the genetic determinants of this condition is particularly relevant considering its progressive economic burden on medical health care. However, this task is particularly challenging for geneticists knowing that multiple genes and environmental factors as well as their interactions contribute to the expression of the metabolic syndrome. In this thesis, we have attempted to shed some light on this question using both candidate gene and genome-wide scan approaches on the metabolic syndrome and its individual components. A total of four candidate genes have been investigated and nine phenotypes underwent genome-wide searches. Throughout this work, some polymorphisms located within the candidate genes have been associated with specific phenotypes and many genomic regions containing susceptibility genes have been identified.

In chapters 1 through 3, we used a candidate gene approach to investigate the effect of a common genetic variation, L162V, in the PPARα gene. PPARα is responsible for the translation of nutritional and metabolic stimuli into changes in gene expression214 and is thus an excellent candidate gene in the context of the metabolic syndrome. In addition, PPARα mediates the effects of fibrates, a class of drug recognized to regulate lipoprotein-lipid metabolism215. In chapter 1, we have documented a part of the interindividual variation in the response to fibrate therapy depending on the presence or the absence of the PPARα L162V mutation. After the 6-month intervention therapy with gemfibrozil, carriers of the V162 allele had a greater increased in HDL2-cholesterol compared to non-carriers. These results and others216 suggest potentially greater benefits of fibrate treatment among individuals carrying the PPARα V162 allele. In chapter 2, we have reported that the same mutation was associated with lower values of adiposity phenotypes in participants of QFS. This observation makes biological sense considering the functional differences between the leucine- and the valine-containing protein products216,217. Rodent experiments have demonstrated that deactivation of PPARα increases body fatness218, whereas its activation by fibrate treatment prevents weight gain and reduces adipose tissue219,220. Accordingly, it is possible that the reduced adiposity values observed in subjects carrying the PPARα V162 allele are explained by a greater activity of the valine-containing protein. In chapter 3, it has been shown that the PPARα L162V mutation acts individually and in interaction with the PPARγ P12A mutation to modulate glucose/insulin parameters following an oral glucose tolerance test. This chapter has demonstrated that genetic polymorphisms in candidate genes encoding proteins with overlapping functions can interact and account for a significant contribution to the final manifestation of the trait. Chapters 1, 2 and 3 constitute a perfect example of the complex genetic architecture underlying components of the metabolic syndrome. Indeed, taken together, these chapters suggest that the same mutation acts individually, interacts with other genetic variants and also influences the response to a treatment. The multifaceted effects of a single mutation demonstrate again the complex puzzle that geneticists must face.

In chapter 4, we have investigated the association between the PLTP gene and obesity-related phenotypes. This study was motivated by genome-wide scans in human and mouse pointing to the PLTP locus as a candidate region for obesity221-223. In addition, important functions governing lipid metabolism have been ascribed to PLTP224,225. Two intronic SNPs were genotyped and tested for their association with several indices of adiposity in QFS. Both single locus and haplotype association tests for family-based study revealed significant associations. Accordingly, this thesis reports two genes associated with obesity, namely PLTP and PPARα. Of course, independent replications for both of these genes will be required to confirm these significant associations.

Chapters 5 and 6 concern genome-wide scans performed on plasma lipid and lipoprotein concentrations. These two studies revealed the existence of multiple loci influencing blood lipids and lipoproteins. Indeed, evidence of linkage has been found on chromosome regions 1q43, 11q13-q24, 15q26.1, and 19q13.32 for LDL-cholesterol, 12q14.1 for HDL-cholesterol, 2p14, 11p13, and 11q24.1 for triglycerides, 18q21.32 for LDL-apoB, and 3p25.2 for apoAI. Some of these regions have been linked to lipid-related traits before, whereas others represent new findings. Other studies will be required to identify the causal genes within these regions. Chapter 6 also summarizes the loci providing evidence of linkage from all the previous published genome-wide scans carried out on blood lipid-related phenotypes. This exercise has been undertaken to make an update in the field and help investigators in positioning new findings without having to digest the heavy literature anew. A cumulative of 152 linkage signals have been gathered in this exercise. Although it may ease the interpretation of the next genome-wide scans on blood lipid-related phenotypes, displaying all these loci on the same map also revealed that a large portion of the genome is now covered with a least suggestive evidence of linkage. With this work, we were unable to achieve a coherent and comprehensive picture of the loci contributing to blood lipids and lipoproteins. This, again, demonstrates the difficulty of finding genes influencing complex traits.

