APOA5 variants and metabolic syndrome in Caucasians.

Apolipoprotein A5 (APOA5) gene variants were reported to be associated with two components of metabolic syndrome (MetS): higher TG levels and lower HDL levels. Moreover, a recent Japanese case-control study found variant −1131T>C associated with MetS itself. Thus, our study systematically analyzed the APOA5 gene for association with lipid parameters, any other features of MetS, including waist circumference, glucose-related parameters, blood pressure, uric acid, and MetS itself in Caucasians. Ten polymorphisms were analyzed in a large fasting sample of the population-based Cooperative Health Research in the Region of Augsburg (KORA) survey S4 (n = 1,354; southern Germany) and in a second fasting sample, the Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (SAPHIR) study (n = 1,770; Austria). Minor alleles of variants −1131T>C, −3A>G, c.56C>G, 476G>A, and 1259T>C were significantly associated with higher TG levels in single polymorphism (P < 0.001) and haplotype (P ⩽ 6.6 × 10−6) analysis. Besides associations with lower HDL levels in SAPHIR (P ⩽ 0.001), there were no significant findings with any other features of MetS. Variant c.56C>G was associated with higher risk for MetS [odds ratio (95% confidence interval) = 1.43 (1.04, 1.99), P = 0.03 for KORA and 1.48 (1.10, 1.99), P = 0.009 for SAPHIR). Our study confirms the association of the APOA5 locus with TG and HDL levels in humans. Furthermore, the data suggest a different mechanism of APOA5 impact on MetS in Caucasians, as variant c.56C>G (not analyzed in the Japanese study) and not −1131T>C, as in the Japanese subjects, was associated with MetS.

Abstract Apolipoprotein A5 (APOA5) gene variants were reported to be associated with two components of metabolic syndrome (MetS): higher TG levels and lower HDL levels. Moreover, a recent Japanese case-control study found variant 21131T.C associated with MetS itself. Thus, our study systematically analyzed the APOA5 gene for association with lipid parameters, any other features of MetS, including waist circumference, glucose-related parameters, blood pressure, uric acid, and MetS itself in Caucasians. Ten polymorphisms were analyzed in a large fasting sample of the populationbased Cooperative Health Research in the Region of Augsburg (KORA) survey S4 (n 5 1,354; southern Germany) and in a second fasting sample, the Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (SAPHIR) study (n 5 1,770; Austria). Minor alleles of variants 21131T.C, 23A.G, c.56C.G, 476G.A, and 1259T.C were significantly associated with higher TG levels in single polymorphism (P , 0.001) and haplotype (P < 6.6 3 10 26 ) analysis. Besides associations with lower HDL levels in SAPHIR (P < 0.001), there were no significant findings with any other features of MetS. Variant c.56C.G was associated with higher risk for MetS [odds ratio (95% confidence interval) 5 1.43 (1.04, 1.99), P 5 0.03 for KORA and 1.48 (1.10, 1.99), P 5 0.009 for SAPHIR).
Our study confirms the association of the APOA5 locus with TG and HDL levels in humans. Furthermore, the data suggest a different mechanism of APOA5 impact on MetS in Caucasians, as variant c.56C.G (not analyzed in the Japanese study) and not 21131T.C, as in the Japanese subjects, was associated with MetS. Apolipoprotein A5 (APOA5) is considered as important modifying gene for familial combined hyperlipidemia (FCHL), a disorder characterized by higher plasma triglyceride (TG) levels and lower HDL levels (1,2). As these parameters are two components of the metabolic syndrome (MetS) (3), APOA5 may also have a role in this complex disorder. The association of common APOA5 variants with higher TG levels was confirmed in several studies (4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19).
However, the underlying mechanisms are still unclear and cannot be easily explained by an impact on APOA5 expression. Although initial overexpression and knockout mouse models revealed an inverse relationship between the protein and TG levels (20), studies in humans revealed a positive correlation between APOA5 and TG levels (21,22).
Two major haplotypes associated with higher TG levels are tagged by the rare genotypes of variant 21131T.C and coding variant c.56C.G, which results in the substitution of tryptophan for serine at residue 19 within the predicted signal peptide responsible for APOA5 secretion (20). Thus, c.56C.G is the only common variant with known influence on APOA5 expression (23).
Recently, variant 21131T.C was identified as the strongest signal for an association with MetS among 158 analyzed DNA polymorphisms in 133 candidate genes in a study of 1,788 unrelated elderly (50-75 years) Japanese individuals admitted in one of five participating hospitals (24). Confirmation and extension of association findings to other populations are mandatory, especially if the original finding was based on a large number of polymorphisms investigated, with its increased probability of false-positives. Furthermore, findings in Japanese populations cannot necessarily be extrapolated to Caucasian populations.
Thus, our study systematically analyzed the genetic information regarding the APOA5 gene locus for association with lipid parameters and other parameters of MetS in two large fasting samples: 1,354 Caucasian subjects of the population-based Cooperative Health Research in the Region of Augsburg (KORA) survey S4 and 1,770 subjects of the Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (SAPHIR) study.

