¹Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA

²Laboratory of Statistical Genetics, Rockefeller University, New York, NY, USA

³Department of Human Genetics, Roche Molecular Systems, Alameda, CA, USA

⁴Hospital Universitario San Carlos, Ciudad Universitaria, Madrid, Spain

⁵F Hoffmann-La Roche Ltd, Basel, Switzerland

Correspondence to: K Lindpainter, MD, MPH, VP Research, Director, Roche Genetics, F Hoffmann-La Roche, Bldg 93/532, CH-4070 Basel, Switzerland. Tel: +41-61-688.0254 Fax: +41-61-688.1929 E-mail: klaus.lindpaintner@roche.com

^aContributed equally to the study

The complexity of recognizing the potential contribution of a number of possible predictors of complex disorders is increasingly challenging with the application of large-scale single nucleotide polymorphism (SNP) typing. In the search for putative genetic factors predisposing to coronary artery restenosis following balloon angioplasty, we determined genotypes for 94 SNPs representing 62 candidate genes, in a prospectively assembled cohort of 342 cases and 437 controls. Using a customized coupled-logistic regression procedure accounting for both additive and interactive effects, we identified seven SNPs in seven genes that, together, showed a statistically significant association with restenosis incidence (P <0.0001), accounting for 11.6% of overall variance observed. Among them are candidate genes for cardiovascular pathophysiology (apolipoprotein-species and NOS), inflammatory response (TNF receptor and CD14), and cell-cycle control (p53 and p53-associated protein). Our results emphasize the need to account for complex multi-gene influences and interactions when assessing the molecular pathology of multifactorial medical entities.

The Pharmacogenomics Journal (2002) 2, 197-201. doi:10.1038/sj.tpj.6500101

restenosis; angioplasty; SNPs; risk factors

INTRODUCTION

The utility of percutaneous transluminal coronary angioplasty (PTCA) is limited by a high incidence of restenosis following the procedure ('post-PTCA restenosis')¹ that occurs in 30-40% of all patients. Like most common diseases, the occurrence of coronary artery restenosis following PTCA is considered a complex multifactorial disease process. Experimental and clinical studies have implicated vascular remodeling,² however, the specific molecular and cellular mechanisms of the processes involved remain largely unknown. None of a large number of clinical risk factors recognized for atherosclerosis or ischemic cardiovascular disease (with the possible exception of diabetes mellitus³) has been found associated with the occurrence of restenosis. To evaluate the possible effects of genetic predisposition, a number of studies focusing on single-gene variants have been carried out to date, but have remained inconclusive due to a number of limitations, including study size and design. In the absence of any known strong predictors for post-PTCA restenosis, we selected this disease entity as a test case for a more comprehensive approach, with regard to both breadth of genotyping and analytical algorithms. In a cohort of post-PTCA patients three times larger than the largest sample studied so far,⁴ a panel of 94 polymorphic markers was tested from 62 'candidate' genes (representing, broadly, genes related to lipid and homocysteine metabolism, vascular tone regulation, cell-matrix adhesion and interaction, matrix integrity, inflammatory and immune responses, thrombosis/coagulation, cellular homeostasis/cytoprotection mechanisms, and cell cycle/apoptotic pathways; see Table 1).

RESULTS

Baseline characteristics of the study population are shown in Table 2. As expected, there were no significant differences between the cases (restenosis) and the controls (no restenosis) with regard to age, blood pressure status, plasma lipid profile, smoking status, or family history of coronary heart disease. Only diabetes mellitus was significantly more common in cases than controls (²_2df = 6.66, P = 0.011). Genotype frequencies for the 94 bi-allelic markers tested were compared by ² analysis. The corresponding uncorrected P values are provided (http://research.bwh.harvard.edu/rdbook/SNP_Table.doc). Of note, only four polymorphisms were found to attain a P-value of <0.05, and among them only one of <0.01.

