The causal effects of alcohol on lipoprotein subfraction and triglyceride levels using a Mendelian randomization analysis: The Nagahama study

Atherosclerosis 257 (2017) 22e28

Contents lists available at ScienceDirect

Atherosclerosis journal homepage: www.elsevier.com/locate/atherosclerosis

The causal effects of alcohol on lipoprotein subfraction and triglyceride levels using a Mendelian randomization analysis: The Nagahama study Yasuharu Tabara a, *, Hidenori Arai b, Yuhko Hirao c, Yoshimitsu Takahashi d, Kazuya Setoh a, Takahisa Kawaguchi a, Shinji Kosugi e, Yasuki Ito c, Takeo Nakayama d, Fumihiko Matsuda a, on behalf of the Nagahama study group a

Center for Genomic Medicine, Kyoto University Graduate School of Medicine, Kyoto, Japan National Center for Geriatrics and Gerontology, Obu, Japan c Research and Development Center, Denka Seiken Co., Ltd., Tokyo, Japan d Department of Health Informatics, Kyoto University School of Public Health, Kyoto, Japan e Department of Medical Ethics and Medical Genetics, Kyoto University School of Public Health, Kyoto, Japan b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 October 2016 Received in revised form 2 December 2016 Accepted 8 December 2016 Available online 22 December 2016

Background: Light-to-moderate alcohol consumption may increase circulating high-density lipoprotein cholesterol (HDL-C) levels and decrease low-density lipoprotein cholesterol (LDL-C) levels. However, the effect of alcohol on biologically important lipoprotein subfractions remains largely unknown. Here we aimed to clarify the effects of alcohol on lipoprotein subfractions using a Mendelian randomization analysis. Methods: The study subjects consisted of 8364 general Japanese individuals. The rs671 polymorphism in aldehyde dehydrogenase 2 gene, a rate-controlling enzyme of alcohol metabolism, was used as an instrumental variable. Lipoprotein subfractions were measured by a homogeneous assay. Results: The biologically active *1 allele of the ALDH2 genotype was strongly associated with alcohol consumption in men (p < 0.001). In a regression analysis adjusted for possible covariates, the *1 allele was positively associated with HDL-C even in a sub-analysis for HDL subfractions (HDL2-C: b ¼ 0.082, p < 0.001; HDL3-C: b ¼ 0.195, p < 0.001). In contrast, the *1 allele was inversely associated with total LDLC levels (b ¼ 0.049, p ¼ 0.008), while its association with large-buoyant LDL-C (b ¼ 0.124, p < 0.001) and small-dense LDL-C (b ¼ 0.069, p < 0.001) was opposite. Therefore, the ratio of small-dense LDL to large-buoyant LDL exhibited a linear increase with the number of *1 alleles carried (b ¼ 0.127, p < 0.001). Furthermore, the *1 allele was inversely associated with triglyceride levels in an analysis adjusted for LDL subfractions (b ¼ 0.097, p < 0.001), but not for the total LDL (b ¼ 0.014, p ¼ 0.410). Conclusions: Alcohol may increase HDL-C levels irrespective of the particle size. Moreover, alcohol may decrease the total LDL-C, although the proportion of atherogenic small-dense LDL-C increased partially due to a potential inter-relationship with decreased triglyceride levels. © 2016 Elsevier Ireland Ltd. All rights reserved.

Keywords: High-density lipoprotein Low-density lipoprotein Lipoprotein subfractions Alcohol consumption Mendelian randomization analysis

1. Introduction It has been well documented that light-to-moderate alcohol consumption can reduce the risk of developing cardiovascular disease [1]. Although the mechanism for the favorable effect of

* Corresponding author. Center for Genomic Medicine, Kyoto University Graduate School of Medicine, Shogoinkawara-cho 53, Sakyo-ku, Kyoto, 606-8507, Japan. E-mail address: [email protected] (Y. Tabara). http://dx.doi.org/10.1016/j.atherosclerosis.2016.12.008 0021-9150/© 2016 Elsevier Ireland Ltd. All rights reserved.

alcohol consumption has not been fully elucidated, an improvement of the lipoprotein profiles, particularly increased levels of circulating high-density lipoprotein cholesterol (HDL-C) [2,3], is thought to provide a potential explanation. In addition to the HDL-C increasing effect of alcohol, previous epidemiological studies [3e5] (including our own [6]) have found that alcohol may also have a low-density lipoprotein cholesterol (LDL-C) lowering effect. Since the cardioprotective activity of HDL, as well as the atherogenic properties of LDL, may differ based on their respective subfractions

Y. Tabara et al. / Atherosclerosis 257 (2017) 22e28

[7e9], the effect of alcohol on the circulating levels of lipoprotein subfractions should be clarified to further understand the mechanisms behind the beneficial effects of alcohol. The effect of alcohol on triglyceride (TG) levels is another inconclusive aspect, with studies reporting both reduced [10,11] and increased [12] TG levels in alcohol drinkers. One challenge in clarifying the effect of alcohol in an epidemiological study setting is the other confounding clinical factors that are potentially affected by alcohol. A Mendelian randomization analysis (MRA) is a method used to clarify the causality of the risk factor of interest for a given outcome by using a genotype robustly associated with the risk factor as an instrumental variable [13]. Given that genes are randomly assigned during meiosis, a genetic variant is fundamentally independent of typical confounding factors. Most East Asians, including Japanese, have the inactive allele (*2 allele) for the aldehyde dehydrogenase 2 (ALDH2) gene, a ratecontrolling enzyme in ethanol metabolism. Therefore, individuals that are homozygotes for *2 alleles tend to be non-drinkers, while the daily alcohol consumption of homozygotes for the enzymatically active *1 allele are reported to be approximately double than that of heterozygotes [6,14e16]. A MRA using the ALDH2 genotype in Japanese individuals is a convincing approach to clarify the effects of alcohol on the circulating levels of lipoprotein subfractions and triglycerides. In the present study, to clarify the pleiotropic effect of alcohol, we assessed the causality between alcohol intake and the circulating levels of lipoprotein subfractions and triglycerides in a large general Japanese population by a MRA using the ALDH2 genotype as an instrumental variable.