Study populations
The KORA survey S4 is a population-based study of adults performed in southern Germany (25). The KORA survey S4 sample included 1,354 fasting subjects in the 55-74 year age group with an oral glucose tolerance test (OGTT) (26). Newly diagnosed type 2 diabetes was identified in 120 participants by OGTT data. Major population stratification within the KORA studies is unlikely, as there was no major population stratification found between the three large population-based studies from Germany (KORA, Study of Health in Pomerania, and Population Genetic Cohort) (27).
The SAPHIR study is an observational study conducted in Austria from 1999 to 2002 involving 1,770 unrelated fasting subjects: 663 women aged 50-70 years and 1,107 men aged 40-60 years (28,29). Study participants were recruited by health screening programs at large companies in and around the city of Salzburg. This sample included 57 participants with type 2 diabetes. Individuals from both samples were of Caucasian origin.

Laboratory measurements
Fasting serum was used for all laboratory measurements in both studies. In the KORA S4 study, TG was assessed with the Boehringer glycerol phosphate oxidase-p-aminophenazone assay. Total cholesterol was measured by enzymatic methods (cholesterol oxidase-p-aminophenazone; Roche Diagnostics), HDL cholesterol was measured after precipitation with phosphotungstic acid/Mg 21 (Roche Diagnostics), and LDL cholesterol was measured after precipitation with dextran sulfate (Quantolip LDL; Immuno AG). Details on further laboratory measurements of the KORA S4 study are described elsewhere (30).
In the SAPHIR study, TG, total cholesterol, HDL cholesterol, and LDL cholesterol were determined using commercially available assays (Hoffmann-LaRoche GmbH, Vienna, Austria). Details on further laboratory measurements of the SAPHIR study are described elsewhere (31).

Definition of MetS
Diagnosis of MetS in both study samples was based on the definition proposed by the National cholesterol education program (NCEP) Adult Treatment Panel III, including medication (32). Participants were thus diagnosed with MetS if they had three or more of the following five components: 1) waist circumference of .102 cm for men or .88 cm for women; 2) fasting serum TG concentration of >1.7 mmol/l (150 mg/dl) or drug treatment for increased TG; 3) serum HDL cholesterol concentration of ,1.0 mmol/l (40 mg/dl) for men or ,1.3 mmol/l (50 mg/dl) for women or drug treatment for reduced HDL; 4) systolic blood pressure of >130 mmHg or diastolic blood pressure of >85 mmHg or drug treatment for hypertension; and 5) fasting plasma glucose level of >6.1 mmol/l (110 mg/dl) or drug treatment for increased glucose.

Genotyping
Genomic DNA of KORA and SAPHIR participants was extracted from blood leukocytes using the Puregene TM DNA Isolation Kit (Gentra Systems) according to the manufacturer's recommendations.
Genotyping of 10 APOA5 variants (29655A.G, 24904C.T, 21131T.C, 21099G.A, 23A.G, c.56C.G, 476G.A, c.457G.A, 1764C.T, and 1259T.C) in KORA S4 and of 21131T.C and 1259T.C in SAPHIR was carried out in the same laboratory by means of matrix-assisted laser desorption ionization-time of flight mass spectrometry analysis of allele-dependent primer extension products as described elsewhere (33). Variant c.56C.G was genotyped in SAPHIR by a 5 ¶ nuclease allelic discrimination (Taqman) assay within the Genotyping Unit of the Gene Discovery Core Facility at the Innsbruck Medical University in Austria.