Subsequently, we carried out a two-stage analysis⁵ to initially select a set of influential markers, followed by modeling of gene-phenotype effects and gene-gene interactions. In what we believe is the first application of a nested bootstrap (re-sampling) procedure to genetic markers, the selection stage was carried out as follows. For each SNP, ² was computed for a 2´3 table, with case-control status and genotype as the dependent and independent variables. Simply picking a marker based on its significant ², however, would ignore the joint effects of all markers (each marker by itself does not have a strong effect). With the selection procedure thus based on univariate ² values, the sub- selection of markers to be used for the later stages of the analysis was determined using an approach described in more detail elsewhere.⁵ In brief, a statistic was calculated as the sum of the ² values over all marker loci, where the statistical significance of this sum was evaluated by 1000 bootstrap samples taken under the null hypothesis of no association. Markers were deleted from the sum stepwise, based on ascending ² value. At each step the significance level was re-evaluated for the remaining sum until the significance level of the sum dropped below 5%. This pre-selected a set of influential markers from the original data. To validate this pre-selection procedure, we replicated the original dataset in 499 bootstrap samples retaining the association between markers and disease status. The pre-selection procedure described above was then carried out for each of the 499 replicated samples. Thus, a potentially different set of markers was pre-selected in each sample. Markers that were pre-selected in at least 2/3 of all 500 samples were then selected for modeling. Seven marker loci were thus found (APOC3 C1100T; CBS 1278T; CD14 C(-260)T; NOS3 E298D; TP53 P72R; MDM2 NlaIV site; TNFR1 A845G).

In stage 2, modeling was carried out by logistic regression, a method commonly used to assess the effects of risk factors on outcome variables, but so far applied to genetic analysis only rarely, and only to small datasets.^6,7 A sequence of backward and forward stepwise regression analyses incorporated two-way and three-way interaction effects into the model. The resulting final model contains the seven markers through linear and quadratic main effects and interaction terms that cumulatively account for 11.6% of the total phenotypic variance (Table 3). The overall significance level of the model is P <0.0001. All these stage 2 calculations were carried out with the commercially available S-PLUS^Ò 2000 software. Note that one marker, NOS3, occurs only through interaction with other markers, that is, its direct main effect is negligible (Table 3).

Under the assumption that non-genetic factors contribute significantly to total variance¾given that the phenotype arises in response to procedural intervention¾this regression formula likely explains a substantial fraction of overall genetic variance. The fact that the only other known predictor for restenosis, diabetes mellitus, only accounted for 6.7% of the variance in the present study emphasizes the significance of our finding.

DISCUSSION

Experimental and human studies have indicated plaque retention, vaso-constrictive remodeling, thrombus formation, and neointimal proliferation as the primary effector mechanisms likely to contribute to the pathophysiology of the condition.⁸ Furthermore, inflammatory and immune responses, as well as apoptotic processes have also been implicated^9,10,11 in the underlying disease mechanism. Our findings represent the first direct clinical-epidemiological evidence for the actual relevance of these mechanisms by highlighting molecular variants of seven genes that belong to the postulated pathogenic pathways (apolipoprotein CIII, cystathionine beta-synthase, monocyte differentiation antigen CD14, endothelial nitric oxide synthase 3, tumor suppressor protein p53, p53-associated protein, and tumor necrosis factor receptor type 1) and that we found to be significantly associated with the incidence of restenosis. In the absence of data on the functional consequences of the implicated polymorphisms, further speculation is presently not justified; however, our findings prioritize these genes, among all the ones implicated in the literature so far, as candidates for functional-genetic studies.

The present study emphasizes the importance of taking a multiplicity of potentially involved gene variants into account when studying the genetic epidemiology of common complex traits. Concomitantly, it highlights the need to apply powerful, customized analytical algorithms to the data sets thus generated. Of note, although the present findings account only for a fraction of the overall genetic variance of the trait, these results were derived from a modest number of genes, indicating the value of a candidate gene-driven approach. Whereas we recognize that selecting individual markers from a large set may be due to random chance (so-called multiple testing problem), we believe that the validation procedure chosen, using 499 bootstrap replicates, will largely eliminate this problem. At the same time, these results call for a more comprehensive selection of candidate genes and their variants, or for additional approaches not restricted to the bias of pre-selected candidate genes. While the current results will have limited practical applicability, we believe they point the way towards a future feasibility of creating comprehensive risk factor panels, and their application to medical practice.

MATERIALS AND METHODS

Study Design

Between March 1995 and March 1997, 909 consecutive patients who were undergoing PTCA at the Hospital San Carlos, Universidad Complutense in Madrid, Spain, were enrolled in a comprehensive follow-up study specifying symptom-independent angiographic follow-up 6 months later. All patients gave appropriate signed informed consent for both the clinical procedures and genotype testing. A total of 779 consecutive subjects fulfilled the enrollment criteria for the present study: successful PTCA, follow-up cardiac catheterization, interpretable quantitative coronary angiography (QCA) measurements at baseline (prior to undergoing PTCA), immediately post-PTCA (demonstrating the initial effect of the procedure) and at follow-up cardiac catheterization 6 months later,^12,13 and the availability of a blood sample for genotype determination. Restenosis was defined as >50% reduction in the luminal diameter at the site of previous PTCA, as compared with the immediate post-PTCA angiographic record; 342 patients were classified as cases. Of note, no intra-coronary stents were deployed in any of the 779 participants in the present investigation.

