Background: Interleukin (IL)-18 is a proinflammatory cytokine that has been implicated in several diseases, including atherosclerosis, and increased circulating IL-18 concentrations increase risk of future coronary heart disease (CHD). We evaluated the effect of common variation within the IL18 gene on concentrations of circulating IL-18.
Methods: We measured IL-18, by ELISA, in the population-based study group [Carotid Ultrasound Disease Assessment Study (CUDAS)] and a predominantly male cohort with premature cardiovascular disease [Carotid Ultrasound in Patients with Ischaemic Heart Disease (CUPID)]. Using a tagging single-nucleotide polymorphism (SNP) approach that captured >90% of genetic variation, we identified 4 common (>10%) haplotypes.
Results: A common SNP was associated with differences in IL-18 concentrations; in CUDAS individuals carrying 2 copies of the rare allele, concentrations were 13% higher than in those with no copies (P = 0.002). Haplotypes were also associated with significant differences in IL-18 concentrations in CUDAS and CUPID. Haplotype GTATA (frequency 23%) was associated with significantly lower IL-18 than others. In CUDAS, those carrying 2 copies had IL-18 concentrations 15% lower than those carrying no copies (P = 0.002); in CUPID, the difference was 22% (P = 0.004). These associations remained significant after adjustment for age, sex, hypertension, HDL cholesterol, waist-to-hip ratio, and alcohol consumption. Despite being associated with differences in IL-18 concentrations, the haplotypes did not occur at different frequencies in those with or without carotid atherosclerotic plaques.
Conclusions: Variation within IL18 affects IL-18 concentrations in healthy and diseased individuals and thus may influence the pathophysiology of plaques at all stages of CHD progression.
Interleukin (IL)1 -18, a pleiotropic cytokine involved in both the innate and the adaptive immune response, is widely expressed in monocytes/macrophages, adipocytes, keratinocytes, Kupffer cells, and osteoblasts(1). Originally identified as an interferon-γ–inducing factor(2), it stimulates interferon-γ production in T lymphocytes and natural killer cells, both of which are key components of atherosclerotic plaque progression and stability(3). Indeed, increased IL-18 expression is seen in atherosclerotic plaques and is associated with plaque instability(4). The use of animal models has further demonstrated the influence of IL-18 and the benefit of its inhibition [through IL-18 binding protein (IL-18BP), its intrinsic inhibitor] on plaque progression and composition(5)(6)(7)(8).
Although IL-18 experimental data appear to be consistent, clinical studies have proven inconsistent and controversial. A large study in healthy, middle-aged European men has shown that total plasma IL-18 concentrations are an independent predictor of coronary events(9), and importantly, variation within the IL-18 gene (IL18)2 has been shown to influence circulating concentrations of IL-18 and clinical outcome in patients with coronary heart disease (CHD)(10). However, recent findings from the Monitoring of Cardiovascular Disease (MONICA)/Cooperative Health Research in the Augsburg Region (KORA) Ausburg Case-Cohort Study show no association between IL-18 and risk of future CHD in apparently healthy individuals(11). Following on from the experimental data outlined above, several groups have sought to clarify the role of IL-18 and IL-18BP in the development of plaque instability in cross-sectional studies(12)(13)(14), but again with inconsistent results.
The exact nature of IL-18’s influence on CHD risk remains unclear. IL-18 shares similar downstream signaling pathways with several other cytokines [interaction with MyD88(15), nuclear translocation of NF-κB(16)], but this may not be the case for all cell types(17). The fact that administration of IL-18 does not induce fever in animals(18), yet both IL-6 and IL-1 do(19), suggests substantial differences in site and method of action between the cytokines. Also, IL-18 was found to be predictive of future CHD events independent of C-reactive protein(9), which further supports this hypothesis. Therefore, if IL-18 is part of an alternative inflammatory pathway in CHD, knowledge of important genetic effects may improve risk prediction beyond that of other inflammatory mediators.
The purpose of this study was to investigate the effect of genetic variation within IL18 on circulating IL-18 concentrations in a large cross-sectional community-based study and in a medically stable cohort with angiographically demonstrated CHD, thereby examining the impact of genetic variation within IL18 and its potential importance in determining CHD risk.
