Microsomal triglyceride transfer protein -164 T > C gene polymorphism and risk of cardiovascular disease: results from the EPIC-Potsdam case-cohort study

Background The microsomal triglyceride transfer protein (MTTP) is encoded by the MTTP gene that is regulated by cholesterol in humans. Previous studies investigating the effect of MTTP on ischemic heart disease have produced inconsistent results. Therefore, we have tested the hypothesis that the rare allele of the -164T > C polymorphism in MTTP alters the risk of cardiovascular disease (CVD), depending on the cholesterol levels. Methods The -164T > C polymorphism was genotyped in a case-cohort study (193 incident myocardial infarction (MI) and 131 incident ischemic stroke (IS) cases and 1 978 non-cases) nested within the European Prospective Investigation into Cancer and Nutrition (EPIC)–Potsdam study, comprising 27 548 middle-aged subjects. The Heinz Nixdorf Recall study (30 CVD cases and 1 188 controls) was used to replicate our findings. Results Genotype frequencies were not different between CVD and CVD free subjects (P = 0.79). We observed an interaction between the -164T > C polymorphism and total cholesterol levels in relation to future CVD. Corresponding stratified analyses showed a significant increased risk of CVD (HRadditve = 1.38, 95% CI: 1.07 to 1.78) for individuals with cholesterol levels <200 mg/dL in the EPIC-Potsdam study. HRadditive was 1.06, 95% CI: 0.33 to 3.40 for individuals in the Heinz Nixdorf Recall study. A borderline significant decrease in CVD risk was observed in subjects with cholesterol levels ≥200 mg/dL (HRadditve = 0.77, 95% CI: 0.58 to 1.03) in the EPIC-Potsdam study. A similar trend was observed in the independent cohort (HRadditve = 0.60, 95% CI: 0.29 to 1.25). Conclusions Our study suggests an interaction between MTTP -164T > C functional polymorphism with total cholesterol levels. Thereby risk allele carriers with low cholesterol levels may be predisposed to an increased risk of developing CVD, which seems to be abolished among risk allele carriers with high cholesterol levels.

Therefore, in the framework of the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study we investigated the association of the -164T > C variant, used as proxy for the three loci, with cardiovascular disease (CVD), myocardial infarction (MI) and ischemic stroke (IS) and the presence of an interaction between -164T > C polymorphism and total cholesterol in relation to CVD (MI and IS). We hypothesized that in carriers of the MTTP -164 C-allele lower cholesterol levels are associated with increased susceptibility to MI and IS.

Ethics statement EPIC-Potsdam study
The Ethics Committee of the Medical Association of the State of Brandenburg approved the study procedures and a written informed consent was obtained from all participants.

Heinz Nixdorf Recall study
The study was approved by the local ethics committees, was conducted in accordance with German Good Epidemiologic Practice (GEP) including extended quality management procedures and re-certifications according to DIN ISO 9001:2001. Informed consent was obtained from all participants. Information on genotype, sex, age, diabetes, anti-hyperlipidemic drugs and CVD was available for up to 1 513 of 4 814 participants.

Study population EPIC-Potsdam study
Between 1994 and 1998, as a part of the large-scale European prospective cohort study EPIC, the EPIC-Potsdam study enrolled from the general population 16 644 women (aged 35 to 65 years) and 10 904 men (aged 40 to 65 years for), for a total of 27 548 participants [29].
At baseline, self-administered questionnaires on diet and lifestyle, PC-guided interviews with additional questions on prevalent diseases, blood pressure and anthropometric measurements were collected following standard procedures [30].
Every two years, information on incident diseases and changes in lifestyle are collected by self-administered questionnaires [31], with response rates during follow-up exceeding 90% at all occasions.
A case-cohort study consisting of all incident cases identified during a mean follow-up of 8 ± 2.2 years [32] and a subcohort of 2 500 individuals randomly selected from the EPIC-Potsdam cohort [33], was used to assess the association of the -164 T > C variant with CVD risk (including both MI and IS).
With this type of study, the results are expected to be representative of the entire cohort [34,35]. After the exclusion of participants with prevalent MI and stroke at the baseline, 269 individuals with incident MI and 246 individuals with incident stroke were identified (199 IS, 41 hemorrhagic strokes, and 6 strokes with undefined pathogenesis). Among the subcohort, non-ischemic strokes were not considered as cases, while for individuals who experienced both MI and IS only the first event was considered [36]. After exclusion of prevalent CVD and missing follow-up dates, 2 368 participants remained to be in the subcohort. After further exclusion of subjects with a history of diabetes at the baseline and/or treated with anti-hyperlipidemic drugs, and those with missing MTTP genotype, biochemical or covariates data the final casecohort consisted of 2 302 individuals (324 CVD cases: 193 MI and 131 IS, and 1 978 non-cases). Among CVD cases, 37 were part of the subcohort.

