METHODS

Patients

Patients (n=318, aged 35-73) with signs of severe coronary atherosclerosis tested by coronary angiography (>70% stenosis in one or more arteries, and clinical signs of stable or unstable angina pectoris, typical ECG abnormalities) were enrolled. All patients had received coronary artery bypass grafting (CABG) by open heart surgery at the National Institute of Cardiology in 1995 and 1996. Healthy controls (n=248, aged 35-73) were randomly recruited from the same areas as the cases, and stratification by age and sex was used to match approximately the age and sex distribution of the controls with that of cases. The control group represents asymptomatic and apparently clinically disease free subjects (no symptoms of CAD, normal ECG, normal blood pressure). All subjects completed a series of questionnaires including questions on family and personal history of CAD. In cases, an additional questionnaire was completed, including items on previous MI and occurrence of angina pectoris. Clinical and biological features of the cases and controls are shown in table 1.

The study was approved by an institutional review committee and the subjects gave informed consent. The investigation conforms to the principles outlined in the Declaration of Helsinki.

Laboratory analysis

Total genomic DNA was extracted from white blood cells using the method of Miller et al.¹⁵ The TNFα −238 and −308 polymorphisms were determined by DNA amplification by PCR using the primers suggested by Day et al.¹ The PCR products were digested at 37°C with MspI to detect the −238 polymorphism and NcoI to detect the −308 polymorphism. The products were separated on a 4% agarose gel and stained with ethidium bromide. Bf allotypes were determined as described elsewhere.¹⁷ C4 typing was performed according to Sim and Cross.¹⁸

Serum lipid parameters (quantified by standard enzymatic procedure) and blood glucose levels were measured after overnight fasting. In patients, the blood was drawn six months after the operations.

Statistical methods

Allele frequencies were calculated by allele counting and given with an estimate of the standard error (SE). Data were analysed using MedCalc and Arlequin software.¹⁹ Distributions of the simultaneous occurrence of the alleles were tested using a likelihood ratio test. In the case of linkage disequilibrium (LD) between a pair of loci (significance level 0.05), coefficients were computed. LD coefficients (|D′|) are the ratio of the unstandardised coefficients to the maximum value they can take.²⁰ Maximum likelihood frequencies were computed using an expectation maximisation algorithm.²¹ Hardy-Weinberg equilibrium was tested by using a χ² goodness of fit test. Fisher's exact test was used to test for differences in distributions of alleles and allelic combinations between the groups. Confidence intervals were calculated at the 95% level. ANOVA test was used to estimate the impact of the polymorphisms on the quantitative traits.

RESULTS

Frequencies of the genotypes and alleles

The frequencies of the genotypes and alleles in question in CAD patients and controls are presented in table 2. Significant differences were observed in the frequencies of C4B*Q0 between CAD patients and controls. The results were overall in Hardy-Weinberg equilibrium, with the exception of the C4B*Q0 allele distribution in CAD patients, where there was deviation (χ²=12.41, p=0.0004) from the equilibrium. Analysis of the genotype distribution of this polymorphism showed that the deviation from the HWE was exclusively because of an increased frequency of the C4B*Q0/C4B*Q0 homozygotes among patients (p=0.02, odds ratio (OR)=3.8, 95% confidence interval 1.2-13.2), suggesting a recessive mode of action of this polymorphism on CAD.

Study of association between alleles tested and known risk factors for CAD

The mean values of several quantitative variables were compared between the different genotypes in both groups. No significant association with the lipid parameters (total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides) and blood pressure was observed (data not shown). The mean body mass index (BMI) in carriers of TNFα −308A was also compared with that of non-carriers, as was done by Herrmann et al.⁵ In contrast to their results, we have not found significant differences between carriers and non-carriers and no association was found between obesity and the TNFα −308A allele (data not shown).

The mean blood glucose levels of carriers of TNFα −238A were lower than that of non-carriers, 5.52 (SD 0.57) mmol/l v 6.38 (SD 2.01) mmol/l (p=0.05), but the TNFα −238A allele frequency in subjects with diabetes did not differ significantly from that of subjects without diabetes (4.3% and 4.5% in subjects with and without diabetes, respectively).