Materials and Methods
Individuals were selected from 2 Western Australian studies: the Carotid Ultrasound Disease Assessment Study (CUDAS)(20) and the Carotid Ultrasound in Patients with Ischaemic Heart Disease (CUPID) study(21). Both groups are predominantly European-Australian. The CUDAS group consists of 1109 participants with an equal male-to-female ratio and equal numbers in each age decile between 20 and 70 years. Participants took part in the original 1989 Australian National Heart Foundation Perth Risk Factor Prevalence Survey and agreed to attend the study clinic between June 1995 and December 1996, having not previously had carotid artery surgery. The CUPID cohort was collected in 1995 and consisted of 556 individuals (487 men) between 26 and 60 years of age with premature CHD. All CUPID participants were medically stable at the time of data collection but had a history of angina, unstable angina, or myocardial infarction and angiographically demonstrated CHD with at least 1 coronary vessel with >50% stenosis. Written informed consent was obtained from all study participants, and the Institutional Ethics Committee of the University of Western Australia approved the study protocol.
A self-administered questionnaire was used to report the prevalence of smoking, physician-diagnosed hypertension, diabetes, angina pectoris, myocardial infarction, stroke, and medication use among study participants. We calculated smoking lifetime exposure by pack-years, measured resting systolic and diastolic blood pressures with a mercury column manometer, and recorded anthropometric measurements (waist and hip circumferences, height, and weight). A fasting venous blood sample was collected from each participant, and sera were stored at −70 °C until analysis. We performed bilateral carotid B-mode ultrasound using a 7.5-MHz annular phased-array transducer on an Interspec (Apogee) CX 200 ultrasound machine(20)(21). The ultrasound study was used to determine the presence of focal carotid plaque and measure the mean common carotid artery intima-medial wall thickness (IMT) as described(20)(21). Presence of an atherosclerotic plaque was defined as a clearly identified area of focal increased thickness (>1 mm) of the intima-media layer.
IL18 single-nucleotide polymorphism identification and genotyping
Tagging single-nucleotide polymorphism identification.
We used IL-18 resequencing data from Innate Immunity PGA (IIPGA,http://innateimmunity.net) in conjunction with a haplotype-tagging single-nucleotide polymorphism (tSNP) program, tagSNPs.exe(22). A previous report gives greater detail of how the single-nucleotide polymorphisms (SNPs) were selected(23).
We carried out SNP genotyping using TaqMan protocols (for primer and probe pairs, see the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol53/issue12) using a 5-μL reaction. Samples were cycled for 40 cycles on MJ Research cyclers and read using a Victor2 fluorometer (Perkin-Elmer). Undetermined samples were regenotyped using tetra-primer amplification refraction mutation system-PCR or restriction fragment length polymorphism protocols (for primer and probe pairs, see online Data Supplement). Genotypes for all SNPs were recorded for >99% of individuals in CUDAS and >98% in CUPID.
We measured serum IL-18 by use of a commercially available ELISA method (MBL Co. Ltd.) as described(9). The ELISA uses monoclonal antibodies against 2 different epitopes of human IL-18, one of which is peroxidase-conjugated. The within-run CV was 5.4% at a mean value of 400 μg/L (28 samples); between-run CV was 8.2% at 298 μg/L (9 samples) and 7.8% at 496 μg/L (9 samples). Total cholesterol, HDL, and triglyceride concentrations were measured enzymatically(24).
The primary quantitative outcome variable of the association analyses was IL-18 concentration. The principal explanatory variables were the 5 genotyped IL18 polymorphisms and inferred haplotypes. The SNP genotypes were coded into 3 classes and analyzed categorically, with the most common homozygous genotype as a reference category to which the effect of the other 2 genotypes was compared. Haplotypes were recoded as independent variables into 3 classes (0, 1, or 2), representing the number of copies of each haplotype that made up an individual’s diplotype. They were then analyzed categorically with zero copies as the baseline. The genetic effect types of the mutations for SNPs and haplotypes were determined to be additive, dominant, or recessive by analyzing the trend of the β-coefficients for each category.
We tested each SNP for departure from Hardy–Weinberg equilibrium (HWE) using a modified Markov-chain random walk algorithm(25). We analyzed pairwise linkage disequilibrium (LD) by use of a likelihood-ratio test, whose empirical distribution was obtained via a permutation procedure(26). Lewontin disequilibrium coefficient D′ was calculated for each pairwise comparison(27).
