Human Molecular Genetics, 2002, Vol. 11, No. 17 2015-2023
© 2002 Oxford University Press
Specific haplotypes of the P-selectin gene are associated with myocardial infarction
1INSERM U525 and 2INSERM U436, Faculté de Médecine, Hôpital Pitié-Salpêtrière, 91 Bld de l'Hôpital, 75634 Paris, France
Received April 22, 2002; Revised May 27, 2002; Accepted June 24, 2002
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIAL AND METHODSREFERENCES |
|---|
P-selectin is a cellular adhesion molecule that may be involved in the development of atherosclerosis and its complications. We have previously identified thirteen polymorphisms of the P-selectin gene among which five were located in the coding region of the gene (S290N, N562D, V599L, T715P, T741T (A/G)). These polymorphisms were tested individually for association with myocardial infarction (MI) and only the T715P polymorphism was shown to be associated with MI. We here extend this work by performing a haplotype analysis which enables us to assess the consequences on the phenotype of the co-presence of several variants on the same chromosome. For this purpose, a new maximum likelihood method was developed for estimating simultaneously haplotype frequencies and haplotype–phenotype effects. While haplotypes defined by the polymorphisms located in the promoter region of the gene were unrelated to MI, those defined by the polymorphisms in the coding region were globally associated with MI in a sample of 582 cases and 630 controls from the Etude Cas-Témoin sur l'Infarctus du Myocarde. Detailed haplotype analysis confirmed the protective effect of the P715 allele but additionally revealed that the presence of two asparagine codons at sites S290N and N562D was associated with a higher risk of MI, consistenly in France and Northern Ireland, but only when they were carried by the same haplotype. This finding illustrates the complexity of the relationship between gene variability and disease and the necessity to explore in detail the polymorphisms of candidate genes.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIAL AND METHODSREFERENCES |
|---|
P-selectin is a cellular adhesion molecule belonging to the lectin family which is mainly expressed by platelets and endothelial cells and plays a major role in the initial phases of leucocytes adhesion to the endothelium and in the interaction between leukocytes and platelets (1). This protein is involved in the recruitment of leukocytes on the activated vessel wall during inflammation and is suspected to play an important role in the early stages of atherosclerosis and in its complications. Studies in atherosclerotic prone ApoE-/- mice have shown that inactivation of the P-selectin gene was associated with a markedly reduced atherosclerotic plaque area and monocyte recruitment to sites of neointima formation after carotid artery injury (2). In addition, elevated levels of circulating P-selectin have been observed in patients affected with various cardiovascular disorders including coronary artery disease (3–5). The putative contribution of the leucocyte–endothelium adhesion cascade to atherosclerosis in humans would be greatly reinforced if genetic variants of components of this system demonstrated an association with clinical complication of atherosclerosis.
We recently performed a molecular screening of the P-selectin gene which led to the identification of several polymorphisms (6). Among them, four were non-synonymous and one (T715P) was associated with myocardial infarction (MI) in the ECTIM Study (Etude Cas-Témoin sur l'Infarctus du Myocarde), a case-control study of myocardial infarction conducted in Northern Ireland (Belfast) and France (6). The results were obtained using the genotypic information at each site. This approach is not fully efficient because it does not use the whole genotypic information available within a gene. Testing associations with several polymorphisms simultaneously, by regression analysis for example, may reveal genetic influences that are undetectable by univariate analysis (7). However, except in simple situations, this strategy is unable to account for linkage disequilibrium (LD) among the polymorphisms and cannot estimate the consequence on the phenotype of the co-presence of several variants on the same chromosome (a haplotype). As haplotypes may have a particular significance with regard to functionality or as markers for unknown functional variants, it appears more and more evident that, in order to better characterize the role of a candidate gene, the full haplotypic information should be exploited (8,9).
In the absence of family data, haplotypes are not readily deducible from genotypes, except for individuals who are homozygous or heterozygous at only one site. Molecular determination of haplotypes is sometimes possible but faces many limitations. For example, PCR-based techniques can only be used for molecular haplotyping of polymorphisms located in relatively close proximity (
3 kb). Alternative approaches have been developped (10,11), however, they are time-consuming, expensive, require access to live cells and are therefore not applicable to large studies. To circumvent the difficulty of molecular haplotyping, several methods based on statistical inference (12–16) have been proposed to estimate haplotype frequencies from genotypic data. These techniques are generally used for comparison of haplotypes frequencies between groups (e.g. cases and controls) and they provide a global test of significance for the comparison of all haplotypes simultaneously. If a difference is observed, no haplotype–disease association parameters are estimated, this precludes the identification of the responsible haplotype(s), and does not allow specific hypotheses such as interaction between two or more loci to be tested. Moreover including covariates and testing for haplotypexcovariate interactions, which are a common requisite when dealing with complex diseases, is not possible by these techniques.
We present here a new strategy for investigating haplotype–phenotype associations. The maximum likelihood (ML) method of inference that we have developed is quite general, it circumvents all limitations mentioned above and can be applied to qualitatitive and quantitative traits as well. The new method has been used for a reanalysis of the P-selectin gene polymorphisms and its reliability checked by comparing its results to those derived from direct molecular haplotyping.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIAL AND METHODSREFERENCES |
|---|
Allele frequencies and pairwise linkage disequilibrium of the P-selectin gene polymorphisms
Allele frequencies of the nine studied polymorphisms are reported in Table 1 for cases and controls, separately in Belfast and in France. All genotype distributions were compatible with Hardy–Weinberg equilibrium in both populations. Among the nine polymorphisms, only T715P exhibited a significant difference in allele frequency between Belfast and France (P<0.002), the P715 allele being less frequent in France than in Belfast. This polymorphism was also significantly associated with myocardial infarction (MI). The P715 allele was less frequent in cases than in controls, in Belfast (0.10 versus 0.17; P=0.015) and in France (0.08 versus 0.10; P=0.059), (P=0.003 on pooled samples stratified by country). The pairwise LD matrix shows that two main blocks of LD are present within the P-selectin gene (Fig. 1). The 5' block, includes the promoter polymorphisms and the S290N polymorphism, whereas the 3' block includes all the other coding polymorphisms. The two blocks were in weak LD with each other. For example, the pairwise LD between the C-2123G and S290N polymorphisms was -0.95, that between S290N and N562D was 0.037 and that between N562D and T715P was -1. As a result of this LD pattern, haplotype analysis was performed in each block separately. However, the S290N polymorphism was also included in the 3' block because its co-occurrence on the same chromosome with other non-synonymous polymorphisms may be functionally significant.