The introduction synthesizes the accumulating evidence of the complex genetic etiology underlying LDL particle heterogeneity. Genetic epidemiology studies have clearly shown a genetic contribution to the LDL subclass phenotypes. The results from chapter 7 have confirmed this finding by showing high familial resemblance for LDL-PPD in 681 individuals participating in QFS226. Indeed, an ANOVA comparing between- versus within-family variance indicated that there was about two times more variance between families than within families. Thus, results from the QFS suggested that the family lines accounted for close the 50% (47-49% depending on covariates adjustment) of the variance in LDL-PPD phenotype. In addition, the pattern of familial correlations revealed no spouse correlation but significant parent-offspring and sibling correlations for the LDL-PPD phenotypes, suggesting that genetic factors are the major determinants of the familial aggregation.

Heritability studies from previous reports (see Table 4) have shown that at least 30% to 60% of the variation in LDL size is attributable to genetic factors. Heritability estimates for LDL-PPD in the QFS study fall within this range226. In chapter 7, three LDL-PPD phenotypes based on three different adjustment procedures have been constructed: LDL-PPD1 adjusted for age, LDL-PPD2 adjusted for age and BMI, and LDL-PPD3 adjusted for age, BMI and triglyceride levels. Heritability estimates for the three phenotypes were 58.8, 58.4 and 52.0%, respectively. The high heritabilities obtained may be explained by the design of the study. Indeed, in this case, heritability is defined as the proportion of variance due to additive familial effects, including both genetic and nongenetic sources of variance. Although, the pattern of familial correlations in the QFS study suggested that the familial resemblance is mostly attributable to genetic factors, heritability estimates derived from this cohort may be considered as upper bound estimates for LDL-PPD.

Complex segregation analyses have consistently demonstrated the existence of a single gene with major effect (see Table 5). Again this finding has been confirmed in the QFS (Chapter 9). All hypotheses of no familial resemblance, no major effect, and no multifactorial effect have been clearly rejected in that study, suggesting that both the major and the multifactorial effects were significant. Tests on the transmission probabilities have been also carried out, and the environmental hypothesis (equal τ‘s) has been rejected whereas the Mendelian τ‘s was not. The putative gene accounted for 24%, 24% and 52% of the phenotypic variance of the age-adjusted, age-BMI-adjusted and age-BMI-triglyceride-adjusted LDL-PPD, respectively. In addition, another 22-34% of the variance was attributable to residual polygenic and familial environmental factors.

Taken together, it seems clear from a genetic epidemiology perspective that LDL size is under the influence of genetic factors. The results obtained from the QFS have simply reinforced this fact by demonstrating: 1-high familial aggregation, 2-significant heritability, and 3-the existence of a major gene effect. This consistency observed between studies has clearly stimulated the search for the causal genetic variants.

However, searching the DNA-based variations responsible has proved to be a difficult task owing to inconsistency and lack of replications among studies (see Tables 6 and 7). Indeed, linkage and association studies with candidate genes have produced some of the expected results, but in general the effect of positive hits does not seem to be uniform in all populations and environmental backgrounds. Genome-wide linkage scans have been undertaken to fill the gap and have produced interesting leads that need to be followed-up. In chapter 8, a genome-wide scan has been carried out in 681 subjects from 236 nuclear families participating in the QFS227. The strongest evidence of linkage was found on chromosome 17q23, with a LOD score of 6.76 for the phenotype adjusted for age, body mass index and triglyceride levels. Other chromosomal regions provided LOD > 2.0, including 1p33-p31, 2q33-q36, 4p15-q13, 5q13-q14 and 14q23-q32. Thus, this genome scan gives strong evidence for the presence of a major quantitative trait locus (QTL) located on 17q, but also demonstrated the multilocus nature of LDL size.