Sex was analyzed as a dichotomous variable. All other measurements were analyzed as continuous, gaussian-distributed variables. IL-18 concentration was not gaussian distributed and was therefore logarithmically (base 10) transformed before analysis. Multivariate analyses used generalized linear models (linear and logistic regression)(28) to model the effects of multiple covariates on continuous and dichotomous outcomes. We used both forward and backward stepwise variable selection procedures to determine a useful subset of independent predictors on the outcomes of interest. Checks of goodness of model fit included examination of Akaiki information criteria to determine the models that best fit the data. In the multivariate analysis presented here those predictors were age, sex, hypertension, HDL, waist-to-hip ratio, and alcohol consumption. Checks of goodness of fit(29) included an investigation of the need for interaction or polynomial terms. Haplotypes were inferred for individuals with ambiguous phase, and haplotype frequencies were estimated using an expectation-maximization algorithm as described by Excoffier and Slatkin(30). Using this approach, it was not possible to statistically test for differences in haplotype frequencies between the study groups, as the frequencies are estimated and not directly observed. We used SimHap (www.genepi.com.au/projects/simhap), R(31), and JLIN(32) to manage and analyze data. P values were derived via empirical simulation where possible. Statistical significance was defined at a nominal 5% level. No adjustments were made for multiple testing, because this has been suggested to lead to errors in interpretation(33).
The baseline characteristics of both study groups are shown in Table 1⇓ . CUPID is a diseased cohort, while CUDAS is a population-based sample; as such, the proportion of smokers and those with metabolic syndrome is far greater in CUPID than CUDAS. Furthermore, the number of individuals using cholesterol-lowering and antihypertensive medication was far greater in CUPID than CUDAS, and therefore mean lipid concentrations and blood pressure are similar in the 2 groups. That concentrations of IL-18 were significantly higher in CUPID than CUDAS is likely to be accounted for by an increased prevalence of obesity and metabolic syndrome in CUPID; IL-18 concentrations strongly associate with obesity and metabolic syndrome risk traits(24)(34).
tSNP selection and validation
A tSNP set comprising SNPs −9731 G>T, −5848 T>C, +4860 A>C, +8855 T>A, and +11015 A>C was selected (rs1946519, rs2043055, rs549908, rs360729, and rs3882891), based on haplotypes derived from resequencing data of white individuals(23). Genotypes for all 5 SNPs were determined in all study groups. LD was significant and high across the whole region, as shown in Fig. 1⇓ (all D′ >0.72 in CUDAS). +4860 A>C was the only exonic SNP found within IL18 by IIPGA; however, it is nonsynonymous—no other potentially functional SNPs were found.
allele and haplotype frequency
The genotype distribution of +8855 T>A differs significantly from that expected by HWE, with fewer heterozygotes observed than expected in both CUDAS and CUPID (Table 2⇓ ). This may be due to the clustering methods used in TaqMan genotyping. The intermediate cluster (relating to heterozygotes) tends to be less focused, therefore making undetermined heterozygote samples more likely. The distribution of +8855 genotypes in previous population-based studies we have genotyped(23) have also shown that it has a tendency todeviate from HWE; therefore, we elected to analyze the data cautiously and have included and excluded +8855 in the analyses to follow.
As shown in Table 2⇑ , there were no significant differences in allele frequency between CUDAS and CUPID.
The 5 IL18 polymorphisms generated 4 common haplotypes (frequency >10%) accounting for 93% of the inferred haplotypes for both cohorts. There was no substantial difference in haplotype frequencies or organization between study groups (Fig. 1⇑ ). As expected from the observation of no difference in allele frequency, it appears there is no difference in genetic architecture between the study groups.
il-18 concentrations: single snp analysis
There were no consistent single SNP associations with IL-18 concentrations across both studies (Table 3⇓ ), and none of the SNPs explained substantial proportions of the variance in IL-18 concentrations in either study (CUDAS: all SNPs <0.9%; CUPID: all SNPs <1.2%). In CUDAS, IL-18 concentrations were only significantly different by −5848 genotype (P = 0.003), with the rare allele associated with higher concentrations than the common allele. In CUPID, however, −5848 was not associated with IL-18 concentrations globally (P = 0.14), but both heterozygotes and rare-allele homozygotes had IL-18 concentrations higher than common-allele homozygotes, though not significantly (P = 0.05 and 0.74, respectively). Furthermore, both −9731 and +11015 were significantly associated with differences in IL-18 concentrations (P = 0.02 and 0.01, respectively); in both cases the rare allele was associated with higher IL-18 concentrations. These associations remained significant in multivariate analysis.
il-18 concentrations: haplotype analysis
To ensure sufficient statistical power in the following analyses, analysis was restricted to common haplotypes observed at a frequency >10%.