Genotype Determination

Cardiovascular disease (CVD) and inflammation/asthma linear array panels (Roche Molecular Systems, Alameda, CA, USA): A general description of the protocol¹⁴ used has previously been reported. In brief, each DNA sample is amplified in multiplex polymerase chain reactions (PCRs) using biotinylated primers. The PCR product pool is then hybridized to the corresponding panel of sequence-specific oligonucleotide probes that have been immobilized in a linear array on nylon membrane strips. A colorimetric detection method based upon incubation with streptavidin-horseradish peroxidase conjugate of hydrogen peroxide and 3,3',5,5'-tetramethylbenzidine as substrates was used. A total of 74 markers within 47 genes were examined (URL:http://research.bwh/harvard.edu/rdbook/ca16ref.htm) using this system (Table 1).

Additional polymorphisms examined: An additional 20 polymorphisms (indicated by asterisks in Table 1) were assayed using previously published protocols.^{15,16,17,18,19,20,21,22,23,24,25,26,27,28}

To confirm genotype assignments, the PCR procedure was performed in replicate on all samples. Scoring was carried out by two independent observers. Disagreements (<2%) were resolved by a further joint reading, and where necessary, a repeat genotyping reaction. All results were scored blinded as to case-control status.

Power Calculations

The least significant relative risk was used that we would be able to recognize in the current study, assuming 80% power, based on a univariate model of analysis, on the present sample size, an additive mode of inheritance, and an alpha error of 0.05, for rare and common allele frequencies. If the minor allele frequency is 0.01, the least significant relative risk detectable is >3.3; if the minor allele frequency is 0.50, the least significant relative risk detectable is >1.3.

DUALITY OF INTEREST

None declared.

Acknowledgements

Supported by a Grant HG00008 from the National Institute of Health (JO), and a Research Development Award K04-HL-03138-01 from the National Heart, Lung, and Blood Institute (KL).

1 McBride W, Lange RA, Hillis LD. Restenosis after successful coronary angioplasty. Pathophysiology and prevention. N Engl J Med 1988; 318: 1734-1737. MEDLINE

2 Mintz GS, Popma JJ, Pichard AD, Kent KM, Satler LF, Wong C et al. Arterial remodeling after coronary angioplasty: a serial intravascular ultrasound study. Circulation 1996; 94: 35-43. MEDLINE

3 Van Belle E, Bauters C, Hubert E, Bodart JC, Abolmaali K, Meurice T et al. Restenosis rates in diabetic patients. A comparison of coronary stenting and balloon angioplasty in native coronary vessels. Circulation 1997; 96: 1454-1460. MEDLINE

4 Samani NJ, Martin DS, Brack M, Cullen J, Chauhan A, Lodwick D et al. Insertion/deletion polymorphism in the angiotensin-converting enzyme gene and risk of restenosis after coronary angioplasty. Lancet 1995; 345: 1013-1016. MEDLINE

5 Hoh J, Wille A, Zee R, Cheng S, Reynolds R, Lindpaintner K et al. Selecting SNPs in two-stage analysis of disease association data: a model-free approach. Ann Hum Genet 2000; 64: 413-417. MEDLINE

6 Chen CC, Lu RB, Chen YC, Wang MF, Chang YC, Li TK et al. Interaction between the functional polymorphisms of the alcohol-metabolism genes in protection against alcoholism. Am J Hum Genet 1999; 65: 795-807. MEDLINE

7 Rice JP, Rochberg N, Neuman RJ, Saccone NL, Liu KY, Zhang X et al. Covariates in linkage analysis. Genet Epidemiol 1999; 17 (Suppl 1): S691-S695. MEDLINE

8 Casterella PJ, Teirstein PS. Prevention of coronary restenosis. Cardiol Rev 1999; 7: 219-231. MEDLINE

9 Libby P, Schwartz D, Brogi E, Tanaka H, Clinton SK. A cascade model for restenosis. A special case of atherosclerosis progression. Circulation 1992; 86 Suppl 6: III47-III52.