Investigation of allelic combinations

We investigated with the exact p test whether the distribution of the different combinations of alleles corresponded to the expected values. Deviation from the equilibrium was found in patients with CAD in TNFα −308A and C4A*Q0 alleles (exact p<0.001, linkage disequilibrium (LD) coefficient (±|D′|) was 0.61, that is, the observed disequilibrium is 61% of the theoretical maximum disequilibrium value) and in TNFα −238A and C4A*6 alleles (p<0.001, ±|D′|=0.71) and in both patients and controls in C4B*Q0 and Bf*F alleles (p<0.001, ±|D′|=0.3 in cases; p<0.05, ±|D′|=0.2 in controls). The linkage between TNFα −308A and C4A*Q0 alleles was supported by the findings that five patients were homozygous for the C4A*Q0 allele and three of them had AA (60%) and one GA genotype at the TNFα −308 position. Complete negative LD (±|D′|=−1; the two alleles are never present on the same haplotype) was found in patients between TNFα −308A and C4A*6 and between TNFα −238A and C4AQ*0 and in both patients and controls between C4A*Q0 and C4B*Q0.

We examined the impact of these allelic combinations on the susceptibility to CAD. We computed the maximum likelihood frequencies of the different allelic combinations with an expectation maximisation algorithm and compared them between cases and controls. The estimated frequency of the TNFα −308A + C4A*Q0 allelic combinations and that of the TNFα −238A + C4A*6 (table 3) were significantly higher in CAD patients than in controls (8% v 3.8%, and 3.3% v 0.8%, respectively). In contrast, the frequency of the TNFα −308A without C4A*Q0 (pooled all C4A alleles, which is not C4A*Q0) was higher in controls (6.8% v 10.5%). The prevalence of other allelic combinations did not differ between the two groups (data not shown). If the ratios between the frequency of patients with TNFα −308A + C4A*Q0 alleles and with TNFα −308A without C4A*Q0 were compared in cases and in controls, the p value was 0.0005 and the OR 3.2 (95%CI 1.7-6.3), indicating that the risk of having CAD was 3.2 times higher in patients with TNFα −308A+C4A*Q0 alleles than with TNFα −308A without C4A*Q0. When the same ratio between patients with TNFα −308A + C4A*Q0 alleles and with TNFα −308G + C4A*Q0 were compared in cases and in controls, the p value was 0.008 and the odds ratio 2.5 (95%CI 1.3-4.9), indicating that the risk of having CAD was 2.5 times higher in patients with TNFα −308A + C4A*Q0 than in patients with TNFα −308G + C4A*Q0 allelic combination. Similar trends were observed in respect of TNFα −238A+C4A*6 (table 3), although because of the low prevalence of C4A*6, the comparison of the ratio between patients with TNFα −238A + C4A*Q0 alleles and with TNFα −238G + C4A*6 did not show a significant difference (p=0.06). These results suggest that the two allelic combinations (TNFα −308A + C4A*Q0 and TNFα −238A+C4A*6) give higher susceptibility to CAD than either allele alone.

Differences in the frequencies of alleles and allelic combinations between CAD patients with or without myocardial infarction in their case history

To investigate the role of these alleles and allelic association in CAD, we analysed the clinical status and case histories of the patients. The CAD patients were divided into two groups based on the occurrence of myocardial infarction in their case history before the bypass surgery. The frequencies of TNFα −308A and the C4A*Q0 alleles were significantly higher in patients with MI than without MI (table 4). The TNFα −308A and the C4A*Q0 alleles occurred together significantly more frequently in patients with preoperative MI than without preoperative MI, while there was no such association in the case of TNFα −238A + C4A*6 allelic combination.

There was no association between other symptoms (extent of stenosis, pre- and postoperative angina pectoris, thrombosis, left or right heart failure, hypertension, embolism, syncope, diabetes) and these alleles or allelic combinations (not shown).