Statistical analysis
Violation of Hardy-Weinberg equilibrium (HWE) was tested by Pearson's Chi-square test. Quantitative traits that were normally distributed on the original or logarithmic scale were analyzed by linear regression adjusted for age, sex, and body mass index (BMI) with the genotype included 1) model-free (2 degrees of freedom), 2) in an additive model, or 3) in a dominant model.
The primary analysis consisted of four lipid-related traits: TG, LDL, HDL, and total cholesterol levels. Participants taking lipid-lowering drugs (n 5 155) were excluded from this quantitative analysis. For the KORA sample, the significance level was corrected for multiple testing of lipid traits and the number of effective loci [a 5 0.05/(4 3 7) 5 0.0017], which was calculated by spectral decomposition of the correlation matrix of all variants analyzed (34). In the replication step, variants that were significantly associated according to the corrected significance level in KORA S4 were analyzed in the SAPHIR study using the same statistical models. The significance level applied in SAPHIR was corrected analogous to KORA S4.
The secondary analysis investigated further traits of MetS using the same statistical models. Subjects with antidiabetic medication were excluded from the analysis of fasting glucose, 2 h glucose (from OGTT), and fasting insulin. For the analysis of blood pressure, subjects using antihypertensive medication were excluded. Finally, associations of the genotypes with MetS were assessed by logistic regression adjusted for age and sex in all 1,354 (KORA) and 1,770 (SAPHIR) fasting subjects. The significance level for the analysis of MetS was corrected for the number of effective loci (KORA, a 5 0.05/7 5 0.007; SAPHIR, a 5 0.05/2 5 0.025). Single variant or haplotype analysis was carried out using SAS (version 9.1; Cary, NC) or the statistical software R (version 2.3.1; haplo.glm procedure), respectively. For details see the supplementary data. Haplotypes with P , 0.0125 computed as 0.05/4 were considered statistically significantly associated.

RESULTS
Characteristics of the KORA and SAPHIR samples are shown in Table 1 stratified by sex and in supplementary Genotyping results of the 10 analyzed APOA5 variants are presented in Table 2. The location and linkage disequilibrium (LD) structure of the analyzed variants are presented in supplementary Fig. I. None of the variants violated HWE (P > 0.05). The genotyping success rates were .94.0% with a discordance of ,0.5% in 210 routine duplicates in KORA S4 and .95.7% with no discordance in 64 routine duplicates in SAPHIR. Genotype distribution combined for 21131T.C and c.56C.G is presented in supplementary Table II with and without the exclusion of subjects taking lipid-lowering drugs.

Primary analysis of lipids
TG levels were significantly different between the genotype groups of variants 1259T.C, 23A.G, 21131T.C, and 476G.A (P < 0.001) and borderline significantly different for c.56C.G. Because 21131T.C is highly correlated with 23A.G (r 2 5 0.98), 476G.A (r 2 5 0.95), and 1259T.C (r 2 5 0.84), Table 3 presents results for 21131T.C and c.56C.G only. In the additive model, variant 21131T.C showed a statistically significant association with 13.30% (P 5 5.0310 24 ) higher TG levels per copy of the minor allele for 21131T.C. Variant c.56C.G showed a borderline significant difference for TG levels between genotype groups (P 5 0.009, F-test), with b-estimates in the same range as for variant 21131T.C. However, this association was insignificant after correction for multiple testing [additive model: percentage change of geometric mean (PCGM) 5 12.3%; P 5 0.005]. No significant associations were observed with total, LDL, and HDL cholesterol.
Variants 21131T.C, 1259T.C (data not shown), and c.56C.G were selected for replication in the SAPHIR study (Table 3). A minor allele of variant c.56C.G was significantly associated with higher TG levels in the dominant model (PCGM 5 22.08%; P 5 6.7 3 10 27 ). Minor alleles of all three variants were significantly associated with HDL, LDL, or total cholesterol levels ( Table 3; 1259T.C data not shown).

APOA5 haplotype analysis
Haplotype analysis performed in the KORA study revealed six haplotypes with frequencies of .5% ( Table 4). The most frequent haplotype was used as a reference. Haplotype APOA5_3, containing the minor alleles of variants 23A.G, 21131T.C, 476G.A, and 1259T.C, was significantly associated with higher TG levels (P 5 0.002). Haplotype APOA5_5, tagged by variant c.56C.G, was also significantly associated with higher TG levels (PCGM 5 15.2%; P 5 0.003). This association was stronger in women (PCGM 5 22.5%; P 5 0.002) than in men.