10 Biasucci LM, Liuzzo G, Buffon A, Maseri A. The variable role of inflammation in acute coronary syndromes and in restenosis. Semin Interv Cardiol 1999; 4: 105-110. MEDLINE

11 Best PJ, Hasdai D, Sangiorgi G, Schwartz RS, Holmes DR Jr, Simari RD et al. Apoptosis. Basic concepts and implications in coronary artery disease. Arterioscler Thromb Vasc Biol 1999; 19: 14-22. MEDLINE

12 Mancini GB, Simon SB, McGillem MJ, LeFree MT, Friedman HZ, Vogel RA. Automated quantitative coronary arteriography: morphologic and physiologic validation in vivo of a rapid digital angiographic method. Circulation 1987; 75: 452-460. MEDLINE

13 Alfonso F, Macaya C, Goicolea J, Iniguez A, Hernandez R, Zamorano J et al. Intravascular ultrasound imaging of angiographically normal coronary segments in patients with coronary artery disease. Am Heart J 1994; 127: 536-544. MEDLINE

14 Cheng S, Grow MA, Pallaud C, Klitz W, Erlich HA, Visvikis S et al. A multilocus genotyping assay for candidate markers of cardiovascular disease risk. Genome Res 1999; 9: 936-949. MEDLINE

15 Stevens PA, Brown MJ. Genetic variability of the ET-1 and the ETA receptor genes in essential hypertension. J Cardiovasc Pharmacol 1995; 26 Suppl 3: S9-S12. MEDLINE

16 Hubacek JA, Rothe G, Pit'ha J, Skodova Z, Stanek V, Poledne R et al. C(-260)T polymorphism in the promoter of the CD14 monocyte receptor gene as a risk factor for myocardial infarction. Circulation 1999; 99: 3218-3220. MEDLINE

17 Mastuura S, Kishi F. Investigation of the polymorphic Avall site by a PCR-based assay at the human CD18 gene locus. Hum Genet 1994; 93: 721. MEDLINE

18 Behague I, Poirier O, Nicaud V, Evans A, Arveiler D, Luc G et al. Beta fibrinogen gene polymorphisms are associated with plasma fibrinogen and coronary artery disease in patients with myocardial infarction. The ECTIM Study. Etude Cas-Temoins sur l'Infarctus du Myocarde. Circulation 1996; 93: 440-449. MEDLINE

19 Zhang B, Ye S, Herrmann SM, Eriksson P, de Maat M, Evans A et al. Functional polymorphism in the regulatory region of gelatinase B gene in relation to severity of coronary atherosclerosis. Circulation 1999; 99: 1788-1794. MEDLINE

20 van den Velden PA, Reitsma PH. Amino acid dimorphism in IL1A is detectable by PCR amplification. Hum Mol Genet 1993; 2: 1753. MEDLINE

21 di Giovine FS, Takhsh E, Blakemore AI, Duff GW. Single base polymorphism at -511 in the human interleukin-1 beta gene (IL1 beta). Hum Mol Genet 1992; 1: 450. MEDLINE

22 Pociot F, Molvig J, Wogensen L, Worsaae H, Nerup J. A Taql polymorphism in the human interleukin-1 beta (IL-1 beta) gene correlates with IL-1 beta secretion in vitro. Eur J Clin Invest 1992; 22: 396-402. MEDLINE

23 Bergholdt R, Karlsen AE, Johannesen J, Hansen PM, Dinarello CA, Nerup J et al. Characterization of polymorphisms of an interleukin 1 receptor type 1 gene (IL1RI) promotor region (P2) and their relation to insulin-dependent diabetes mellitus (IDDM). The Danish Study Group of Diabetes in Childhood. Cytokine 1995; 7: 727-733. Article MEDLINE

24 Tarlow JK, Blakemore AI, Lennard A, Solari R, Hughes HN, Steinkasserer A et al. Polymorphism in human IL-1 receptor antagonist gene intron 2 is caused by variable numbers of an 86-bp tandem repeat. Hum Genet 1993; 91: 403-404. MEDLINE

25 Aldred MA, Wright AF. PCR detection of existing and new polymorphism at the TIMP locus. Nucleic Acids Res 1991; 19: 1165. MEDLINE

26 Vinasco J, Beraun Y, Nieto A, Fraile A, Pareja E, Mataran L et al. Heat shock protein 70 gene polymorphisms in rheumatoid arthritis. Tissue Antigens 1997; 50: 71-73. MEDLINE

27 Sjalander A, Birgander R, Kivela A, Beckman G. p53 polymorphisms and haplotypes in different ethnic groups. Hum Hered 1995; 45: 144-149. MEDLINE

28 Heighway J, Mitchell EL, Jones D, White GR, Santibanez Koref MF. A transcribed polymorphism and sub-localization of MDM2. Hum Genet 1994; 93: 611-612. MEDLINE

Table 1 Polymorphisms assessed

Table 2 Baseline characteristics of the study population

Table 3 Results of the coupled-logistic regression