Plasminogen Activator Inhibitor-1 Polymorphism (4G/5G) Predicts Recurrence in Nonhyperlipidemic Postinfarction Patients
Objective— Nonhyperlipidemic postinfarction patients are at high risk for recurrent coronary events by virtue of incident myocardial infarction (MI); however, few studies assess risk beyond incident MI. The aim of this study was to assess such risk as a function of 37 atherosclerosis-associated genetic polymorphisms and 17 blood marker variables.
Methods and Results— Screening of polymorphisms in nonhyperlipidemic postinfarction patients revealed significant risk only for the 4G/5G insertion/deletion polymorphism in the promoter of the plasminogen-activator inhibitor-1 (PAI-1) gene. Outcome event mapping, an exploratory data analysis tool, was then applied to define a subgroup (182 patients from total study population of 846 nondiabetic patients) exhibiting maximal functional dependence of risk on the PAI-1 polymorphism. Cox multivariable regression analyses within the subgroup adjusted for significant clinical covariates and medication use as a function of the PAI-1 polymorphism and 17 atherosclerosis-associated blood markers revealed significant risk for patients homozygous for the 4G allele (hazard ratio 4.30, 95% CI 1.98 to 9.33, P=0.00023), and lack of significant risk-association with any blood marker.
Conclusions— In a subgroup of normolipidemic postinfarction patients, only the PAI-1 4G/5G polymorphism was associated with recurrent risk from a set of atherosclerosis-associated genetic polymorphisms and blood markers.
- plasminogen activator inhibitor-1 (PAI-1), 4G/5G polymorphism
- risk factors
- multivariable analysis
Although hypercholesterolemic postinfarction patients are at high risk for recurrent coronary events, normocholesterolemic postinfarction patients suffer recurrent events as well.1 Beyond incident myocardial infarction (MI), traditional and nontraditional risk factors play a role2,3 although little specific information is available especially for normolipidemic postinfarction patients.4 In a previous report on nondiabetic patients of the Thrombogenic Factors and Recurrent Coronary Events (THROMBO) postinfarction study, when lipoprotein-associated phospholipase A2 (Lp-PLA2) was included in models, it resulted in replacement of apolipoprotein B (apoB) as the only risk factor of recurrent coronary events from a set of thrombogenic, inflammatory, and metabolic blood markers.5 To determine predictors of risk for recurrent events in normolipidemic postinfarction patients, we studied nondiabetic THROMBO patients (n=846) as a function of a set of 37 genetic markers associated with cardiovascular disease (CVD) followed by studies with a set of 17 thrombogenic, inflammatory, and metabolic blood markers.5
Preliminary screening of genetic markers in normolipidemic (cholesterol <5.17 mmol/L, triglycerides <1.69 mmol/L) study patients revealed most significant risk for the 4G/5G insertion/deletion polymorphism (4 or 5 sequential guanosines, respectively) of plasminogen activator inhibitor-1 (PAI-1), a major inhibitor of fibrinolysis. The polymorphism is located in the gene promoter region (chromosome 7) and results in allele-specific responses to multiple agents.6 Accordingly, studies have shown the 4G allele to be associated with both higher levels of PAI-1 and CVD risk.6 The aim of the current work was to define in a bivariate sense, beyond simple univariate cut points for cholesterol and triglyceride levels, the normolipidemic subgroup exhibiting maximal functional dependence of risk on the PAI-1 polymorphism; not only to precisely determine polymorphism/lipid interactions maximizing polymorphism-associated risk but also to provide a high level of statistical confidence given the multiple comparisons inherent in the screening protocol. Additionally, we sought to assess risk within the subgroup potentially associated with a set of thrombogenic, inflammatory, and metabolic blood markers. We did this by graphically analyzing recurrent coronary event rates as a function of serum total cholesterol and triglyceride levels using outcome event mapping,7–9 an exploratory data analysis tool that allows identification of patient subgroups based on risk.