2. Materials and methods 2.1. Study participants We analyzed a dataset of the Nagahama Prospective Cohort for Comprehensive Human Bioscience (the Nagahama Study). The study participants in this cohort were recruited from 2008 to 2010 from the general population of Nagahama City, a largely rural city consisting of 125,000 inhabitants located in central Japan. The inclusion criteria for this cohort consisted of the following: community residents aged from 30 to 74 years, living independently in the community, and without physical impairment or dysfunction. Among a total of 9804 participants, the remaining 9769 individuals after the exclusion of the following conditions: individuals that withdrew consent to participate in this study (n ¼ 9), and that proved to have a different ethnic background by genetic analysis (n ¼ 26). The baseline measurements were performed at the time of recruitment. Of the 9769 potential participants, those meeting any of the following conditions were excluded from this study: pregnant women (n ¼ 43), individuals undergoing insulin therapy (n ¼ 25) or lipid-lowering medications (n ¼ 1171), an unsuccessful measurement or extreme deviation of lipoprotein cholesterol subfractions (n ¼ 25) or other clinical parameters required for this study (n ¼ 107), and the unavailability of a ADLH2 rs671 genotype (n ¼ 34). We did not consider absorption disorders in this study population. A final total of the remaining 8364 participants were considered as this study population. All study procedures were approved by the ethics committee of the Kyoto University Graduate School of Medicine and by the Nagahama Municipal Review Board. Written informed consent was obtained from all participants.

23

2.2. Basic clinical parameters The basic clinical parameters used in this study were obtained from the baseline measurements. Circulating levels of triglyceride (Determiner C-TG, Kyowa Medex, Co., Ltd. Tokyo, Japan) and total cholesterol (Determiner C-TC) levels were measured using serum samples. The history of cardiovascular diseases (CVD), menopausal status, smoking habits, amount of alcohol consumed in a single sitting, and medication use were obtained using a structured, selfadministered questionnaire. The level of alcohol consumption was described in Japanese traditional units (Go), in which 1 Go corresponds to 22 g of ethanol. A peripheral blood sample from a number of participants was collected under non-fasting (<5 h, n ¼ 693) or near-fasting (5e11 h, n ¼ 3831) conditions. 2.3. Measurement of lipoprotein subfractions Circulating levels of lipoprotein subfractions were measured using a plasma sample stored at 80  C. To directly measure HDL3C, TG-rich lipoproteins and LDL were first digested using sphingomyelinase, and the cholesterol released from these lipoproteins was then eliminated using cholesterol esterase/oxidase and catalase. Cholesterol in HDL3 was then measured via a standard peroxidase method following enzymatic treatment with a polyoxyethylene styrenated phenyl ether derivative that specifically acts on HDL3 (HDL3-EX, Denka Seiken Tokyo, Japan) [17]. HDL2-C levels were then calculated by subtracting the HDL3-C from the total HDL-C measured using a commercially available assay kit (HDL-EX, Denka Seiken). A strong collinearity between the HDL subfraction levels measured using this method and those measured by an objective standard method via ultracentrifugation has been reported elsewhere [17]. Small dense LDL cholesterol (sdLDL-C) was measured using a standard cholesterol assay following enzymatic treatment with a polyoxyethylene benzylphenyl ether derivative for eliminating TGrich lipoproteins and HDL, as well as a sphingomyelinase that specifically reacts with large buoyant LDL (lbLDL) (sdLDL-EX, Denka Seiken) [18]. The total LDL cholesterol was measured using a commercially available assay kit (LDL-EX (N), Denka Seiken). lbLDLC levels were calculated by subtracting the sdLDL-C from the total LDL-C. The accuracy of the LDL subfraction levels measured using this method has been reported elsewhere [18]. 2.4. DNA extraction and genotyping Genomic DNA was extracted from the peripheral blood using a conventional phenol-chloroform method. The ALDH2 rs671 genotype was analyzed using a series of BeadChip DNA arrays (Illumina, San Diego, CA, USA), or via a TaqMan probe assay (Applied Biosystems Co., Ltd., Foster City, CA, USA) using commercially available primers and probes purchased from the Assay-on-Demand system (C_11703892_10). The fluorescence of TaqMan-PCR products was measured using an ABI PRISM 7900HT sequence detector (Applied Biosystems). Alleles for G (Glu) and A (Lys) were considered as the *1 and *2 alleles, respectively. 2.5. Statistical analysis Differences in the numerical variables were assessed using an analysis of variance, while the frequency differences were assessed using a chi-squared test. Associations of the ALDH2 genotype with plasma markers were evaluated via a linear regression analysis with an additive genetic model (*1*1 ¼ 2, *1*2 ¼ 1, *2*2 ¼ 0). Statistical analyses were performed using commercially available statistical software (JMP ver. 9.03; SAS Institute, Cary, NC, USA). Null