DISCUSSION

In this study we investigated the distribution and association of six alleles in the MHC in CAD patients undergoing CABG and in controls. There is some evidence that complement plays an important role in the establishment of atherosclerosis. Complement C4 is a precursor of a subunit of the enzyme complex C3 convertase and is encoded by two closely related genes. The protein products of these loci are called C4A and C4B. Both genes are highly polymorphic and there is a relatively high frequency of the non-expressed variants, termed C4A*Q0 and C4B*Q0.¹¹ These alleles are associated with several autoimmune diseases.^12,13 The rare, haemolytically inactive C4A*6 allotype was reported to be associated with rheumatic heart disease.¹⁴

Our present findings indicate that the susceptibility of homozygous carriers of the C4B*Q0 allele to severe CAD is higher than that of non-carriers. Previously we found the prevalence of C4B*Q0 to be markedly lower in healthy, elderly, Hungarian people, particularly in men, as compared to healthy, young subjects.²² We explained this observation by an increased morbidity and mortality from some diseases in middle aged carriers of the C4B*Q0 allele. This assumption was supported by our more recent observation indicating an increased frequency of the C4B*Q0 allele in 60-79 year old myocardial infarction patients as compared with age matched, healthy controls.²³

Little is known about the mechanism of this greater susceptibility to cardiovascular disease associated with the C4B*Q0 allele. Here we looked for a possible link between the C4B*Q0 allele and some characteristics and symptoms of CAD patients undergoing CABG. We found, however, no differences in these factors between carriers and non-carriers of this allele.

Several studies have shown association between atherosclerosis and certain bacterial and viral pathogens. The most compelling evidence for an infectious factor in atherosclerosis is related to Chlamydia pneumoniae.²⁴ The complement system plays a principal role in the defence against bacterial infection. Therefore, it can be assumed that homozygous carriers of the silent C4B*Q0 allele have an impaired capacity to eliminate or mitigate Chlamydia pneumoniae infection. Finally, it cannot be ruled out that C4B*QO is a marker of a known or unknown gene in the MHC with linkage disequilibrium, or there are still unknown interactions with products of genes at other linked loci, which increase the susceptibility to the disease.

TNFα is an inducible cytokine with a wide range of proinflammatory and immunoregulatory actions. Through its effect on lipid metabolism,²⁵ obesity,⁵ insulin resistance,¹⁶ and endothelial function,²⁶ and stimulation of growth factors and adhesion molecules,²⁷ it could be involved in cardiovascular pathophysiology. The large and stable interpersonal differences in TNFα production indicate a genetic background. Wilson et al²⁸ raised considerable interest with their report that the −308A allele in the promoter region is transcribed in vitro at seven times the rate of the −308G allele. Moreover, the −308A allele has also been found to correlate with enhanced spontaneous and stimulated TNFα production in vivo.²⁹ Several studies have investigated TNFα polymorphisms in diseases in which dysregulation of TNFα production might have played a role and several of them found association between TNFα −238 and −308 promoter polymorphisms and some diseases.^5,6,16 Herrmann et al⁵ investigated patients with coronary heart disease and found no association between polymorphisms in TNFα and susceptibility to the disease. The frequencies of TNFα −308A and −238A in patients with CAD undergoing CABG did not differ significantly from those of in controls in our present study either.

Investigation of the distributions of the different combinations of alleles showed some deviation from the calculated values, which was not totally unexpected, since all alleles are in the MHC within a stretch of a few hundred kilobases and linkage disequilibrium is one of the characteristic features of the MHC. However, it must be noted that the C4A genotypes were obtained by protein analysis, and it is possible that the quantitative null alleles were caused by multiple nucleotide changes resulting in the pooling of several different alleles as a single null allele. Therefore, the finding that there is linkage disequilibrium in CAD patients between TNFα −308A and C4AQ*0 does not necessarily mean that the two variants are in one haplotype, but together they are over-represented in these patients, suggesting that there might be a connection between the coincidence of the simultaneous occurrence of these variants and the development of CAD. Haplotype analysis would be needed to clarify whether there is functional importance that the alleles are in cis or trans positions.

At present it is not possible to explain the higher simultaneous occurrence of these two allelic combinations in patients with severe CAD, since apart from the higher occurrence of preoperative MI in patients with TNFα −308A + C4A*Q0 alleles, there were no clinical or laboratory parameters that differed in patients with or without these haplotypes. However, it can be hypothesised that since each allele was found to be involved in immunological disturbances, which play important roles in CAD, the simultaneous occurrence of the alleles increases the susceptibility to the development of the disease. It is also possible that genes linked with these alleles are also involved, since there are other candidate genes in the close vicinity,^8,30 including lymphotoxin α and β (related to TNFα), heat shock protein 70 (putative role in autoimmune inflammation), allograft inflammatory factor (allograft rejection), leucocyte specific transcript-1 (involved in macrophage activation), and several other genes with uncertain or unknown functions.