Nondiabetic patients (n=846) of the THROMBO postinfarction study comprised the study population including 749 patients with genotyping results for the PAI-1 4G/5G polymorphism. The study was carried out with approval of and according to guidelines of the Research Subjects Review Boards of participating centers. Recurrent coronary outcome events included cardiac death, MI, or unstable angina (hospitalization during follow-up with an increase in either frequency or duration of angina symptoms, or development of new angina at rest—whichever occurred first with both requiring ischemic ECG changes without enzyme elevation). Average patient follow-up was 26 months.
Fasting sera were drawn 2 months after index MI and were analyzed as described previously5,10 for the set of blood markers: apoB, total cholesterol, lipoprotein-associated phospholipase A2 (Lp-PLA2), apoA-I, HDL, triglyceride, LDL peak particle diameter (LDL-PPD), glucose, insulin, PAI-1, lipoprotein(a) (Lp(a)), C-reactive protein (CRP), von Willebrand factor antigen (vWF), fibrinogen, D-dimer, factor VII, and factor VIIa. Gradient gel electrophoresis was used to fractionate LDL particles according to size9 with approximate size ranges (nm) as follows: LDL1, 26.4 to 29; LDL2, 25.5 to 26.4; LDL3, 24.2 to 25.5; and LDL4, 21.0 to 24.2.
Selection of Polymorphisms and Genotyping
Briefly and as described previously,11 genetic variants in genes coding for proteins involved in coagulation, oxidative stress, inflammation, vascular reactivity, insulin resistance, apolipoproteins, and triglyceride-rich lipoprotein metabolism were selected based on: pathophysiologic associations, polymorphism location within the gene, available information on gene regulation, and reported associations with CVD. Genotyping of candidate polymorphisms was performed as described previously.11 In particular, genotyping for PAI-I 4G/5G insertion polymorphism (refSNP ID: rs34857375) was performed by melting curve analysis (Light Cycler, Roche Diagnostics) based on G-nucleobase quenching using a 3′-FAM labeled probe (CTCCCCACGTGTC) spanning the polymorphic locus, in which the 5G variant results in a 3′ 2-bp mismatch. This assay was validated in selected samples of each genotype by comparison with polymerase chain reaction (PCR)-restriction fragment length polymorphism (RFLP) gel electrophoresis assay (amplified by GCCCTCAGGGGCACAGAGAGAGTCTGGCCA and GCAATGCAGCCAGCCACGTG primers to give a 163-bp product that when digested by BslI endonuclease gives 107/56-bp [4G] or 74/56/34 [5G] products).
Graphical and statistical analyses were performed with Statistica 7.0 (StatSoft Inc). Variables were adjusted for age using linear regression. Significant differences (P<0.05 level, unless stated otherwise) between and among populations were assessed using the Mann–Whitney U test and Kruskal-Wallis test using the Bonferroni correction for post-hoc testing; comparison of proportions and genotypic distributions were performed using the Chi-square test.
Outcome event mapping7–9 was used to graphically delineate risk over a bivariate risk parameter domain (in this case, cholesterol and triglyceride levels). Briefly, 3-dimensional scatter plots of patients were generated with outcome plotted on the z axis (without and with recurrent events coded as 0 and 1, respectively) and cholesterol and triglyceride levels (transformed to ranks to more evenly distribute patient values over the bivariate risk domain) plotted over the x-y plane. A smoothing algorithm was then applied to give a surface (outcome event map) with height over the cholesterol/triglyceride plane corresponding to estimated outcome rate. To define regions of polymorphism-associated risk (regions with differential outcome rates for the dichotomized polymorphism), separate outcome maps were generated for each of the 2 polymorphism variants and then compared.
Kaplan-Meier analysis with log rank statistic and Cox multivariable proportional-hazards regressions (blood markers treated as continuous variables; PAI-1 polymorphism as a binary variable) were used to follow outcomes over time. Univariate Cox regressions were run to assess significance levels for clinical covariates (P<0.1; sex, race, smoking, prior MI, pulmonary congestion, ejection fraction, and claudication), blood markers (P<0.05), and the PAI-1 polymorphism (P<0.05). Multivariable models were then formulated as simultaneous entry of all univariate significant variables. Medication effects (P<0.05; statins, beta blockers, aspirin, calcium channel blockers, nitrates, ACE-inhibitors, and oral anticoagulants) were assessed by individual addition to multivariable models.