24

Y. Tabara et al. / Atherosclerosis 257 (2017) 22e28

hypotheses were rejected at the p < 0.05 level of significance. 3. Results The clinical characteristics of the study subjects are shown in Table 1, and the differences in the parameters among the ALDH2 genotypes are summarized in Table 2. Alcohol consumption in men was markedly different among the genotypes. Although both systolic and diastolic blood pressure (BP) was also significantly different among the genotypes, this difference disappeared after adjusted for alcohol consumption (systolic BP, p ¼ 0.397; diastolic BP, p ¼ 0.646). Significantly higher HDL-C and lower LDL-C in *1 allele carriers was also attributed to the higher alcohol consumption of the carriers (alcohol consumption adjusted p values: HDL-C, p ¼ 0.825; LDL-C, p ¼ 0.887). The ALDH2 genotype was thus a credible proxy for alcohol consumption. Similar results were not observed in women, possibly due to lower alcohol consumption. The association between the ALDH2 *1 allele and the various lipoprotein subfractions are summarized in Table 3. A clear association was observed in men, and the *1 allele was positively associated with both HDL2-C and HDL3-C, but not the HDL2-C to HDL3-C ratio. In contrast, the *1 allele was inversely associated with the LDL-C levels. However, in a separate analysis by LDL subclass, association of the *1 allele with lbLDL-C and sdLDL-C was opposite, which resulted in a positive association between the *1 allele and the sdLDL to lbLDL ratio. Fig. 1AeC present the covariate-adjusted mean for sdLDL, lbLDL, and sdLDL/LDL in each ALDH2 genotype. An occupation is a social factor that may be involved in alcohol consumption. In this study population, self-employed consumed more alcohol than employee and individuals without a regular occupation in both men (employee: 1.2 ± 1.2, self-employed: 1.4 ± 1.2, none: 1.2 ± 1.1, p ¼ 0.010) and women (employee: 0.3 ± 0.7, self-employed: 0.4 ± 0.9, none: 0.2 ± 0.6, p < 0.001). Although these associations were independent of the ALDH2 genotype (men: p ¼ 0.004; women: p < 0.001), results of the

regression analysis (Table 3) did not change substantially after further adjustment for an occupation (men, HDL-C: p < 0.001, HDL2-C: p < 0.001, HDL3-C: p < 0.001, HDL2/HDL3: p ¼ 0.092, LDLC: p ¼ 0.008, lbLDL-C: p < 0.001, sdLDL-C: p < 0.001, sdLDL/lbLDL: p < 0.001; women, HDL-C: p ¼ 0.001, HDL2-C: p ¼ 0.049, HDL3-C: p < 0.001, HDL2/HDL3: p ¼ 0.035, LDL-C: p ¼ 0.049, lbLDL-C: p ¼ 0.001, sdLDL-C: p ¼ 0.218, sdLDL/lbLDL: p ¼ 0.043). Table 4 summarizes the results of the association analysis between the ALDH2 genotype and the triglyceride levels in men. In the analysis adjusted for basic covariates (Model 1), as well as for the total LDL-C (Model 2), no significant association was observed between the ALDH2 genotype and triglyceride levels. However, when LDL-C was separately included in the same model as lbLDL-C and sdLDL-C (Model 3), the ALDH2 genotype was identified as a significant and inverse determinant for triglyceride levels. Similar results were also found in the analysis adjusted for the sdLDL/lbLDL ratio (Model 4), but not the HDL2/HDL3 ratio (Model 5). The covariate adjusted mean triglyceride levels by the ALDH2 genotype are depicted in Fig. 1D. When alcohol consumption was included in Model 3 instead of the ALDH2 genotype, alcohol consumption also exhibited an inverse association with triglyceride levels (b ¼ 0.148, p < 0.001). 4. Discussion In this MRA using the ALDH2 genotype as an instrumental variable, we clarified a causal relationship between alcohol consumption and increased HDL-C levels irrespective of its lipoprotein size in men, but not in women. We also clarified the causality of alcohol consumption in lowering the total LDL-C levels, while the LDL profiles might be worsened by increasing the proportion of atherogenic sdLDL, partly due to an inter-correlation with decreased TG levels. The results of the present study confirmed the HDL-increasing effect of alcohol consumption, and the effect was somewhat

Table 1 Clinical characteristics of the study participants.

Age (years old) Menopause (%) BMI (kg/m2) Current smoker (%) Brinkman index Habitual drinker (%) Alcohol consumption (Go/sitting) Occupation (employee/self-employed/none, %) Systolic BP (mmHg) Diastolic BP (mmHg) Antihypertensive medication (%) Glucose (mg/dl) Insulin (mU/mL) g-glutamyl transpeptidase (IU) Triglyceride (mg/dl) Cholesterol (mg/dl)

ALDH2 rs671 *1 allele frequency (%)

Total cholesterol HDL-C HDL2-C HDL3-C LDL-C lbLDL-C sdLDL-C

Men

Women

(2756)

(5608)

54.7 ± 13.8

50.8 ± 13.0 55.0 21.6 ± 3.2 7.0 31 ± 101 19.8 0.3 ± 0.7 1961/560/3087 119 ± 17 73 ± 11 10.6 88 ± 10 4.0 [2.8e5.7] 17 [13e24] 70 [52e100] 209 ± 35 69.3 ± 15.3 45.9 ± 13.0 23.4 ± 3.7 123.4 ± 33.2 90.2 ± 22.2 33.1 ± 15.3 73.5