|
|
Associations of P-selectin gene haplotypes with myocardial infarction
Detailed results of the haplotype analysis of the 3' block are given in Table 2. As a consequence of the strong LD between the N562D, V599L and T715P polymorphisms, only four haplotypes, –NVT–, –NVP–, –NLT– and –DVT– were inferred using these three polymorphisms. These four haplotypes were associated with both the S and N codons at the S290N locus, suggesting that recombination occurred between the S290N and the N562D polymorphisms. The G allele of the T741T polymorphism was present only on two of these haplotypes, –NVT– and –NLT–, leading to the final estimation of 10 haplotypes with a frequency
1%. These haplotypes accounted for 99.7% of all haplotypes in Belfast and 98.6% in France. Haplotype frequencies ranged from 0.01 to 0.39 and their distributions were significantly different between Belfast and France (P<10-4). In particular, the SNVPA haplotype was more frequent in Belfast than in France both in cases and in controls, (0.13 versus 0.08; P<10-4). The test of a global haplotypic association with MI was significant both in Belfast (
2=18.93 with 8 d.f.; P=0.015) and in France (2=17.02 with 9 d.f.; P=0.048). In Belfast, the SNVPA haplotype was associated with a lower risk of MI (OR=0.47; P=0.009) by comparison to the reference SNVTA haplotype which combines the most frequent alleles at each polymorphic site. This association mainly reflects the lower frequency of the P715 allele in cases than in controls. More unexpectedly, by comparison to the SNVTA haplotype, the NNVTA haplotype was associated with an increased risk of MI in Belfast (OR=2.84; P=0.044) and France (OR=2.09; P=0.011). No significant association was observed for the other haplotypes.
|
Using the inference method, it is also possible to investigate the effect of each polymorphism on different haplotypic background. The odds ratio for MI associated with each non-synonymous polymorphism was estimated according to the haplotypic background conferred by the other non-synonymous polymorphisms. For ease of presentation, results are reported here on the pooled samples after adjusting for population since estimates were quite similar in both populations. The N562 and D562 alleles being present on both the S-VT and N-VT haplotypes, the effect of the N562D polymorphism could be tested on these two different haplotypic backgrounds. When carried by the S-VT haplotype, the D562 allele was associated with an OR of 1.04 (P=0.70) while, when carried by the N-VT haplotype, the OR was 0.37 (P=0.0007). The test of homogeneity of these two ORs was significant (
2=9.24 with 1 d.f.; P=0.002). Since the S-VT and N-VT haplotypic backgrounds differ by the codon at position 290, this result suggests an interaction between the amino acids present at positions 290 and 562 of the P-selectin protein on the risk of MI or alternatively that these two polymorphisms define a haplotype that is in tight LD with an unknown functional variant. Similar analyses were performed for the three other non-synonymous polymorphisms (Fig. 2). The OR for MI associated with the N290 allele differed significantly according to haplotypic background (2=9.26 with 3 d.f.; P=0.026 for the test of homogeneity of the four ORs). In particular, the N290 allele was associated with an increased OR when it was carried by the –NVT haplotype (OR=1.92; 2=7.87 with 1 d.f.; P=0.005) while it was associated with a reduced OR when it was carried by the –DVT haplotype (OR=0.68; 2=3.63 with 1 d.f.; P=0.057). This is the symmetric way of looking at the interaction between the D562 and N290 alleles mentioned above.
|
We performed further analyses to evaluate whether the hypothesis of additivity on a logistic scale of the main haplotype effects was compatible with the data. No significant deviation from additivity was observed. In other words, the data are compatible with the combined effect of two haplotypes (a genotype) on the risk of MI being multiplicative. One should be aware however that the power of such a test is quite low.
A similar haplotype analysis was performed on the polymorphisms of the 5' block. The results of this analysis demonstrated no haplotypic association of these polymorphisms with MI so are not reported here but are available online (http://genecanvas.idf.inserm.fr/).
Checking the reliability of the statistical inference by molecular haplotyping of the non-synonymous polymorphisms of the P-selectin gene
In order to check the reliability of the results provided by the inference ML method for the four non-synonymous polymorphisms, molecular haplotyping of the P-selectin gene was realized. Amongst the 1212 studied individuals, 288 were heterozygous for at least two of the three loci N562D, V599L and T715P and were therefore selected for molecular haplotyping. Only 222 (success rate=77%) individuals were successfully haplotyped. For the 66 remaining individuals, ambiguity could not be resolved because of lack of DNA or unsuccessful assay. The S290N polymorphism was too distant from the three others to allow molecular haplotyping. However, considering that the molecular haplotypes of the N562D, V599L and T715P polymorphisms defined a multiallelic marker, haplotypes combining this marker with the S290N polymorphism could be deduced unambiguously for all individuals who were either homozygous at both sites or heterozygous at one site only. Finally, complete haplotypic information at the four non-synonymous polymorphisms deduced either from genotypic data or from molecular haplotyping was available for 937 (77%) of the 1212 initial individuals. Haplotype frequencies and haplotypic ORs estimated in this ‘haplotyped sample’ were almost identical to those estimated by the ML inference method (Table 3).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIAL AND METHODSREFERENCES |
|---|
The raw association between a polymorphism and a phenotype incorporates not only the potential effect of the polymorphism itself but also the potential effect of all polymorphisms which are in LD with it. The haplotype-based strategy proposed here provides a powerful tool to differentiate haplotype from single polymorphism effects and to test whether the effect of a polymorphism depends on the haplotypic background by which it is carried. Compared to a traditional approach based on genotypes such as that used in our former work on the CETP gene polymorphisms (7), it constitutes an important step towards a better characterization of the relationship between the variability of a gene and a phenotype. With regard to P-selectin gene polymorphisms and MI this analysis was able to detect specific associations that were undetectable by classical analysis or by a global comparison of haplotype frequencies between cases and controls.