The TNFα −308A and C4A*Q0 alleles separately and together occurred at higher frequency in CAD patients with preoperative MI in their case history than in patients without MI. It is well known that the TNFα and the complements are involved in local inflammatory reactions in myocardial infarction.^10,31 Several studies reported raised TNFα levels in patients after MI, expressed by cardiac myocytes and macrophages migrated into the myocardium.³¹ Complement activation is a key event mediating the deleterious effects of the local inflammatory response occurring in the infarcted myocardium. A partial C4 deficiency, which may include defective handling of immune complexes, can also be an additional risk factor in MI. In addition, inflammation is known to increase the probability of rupture of the vulnerable plaques.³² Therefore it is reasonable to assume that the inflammation associated TNFα −308A and C4A*Q0 alleles may facilitate plaque rupture in MI. Besides, the connection between these alleles and high relative risk of CAD and MI correlates well with the recent findings that the imbalance of inflammatory processes (CRP, IL6, and IL1β) increases the risk of future cardiovascular disease significantly.³³

Our study shows the importance of investigation of allelic association in the search for disease susceptibility genes. Several studies have investigated the TNFα polymorphisms in diseases in which dysregulation of TNFα production might have played a role and several of them found no association between TNFα −308A and −238A alleles and the suspected diseases.^5,34 It is possible that these negative results would change drastically if the C4A alleles were also considered.

Retrospective case-control studies may suffer from several biases, which may lead to false positive and false negative results. We have matched our patient and control groups for age, sex, and ethnicity to reduce this possibility. On the other hand, a survival bias cannot be avoided in a disease association study and prospective studies will be necessary to confirm the role of these alleles in CAD. Moreover, in this study only those CAD patients were analysed who were sent for CABG. Because only the most severe CAD patients undergo CABG, this is a selected population; thus it is possible that the conclusions of this study may not be extended to CAD patients in general, but only to the most severe cases. Furthermore, it is also possible that patients referred for CABG with MI in their case histories differ from patients with MI in general. This could be an explanation of the differences between the results of this study and those of a recent report by Nityanand et al,³⁵ where no association of C4A*Q0 and MI was found.

Another important concern might be that the gene pool of the cases and the controls differs, which could account for the associations observed. In the quest of alleles contributing to the susceptibility to the disease, several other polymorphisms were investigated in these populations on other chromosomes as well^36–38 (Szalai et al, unpublished data). The distributions of the vast majority of these alleles did not differ between cases and controls, for example, ACE D, 53.1%, 54.2%; apoE4, 10.1%, 9.2%; factor V Leiden mutation of the blood coagulation system, 4.8%, 4.1%; PlA2 allele of the platelet glycoprotein IIb/IIIa receptor, 11.3%, 13.5%; F allele of the C3 component of complement, 16.5%, 16.7%; chemokine receptor 5 CCR5Δ32, 10.5%, 11.7%; CCR264I, 12.3%, 11.3%; stromal derived factor 1-3′A, 19.1%, 20.4%; RANTES-28G, 4.2%, 3.3%; RANTES-403A, 20.8%, 17.8%; methylenetetrahydrofolate reductase 677T, 34.8%, 37.2%; apo(a) (TTTTA)_n repeat polymorphism, mean n= 8.6 (SD 1.1) v 8.5 (SD 0.8), in cases and controls, respectively. Altogether, 32 polymorphisms were investigated and only two differences have been found between the two populations (MCP-1 −2518³⁶ and C4B*Q0 (this report)), both of which could contribute to the susceptibility to CAD. The careful selection of our patients and these results ensure that the differences in the gene pool between cases and controls are as minimal as possible.

In summary, according to this study, homozygous carriers of the C4B*Q0 allele have an increased risk of developing severe CAD. The simultaneous occurrences of the TNFα −308A + C4A*Q0 and the TNFα −238A + C4A*6 alleles are higher in CAD patients undergoing CABG than in healthy controls. Among CAD patients, carriers of the TNFα −308A + C4A*Q0 allelic combination have a higher risk of myocardial infarction.