23.3 ± 3.0 32.6 421 ± 419 62.5 1.3 ± 1.2 1295/746/715 130 ± 16 80 ± 11 18.0 93 ± 17 4.2 [2.8e6.5] 33 [23e53] 98 [69e141] 203 ± 34 59.1 ± 14.7 36.5 ± 12.0 22.6 ± 4.1 127.0 ± 33.7 86.0 ± 23.7 41.0 ± 18.9 72.8

Values are expressed as the mean ± standard deviation, frequency, or median and interquartile range (insulin, g-glutamyl transpeptidase, and triglyceride). Aldehyde dehydrogenase 2 (ALDH2) rs671 allele; *1: G (Glu) allele; BMI, body mass index; BP, blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein, lbLDL, large buoyant LDL; sdLDL, small-dense LDL. Alcohol consumption was calculated based on the amount consumed in a single sitting and was described in Japanese traditional units of alcohol (Go), where 1 Go corresponds to 22 g of ethanol.

Y. Tabara et al. / Atherosclerosis 257 (2017) 22e28

25

Table 2 Differences in the clinical parameters by the ALDH2 genotype. ALDH2 rs671 genotype

Men (n) Age (years) BMI (kg/m2) Alcohol consumption (Go/sitting) Brinkman Index Systolic BP (mmHg) Diastolic BP (mmHg) Glucose (mg/dl) Insulin (mU/mL) Triglycerides (mg/dl) Total cholesterol (mg/dl) HDL-C (mg/dl) LDL-C (mg/dl) Women (n) Age (years) BMI (kg/m2) Alcohol consumption (Go/sitting) Brinkman Index Systolic BP (mmHg) Diastolic BP (mmHg) Glucose (mg/dl) Insulin (mU/mL) Triglycerides (mg/dl) Total cholesterol (mg/dl) HDL-C (mg/dl) LDL-C (mg/dl)

p

*1*1

*1*2

*2*2

1450 54.4 ± 13.9 23.3 ± 3.0 1.8 ± 1.1 404 ± 403 131 ± 17 81 ± 11 93 ± 16 4.1 [2.8e6.3] 98 [69e114] 204 ± 34 61.0 ± 15.4 125.9 ± 34.0 3.015 51.3 ± 13.0 21.6 ± 3.2 0.5 ± 0.8 32 ± 105 119 ± 17 74 ± 11 88 ± 10 4.0 [2.8e5.7] 70 [52e99] 209 ± 36 69.9 ± 15.5 122.7 ± 33.9

1113 54.9 ± 13.8 23.3 ± 3.0 0.8 ± 1.0 436 ± 434 129 ± 16 80 ± 11 93 ± 17 4.1 [2.9e6.5] 96 [68e138] 202 ± 33 57.3 ± 13.8 127.4 ± 33.1 2209 51.1 ± 13.1 21.5 ± 3.2 0.1 ± 0.4 28 ± 97 119 ± 18 73 ± 11 88 ± 10 3.9 [2.8e5.7] 71 [51e100] 210 ± 35 68.7 ± 15.1 124.1 ± 32.8

193 55.7 ± 13.5 23.4 ± 3.1 0.0 ± 0.2 462 ± 438 127 ± 17 78 ± 12 92 ± 15 5.1 [3.4e7.5] 105 [75e141] 208 ± 36 56.0 ± 13.3 133.9 ± 34.7 384 50.5 ± 13.1 21.6 ± 3.3 0.0 ± 0.1 33 ± 95 119 ± 17 74 ± 11 88 ± 11 4.3 [2.9e6.1] 69 [52e101] 208 ± 34 68.0 ± 15.1 123.3 ± 29.3

0.373 0.882 <0.001 0.059 <0.001 <0.001 0.520 0.099 0.605 0.085 <0.001 0.008 0.585 0.733 <0.001 0.367 0.777 0.215 0.307 0.173 0.975 0.556 0.007 0.285

Values are expressed as the mean ± standard deviation or median and interquartile range (insulin and triglyceride). Aldehyde dehydrogenase 2 (ALDH2) rs671 allele; *1: G (Glu) allele; *2: A (Lys) allele; BP, blood pressure; BMI, body mass index; HDL, high-density lipoprotein; LDL, lowdensity lipoprotein. Statistical significance was assessed using an analysis of variance.

Table 3 Association of the ALDH2 rs671 *1 allele with lipoprotein levels. Men

HDL-C (mg/dl) HDL2-C (mg/dl) HDL3-C (mg/dl) HDL2/HDL3 LDL-C (mg/dl) lbLDL-C (mg/dl) sdLDL-C (mg/dl) sdLDL/lbLDL

Women

Coefficient

b

SE

p

Coefficient

b

SE

p

2.878 1.606 1.272 0.018 2.646 4.738 2.092 0.062

0.122 0.083 0.195 0.026 0.049 0.124 0.069 0.127

0.405 0.320 0.117 0.011 0.992 0.700 0.443 0.006

<0.001 <0.001 <0.001 0.090 0.008 <0.001 <0.001 <0.001

1.000 0.496 0.503 0.018 1.188 1.488 0.300 0.012

0.040 0.024 0.085 0.023 0.022 0.042 0.012 0.046

0.297 0.247 0.077 0.009 0.615 0.441 0.242 0.002

0.001 0.044 <0.001 0.038 0.053 0.001 0.216 <0.001

Aldehyde dehydrogenase 2 (ALDH2) rs671 allele; *1: G (Glu) allele; HDL, high-density lipoprotein; LDL, low-density lipoprotein. Unstandardized coefficient, standardized coefficient (b), and standard error (SE) were calculated via a linear regression analysis with an additive genetic model. Adjusted factors consted of age, body mass index, menopausal status, Brinkman index, serum levels of glucose, insulin, triglyceride, LDL (for HDL) or HDL (for LDL) cholesterol levels, and fasting duration.