Clinical and laboratory characterization of the postinfarction study population are essentially the same as reported previously7,9 with notable features including that: 77.7% were male; 78.0% were white; and they were overweight (mean BMI, 27.7±4.93 kg/m2). Mean blood marker levels outside standard reference ranges included HDL (1.01±0.29 mmol/L; normal ≥1.03 mmol/L), triglycerides (2.25±1.29 mmol/L; normal <1.69 mmol/L), CRP (4.37±6.82 mg/L; normal <1.0 mg/L), and D-dimer (510±635 μg/L; normal <200 μg/L).
Samples were typed for 37 CVD-associated polymorphisms (Table 1).11 Recurrence rates for dichotomized polymorphism variants were then compared in normolipidemic patients (n=239; total cholesterol <5.17 mmol/L [200 mg/dL] and triglycerides <1.69 mmol/L [150 mg/dL]). Only the 4G/5G polymorphism of PAI-1 showed a statistically significant difference in outcome rates (12.1% in 4G/5G plus 5G/5G patients versus 25.0% in 4G/4G patients, P=0.039).
4G/5G PAI-1 Polymorphism
Genotyping results were available for 749 of the 846 total study patients with frequencies in Hardy-Weinberg equilibrium as follows: 5G/5G - 200 (26.7%), 4G/5G - 368 (49.1%), and 4G/4G - 181 (24.2%). PAI-1 levels as a function of genotype variant were: 5G/5G - 22.4±18.9 μg/L, 4G/5G - 27.2±26.2 μg/L, and 4G/4G - 30.2±23.8 μg/L (Kruskal-Wallis, P=0.0057). Significance level in post-hoc testing for difference in PAI-1 levels between 4G/5G patients and 4G/4G patients (P=0.033) was smaller than the corresponding value between 4G/5G and 5G/5G patients (P=0.15) supporting dichotomization as: 5G/5G plus 4G/5G patients versus 4G/4G patients. Dichotomization as such resulted in significant differences between homozygous 4G patients and patients with the 5G allele as: higher proportion of White patients (92.3% versus 75.7%, P=0.000001), and higher levels of PAI-1 (30.2±23.8 μg/L versus 25.5±24.0 μg/L, P=0.004), and triglycerides (2.44±1.37 mmol/L versus 2.20±1.26 mmol/L, P=0.015).
Interaction of Cholesterol and Triglycerides
Risk of recurrence as a function of cholesterol and triglyceride levels is demonstrated by outcome event mapping (Figure 1). The contour map reveals a major high-risk region (high cholesterol/high triglycerides), several minor high-risk regions, and 2 low-risk regions (low cholesterol/low triglycerides and low cholesterol/high triglycerides).
To delineate regions of differential PAI-1 polymorphism-associated risk, mappings were generated for 5G carriers (Figure 2A) and for 4G/4G patients (Figure 2B). Both mappings show a high-risk region (high cholesterol/high triglycerides) corresponding to the high-risk subgroup of Figure 1. This suggests lack of PAI-1 polymorphism-associated risk in this subgroup. In contrast, the normolipidemic region shows high risk for 4G/4G patients (Figure 2B) and low risk for 5G carriers (Figure 2A). This suggests PAI-1 polymorphism-associated risk for normolipidemic patients.
Normolipidemic Subgroup Characterization
To define a normolipidemic subgroup beyond simple univariate cut points for cholesterol and triglyceride levels that exhibits maximal functional dependence of risk on the PAI-1 polymorphism, Figure 3 shows the contour mapping for 4G/4G patients with iso-contour lines of risk at the outcome event rate (0.23) that just separates normolipidemic and hyperlipidemic higher-risk subgroups. Superimposed on this is the scatter plot of all patients in the study group. Thus, 182 patients were identified in the normolipidemic subgroup (21.5% of total study population; 161 with PAI-1genotyping); and 136 patients in the hyperlipidemic subgroup (16.1% of total study population; 118 with PAI-1 genotyping).