Thirteen polymorphisms of the P-selectin gene had been identified previously by our group and tested individually for association with MI in the ECTIM study (6). Due to some redundancy or very low allele frequency, only nine of these polymorphisms were incorporated in the present haplotype analysis. Other polymorphisms of the P-selectin gene not identified by our screening have been reported (17), including four non-synonymous variants (V168M, P260L, M324V and S459F) with allele frequencies lower than 5%. We tested these polymorphisms in a sample of 190 chromosomes from individuals of European origin but failed to identify any carrier of the L260, V324 and F459 variants. Due to the multi-ethnic origin of the DNA samples used in the study reporting these polymorphisms (17), it cannot be ruled out that these three variants are absent or very rare in populations of European descent. The V168M polymorphism on the contrary was present in our sample of 190 chromosomes. This polymorphism was genotyped in the ECTIM study. Its allele frequencies were 0.96/0.04 and 0.95/0.05 in Belfast and France, respectively, without any difference between cases and controls. As a consequence of its LD pattern with the other polymorphisms and of its low allele frequency, no further information was gained by including this polymorphism in the haplotype analysis (data not shown). Detailed information on this polymorphism can be found on our internet site (http://genecanvas.idf.inserm.fr/).
From the pairwise LD matrix of the P-selectin gene polymorphisms, it was apparent that there were two main blocks of LD within the P-selectin gene. LD being strong within blocks and weak between blocks, haplotype analysis within blocks instead of over the whole gene reduced considerably the number of potential haplotypes and generated more interpretable results. Haplotype frequencies in the 5' block were homogeneous between cases and controls and none of the 5' polymorphisms was significantly associated with MI whatever its haplotypic background (data not shown). The S290N polymorphism was also considered in the haplotype analysis of the 3'block because it could affect the function of the P-selectin protein in specific ways depending on the particular combination of amino acids. The SNVP haplotype, which was the only frequent haplotype carrying the P715 variant, was more frequent in Belfast (0.134) than in the French population (0.080) and appeared to be protective in both countries. Independently of this association, the results show a consistent interaction between the S290N and N562D polymorphisms suggesting that the simultaneous presence of the N290 and N562 amino acids on the –VT haplotypic background may alter the function of the P-selectin protein and result in an increased risk of MI. It is important to point out why our previous univariate analyses failed to identify these associations. Any raw allelic OR associated with a given allele is actually a weighted combination of all corresponding ORs that can be estimated on specific haplotypic backgrounds, the weights depending on the haplotype frequencies. For example, the raw allelic OR associated with the D562 allele is a weighted combination of the ORs associated with the D562 allele when it is on the S-VT haplotype (OR=1.04) or on the –NVT– haplotype (OR=0.37). This combination yields a raw allelic OR of 0.86 (OR=0.86 [0.72–1.04], 2=2.41; P=0.12), explaining the lack of marginal association with the N562D polymorphism. Incidentally, we note that if a polymorphism exhibits different degrees of association with a phenotype according to its haplotypic background this may lead to inconsistent marginal associations in populations where the frequencies of the background haplotypes differ.
The S290N and N562D polymorphisms are both located within the consensus repeat domain of the P-selectin protein. This domain has been shown to be of particular importance for the binding of P-selectin to its ligand on leukocytes (18,19). It may therefore be hypothesized that the presence of an asparagine at positions 290 and 562 of P-selectin results in a protein that is more efficient to recruit leukocytes to the endothelium. In vitro functional investigations are currently under way to test this hypothesis. If this hypothesis was validated, it would imply an important contribution of the endothelial cell–leukocyte interaction mechanism in coronary heart disease.
Unlike most available techniques, the new ML method proposed here can deal with both quantitative and binary phenotypes and may incorporate covariates or interaction terms; in addition, it can express any haplotype–phenotype association in terms of association parameters that can then be used for hypothesis testing. Coincidentally, a similar approach based on score tests was proposed at a similar time by other authors (20). Instead of maximizing the full likelihood of the data as we have done here, these authors used a procedure based on a score statistic, which is asymptotically equivalent to the likelihood-ratio test statistic (20). According to the authors, this method has the advantage of being less computationally intensive than ML procedures. However, it seems less flexible than our ML method to test complex hypotheses such as interactions among polymorphisms carried by a given haplotype, or interactions between haplotypes and environmental covariates. Also, probably for easiness of computation, the proposed score-based method assumed independence between haplotypes and environmental factors and it is not known how the release of this condition would affect the computational procedure.
As with the score-based method (20), results obtained by our ML method on rare haplotypes should be interpreted with caution since in that case haplotype frequencies may be inaccurately estimated (21) and the asymptotic properties of the ML method may not be valid. In that case, empirical P-values obtained from permutation tests or simulations may be preferred. Simulation studies are under way to investigate in detail the statistical properties of our method, both in terms of power and type I error and to compare it with other techniques. However, it should be noted that the conclusions drawn from the present analysis mainly involved relatively common haplotypes, for which asymptotic properties of ML theory should hold.
To our knowledge, the performance of ML methods for estimating haplotype frequencies have mainly been investigated on simulated data (15,16,22). In the present paper, the main results provided by our inference ML model were compared to those derived from the observed haplotypes characterized by molecular haplotyping. The results of both analyses were very close, indicating that in the sample where haplotype frequencies could be directly evaluated, the statistical inference was quite satisfactory. Note, however, that the proportion of individuals for whom haplotypes could not be deduced from genotypes and who, consequently, contributed to the differences between the two methods, represented only 24% (222/937) of the ‘haplotyped sample’.