Heinz Nixdorf Recall study
For the replication we analyzed data from the Heinz Nixdorf Recall (Risk Factors, Evaluation of Coronary Calcium and Lifestyle) study comprising 4 814 participants aged 45-75 years. The participants were randomly selected from registration lists of the densely populated Ruhr metropolitan area in Germany between 2000 and 2003. The rationale and design of the study have been described in detail [37]. Between 2006 and 2008 incident cases were identified during the 5-year follow-up examination.
The genotyping of the MTTP SNP -I128T (rs3816873) was already available in a random selected sample of n = 1 513 Heinz Nixdorf Recall participants out of 4 814. After exclusion of participants with a history of CVD and/or diabetes at baseline and/or treated with anti-hyperlipidemic drugs, the final replication cohorts consisted of n = 1 218 individuals (30 CVD and 1 188 non-cases).

Ascertainment of CVD EPIC-Potsdam study
As described elsewhere [32], all possible cases of MI or stroke were identified by self-report or by death certificate in one of the four follow-up questionnaires and further verified by contacting the patients' attending physician or by review of death certificates according to WHO MONICA criteria.
Primary endpoints for this study were based on unequivocally documented incident coronary events that met predefined study criteria. We considered a myocardial infarction event based on symptoms, signs of electrocardiography, and enzymes (levels of creatine kinase (CK-MB)) as well as troponin T or I, and necropsy as 1) non-fatal acute myocardial infarction and 2) coronary death, which occurred between the baseline examination and five years after study entry [38,39]. For all primary study endpoints, hospital and nursing home records including electrocardiograms, laboratory values, and pathology reports were collected. For deceased subjects, death certificates were collected and interviews with general practitioners, relatives and eyewitnesses were undertaken if possible. Medical records were obtained in 100% of all reported endpoints. An external criteria and endpoint committee blinded for conventional risk factor status and CAC scores reviewed all documents and classified the endpoints thereafter. Due to the small number of cases only total CVD were considered for the replication analysis.

Other measurements
Prevalent diabetes mellitus was identified by a physician and based on self-reported medical diagnoses, medication records and dieting behavior. Prevalent hypertension was defined as follows: systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg or self-reporting of a diagnosis or use of antihypertensive medication. Education, lifestyle characteristics (including alcohol consumption), regular sport activity and smoking history were assessed at baseline by trained interviewers during a PCguided interview. Trained personnel took anthropometric and blood pressure measurements.

Biochemical analyses EPIC-Potsdam study
At baseline, 30 ml of venous blood was taken from the respected participant (non-fasting or fasting blood) and, after fractionation into serum, plasma (collected on citrate, 10% of total volume), leukocytes, and erythrocytes immediately stored at -196°in liquid nitrogen [31]. All biomarkers were determined in 2007 in the Department of Internal Medicine, University of Tübingen. Plasma glucose, high-density lipoprotein cholesterol (HDL), total cholesterol and triglycerides were measured with an automatic analyzer (ADVIA 1650, Siemens Medical Solutions, Erlangen, Germany). LDL-cholesterol was calculated using Friedewald's formula [40]. To account for citrate's dilution factor concentrations of total, HDL-, LDL-cholesterol and triglycerides were multiplied by 1.1.

Heinz Nixdorf Recall study
At baseline, plasma cholesterol levels were measured with an automatic analyzer (ADVIA 1650, Siemens Medical Solutions, Erlangen, Germany).