Clinical and laboratory characterization of the normolipidemic subgroup for homozygous 4G patients versus 5G carriers (Table 2) demonstrated significantly higher rate of recurrent coronary events, less aspirin use, and lower levels of Lp(a) and D-dimer in homozygous 4G patients. Recurrent rates in the hyperlipidemic subgroup did not differ according to PAI-1 polymorphism variant (37.9% versus 22.7%, P=0.22).
PAI-1 Polymorphism and Recurrent Coronary Event Risk
Kaplan-Meier analysis (Figure 4) was used to demonstrate risk of recurrence in the 2 subgroups as a function of the PAI-1 polymorphism. Figure 4 shows for normolipidemics significantly worse outcome for homozygous 4G patients in comparison to 5G carriers (P=0.0002). In contrast, outcome was not significantly different (P=0.12) in hyperlipidemics; even though, homozygous 4G patients demonstrated similar trends of poor outcome as in normolipidemics suggesting 4G allele-associated risk in both subgroups. Thus, lack of PAI-1 polymorphism-associated risk in the hyperlipidemic subgroup was not attributable to lack of 4G allele-associated risk but rather to comparable risk even in 5G carriers.
Multivariable Risk Models in Normolipidemics
The PAI-1 polymorphism and set of blood markers were examined as risk predictors in the normolipidemic subgroup using Cox multivariable proportional hazards regression adjusted for significant clinical covariates and medication use. The PAI-1 polymorphism was treated as a dichotomous variable (5G carriers coded as 0, 4G/4G patients coded as 1) and blood markers as continuous variables. The only statistically significant result of univariate Cox modeling was for the PAI-1 polymorphism (hazard ratio [HR] 4.02, 95% CI 1.86 to 8.69, P=0.00039); none of the blood markers or clinical covariates achieved statistical significance (although race approached significance, P=0.15). Thus, multivariable modeling with individual entry for medication effects demonstrated significant association with recurrence only for the PAI-1 polymorphism (HR 4.30, 95% CI 1.98 to 9.33. P=0.00023) and calcium-channel blockers (HR 2.47, 95% CI 1.06 to 5.76, P=0.037) without interaction between the 2 (P=0.64). The high level of significance in the hazard ratio of the PAI-1 polymorphism substantiates the validity of the finding of risk in homozygous 4G patients especially in view of the problem of multiple comparisons inherent in the original polymorphism screening protocol. Furthermore, multivariate modeling with an independent variable representing the 3 genotypes of the PAI-1 polymorphism gave 1.97(1.08 to 3.58) for hazard ratio and 95%CI with a probability value of 0.027 indicative of a dosage effect on risk of the 4G allele.
Because of a potential connection of race with risk as indicated above, a multivariable model was constructed forcing race (black coded as 0; white coded as 1) into the model. Results indicated little effect of race on PAI-1 polymorphism-associated risk as follows: race (HR 2.61, 95% CI 0.61 to 11.26, P=0.20), calcium channel blockers (HR 2.78, 95% CI 1.17 to 6.61, P=0.02), and PAI-1 polymorphism (HR 4.14, 95% CI 1.89 to 9.05, P=0.00038). This was further substantiated by results of univariate Cox modeling of the polymorphism in white subgroup patients (n=132) (HR 4.14, 95% CI 1.85 to 9.24, P=0.00054). Low patient number (n=29) precluded corresponding analysis for Black patients of the subgroup. Additionally, regarding the previous finding of less aspirin use in homozygous 4G subgroup patients versus 5G carriers, outcome rates for homozygous 4G patients on and not on aspirin were 34.8% and 41.7%, respectively, but the difference was not significant (P=0.69). Further, forcing aspirin use into multivariable models left polymorphism-associated risk essentially unchanged and inclusion of an interaction term between aspirin use and polymorphism was not statistically significant.
The 4G Allele and Potentiation of Risk by LDL
In view of reported 4G/5G allele-specific effects by agents including lipoproteins,6 assessment of LDL particle size on potentiation of PAI-1 polymorphism-associated risk was performed using LDL particle subfractionation according to size in the normolipidemic subgroup as a function of outcome and the PAI-1 polymorphism. Percentage changes in fractional absorbance for patients with outcome events relative to patients without outcome events for carriers of the 5G allele were +0.2% (LDL1), +3.8% (LDL2), −14.3% (LDL3), +9.5% (LDL4); whereas for patients homozygous for the 4G allele, corresponding values were +23.0% (LDL1), −30.3% (LDL2), −37.3% (LDL3), −14.9% (LDL4). Results were statistically significant only for LDL1 in patients homozygous for the 4G allele signifying higher proportion of large LDL particles in patients with recurrence versus those without recurrence.