Haplotypes were significantly associated with differences in IL-18 concentrations in CUDAS (Table 4⇓ ). More specifically, the GTATA haplotype (hGTATA) was associated with significantly lower IL-18 concentrations (P <0.001) and explained 1.7% of variance in IL-18 concentrations. In those carrying 2 copies of the haplotype, IL-18 concentrations were 15% lower than in those carrying no copies (P = 0.002). The effect appeared to be additive, with those individuals carrying 1 copy of the haplotype having concentrations midway between those carrying 2 copies and no copies. hGCATA also appeared to have an effect on IL-18 concentrations, with those carrying 2 copies of the haplotype having significantly higher IL-18 concentrations than those carrying no copies (P = 0.02). Globally, hGCATA explained 0.4% of the variance in IL-18 concentrations and was marginally associated with differences in IL-18 concentration (P = 0.04). The effect appeared to be additive, although for individuals carrying only 1 copy of hGCATA the association did not reach statistical significance (P = 0.08). Neither of these associations was attenuated in multivariate analysis or in analysis that excluded +8855 (data not shown).
Both hGTATA and the TTATC haplotype (hTTATC) were associated with statistically significant differences in IL-18 concentrations (P <0.001 and P = 0.001, respectively; Table 4⇑ ), and explained 3.0% and 2.1% of the variance in IL-18 concentrations, respectively. Furthermore, both associations were additive. Individuals with 2 copies of hGTATA had IL-18 concentrations, on average, 22% lower than those with no copies of the haplotype (P = 0.004), whereas those carrying 1 copy of the haplotype had concentrations 12% lower than those with no copies (P <0.001). hTTATC operated similarly but in the opposite direction, with those carrying 2 copies having estimated geometric mean IL-18 concentrations 55% higher than those carrying no copies (P = 0.02), whereas those with 1 copy had concentrations 15% higher than those with no copies (P = 0.003). Both associations were not attenuated in multivariate analysis, or in analysis that excluded +8855 (data not shown).
carotid plaque and mean imt haplotype analysis
All participants in both studies were assessed for carotid IMT and evidence of plaques by ultrasound. Because IL-18 is produced in atherosclerotic plaques and related to plaque instability(4)(8), it is possible that genetic variants that alter IL-18 concentrations could affect plaque initiation and progression. When this hypothesis was tested in CUDAS and CUPID, none of the IL18 haplotypes were associated with risk of carotid plaque or mean IMT (data not shown). However, hTTATC, shown to be associated with higher IL-18 concentrations, was borderline significantly associated with a higher risk of carotid plaque in CUPID (P = 0.05).
The data presented here demonstrate that genetic variation within the IL18 gene affects IL-18 protein concentrations and may therefore affect IL-18 concentrations at the site of disease.
In single SNP analysis, only 1 SNP, −5848, was associated with differences in serum IL-18 concentrations. −5848C was seen only on 1 common haplotype, hGCATA, and as might be expected, there was evidence that this haplotype was associated with significantly higher IL-18 concentrations. By contrast, −5848T was seen on 3 common haplotypes. One of these, hGTATA, was associated with significant differences in IL-18 concentrations in both study groups. This haplotypic effect was of greater impact, in terms of difference in IL-18 concentrations, than that seen with hGCATA and was of greater statistical significance. Of interest, hGCATA was not significantly associated with differences in IL-18 concentrations in CUPID (P = 0.88), suggesting that the effect observed in the healthy CUDAS group may be attenuated or masked in diseased individuals.
Further evidence for hGTATA’s effect on IL-18 concentrations comes from the AtheroGene cohort. Tiret et al.(10) highlighted the role of the IL18 gene in cardiovascular disease, demonstrating that IL18 haplotypes caused variation in IL-18 serum concentrations and were associated with cardiovascular mortality. The 5 IL18 tSNPs genotyped here were not the same as those used by Tiret et al.(10). Because both are studies of whites, however, it can be deduced, with IIPGA IL18 resequencing (http://innateimmunity.net/), that hGTATA marks the same haplotype previously found to be associated with lower IL-18 concentrations and a protective effect on risk. Further data have demonstrated that this haplotype has a significant effect on IL18 mRNA concentrations in transformed lymphoblastoid cell lines(35), although the exact functional mechanism was not elucidated. Furthermore, Frayling et al.(36) studied the relationship of the rs5744256 SNP on IL-18 production and found the rare allele to be associated with lower IL-18 concentrations. Again, using the IIPGA gene-wide haplotypes, we find that the rare allele is seen only on 1 common haplotype, which was represented by hGTATA in this study. Therefore the same haplotype has been associated with lower IL-18 concentrations in 4 separate studies by 3 separate laboratories, as well as differences in body mass index in a further 2 disease cohorts(23). The SNP effect of −5848 may itself be functional, or it is most probably derived from the haplotype effect, with the true functional SNP being elsewhere on the haplotype, as none of the other haplotypes carrying −5848T showed a consistent effect. Furthermore, hGCATA may also be mirroring hGTATA’s association, in a reverse direction.