stronger for HDL3 than for HDL2. Although previous small scale conventional epidemiological studies [20,21] have also reported large increases in HDL3-C in drinkers, this is the first study to clearly show a predominant association of alcohol consumption with HDL3 by MRA in a large-scale general population. Another MRA in the atherosclerosis Risk in Community (ARIC) cohort which analyzed several SNPs in alcohol dehydrogenase (ADH) genes reported a weak association of the genotypes with HDL2-C but not HDL3-C [22]. However, the association of interracially common alcohol-related genotypes such as ADH1B with alcohol consumption was found to be weaker than that of ALDH2 [23]. Further, given an insignificant association of the ADH1b genotype with alcohol intake in Japanese [24], our present findings cannot be simply compared with results from ethnically different populations. A plausible mechanism for the predominant association of alcohol is the increased circulating levels of apolipoprotein A-I

(apoA-I) and A-II [25], major apolipoproteins in the HDL particle. Since functional interactions between the apoA-I and ABCA1 transporter is an initial step of small HDL formation, increasing apoA-I levels, as well as the increased transport rate of apoA-I and apoA-II [25], might provide a better explanation for the HDLincreasing effect of alcohol consumption, rather than the changes in activities of alcohol metabolism enzymes (e.g., decreased cholesteryl ester transfer protein (CETP) activity and increased lipoprotein lipase (LPL) activity [19]). CETP plays a pivotal role in the exchange of cholesterol esters contained in large HDL with TG in apo B-rich lipoproteins, which is the first step in taking up TGenriched HDL by the liver. Decreases in CETP activity were thus suggested to increase HDL2 and decrease HDL3 in a dosedependent manner [26,27], which is an inconsistent phenomenon with the present results. LPL catabolizes TG, and therefore, indirectly suppresses the CETP-mediated HDL metabolism pathway.

26

Y. Tabara et al. / Atherosclerosis 257 (2017) 22e28

Fig. 1. Adjusted mean cholesterol levels for the ALDH2 rs671 genotype in men. The values are mean adjusted for the following factors: (A, B, and C) age, body mass index, Brinkman index, serum levels of glucose, insulin, triglycerides, HDL-C levels, and fasting duration; (D) age, body mass index, Brinkman index, serum levels of glucose and insulin, and plasma levels of sdLDL-C and lbLDL-C, and fasting duration. The number of subjects in each subgroup are shown in the columns. Aldehyde dehydrogenase 2 (ALDH2) rs671 allele; *1: G (Glu) allele.

Table 4 Association of the ALDH2 rs671 *1 allele with triglyceride levels in men. Triglyceride levels (mg/dl) Model 1

ALDH2 genotype (per *1 allele) LDL-C (mg/dl) sdLDL-C (mg/dl) lbLDL-C (mg/dl) sdLDL/lbLDL ratio HDL2/HDL3 ratio

Model 2

Model 3

Model 4

Model 5

b

p

b

P

b

p

b

p

b

p

0.006

0.724

0.014 0.158

0.410 <0.001

0.097

<0.001

0.094

<0.001

<0.001

0.997

0.670 0.361

<0.001 <0.001 0.702

<0.001 0.364

<0.001

Statistical significance was assessed by a linear regression model adjusted for age, body mass index, Brinkman index, serum levels of glucose and insulin, and fasting duration. b indicates standardized coefficient. The ALDH2 rs671 genotype was included by an additive genetic model. A factor that showed the maximum variation inflation factor (VIF) among these models to be insulin (Model 5, VIF ¼ 1.36).

The involvement of other enzymes, (i.e., lecithin-cholesterol acyltransferase, phospholipid transfer protein, and hepatic lipase) has also been suggested, and the contribution of these factors, as well as that of CETP and LPL, has been reported to be minimal in light-tomoderate drinkers [19]. Given that our study population was a general population in which the average alcohol consumption of men was less than 30 g per sitting, changes in the enzymatic activity might not be a superior explanation for the HDL-increasing effect of alcohol consumption. The total LDL-C was lower in the *1 allele carriers, which is in agreement with our previous report [6] that demonstrated LDL-C as well as the number of LDL particles were lower in the *1 allele carriers. Similar results were also reported from an epidemiological study in Koreans, namely lower LDL-C and higher HDL-C levels in men carrying the enzymatically active *1 allele, but not in women [28]. Although conflicting results have been reported from conventional observational studies [29,30], the results of these studies cannot be simply compared to that of MRA due to the difficulty in controlling clinical and lifestyle factors which could be affected by alcohol consumption. Further, several MRA in Chinese population [31,32] failed to find a relationship between the ALDH2 rs671 genotype and LDL-C levels. Although the reason for this discrepancy are uncertain, possibilities are the inclusion of individuals using lipid lowering drugs, and substantially lower LDL-C levels in the Chinese population. Details regarding the mechanism by which alcohol intake decreases the circulating total LDL-C levels remains uncertain. Given