Genetic analyses EPIC-Potsdam study
DNA extraction was performed using a commercial kit (Qiagen, Hilden, Germany). In 2009 at the Max Delbrück Center for Molecular Medicine, Berlin, Germany, the MTTP SNP -164T > C (rs1800804) was genotyped by TaqMan technology (Applied Biosystems, Foster City, CA, USA) using 5 ng of whole-genome amplified DNA per sample. The call rate for the SNP assay exceeded 98%.

Statistical analysis
For both EPIC-Potsdam and Heinz Nixdorf Recall studies statistical analyses were performed with the use of SAS software package, release 9.2 (SAS Institute, Cary, NC).

EPIC-Potsdam study
Data on triglyceride measurements were transformed into natural logarithms to reduce skewness and data were reported as geometric means and 95% confidence interval (CI).
The deviation from Hardy-Weinberg equilibrium (HWE) was measured using the χ 2 test.
The HWE was tested in the subcohort. Age and sex adjusted analysis of variance was used to describe general characteristics according to -164T > C genotype. Data were reported as means and standard error (SE). P for trend was calculated from age and sex adjusted linear regression model. To investigate the impact of total cholesterol on the associations between genotype and CVD we performed stratified analyses according to low (<200 mg/dL) and borderline-high/high (≥200 mg/dL) cholesterol levels as defined in the Adult Treatment Panel III (ATP III) report [41]. Multiplicative interaction between cholesterol levels and genotype as well as sex and genotype in relation to CVD was tested with crossproduct term. As a measure of additive interaction between cholesterol levels and genotype we further calculated the Synergy Index (SI) and the relative excess risk due to interaction (RERI) and their 95% confidence interval (CI) [42], as suggested by Rothman [43]. SI > 1 and RERI > 0 suggest a positive interaction (superadditive effect); SI < 1 and RERI < 0 suggest a negative interaction (subadditive effect); SI = 1 and RERI = 0 suggest absence of interaction.
Both the multiplicative and additive interactions were also tested for triglycerides, HDL-and LDL-cholesterol. The choice of the cut-off values for these biomarkers was also based on the ATP III criteria (lower and higher than 130 mg/dL for LDL-cholesterol; lower and higher than 40 mg/dL and lower and higher than 50 mg/dL for HDLcholesterol, respectively, in men and women; lower and higher than 150 mg/dL for triglycerides) [41].
Cox proportional-hazard regression modified according to the Prentice method [34] was used to compute the age and sex adjusted hazard ratio (HR) and 95% CI for the associations between MTTP -164T > C and risk of MI, IS and total CVD (combined endpoints). In the counting processes age was the underlying time variable with "entry time" defined as age at baseline and "exit time" as age at CVD event (MI or IS) or censoring. Associations between MTTP -164T > C and CVDs were tested in three models: additive, dominant and recessive. Furthermore, competing risk analyses were performed to test whether the associations of MTTP -164T > C with cardiovascular events differed between MI and IS, as described by Lunn and McNeil [44].
The power to detect an association between the MTTP -164T > C SNP and CVDs (MI, IS and combined endpoints) was computed with Quanto (http://hydra.usc.edu/ GxE/) [45], in relation to a desirable power = 80%, assuming α = 0.05 and a disease prevalence of 3.1% for CVD (n = 847), 2.0% for MI (n = 544) and 1.1% for IS (n = 303), reflecting the baseline prevalence data in the EPIC-Potsdam population. The detectable odds ratio per risk allele equals 1.30, 1.39 and 1.47, respectively for CVD, MI and IS.

Heinz Nixdorf Recall study
Cox proportional-hazard regression was used to compute the age and sex adjusted hazard ratio (HR) and 95% CI for the associations between MTTP -I128T and risk of total CVD. Associations between MTTP -I128T and CVDs were tested using the additive model.