hTTATC was also associated with significant differences in plasma IL-18 concentrations in CUPID. However, concentrations associated with this haplotype in CUDAS did not replicate the pattern seen in CUPID. This may suggest that this genetic association, which seems distinct from that associated with hGTATA, appears only when the IL-18 system is “stressed”, by either advanced disease or other parameters that differ in CUPID compared with CUDAS (e.g., smoking, medication, or metabolic syndrome).
The lack of any strong association between IL18 genetic variation and presence of carotid plaque or carotid IMT, despite IL-18 concentrations being associated with carotid IMT in previous studies(37) but not those studied here(34), is suggestive of several explanations. First, assuming IL-18 to be causal in CHD, it may be that environmental/disease-associated variables are playing a far greater role in determining the influence of IL-18 on disease than the genetic variants studied here (this seems especially likely because the proportion of variance in IL-18 concentrations explained by genetic variation in IL18 is low), or that IL-18 production at the site of disease is not regulated in the same manner as systemic IL-18 production. Presence of carotid plaque was seen to be associated only with hTTATC. Because hTTATC was associated with some of the highest IL-18 concentrations observed, this association might reflect the need for higher IL-18 concentrations to affect plaque progression than that produced by the genetic variants described here. Second, causality for IL-18 in CHD has not yet been established, so it is entirely possible that IL-18 is simply reflecting burden of disease and is not directly causing it (reverse causality).
Several studies have shown IL-18BP to be a major determinant of free IL-18 concentrations(38)(39)(40), and the absence of IL-18BP measures in these analyses is a limitation. Without such measures, it is not possible to infer that the differences observed here are strong enough to affect free IL-18 concentrations, and whether they may affect disease progression. With regard to assessing functionality, it is our belief that the functional SNP lies on the haplotype but has not been genotyped directly. Because of the tSNP methodology, we have no means of identifying functional SNPs on hGTATA, and the task has added difficulty given the high degree of LD between the tSNPs and other SNPs within the gene.
In conclusion, we show that common genetic variation within IL18 is associated with interindividual differences in concentrations of IL-18 protein. Such an association may prove important in individual prediction of CHD risk (and other diseases in which IL-18 has been implicated).
Grant/funding support: S.R.T. was supported by a British Heart Foundation studentship FS/04/039, and S.E.H. received support from Programme Grant 2000/015. P.A.M. was supported by an Australian Postgraduate Award. The study was supported by a grant-in-aid from the National Health and Medical Research Council (211980). C.C. and J.P.B. were supported by HeartSearch (Perth, Western Australia).
Financial disclosures: None declared.
1 Data are mean (SD) unless noted otherwise. NA, not applicable.
2 P <0.05.
1 Data are estimated geometric mean IL-18 pg/mL (95% CI). P1, univariate P value compared to reference; P2, multivariate P value compared to reference.
1 Data are estimated geometric mean IL-18 pg/mL (95% CI). P1, univariate P value compared to reference; P2, multivariate P value compared to reference.
↵1 Nonstandard abbreviations: IL, interleukin; IL-18BP, IL-18 binding protein; CHD, coronary heart disease; CUDAS, Carotid Ultrasound Disease Assessment Study; CUPID, Carotid Ultrasound in Patients with Ischaemic Heart Disease; IMT, intima-media thickness; IIPGA, Innate Immunity PGA; IMT, intima-medial wall thickness; tSNP, tagging single-nucleotide polymorphism; SNP, single-nucleotide polymorphism; HWE, Hardy–Weinberg equilibrium; LD, linkage disequilibrium; hGTATA, GTATA haplotype; hTTATC, TTATC haplotype.
↵2 Human gene: IL18, interleukin 18 (interferon-γ–inducing factor).
- © 2007 The American Association for Clinical Chemistry