that the *1 allele was positively associated with sdLDL-C levels, TG is a possible factor underlying the relationship between alcohol intake and LDL-C. High TG levels in plasma increase the amount of TG-rich lipoproteins, and high TG-rich lipoproteins enhance the transfer of TG to LDL by CETP, which in turn increases sdLDL in plasma via the hydrolysis of core TG in LDL particles by the action of hepatic lipase [33]. Since the cholesterol content in sdLDL was lower than lbLDL, increased sdLDL and relatively lower lbLDL may explain the inverse association between alcohol intake and LDL-C levels. Lower TG levels in the *1 allele carriers observed in this study might partially be attributable to the increased sdLDL. It is probable that increased levels of HDL is the cardioprotective mechanism of alcohol consumption. However, the results of the present study raise the question of whether the decreased total LDL-C is a favorable response to alcohol intake, since the quality of LDL was impaired by increasing the proportion of atherogenic sdLDL. Given that people usually consume alcohol with meals (i.e., under conditions of higher plasma TG levels), further careful longitudinal studies focusing attention on the LDL subclasses are needed to precisely evaluate the effect of alcohol. Studies of different populations with different plasma lipid profiles may be helpful, especially since the mean plasma lipid levels in Japanese were reported to be lower than those of individuals living in other Western countries [34]. Several limitations of the present study should be mentioned. First, we did not consider nutrient intake or dietary patterns in this analysis, despite reports that increasing daily alcohol intake affects

Y. Tabara et al. / Atherosclerosis 257 (2017) 22e28

the intake of total energy and a number of nutrients [35], i.e. higher energy intakes from alcohol and consequently lower carbohydrate intakes in drinkers. Effects of the *1 allele on lipoprotein and TG levels might be underestimated because excessive energy intakes in drinker acts on increasing total LDL-C and TG levels and decreasing HDL-C level. However, while this potential influence does merit further investigation, we believe that we controlled for the major confounding factors by adjusting for BMI and smoking habits. Second, as daily alcohol consumption was obtained using a self-administrated questionnaire, there might be differential misclassification in alcohol consumption in individuals carrying *1 allele. However, we used the ALDH2 genotype as a proxy of alcohol consumption, the misclassification might not have substantial effects on the present findings. Third, we used the ALDH2 genotype as an instrumental variable in MRA instead of applying a two-stage least squares (TSL) analysis, the usual method in MRA [36], due to limitations on the accuracy of assessment of daily alcohol intake with the questionnaire. However, the results of both methods were found to be similar in a case using the ALDH2 genotype as a proxy of alcohol intake in Asians [37]. Third, we used a frozen plasma sample to measure lipoprotein subfractions. However, effect of frozen storage for measurement of TG, LDL-C, and HDL-C subfractions has been reported to be negligible when plasma sample was stored at 70  C [38]. Further, the presently observed genotype differences in lipoprotein and TG levels were substantially larger than the denaturation levels that reported to occur during sample storage. In summary, we clarified the causal relationship between alcohol intake and higher HDL-C and lowered LDL-C levels via performing a large-scale MRA in a Japanese general population. The effect of alcohol intake on HDL was stronger for HDL3-C, while the effect on LDL was inversely associated between sdLDL and lbLDL. Our results may help elucidate how alcohol intake affects lipoprotein levels and suggest the importance of considering lipoprotein subclasses to precisely evaluate the effect of alcohol intake on cardiovascular outcomes in epidemiological studies. Conflict of interest Measurements of lipoprotein subfractions were performed by Denka Seiken, Ltd. Financial support This study was supported by a University Grant, The center of Innovation Program, The Global University Project, and a Grant-inAid for Scientific Research from the Ministry of Education, Culture, Sports, Science & Technology in Japan, the Practical Research Project for Rare/Intractable Diseases, the Comprehensive Research on Aging and Health Science Research Grants for Dementia R&D from Japan Agency for Medical Research and Development (AMED), and a research grant from the Takeda Science Foundation. Acknowledgements We are extremely grateful to the Nagahama City Office and nonprofit organization Zeroji Club for their help in performing the Nagahama study. We also thank the editors of Crimson Interactive Pvt. Ltd. for their help in the preparation of this manuscript. References [1] P.E. Ronksley, S.E. Brien, B.J. Turner, K.J. Mukamal, W.A. Ghali, Association of alcohol consumption with selected cardiovascular disease outcomes: a systematic review and meta-analysis, BMJ 342 (2011) d671.