With regard to multiple testing, a conservative approach migth be to follow the conventional statistical strategy which consists of performing a global test of association and only if it is significant, to investigate further individual associations. In the case of P-selectin the global statistical test was significant in both sub-studies conducted in France and Northern Ireland, which was a justification for the more in-depth analysis performed later. However, in general, given the large number of possible haplotypes the global test will have a very low power and significant haplotypic effects may be missed. It is therefore very important before any application of the estimation method to reduce the size and complexity of the data set by discarding redundant polymorphisms, defining blocks of LD that will be investigated independently and possibly also focusing on polymorphisms that may be a priori more biologically significant.
The haplotypic strategy incorporates the complex interactions that may exist among polymorphisms in a genomic region with high LD. This may lead to a considerable reduction in the number of statistical tests to be performed when several polymorphisms are investigated, since the number of haplotypes is, in general, much smaller than the number of terms involving all possible interactions between genotypes. Contrary to a frequent assumption, haplotype analysis, for this reason, does not inflate the number of statistical tests in association studies, but reduces it. This is well illustrated on a global genomic scale by a recent characterization of blocks of LD on human chromosome 21 and derivation of the minimum number of polymorphisms that would be needed for a complete exploration of the chromosome (23).
In conclusion, these results show that an analytical strategy focusing on haplotypes is able to identify potentially relevant genetic effects that are undetected by a genotyped-based analysis. By applying this strategy to the P-selectin gene polymorphisms, it was possible to detect a consistent involvement of the S290N and N562D polymorphisms in MI predisposition both in the French and the Northern Irish populations participating in the ECTIM study. According to their combination on the same P-selectin molecule, the amino acids at these two sites are associated with different risks of MI. These findings illustrate the complexity of the relationship between gene variability and disease and the necessity to explore in details the global polymorphism of candidate genes. For the particular case of P-selectin, our results indicate that at least three polymorphisms (S290N, N562D and T715P) should be genotyped in any study investigating the role of this gene in disease etiology.
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIAL AND METHODSREFERENCES |
|---|
Study population
The study population was the same as that used in our previous analysis of the P-selectin gene polymorphisms (6). Patients aged 25–64 years with MI were recruited between 1989 and 1991 from four World Health Organization Monitoring in Cardiovascular Diseases (MONICA) registers in Northern Ireland (Belfast) and in France (Lille, Strasbourg and Toulouse). Age-matched controls were randomly sampled from the populations covered by the MONICA registers. Informed consent was obtained from all subjects. For the present analysis, controls with coronary heart disease were excluded, leading to a study sample composed of 582 cases and 630 controls for whom full genotypic information was available. The Belfast sample included 176 cases and 166 controls and the French sample 406 cases and 464 controls.
Genotyping
Genotyping of all subjects was performed using allele-specific oligonucleotides (ASOs) (24). All informations for genotyping polymorphisms (PCR primers, probes, conditions of amplification and hybridation) can be found on our internet site (http://genecanvas.idf.inserm.fr). Note that a re-examination of the initial genotyping results revealed that allele frequencies of the L599V polymorphism had been interchanged and that the most frequent allele encodes the valine at this position as reported in public databases. This polymorphism should therefore be read as V599L.
Allele frequencies and pairwise linkage disequilibrium
Thirteen polymorphisms had been identified in our previous molecular screening of the P-selectin gene (6) but only nine were included in the present analysis. Five polymorphisms were located in the 5' region of the gene (C-2123G, A-1969G, T-1817C, C-1576G, and -485I/D) and eight in the exonic regions (P98P, S290N, C557C, N562D, N563N, V599L, T715P and T741T). Due to their very low allele frequency (<1%), the C-1576G and the P98P polymorphisms were not included in the haplotype analysis. In addition, since the C557C and the N563N polymorphisms were completely concordant with the V599L polymorphism, they were excluded from the analysis.
Allele frequencies were estimated by gene-counting and departure from Hardy–Weinberg equilibrium was tested using a 2 with 1 d.f. Allele frequencies were compared between cases and controls by a 2 test. Pairwise LD was estimated by a log-linear model (25) and the extent of disequilibrium was expressed in terms of D' which is the ratio of the unstandardized coefficient to its maximal/minimal value (26).
Haplotype-disease model
The statistical model used in the present analysis is an extension to haplotypes of the standard logistic model used for testing a marker–disease association (27). In the case of a diallelic marker A/a, the genotype of an individual can be coded as a set of two indicator variables associated with the genotypes Aa and aa, the genotype AA being taken as the reference. In a logistic model, the exponential of the regression coefficient associated with each genotype represents the genotypic OR for disease associated with this genotype by comparison to the reference genotype. Under a multiplicative model (i.e. an additive model on the logistic scale), the set of two indicator variables reduces to only one variable coded 0, 1 and 2 for AA, Aa and aa genotypes, respectively. In such a model, the exponential of the regression coefficient associated with the genotype variable represents the allelic OR for disease associated with allele a by comparison to allele A. This particular model implicitly assumes that the OR associated with a given genotype is the product of the allelic ORs associated with each allele making up the genotype. Extension of this model to haplotypes is not straightforward since haplotypes, unlike alleles, are not directly observed. Therefore, we developed a ML method based on genotypic data that combines estimation of haplotype frequencies and of their effects.