Of the 37 polymorphisms in CVD-associated genes screened, only the 4G/5G insertion/deletion polymorphism in PAI-1 was found to be associated with recurrent risk in a subgroup of nondiabetic, normolipidemic MI patients (HR - 4.02, P=0.00039, for 4G/4G versus 4G/5G plus 5G/5G patients). This finding was robust to adjustments for significant clinical parameters and medication use and inclusion of 17 thrombogenic, inflammatory, and metabolic blood markers. This result provides important new information regarding recurrent risk beyond incident MI in normolipidemic postinfarction patients (nearly 20% of study population). Additional results suggested potentiation of risk in 4G/4G patients by large LDL particles.
Multiple studies report association of the 4G allele with increased PAI-1; however, association studies of the 4G allele with CVD risk show mixed results.6,12 This tendency was supported by recent meta-analyses of 7 hemostatic gene markers including the 4G/5G PAI-1 polymorphism that demonstrated weak but significant association of the 4G/4G variant with CVD risk13; however, results limited to 5 large studies demonstrated loss of significance of the association. In contrast, results of the current study demonstrated highly significant polymorphism-associated risk, but this was limited to a subgroup of normolipidemic patients. This finding underscores the potential of heterogeneous distribution of polymorphism-associated risk as a basis for widely discrepant results in previous studies.
Our study also found association of large LDL particles with homozygous 4G subgroup patients having recurrent events. We speculate that this association represents allele-specific interactions of the PAI-1 promoter site with lipoprotein-induced transcriptional factors leading to acutely heightened PAI-1 responses. This is consistent with reported greater influence of the 4G allele on acute regulation of PAI-1 levels versus basal levels in response to multiple agents, especially interleukin (IL)-1 and VLDL.6 This may help to explain lack of association of risk with increased PAI-1 in the current study as levels were determined in a stable period 2 months after index MI. With regard to VLDL, an identified VLDL responsive element adjacent to the 4G/5G polymorphism site is thought to serve as a basis for allele-specific effects of triglyceride-rich lipoproteins on PAI-1 blood levels.14–16 However, regarding our finding in homozygous 4G patients of risk in association with large LDL particles, although native LDL is not thought to be a strong determinant of PAI-1 levels or responses, modified LDL (oxidized, glycated, triglyceride-rich) does appear to be active in this regard.6 Indeed, consistent with our LDL results are findings from a report by Allison et al17 showing that: triglyceride-enriched LDL (large LDL) is a potent activator of PAI-1 protein and mRNA; potency of stimulation correlates with LDL triglyceride content; and triglyceride-enriched LDL induces the same transcription factor as VLDL. Conclusive demonstration of involvement in polymorphism-associated risk of allele-specific interactions with triglyceride-rich lipoproteins affecting acute responses in PAI-1 levels would require additional specifically focused studies.
Results and conclusions of our study were limited by several factors. Although the study population was large, female and Black patient numbers were not high enough to permit performance of relevant subgroup analyses. Further, additional information on diet, ethanol use, exercise, social support, and depression might be helpful. The current work involved studies using single polymorphisms rather than more extensive haplotype analysis (most often used for localization of genetic loci associated with specific effects). However, results of the present work demonstrating highly significant association of the PAI-1 polymorphism with risk along with known location of the polymorphism in the promoter region of the gene and established findings underlying allele-specific gene expression aid in validating our approach. Strengths of the work include use of a set of 37 genetic polymorphisms associated with CVD risk as well as a set of 17 thrombogenic, inflammatory, and metabolic blood markers in a prospective study with long-term follow-up. Additionally, using outcome event mapping, a subgroup of patients defined by maximal functional dependence of risk on a genetic polymorphism was identified that led to a high level of confidence in results especially important given the multiple comparisons inherent in the screening protocol.