General characteristics
The genotype distribution of the -164T > C and -I128T SNPs followed the HWE (P = 0.37 and P = 0.71, respectively for the EPIC-Potsdam subcohort and for the replication Heinz Nixdorf Recall study). There were 1 114, 758 and 143 subjects with genotypes TT, CT, CC respectively (738 men and 1 277 women) observed in the subcohort. The C allele frequency was 0.26 for both studies. According to -164T > C genotype, age-and sex-adjusted baseline characteristics of subjects who did or did not develop cardiovascular events during the follow-up period are shown in Table 1. In particular, subjects with incident CVD (n = 324; 193 incident MI; 131 incident IS) were men, older, smokers and with a lower educational level than individuals who remained free of CVD (n = 1 978) during a mean follow-up of 8.2 years. Furthermore, they had a higher prevalence of abdominal obesity and hypertension, slightly higher total and low density lipoprotein-(LDL-) and lower HDL-cholesterol levels ( Table 1). According to genotype no significant differences in central obesity, obesity and hypertension and socio-demographic characteristics were observed in subjects with or without CVD (Table 1). In contrast, opposite trends were observed for total-, LDL-cholesterol and triglyceride levels. In the group free of CVD carriers of the C-allele showed slightly higher total-and LDL-cholesterol levels (P trend = 0.036 and P trend = 0.026, respectively). In the group of CVD carriers of the C-allele showed lower triglyceride (P trend = 0.016) along with slight lower total-cholesterol levels (P trend = 0.033) ( Table 1).

Association between MTTP-164T > C polymorphism and incident CVD
The association between the MTTP -164T > C polymorphism and CVD events was tested also for MI and IS separately, taking into consideration the additive, dominant and recessive models ( Table 2). Since there were no sex differences in the association between MTTP -164T > C and CVD (P for interaction = 0.86), we combined men and In subjects with cholesterol levels ≥200 mg/dL we observed a borderline inverse association between MTTP -164T > C and CVD in the additive model (HR additive = 0.77, 95% CI: 0.58 to 1.03, P = 0.075) and significant relationships in the dominant model (HR dominant = 0.67, 95% CI: 0.48 to 0.94, P = 0.021; HR dominant = 0.65, 95% CI: 0.43 to 0.97, P = 0.036, respectively, for CVD and MI) ( Table 2).