Let Yi denotes individual i's binary phenotype and Gi be its genotypic vector at the different loci. The likelihood of individual i can be decomposed as 
where c is the number of possible haplotypic pairs Hj=(hj1,hj2) and where P(Gi/Hj)=1 if Gi is compatible with Hj, 0 otherwise (14). The first term is modeled using a logistic formulation assuming an additive model on a logistic scale. This is equivalent to: P(Yi/Hj=(hj1, hj2))=(1 + exp(-(
+ßj1+ßj2)))-1 where 
is the OR for disease associated with haplotype hj1 by comparison to a reference haplotype. The reference haplotype was chosen here as the haplotype combining the most frequent allele at each site. In this model, the OR associated with a given pair of haplotypes is the product of the corresponding haplotypic ORs. However, the hypothesis of multiplicativity of haplotype effects can be tested by including specific interaction terms in the logistic model. Incorporating additional covariates into this logistic model is straightforward and allows one to adjust for environmental factors and to test for haplotypexenvironment interactions. The last term of the sum is a function of haplotype frequencies under the assumption of random mating. The likelihood of the sample is therefore the product of the P(Yi,Gi) over all individuals. We developed a program for the inference ML method and linked it to the GEMINI maximization procedure (28) as we have successfully done for other genetic programs (29,30).
Note concerning quantitative haplotype-phenotype association analysis
The model described above is general enough to also accommodate quantitative traits. In that case, P(Yi/Hj=(hj1, hj2)) is modelled assuming a normal distribution for the trait, leading to E(Yi/Hj=(hj1,hj2))=+ßj1+ßj2 under an additive model.
Scheme of haplotype analysis
Haplotype analysis was carried out using a two-step procedure. A model in which all ßs were fixed to zero was fitted first, leading to the estimation of haplotype frequencies and the intercept . All haplotypes whose frequencies were estimated to be non zero at the first stage were then included in the model at the second step and the corresponding ß coefficients were estimated. A test for a global haplotypic effect on the phenotype can then be obtained by comparing the likelihood of the model with all ßs fixed to 0 to that obtained when ßs were estimated. From this model, it is also possible to test specific hypotheses concerning haplotypic effects by setting appropriate constraints on regression parameters. For example, investigating the heterogeneity of the effect of the N562D polymorphism according to two different haplotypic backgrounds, say –SVT and –NVT, was done by testing ßSNVT-ßSDVT=ßNNVT-ßNDVT. All hypotheses were tested by means of the likelihood-ratio criterion.
Molecular haplotyping
Molecular haplotyping was performed to check the validity of the results obtained by the statistical estimation analysis of the polymorphisms located in the 3' block. Due to the proximity of the N562D, V599L and T715P polymorphisms, molecular haplotyping using a PCR-based method was feasible for these three sites. 20 bp oligonucleotides whose 3' last nucleotide was overlapping the polymorphic site were designed as primers, in both allelic forms. Experimental set up of the optimal conditions was carried out on a limited numbers of DNA samples of known haplotypes including homozygotes and single heterozygotes for each locus. All three pairs of two variants (562
599, 599715 and 562715) were studied. For the haplotyping of two given polymorphisms, all four primer pairs were envisaged. However, among these four combinations, the less frequent one could not always be validated as very few individuals, if any, could theoretically carry this haplotype. In addition, among all the remaining combinations, one failed to give a completely specific signal. Finally, optimal specific amplifications could be performed for eight primer pairs. These were sufficient to determine the haplotype combination unambiguously for the three polymorphisms under study, due to the redundancy of the information generated by each couple of variants. Detailed description (primers list, PCR conditions, etc.) of the molecular haplotyping can be found on our internet site (http://genecanvas.idf.inserm.fr/). Complete haplotypic information at the four non-synonymous polymorphisms, either deduced unambiguously from genotypic data or directly derived from molecular haplotyping, was available for 937 (77%) amongst the 1212 initial individuals. In this sample referred to as ‘haplotyped sample’, we applied our ML method to the analysis of P-selectin gene polymorphisms in relation to MI as if haplotypes had not been observed, i.e. by using genotypic data only and the results were compared to those provided by a standard logistic regression analysis using the observed haplotypes.
|
ACKNOWLEDGEMENTS |
|---|
We are grateful to all investigators of the ECTIM study and to Dr A. Mallet for insightful discussions on this work. S.B. is supported by a fellowship from the Ligue Nationale contre le Cancer.
|
FOOTNOTES |
|---|
* To whom correspondence should be addressed: Tel: 33 140779693; Fax: 33 140779728; Email: tregouet{at}idf.inserm.fr
The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIAL AND METHODSREFERENCES |
|---|
1 Vestweber, D. and Blanks, J.E. (1999) Mechanisms that regulate the functions of the Selectins and their ligands. Physiol. Rev., 79, 181–213.
2
Manka, D., Collins, R.G., Ley, K., Beaudet, A.L. and Sarembock, J.J. (2001) Absence of p-selectin, but not intercellular adhesion molecule-1, attenuates neointimal growth after arterial injury in apolipoprotein e-deficient mice. Circulation, 103, 1000–1005.
3
Ikeda, H., Takajo, Y., Ihiki, K., Ueno, T., Maki, S., Noda, T., Sugi, K. and Imaizumi, T. (1995) Increased soluble form of P-selectin in patients with unstable angina. Circulation, 92, 1693–1696.
4
Merten, M. and Thiagarajan, P. (2000) P-selectin expression on platelets determines size and stability of platelet aggregates. Circulation, 102, 1931–1936.
5
Barbaux, S., Blankenberg, S., Rupprecht, H.J., Francomme, C., Bickel, C., Hafner, G., Nicaud, V., Meyer, J., Cambien, F. and Tiret, L. (2001) Association between P-selectin gene polymorphisms and soluble P-selectin levels and their relation to coronary artery disease. Arterioscler. Thromb. Vasc. Biol., 21, 1668–1673.
6
Herrmann, S.M., Ricard, S., Nicaud, V., Mallet, C., Evans, A., Ruidavets, J.B., Arveiler, D., Luc, G. and Cambien, F. (1998) The P-selectin gene is highly polymorphic: reduced frequency of the Pro715 allele carriers in patients with myocardial infarction. Hum. Mol. Genet., 7, 1277–1284.