In summary, results of our study show the 4G/5G polymorphism in the promoter region of the PAI-1 gene to be associated with risk for recurrent coronary events in a subgroup of normolipidemic postinfarction patients. In comparison to 5G allele carriers, subgroup patients homozygous for the 4G allele are at greater risk (HR 4.02) for recurrent events. Further, it was only the PAI-1 polymorphism that predicted risk out of a set of 37 CVD-associated genetic markers and a set of 17 thrombogenic, inflammatory, and metabolic blood markers—a finding especially significant in view of limited information on this important patient group and indicative of a critical role for thrombogenesis in production of risk. Also, results of the current study indicate association of large LDL particles with recurrent events in homozygous 4G patients supporting the notion of risk production through allele-specific interactions with triglyceride-rich lipoproteins. Further studies are needed to confirm the role of the 4G/5G PAI-1 polymorphism as a predictor of risk and especially to characterize phenotypic settings that foster such risk.
We are indebted to the Study Coordinators who enrolled and followed up the patients from the 13 participating centers.
Sources of Funding
This study was supported by research grant HL-48259 from the National Institutes of Health, Bethesda, Md and by a contract from Millennium Pharmaceuticals Inc, Cambridge, Mass.
Original received September 10, 2007; final version accepted December 5, 2007.
Corsetti JP, Rainwater DL, Moss AJ, Zareba W, Sparks CE. High lipoprotein-associated phospholipase A2 is a risk factor for recurrent coronary events in postinfarction patients. Clin Chem. 2006; 52: 1331–1338.
Corsetti JP, Ryan D, Moss AJ, Zareba W, Sparks CE. NAD(P)H oxidase polymorphism (C242T) and high HDL cholesterol associate with recurrent coronary events in post infarction patients. Atherosclerosis. In press.
Moss AJ, Goldstein RE, Marder VJ, Sparks CE, Oakes K, Greenberg H, Weiss HJ, Zareba W, Brown MW, Liang CS, Lichstein E, Little WC, Gillespie JA, VanVoorhees L, Krone RJ, Bodenheimer MM, Hochman J, Dwyer E, Arora R, Marcus FI, Watelet. LFM, Case RB. Thrombogenic factors and recurrent coronary events. Circulation. 1999; 99: 2517–2522.
Moss AJ, Ryan D, Oakes D, Goldstein RE, Greenberg H, Bodenheimer MM, Brown MW, Case RB, Dwyer EM, Eberly SW, Francis CW, Gillespie JA, Krone RJ, Lichstein E, MacCluer JW, Marcus FI, McCarthy J, Sparks CE, and Zareba W. Atherosclerotic risk genotypes and recurrent coronary events after myocardial infarction. Am J Cardiol. 2005; 96: 177–182.
Ding J, Nicklas BJ, Fallin MD, de Rekeneire N, Kritchevsky SB, Pahor M, Rodondi N, Li R, Zmuda JM, Harris TB. Plasminogen activator inhibitor type 1 gene polymorphisms and haplotypes are associated with plasma plasminogen activator inhibitor type 1 levels but not with myocardial infarction or stroke. Am Heart J. 2006; 152: 1109–1115.
Eriksson P, Kallin B, van’t Hooft FM, Bavenholm P, Hamsten A. Allele-specific increase in basal transcription of the plaminogen-activator inhibitor 1 gene is associated with myocardial infarction. Proc Nat Acad Sci USA. 1995; 92: 1851–1855.
Eriksson P, Nilsson L, Darpe F, Hamsten A. Very-low-density lipoprotein response element in the promoter region of the human plasminogen activator inhibitor-1 gene implicated in the impaired fibrinolysis of hypertriglyceridemia. Arterioscler Thromb Vasc Biol. 1998; 18: 20–26.
Allison BA, Nilsson L, Karpe F, Hamsten A, Eriksson P. Effects of native, triglyceride-enriched, and oxidatively modified LDL on plasminogen activator inhibitor-1 expression in human endothelial cells. Arterioscler Thromb Vasc Biol. 1999; 19: 1354–1360.