Secondary analysis: MetS and its features
Results of the secondary analysis are presented in supplementary Table III. In both study samples, no significant associations of gene variants with waist circumference,

DISCUSSION
The present study investigated 10 polymorphisms covering the APOA5 locus for association with lipid parameters, NCEP-defined MetS, and its features in KORA S4. Minor alleles of 21131T.C and variant c.56C.G were associated with higher TG levels, as reported previously (5,20,35,36). Both associations were replicated in the SAPHIR study. Furthermore, associations with lower HDL and higher LDL and total cholesterol levels were found for 21131T.C and c.56C.G in the SAPHIR population but not in KORA S4. Variant c.56C.G was associated with higher risk of MetS in the SAPHIR study, with similar ORs in the KORA study or in combined analysis. Although association with MetS was convincingly statistically significant in SAPHIR, this association did not remain significant in KORA after applying correction for multiple testing.
Especially findings with MetS deserve some attention, as a result of the recent description of an association of 21131T.C with MetS (OR, 1.45-2.07) in a Japanese study investigating a large number of polymorphisms in 133 candidate genes (24). Our results could not confirm this association.
Although the association of APOA5 variants with TG levels could be replicated in different ethnicities, the complexity of polygenic susceptibilities of MetS might be an explanation for the lack of association of 21131T.C with MetS in Caucasians. The comparability of the Japanese case-control study with our population-based study samples could be biased by different study designs. Whereas the age range was similar in both studies, the MetS definition in the Japanese study included a body mass index threshold (>25 kg/m 2 ) instead of waist circumference thresholds. We decided to use waist circumference criteria because mean body mass index was rather high in our populations. Another difference was the selection of control subjects, which were free from any MetS component in the Japanese study. Furthermore, the definition of MetS is controversial, especially concerning different ethnicities (37). Genetic differences were reflected in 21131T.C minor allele frequencies of 35.3% in Japanese versus 7.4% in Caucasians.
We were able to find an association of another functionally relevant polymorphism (c.56C.G) with MetS. This could be attributable to the fact that MetS is a strongly lipiddriven or lipid-influenced syndrome involving three of five components. Two of these components are clear lipid com-   ponents correlated not only with each other but also with other components of MetS. Therefore, it is possible that an association found between a genetic variant and TG and HDL levels extends to an association with MetS only marginally triggered by the other components. This interpretation is also supported by an association of c.56C.G and FCHL, which might share etiological overlap with MetS (38,39).
In most (15,40,41) but not all (42) studies investigating Caucasian FCHL subjects, variant 21131T.C was associated with FCHL. More consistent results were reported for c.56C.G, showing increased transmission of the G allele in FCHL subjects (40,41). However, the major impact on serum TG levels by homozygote c.56C.G found in several studies (14,16,19,40) could only be confirmed in SAPHIR.
In combination with c.56C.G, 21131T.C might have an effect on MetS risk in Caucasians too, as estimated MetS risk was increased in subjects heterozygous for minor alleles of both variants.
Haplotype analysis supports the findings of the single variant analysis. Estimated haplotypes are in agreement with the haplotypes reported by Pennacchio et al. (14), and the additional variants included do not appear to provide further information.
The KORA survey S4, being representative of the general population of Augsburg, Germany, benefits from the high-quality phenotyping of the internationally approved Monitoring Trends and Determinants on Cardiovascular Diseases/KORA surveys (43). Our analyzed sample included fasting values of parameters relevant to MetS. The SAPHIR study, being representative of the working population of Salzburg, Austria, provides fasting measurements as well. Thus, both populations are exceptionally applicable to the detection of associations with MetS (total power of 97.5% to detect OR of 1.5) or related traits (29,44). The ability to include replication in a larger sample ensures the observed associations. Significance level was strictly corrected for multiple testing to avoid false-positive results, which are a major problem in genetic epidemiological association studies.

CONCLUSION
Our investigation of the association between APOA5 variants and features of MetS in one of the largest Caucasian study samples to date provides strong support for an association with TG and HDL levels. The association of variant c.56C.G with MetS might be driven mainly by an association with TG and HDL levels.