27

[2] H. Ueshima, K. Mikawa, S. Baba, S. Sasaki, H. Ozawa, M. Tsushima, A. Kawaguchi, T. Omae, Y. Katayama, Y. Kayamori, K. Ito, Effect of reduced alcohol consumption on blood pressure in untreated hypertensive men, Hypertension 21 (1993) 248e252. [3] S.R. Choudhury, H. Ueshima, Y. Kita, K.M. Kobayashi, A. Okayama, M. Yamakawa, Y. Hirao, M. Ishikawa, Y. Miyoshi, Alcohol intake and serum lipids in a Japanese population, Int. J. Epidemiol. 23 (1994) 940e947. [4] Y. Zhou, R. Yin, Y. Deng, Y. Li, J. Wu, Interactions between alcohol intake and the polymorphism of rs708272 on serum high-density lipoprotein cholesterol levels in the Guangxi Hei Yi Zhuang population, Alcohol 42 (2008) 583e591. [5] J.S. Tolstrup, M. Gronbaek, B.G. Nordestgaard, Alcohol intake, myocardial infarction, biochemical risk factors, and alcohol dehydrogenase genotypes, Circ. Cardiovasc Genet. 2 (2009) 507e514. [6] Y. Tabara, H. Ueshima, N. Takashima, T. Hisamatsu, A. Fujiyoshi, M. Zaid, M. Sumi, K. Kohara, T. Miki, K. Miura, SESSA Research Group; J-SHIPP Study Group. Mendelian randomization analysis in three Japanese populations supports a causal role of alcohol consumption in lowering low-density lipid cholesterol levels and particle numbers, Atherosclerosis 254 (2016) 242e248. [7] S. Mora, J.D. Otvos, N. Rifai, R.S. Rosenson, J.E. Buring, P.M. Ridker, Lipoprotein particle profiles by nuclear magnetic resonance compared with standard lipids and apolipoproteins in predicting incident cardiovascular disease in women, Circulation 119 (2009) 931e939. [8] R.C. Hoogeveen, J.W. Gaubatz, W. Sun, R.C. Dodge, J.R. Crosby, J. Jiang, D. Couper, S.S. Virani, S. Kathiresan, E. Boerwinkle, C.M. Ballantyne, Small dense low-density lipoprotein-cholesterol concentrations predict risk for coronary heart disease: the Atherosclerosis Risk in Communities (ARIC) study, Arterioscler. Thromb. Vasc. Biol. 34 (2014) 1069e1077. [9] H. Arai, Y. Kokubo, M. Watanabe, T. Sawamura, Y. Ito, A. Minagawa, T. Okamura, Y. Miyamato, Small dense low-density lipoproteins cholesterol can predict incident cardiovascular disease in an urban japanese cohort: the suita study, J. Atheroscler. Thromb. 20 (2012) 195e203. [10] V. Rakic, I.B. Puddey, S.B. Dimmitt, V. Burke, L.J. Beilin, A controlled trial of the effects of pattern of alcohol intake on serum lipid levels in regular drinkers, Atherosclerosis 137 (1998) 243e252. [11] B. Klop, Torres do rego a, castro cabezas m. Alcohol and plasma triglycerides, Curr. Opin. Lipidol. 24 (2013) 321e326. [12] K.A. Volcik, C.M. Ballantyne, F.D. Fuchs, A.R. Sharrett, E. Boerwinkle, Relationship of alcohol consumption and type of alcoholic beverage consumed with plasma lipid levels: differences between whites and african americans of the ARIC study, Ann. Epidemiol. 18 (2008) 101e107. [13] D.A. Lawlor, R.M. Harbord, J.A. Sterne, N. Timpson, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology, Stat. Med. 27 (2008) 1133e1163. [14] K. Amamoto, T. Okamura, S. Tamaki, Y. Kita, Y. Tsujita, T. Kadowaki, Y. Nakamura, H. Ueshima, Epidemiologic study of the association of low-Km mitochondrial acetaldehyde dehydrogenase genotypes with blood pressure level and the prevalence of hypertension in a general population, Hypertens. Res. 25 (2002) 857e864. [15] Y. Idewaki, M. Iwase, H. Fujii, T. Ohkuma, H. Ide, S. Kaizu, T. Jodai, Y. Kikuchi, A. Hirano, U. Nakamura, M. Kubo, T. Kitazono, Association of genetically determined aldehyde dehydrogenase 2 activity with diabetic complications in relation to alcohol consumption in Japanese patients with type 2 diabetes mellitus: the Fukuoka Diabetes Registry, PLoS One. 10 (2015) e0143288. [16] S. Maeda, K. Mure, K. Mugitani, Y. Watanabe, M. Iwane, O. Mohara, T. Takeshita, Roles of the ALDH2 and ADH1B genotypes on the association between alcohol intake and serum adiponectin levels among Japanese male workers, Alcohol Clin. Exp. Res. 38 (2014) 1559e1566. [17] Y. Ito, N. Satoh, T. Ishii, J. Kumakura, T. Hirano, Development of a homogeneous assay for measurement of high-density lipoprotein-subclass cholesterol, Clin. Chim. Acta 427 (2014) 86e93. [18] Y. Ito, M. Fujimura, M. Ohta, T. Hirano, Development of a homogeneous assay for measurement of small dense LDL cholesterol, Clin. Chem. 57 (2011) 57e65. [19] E.A. Brinton, Effects of ethanol intake on lipoproteins and atherosclerosis, Curr. Opin. Lipidol. 21 (2010) 346e351. [20] M. Nishiwaki, T. Ishikawa, T. Ito, H. Shige, K. Tomiyasu, K. Nakajima, K. Kondo, H. Hashimoto, K. Saitoh, M. Manabe, Effects of alcohol on lipoprotein lipase, hepatic lipase, cholesteryl ester transfer protein, and lecithin:cholesterol acyltransferase in high-density lipoprotein cholesterol elevation, Atherosclerosis 111 (1994) 99e109. €, Alcohol con[21] P. Sillanaukee, T. Koivula, H. Jokela, T. Pitk€ aj€ arvi, K. Seppa sumption and its relation to lipid-based cardiovascular risk factors among middle-aged women: the role of HDL3 cholesterol, Atherosclerosis 152 (2000) 503e510. [22] K.N. Vu, C.M. Ballantyne, R.C. Hoogeveen, V. Nambi, K.A. Volcik, E. Boerwinkle, A.C. Morrison, Causal role of alcohol consumption in an improved lipid profile: the atherosclerosis risk in communities (ARIC) study, PLoS One. 11 (2016) e0148765. [23] M. Ala-Korpela, N. Lankinen, A. Salminen, T. Suna, P. Soininen, R. Laatikainen, berger, K. Kaski, The inherent P. Ingman, M. Jauhiainen, M.R. Taskinen, K. He accuracy of 1H NMR spectroscopy to quantify plasma lipoproteins is subclass dependent, Atherosclerosis 190 (2007) 352e358. [24] A. Hidaka, S. Sasazuki, K. Matsuo, H. Ito, N. Sawada, T. Shimazu, T. Yamaji, M. Iwasaki, M. Inoue, S. Tsugane, JPHC Study Group. Genetic polymorphisms of ADH1B, ADH1C and ALDH2, alcohol consumption, and the risk of gastric