7 Corbex, M., Poirier, O., Fumeron, F., Betoulle, D., Evans, A., Ruidavets, J.B., Arveiler, D., Luc, G., Tiret, L. and Cambien, F. (2000) Extensive association analysis between the CETP gene and coronary heart disease phenotypes reveals several putative functional polymorphisms and gene-environment interaction. Genet. Epidemiol., 19, 64–80.[Web of Science][Medline]
8 Hodge, S.E., Boehnke, M. and Spence, M.A. (1999) Loss of information due to ambiguous haplotyping of SNPs. Nat. Genet., 21, 360–361.[Web of Science][Medline]
9 Rieder, M.J., Taylor, S.L., Clark, A.G. and Nickerson, D.A. (1999) Sequence variation in the human angiotensin converting enzyme. Nat. Genet., 22, 59–62.[Web of Science][Medline]
10 Douglas, J.A., Boehnke, M., Gillanders, E., Trent, J.M. and Gruber, S.B. (2001) Experimentally-derived haplotypes susbtantially increase the efficiency of linkage disequilibrium studies. Nat. Genet., 28, 361–364.[Web of Science][Medline]
11 Olivier, M., Bustos, V.I., Levy, M.R., Smick, G.A., Moreno, I., Bushard, J.M., Almendras, A.A., Sheppard, K., Zierten, D.L., Aggarwal, A. et al. (2001) Complex high-resolution linkage disequilibrium and haplotype patterns of single-nucleotide polymorphisms in 2.5 Mb of sequence on human chromosome 21. Genomics, 78, 64–72.[Web of Science][Medline]
12 Clark, A.G. (1990) Inference of haplotypes from PCR-amplified samples of diploid populations. Mol. Biol. Evol., 7, 111–122.[Abstract]
13 Clark, A.G., Weiss, K.M., Nickerson, D.A., Taylor, S.L., Buchanan, A., Stengard, J., Salomaa, V., Vartiainen, E., Perola, M., Boerwinkle, E. and Sing, C.F. (1998) Haplotype structure and population genetic inferences from nucleotide-sequence variation in human lipoprotein lipase. Am. J. Hum. Genet., 63, 595–612.[Web of Science][Medline]
14 Excoffier, L. and Slatkin, M. (1995) Maximum-likelihood estimation of molecular haplotype frequencies in diploid population. Mol. Biol. Evol., 12, 921–927.[Abstract]
15 Chiano, M.N. and Clayton, D.G. (1998) Fine genetic mapping using haplotype analysis and the missing data problem. Ann. Hum. Genet., 62, 55–60.[Web of Science][Medline]
16 Stephens, M., Smith, N.J. and Donnelly, P. (2001) A new statistical method for haplotype reconstruction from population data. Am. J. Hum. Genet., 68, 978–989.[Web of Science][Medline]
17 Cargill, M., Altshuler, D., Ireland, J., Sklar, P., Ardlie, K., Patil, N., Shaw, N., Lane, C.R., Lim, E.P., Kalyanaraman, N. et al. (1999) Characterization of single-nucleotide polymorphisms in coding regions of human genes. Nat. Genet., 22, 231–238.[Web of Science][Medline]
18
Patel, K.D., Nollert, M.U. and McEver, R.P. (1995) P-selectin must extend a sufficient length from the plasma membrane to mediate rolling of neutrophils. J. Cell. Biol., 131, 1893–1902.
19 Ruchaud-Sparagano, M.H., Malaud, E., Gayet, O., Chignier, E., Buckland, R. and McGregor, J.L. (1998) Mapping the epitope of a functional P-selectin monoclonal antibody (LYP20) to a short complement-like repeat (SCR4) domain: use of human–mouse chimaera and homologue-replacement mutagenesis. Biochem. J., 332, 309–314.
20 Schaid, D.J., Rowland, C.M., Tines, D.E., Jacobson, R.M., Poland, G.A. (2002) Score tests for association between traits and haplotypes when linkage phase is ambiguous. Am. J. Hum. Genet., 70, 425–434.[Web of Science][Medline]
21 Tishkoff, S., Pakstis, A., Ruano, G., Kidd, K. (2000) The accuracy of statistical methods for estimation of haplotypes frequencies: an example from the CD4 locus. Am. J. Hum. Genet., 67, 518–522.[Web of Science][Medline]
22 Rohde, K. and Fuerst, R. (2001) Haplotyping and estimation of haplotype frequencies for closely linked biallelic multilocus genetic phenotypes including nuclear family information. Hum. Mutat., 17, 289–295.[Web of Science][Medline]
23
Patil, N., Berno, A.J., Hinds, D.A., Barret, W.A., Doshi, J.M., Hacker, C.R., Kautzer, C.R., Lee, D.H., Marjoribanks, C., McDonough, D.P. et al. (2001) Blocks of limited haplotype diversity revealed by high-resolution scanning of human chromosome 21. Science, 294, 1719–1723.
24 Saiki, R.K., Bugawan, T.L., Horn, G.T., Mullis, K.B. and Horn, H.A. (1986) Analysis of enzymatically amplified beta-globin and HLA-DQA alpha DNA with allele-specific oligonucleotide probes. Nature, 324, 163–166.[Medline]
25 Tiret, L., Amouyel, P., Rakotovao, R., Cambien, F. and Ducimetiere, P. (1991) Testing for association between disease and linked marker loci: a log-linear-model analysis. Am. J. Hum. Genet., 48, 926–934.[Web of Science][Medline]
26 Nei, M. (1987) Molecular Evolutionary Genetics. Columbia University Press. New York.
27 Trégouet, D.A., Ducimetière, P. and Tiret, L. (1997) Testing association between candidate-gene markers and phenotype in related individuals, by use of estimating equations. Am. J. Hum. Genet., 61, 189–199.[Web of Science][Medline]
28 Lalouel, J. (1981) GEMINI: a computed program for optimization of general nonlinear functions. Department of Medical Biophysics and Computing, University of Utah, Salt Lake City.