The APOH gene is a particularly interesting candidate gene in the 17q area. In chapter 9, the promoter, the exons ant the exon-intron splicing boundaries have been sequenced in subjects of the QFS cohort. Five genetic variations have been identified, including three missenses mutations. The entire cohort has been then genotyped for genetic association testing. An haplotype family-based association test revealed the existence of an haplotype significantly associated with greater LDL size. This result suggests that the APOH gene is responsible for the genome-wide linkage signal observed on chromosome 17q. However, considering the limitations of association studies (see introduction), independent replication of this finding is essential before reaching conclusions.

Figure 11 summarizes the contribution of the present work in identifying the genes responsible for the large and consistent genetic influences observed on LDL size. Although the genome-wide linkage scan performed in the QFS cohort has been very fruitful in finding quantitative trait loci (QTL) for LDL size, the rate of replication with the previous published scans130,131,151,209 is low. Indeed, the only evidence of replication is observed on chromosome 5 with a QTL observed for LDL median diameter of baboon exposed to a low-cholesterol, high-fat diet209. This genomic region contains the HMG CoA reductase gene which constitutes an interesting candidate gene to test in the near future.

It is becoming obvious that several different genetic loci contributed to the expression of small, dense LDL. This observation suggests that different genetically determined metabolic mechanisms may give rise to the phenotype. For most of the loci identified so far, it is unclear whether the effect is direct or mediated through the interrelationship with other metabolic parameters such as glucose/insulin homeostasis and triglyceride metabolism. The number of false positives reported is difficult to assess but may be important due to publication bias toward positive findings. Accordingly, this summary should be interpreted with caution and awareness since some of the positive loci may eventually prove to be false positives.

Understanding the genetic etiology of small, dense LDL will help to elucidate the complex multifactorial networks involved in the progression of atherosclerosis and its ultimate consequence—CHD. Although searching the genes has been and continues to be a demanding adventure, the challenge may still be ahead in order to identify the combination of genes and environmental circumstances predisposing to small, dense LDL. It should be emphasized, however, that the nongenetic factors influencing the expression of small, dense LDL can be taken to our advantage by treating genetically susceptible individuals with appropriate lifestyle modifications.

Figure 11. Ideogram of human karyotype showing chromosomal locations of genes and QTLs potentially involved in LDL size/density. Results from QFS are in red, while all the other results presented in Figure 10 are in gray. 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.

Finally, chapter 10 was an attempt to summarize this thesis by identifying chromosomal regions harbouring genes contributing to the clustering of the metabolic syndrome-related phenotypes. Factor analysis has been used for that purpose to create a quantitative metabolic syndrome variable representing the common variance among the individual components of the syndrome. Factor analysis has revealed one underlying factor (the metabolic syndrome factor) with high loading for all the metabolic syndrome-related phenotypes, suggesting that common factors account for the observed risk variables clustering. A genome-wide scan on the metabolic syndrome factor revealed the existence of a QTL on chromosome 15q indicating the presence of a gene located in the area contributing to the shared variance among components of the metabolic syndrome. Again, further studies will be required to locate the causal gene.

Many questions arise and have remained unsolved throughout this work. For example, many chromosomal regions containing genes affecting components of the metabolic syndrome have been identified. However, the ultimate goal of QTL mapping is to identify the genes underlying these polygenic traits and to gain a better understanding of them. Except for the 17q region identified for LDL-PPD, no attempt was made in this work to locate the genes causing the linkage signals. In addition, all the association studies reported in this thesis require replication before reaching conclusions. The limitations of association studies are highlighted in the introduction and leave some uncertainties about the results presented. Accordingly, a substantial amount of work has emerged from this thesis and hopefully it will generate a lot of follow-up studies.

The genetic dissection of the metabolic syndrome is a tremendous challenge. The present thesis shed some lights on different aspects of the genetics of the metabolic syndrome, but above all underscores the difficulty of the task. Without new development in finding genes involved in complex human diseases, a long adventure is anticipated before reaching the finish line. However, the recent years have witnessed the development of novel methods and strategies for the genetic dissection of complex human diseases. These emerging new methods and ideas are clearly welcomed to tackle the challenge and fulfill the promise hold by the field of genetics, that is better understanding the pathogenesis of complex diseases and consequently improve prevention strategies, diagnostic tools and therapies.