Discussion
In this study, we anticipated an interaction between total cholesterol levels and the MTTP -164T > C polymorphism with regard to the CVD risk. The presence of a statistically significant interaction confirmed our hypothesis and indicated carriers of the C allele of the MTTP -164T > C Table 2 Hazard rate ratios (HR) and 95% confidence intervals (95% CI) for the associations between MTTP -164 T/C polymorphism, CVD (combined endpoints), MI and IS polymorphism with plasma total cholesterol levels lower than 200 mg/dL had an increased risk of CVD. The association seemed to be stronger for IS than for MI, but differences in the associations were not supported by competing risk analysis. Conversely, the MTTP -164 Callele showed a lower CVD, and MI, risk in participants with cholesterol levels higher than 200 mg/dL. Similar relationships were observed considering LDL-cholesterol with levels lower and higher than 130 mg/dl suggesting that LDL is the driving cholesterol component. However, the value of LDL levels seems to be limited as they had to be estimated based on the Friedewald formula [46]. In fact further studies are needed to replicate these findings.
The association between MTTP -I128T polymorphisms and CVD risk observed in the replication cohort showed a similar trend within the strata of cholesterol levels higher than 200 mg/dL. However, considering that the number of cases in the Heinz Nixdorf Recall study is small further replication studies are needed. To our knowledge, this is the first prospective study showing such an effect of MTTP on risk of IS.
With regard to the association between the MTTP -164T > C polymorphism and cholesterol levels, previous studies observed inconsistent results. Few studies reported a slight cholesterol lowering effect of the rare alleles of the MTTP promoter polymorphisms [7,10]. Ledmyr et al. investigated the association between the MTTP -493 G/T polymorphism and cholesterol in both healthy and hypercholesterolemic individuals, and observed decreased levels of total cholesterol in carriers of the -493 T variant [7,10]. Furthermore, Phillips et al. in a small study including 82 patients with type 2 diabetes mellitus (T2DM) of a Caucasian population found that the subjects heterozygous for the -493 G/T had lower LDL-cholesterol and, in the postprandial phase, higher apoB48 levels in the VLDL fraction. The authors suggested that the -493 G/T polymorphism seemed to confer protection against atherosclerosis in T2DM patients [12]. In contrast, Jou et al. observed that total cholesterol, LDL-, and non HDL-cholesterol levels were higher according to the rare allele of the MTTP -493 G/T polymorphism when disease free young African Americans were investigated [19,20]. Further, Lundahl et al. observed lower serum triglyceride levels in subjects affected by familial hypercholesterolemia and homozygous for the rare allele of the MTTP -493 G/T genotype [8].
Overall, these studies seem to suggest that MTTP regulates lipids differently in the presence or absence of disease, although the occurrence of a possible interaction between the LDL receptor and the MTTP genes is not excluded [7,14]. Our results seem to be in line with these hypotheses. On one hand we observed slightly higher total and LDL-cholesterol levels in subjects free of CVD and homozygous for the rare -164 C allele, and on the other, lower total cholesterol and triglyceride levels according to the rare allele of the MTTP -164T > C polymorphism in the group of future CVD cases.
It has been shown that the C-allele of the MTTP-164 T > C polymorphism is homologous to a putative sterol response element (SRE) binding site and as such confers a reduced MTTP expression [26][27][28]. These findings come from an experimental study in which Hagan et al. demonstrated that human MTTP promoter activity is upregulated by cholesterol [27]. The mechanism based on which cholesterol regulates MTTP gene expression is linked to the presence of a modified SRE in the MTTP promoter [27]. When cholesterol levels are low, the sterol regulatory element binding protein (SREBP) acts as transcription factor, binds to the SRE thereby inhibiting MTTP gene expression [28]. In contrast, in presence of cholesterol the modified SRE likely binds a new SREBP family member thus up-regulating MTTP expression [27,28]. These observations suggest that MTTP gene expression is differently regulated by high and low cholesterol levels.
Despite the lack of significant associations between the -493G > T or -164T > C single nucleotide polymorphisms (SNPs), coronary heart disease and blood lipids observed in two previous studies [24,25], recently Aminoff et al. put forward that carriers of the rare -164C allele are at increased risk of IHD [26]. They substantiated their findings by showing in vivo that the presence of the rare alleles of the -493G > T and -164T > C SNPs confer lower MTTP transcription in the heart, liver and macrophage. This mechanism, in turn, by causing the lipid accumulation in the heart would provoke an increased IHD risk. Indeed, our findings are in line with those of Aminoff et al. though they concluded that the increased IHD risk observed according to the -164C variant was independent of plasma lipids. As mentioned above, because human MTTP promoter activity is positively regulated by cholesterol [27], it is reasonable to assume that subjects with low cholesterol levels have, in general, a lower MTTP gene expression. Thus, in this low risk group carriers of the -164C variant, compared to carriers of the common allele, might be at increased CVD because of their lower MTTP gene expression. At the same time, if one would consider the observed associations as those mimicking MTTP inhibitors, then these findings could further highlight the concerns expressed by Aminoff et al. regarding the long term side effects MTTP inhibition may generate [26,47].
In contrast, in subjects with higher cholesterol levels we observed a reduced, though borderline significant, CVD risk accordingly to the MTTP gene -164 C variant. Our findings suggest that there could be an antagonistic (qualitative) interaction between cholesterol levels and MTTP -164 T > C polymorphism. These observations warrant further investigation.
The main limitation of this study is that the plasma lipoprotein and apolipoprotein levels, which are important in the effect of MTTP in cardiovascular disease, were not measured; our analyses on triglyceride levels were based on both fasting and non-fasting subjects; we estimated the LDL-cholesterol levels based on Friedewald equation. Strength of our study includes its prospective design. Furthermore, all cases of MI and IS were validated by medical records and were derived from a cohort population with a very high follow-up coverage.

Conclusions
The findings of this study suggest that in the subjects investigated an interaction between MTTP -164T > C functional polymorphism with total cholesterol levels predisposes risk allele carriers with low cholesterol levels to an increased risk of developing CVD, which seems to be abolished among risk allele carriers with high cholesterol levels. However, further studies are warranted in order to shed more light on these complex mechanisms.