29 Tiret, L., Rigat, B., Visvikis, S., Breda, C., Corvol, P., Cambien, F. and Soubrier, F. (1992) Evidence, from combined segregation and linkage analysis, that a variant of the angiotensin I-converting enzyme (ACE) gene controls plasma ACE levels. Am. J. Hum. Genet., 51, 197–205.[Web of Science][Medline]
30 Trégouet, D.A., Ducimetiere, P., Bocquet, V., Visvikis, S., Soubrier, F. and Tiret, L. (1999) A parametric copula model for analysis of familial binary data. Am. J. Hum. Genet., 64, 886–893.[Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  K. A. Volcik, D. Catellier, A. R. Folsom, N. Matijevic, B. Wasserman, and E. Boerwinkle SELP and SELPLG Genetic Variation Is Associated with Cell Surface Measures of SELP and SELPLG: The Atherosclerosis Risk in Communities Carotid MRI Study Clin. Chem., June 1, 2009; 55(6): 1076 - 1082. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Park, Y. Jamshidi, D. Vaideanu, M. Bitner-Glindzicz, S. Fraser, and J. C. Sowden Genetic Risk for Primary Open-Angle Glaucoma Determined by LMX1B Haplotypes Invest. Ophthalmol. Vis. Sci., April 1, 2009; 50(4): 1522 - 1530. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. J. Talmud, M. Smart, E. Presswood, J. A. Cooper, V. Nicaud, F. Drenos, J. Palmen, M. G. Marmot, S. M. Boekholdt, N. J. Wareham, et al. ANGPTL4 E40K and T266M: Effects on Plasma Triglyceride and HDL Levels, Postprandial Responses, and CHD Risk Arterioscler Thromb Vasc Biol, December 1, 2008; 28(12): 2319 - 2325. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. P. Reiner, C. S. Carlson, B. Thyagarajan, M. J. Rieder, J. F. Polak, D. S. Siscovick, D. A. Nickerson, D. R. Jacobs Jr, and M. D. Gross Soluble P-Selectin, SELP Polymorphisms, and Atherosclerotic Risk in European-American and African-African Young Adults: The Coronary Artery Risk Development in Young Adults (CARDIA) Study Arterioscler Thromb Vasc Biol, August 1, 2008; 28(8): 1549 - 1555. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Schunkert, A. Gotz, P. Braund, R. McGinnis, D.-A. Tregouet, M. Mangino, P. Linsel-Nitschke, F. Cambien, C. Hengstenberg, K. Stark, et al. Repeated Replication and a Prospective Meta-Analysis of the Association Between Chromosome 9p21.3 and Coronary Artery Disease Circulation, April 1, 2008; 117(13): 1675 - 1684. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. T.E. Wootton, N. L. Arora, F. Drenos, S. R. Thompson, J. A. Cooper, J. W. Stephens, S. J. Hurel, E. Hurt-Camejo, O. Wiklund, S. E. Humphries, et al. Tagging SNP haplotype analysis of the secretory PLA2-V gene, PLA2G5, shows strong association with LDL and oxLDL levels, suggesting functional distinction from sPLA2-IIA: results from the UDACS study Hum. Mol. Genet., June 15, 2007; 16(12): 1437 - 1444. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. P. Mathew, M. V. Podgoreanu, H. P. Grocott, W. D. White, R. W. Morris, M. Stafford-Smith, G. B. Mackensen, C. S. Rinder, J. A. Blumenthal, D. A. Schwinn, et al. Genetic Variants in P-Selectin and C-Reactive Protein Influence Susceptibility to Cognitive Decline After Cardiac Surgery J. Am. Coll. Cardiol., May 15, 2007; 49(19): 1934 - 1942. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 U. W. Preuss, P. Zill, G. Koller, B. Bondy, and M. Soyka D2 dopamine receptor gene haplotypes and their influence on alcohol and tobacco consumption magnitude in alcohol-dependent individuals Alcohol Alcohol., May 1, 2007; 42(3): 258 - 266. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. A. Tregouet and V. Garelle A new JAVA interface implementation of THESIAS: testing haplotype effects in association studies Bioinformatics, April 15, 2007; 23(8): 1038 - 1039. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Debette, J.-C. Lambert, J. Gariepy, N. Fievet, C. Tzourio, J.-F. Dartigues, K. Ritchie, A.-M. Dupuy, A. Alperovitch, P. Ducimetiere, et al. New Insight Into the Association of Apolipoprotein E Genetic Variants With Carotid Plaques and Intima-Media Thickness Stroke, December 1, 2006; 37(12): 2917 - 2923. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Inoue, Y. Mineharu, S. Inoue, S. Yamada, F. Matsuda, K. Nozaki, K. Takenaka, N. Hashimoto, and A. Koizumi Search on Chromosome 17 Centromere Reveals TNFRSF13B as a Susceptibility Gene for Intracranial Aneurysm: A Preliminary Study Circulation, April 25, 2006; 113(16): 2002 - 2010. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Beekman, G. J. Blauw, J. J. Houwing-Duistermaat, B. W. Brandt, R. G. J. Westendorp, and P. E. Slagboom Chromosome 4q25, microsomal transfer protein gene, and human longevity: novel data and a meta-analysis of association studies. J. Gerontol. A Biol. Sci. Med. Sci., April 1, 2006; 61(4): 355 - 362. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Vaxillaire, A. Dechaume, V. Vasseur-Delannoy, S. Lahmidi, V. Vatin, F. Lepretre, P. Boutin, S. Hercberg, G. Charpentier, C. Dina, et al. Genetic Analysis of ADIPOR1 and ADIPOR2 Candidate Polymorphisms for Type 2 Diabetes in the Caucasian Population Diabetes, March 1, 2006; 55(3): 856 - 861. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. M. McCaffery, N. Frasure-Smith, M.-P. Dube, P. Theroux, G. A. Rouleau, Q. Duan, and F. Lesperance Common genetic vulnerability to depressive symptoms and coronary artery disease: a review and development of candidate genes related to inflammation and serotonin. Psychosom Med, March 1, 2006; 68(2): 187 - 200. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. T.E. Wootton, F. Drenos, J. A. Cooper, S. R. Thompson, J. W. Stephens, E. Hurt-Camejo, O. Wiklund, S. E. Humphries, and P. J. Talmud Tagging-SNP haplotype analysis of the secretory PLA2IIa gene PLA2G2A shows strong association with serum levels of sPLA2IIa: results from the UDACS study Hum. Mol. Genet., January 15, 2006; 15(2): 355 - 361. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. L. Gonzalez-Sanchez, M. J. Martinez-Calatrava, M. T. Martinez-Larrad, C. Zabena, C. Fernandez-Perez, M. Laakso, and M. Serrano-Rios Interaction of the -308G/A Promoter Polymorphism of the Tumor Necrosis Factor-{alpha} Gene with Single-Nucleotide Polymorphism 45 of the Adiponectin Gene: Effect on Serum Adiponectin Concentrations in a Spanish Population Clin. Chem., January 1, 2006; 52(1): 97 - 103. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Meirhaeghe, D. Cottel, P. Amouyel, and J. Dallongeville Association Between Peroxisome Proliferator-Activated Receptor {gamma} Haplotypes and the Metabolic Syndrome in French Men and Women Diabetes, October 1, 2005; 54(10): 3043 - 3048. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Evans, V. Salomaa, S. Kulathinal, K. Asplund, F. Cambien, M. Ferrario, M. Perola, L. Peltonen, D. Shields, H. Tunstall-Pedoe, et al. MORGAM (an international pooling of cardiovascular cohorts) Int. J. Epidemiol., February 1, 2005; 34(1): 21 - 27. [Full Text] [PDF]
|
|
|
|
 |
|
 S. Benhamou and A. Sarasin ERCC2 /XPD Gene Polymorphisms and Lung Cancer: A HuGE Review Am. J. Epidemiol., January 1, 2005; 161(1): 1 - 14. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Lecomte, J.-M. Ferraz, F. Zinzindohoue, M.-A. Loriot, D.-A. Tregouet, B. Landi, A. Berger, P.-H. Cugnenc, R. Jian, P. Beaune, et al. Thymidylate Synthase Gene Polymorphism Predicts Toxicity in Colorectal Cancer Patients Receiving 5-Fluorouracil-based Chemotherapy Clin. Cancer Res., September 1, 2004; 10(17): 5880 - 5888. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. Putt, J. Palmen, V. Nicaud, D.-A. Tregouet, N. Tahri-Daizadeh, D. M. Flavell, S. E. Humphries, P. J. Talmud, and on behalf of the EARSII group Variation in USF1 shows haplotype effects, gene : gene and gene : environment associations with glucose and lipid parameters in the European Atherosclerosis Research Study II Hum. Mol. Genet., August 1, 2004; 13(15): 1587 - 1597. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Ninio, D. Tregouet, J.-L. Carrier, D. Stengel, C. Bickel, C. Perret, H. J. Rupprecht, F. Cambien, S. Blankenberg, and L. Tiret Platelet-activating factor-acetylhydrolase and PAF-receptor gene haplotypes in relation to future cardiovascular event in patients with coronary artery disease Hum. Mol. Genet., July 1, 2004; 13(13): 1341 - 1351. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. H. Gibbons, C. C. Liew, M. O. Goodarzi, J. I. Rotter, W. A. Hsueh, H. M. Siragy, R. Pratt, and V. J. Dzau Genetic Markers: Progress and Potential for Cardiovascular Disease Circulation, June 29, 2004; 109(25_suppl_1): IV-47 - IV-58. [Full Text] [PDF]
|
|
|
|
|
|
R. Roussel, D.-A. Tregouet, S. Hadjadj, X. Jeunemaitre, and M. Marre Investigation of the Human ANP Gene in Type 1 Diabetic Nephropathy: Case-Control and Follow-up Studies Diabetes, May 1, 2004; 53(5): 1394 - 1398. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D.-A. Tregouet, S. Ricard, V. Nicaud, I. Arnould, S. Soubigou, M. Rosier, N. Duverger, O. Poirier, S. Mace, F. Kee, et al. In-Depth Haplotype Analysis of ABCA1 Gene Polymorphisms in Relation to Plasma ApoA1 Levels and Myocardial Infarction Arterioscler Thromb Vasc Biol, April 1, 2004; 24(4): 775 - 781. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Y.L. Zee, N. R. Cook, S. Cheng, R. Reynolds, H. A. Erlich, K. Lindpaintner, and P. M. Ridker Polymorphism in the P-selectin and interleukin-4 genes as determinants of stroke: a population-based, prospective genetic analysis Hum. Mol. Genet., February 15, 2004; 13(4): 389 - 396. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. D. Blann, S. K. Nadar, and G. Y.H. Lip The adhesion molecule P-selectin and cardiovascular disease Eur. Heart J., December 2, 2003; 24(24): 2166 - 2179. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M Sancandi, P Griseri, B Pesce, G Patrone, F Puppo, M Lerone, G Martucciello, G Romeo, R Ravazzolo, M Devoto, et al. Single nucleotide polymorphic alleles in the 5' region of the RET proto-oncogene define a risk haplotype in Hirschsprung's disease J. Med. Genet., September 1, 2003; 40(9): 714 - 718. [Full Text] [PDF]
|
|
|
|
 |
|
 N. Tahri-Daizadeh, D.-A. Tregouet, V. Nicaud, N. Manuel, F. Cambien, and L. Tiret Automated Detection of Informative Combined Effects in Genetic Association Studies of Complex Traits Genome Res., August 1, 2003; 13(8): 1952 - 1960. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. H.E.M. Klerkx, M. W.T. Tanck, J. J.P. Kastelein, H. O.F. Molhuizen, J. W. Jukema, A. H. Zwinderman, and J. A. Kuivenhoven Haplotype analysis of the CETP gene: not TaqIB, but the closely linked -629C->A polymorphism and a novel promoter variant are independently associated with CETP concentration Hum. Mol. Genet., January 15, 2003; 12(2): 111 - 123. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||