28

[25]

[26]

[27]

[28]

[29]

[30]

Y. Tabara et al. / Atherosclerosis 257 (2017) 22e28 cancer: the Japan Public Health Center-based prospective study, Carcinogenesis 36 (2015) 223e231. E. De Oliveira, E.R. Silva, D. Foster, M. McGee Harper, C.E. Seidman, J.D. Smith, J.L. Breslow, E a Brinton, Alcohol consumption raises HDL cholesterol levels by increasing the transport rate of apolipoproteins A-I and A-II, Circulation 102 (2000) 2347e2352. L. Lagrost, A. Athias, B. Herbeth, V. Guyard-Dangremont, Y. Artur, F. Paille, P. Gambert, C. Lallemant, Opposite effects of cholesteryl ester transfer protein and phospholipid transfer protein on the size distribution of plasma high density lipoproteins. Physiological relevance in alcoholic patients, J. Biol. Chem. 271 (1996) 19058e19065. K. Hirano, Y. Matsuzawa, N. Sakai, H. Hiraoka, S. Nozaki, T. Funahashi, S. Yamashita, M. Kubo, S. Tarui, Polydisperse low-density lipoproteins in hyperalphalipoproteinemic chronic alcohol drinkers in association with marked reduction of cholesteryl ester transfer protein activity, Metabolism 41 (1992) 1313e1318. Y. Cho, S.Y. Shin, S. Won, C.L. Relton, G. Davey Smith, M.J. Shin, Alcohol intake and cardiovascular risk factors: a Mendelian randomisation study, Sci. Rep. 5 (2015) 18422. E. Perissinotto, A. Buja, S. Maggi, G. Enzi, E. Manzato, E. Scafato, G. Mastrangelo, A.C. Frigo, A. Coin, G. Crepaldi, G. Sergi, ILSA Working Group. Alcohol consumption and cardiovascular risk factors in older lifelong wine drinkers: the Italian Longitudinal Study on Aging, Nutr. Metab. Cardiovasc Dis. 20 (2010) 647e655. A. Onat, G. Hergenc, D. Dursunoglu, S. Ordu, G. Can, S. Bulur, H. Yüksel, Associations of alcohol consumption with blood pressure, lipoproteins, and subclinical inflammation among Turks, Alcohol 42 (2008) 593e601.

[31] S.L. Au Yeung, C. Jiang, K.K. Cheng, B.J. Cowling, B. Liu, W. Zhang, T.H. Lam, G.M. Leung, C.M. Schooling, Moderate alcohol use and cardiovascular disease from Mendelian randomization, PLoS One 8 (2013) e68054. [32] A.E. Taylor, F. Lu, D. Carslake, Z. Hu, Y. Qian, S. Liu, J. Chen, H. Shen, G.D. Smith, Exploring causal associations of alcohol with cardiovascular and metabolic risk factors in a Chinese population using Mendelian randomization analysis, Sci. Rep. 5 (2015) 14005. [33] S. Hirayama, T. Miida, Small dense LDL: an emerging risk factor for cardiovascular disease, Clin. Chim. Acta 414 (2012) 215e224. [34] H. Ueshima, A. Sekikawa, K. Miura, T.C. Turin, N. Takashima, Y. Kita, M. Watanabe, A. Kadota, N. Okuda, T. Kadowaki, Y. Nakamura, T. Okamura, Cardiovascular disease and risk factors in Asia: a selected review, Circulation 118 (2008) 2702e2709. [35] G.A. Colditz, E. Giovannucci, E.B. Rimm, M.J. Stampfer, B. Rosner, F.E. Speizer, E. Gordis, W.C. Willett, Alcohol intake in relation to diet and obesity in women and men, Am. J. Clin. Nutr. 54 (1991) 49e55. [36] A.G. Boef, O.M. Dekkers, S. le Cessie, Mendelian randomization studies: a review of the approaches used and the quality of reporting, Int. J. Epidemiol. 44 (2015) 496e511. [37] D.A. Lawlor, M. Benn, L. Zuccolo, N.M. De Silva, A. Tybjaerg-Hansen, G.D. Smith, B.G. Nordestgaard, ADH1B and ADH1C genotype, alcohol consumption and biomarkers of liver function: findings from a Mendelian randomization study in 58,313 European origin Danes, PLoS One 9 (2014) e114294. [38] M.N. Nanjee, N.E. Miller, Evaluation of long-term frozen storage of plasma for measurement of high-density lipoprotein and its subfractions by precipitation, Clin. Chem. 36 (1990) 783e788.