Bone homeostasis is regulated by a delicate balance between osteoclasts, bone-resorbing cells, and bone-forming osteoblasts. Dysregulation of this process by increased osteoclast activity or reduced osteoblast function results in impaired bone status, as seen in osteoporosis. Osteoporosis is a multifactorial disease developing from interaction between sex, age, lifestyle factors, and genetic components.1, 2
Aldehyde dehydrogenases (ALDHs) are a superfamily of NAD(P)+-dependent enzymes, and, to date, 19 distinct ALDH genes have been identified in the human genome.3, 4 ALDH2 is reportedly expressed in various tissues such as liver, lung, testis, and colon, often with other ALDH family members.5–8 Aldehyde is produced in the first step of alcohol detoxification by the enzymatic activity of alcohol dehydrogenase (ADH) and is detoxified by ALDHs.5 Aldehyde is a substrate of several ALDH family members such as ALDH2, ALDH1A1, and ALDH1B1, and among them ALDH2 plays a major role in aldehyde detoxification.3, 8 A single-nucleotide polymorphism (SNP) in human ALDH2, rs671, results in the amino acid change ALDH2Glu478Lys, referred to as ALDH2*2, which acts as a dominant-negative protein. ALDH2 functions as a homo- or heterotetramer, and all tetramers containing at least one ALDH2*2 subunit are inactive.9 Humans homozygous for the ALDH2*2 allele exhibit no ALDH2 activity, whereas activity in human heterozygotes is as low as 1/16 that seen in wild-type (WT) (ALDH2*1) individuals.9, 10 ALDH2*2 is implicated in defects in alcohol metabolism as well as diseases such as Alzheimer's disease.11–13 Recently, a significant association was found between rs671 and osteoporosis, and that polymorphism greatly increases the risk of bone fracture and osteoporosis.14 More recently, it was reported that Aldh2-deficient mice exhibited reduced trabecular bone formation and bone volume with alcohol consumption.15, 16 Furthermore, it was suggested that the interaction of certain lifestyles with mutated ALDH2 genes causes idiopathic osteonecrosis of the femoral head.17 By contrast, ALDH2*2 reportedly induces increased resistance to ischemia/reperfusion of the heart.18 However, the molecular action of ALDH2*2 on bone metabolism remains largely unknown. In this study, we found that ALDH2*2 transgenic mice (Aldh2*2 Tg) exhibit severe osteoporotic phenotypes and show increased blood aldehyde levels without alcohol consumption. Our results indicate that the dominant negative form of ALDH2, ALDH2*2, promotes increased aldehyde levels, which in turn induce hyperaccumulation of peroxidated lipid and severe osteopenia owing to inhibited osteoblastogenesis, and that such defects can likely be rescued by antioxidant treatment.
Subjects and Methods
Chemicals and antibodies
Acetaldehyde, GW9664, Trolox C, and all reagents used in enzyme assays were obtained from Wako Co. Ltd. (Tokyo, Japan). Human recombinant bone morphogenic protein (BMP) was purchased from R&D Systems, Inc. (Minneapolis, MN, USA). Human mesenchymal stem cell conditioning medium was purchased from Takara Bio Inc. (Tokyo, Japan). Actin and goat anti-rabbit immunoglobulin G (IgG) antibodies were purchased from Sigma-Aldrich (St. Louis, MO, USA), and from Life Technologies Ltd. (Carlsbad, CA, USA), respectively. 4-Hydroxy-2-nonenal (4HNE) antibody was obtained from Calbiochem (Darmstadt, Germany), and phospho-Smad1,5,8 antibody was obtained from Cell Signaling Technology, Inc., (Danvers, MA, USA). ALDH2 antibody was prepared as described.18
ALDH2*2 Tg mice were previously generated as described in prior studies.18 Animals were maintained under specific pathogen-free conditions in animal facilities certified by the Keio University School of Medicine Animal Care Committee. Animal protocols were approved by the Keio University School of Medicine Animal Care Committee.
Cell culture and treatment
Mouse primary osteoblastic cells were isolated from calvaria of newborn mice or from bone marrow cells of 3-month-old mice, and seeded in 96-well plates in osteogenic medium (Differentiation Basal Medium–Osteogenic; Takara Bio Inc.) at 2.5 × 104 and 1.0 × 105 cells/mL, respectively. Osteoblastic MC3T3-E1 cells were cultured at 2.5 × 104 or 5.0 × 105 cells per 96-well or six-well plate, respectively, in alpha MINIMUM Eagle's Medium (α-MEM) containing 10% fetal calf serum (FCS) with or without 300 ng/mL recombinant human BMP2 (R&D Systems). In some experiments, 0.04% acetaldehyde was added to cultures of primary osteoblasts or MC3T3-E1 cells. Osteoblastogenesis was evaluated by staining with alkaline phosphatase (ALP), Von Kossa, and Alizarin Red. For ALP staining, cells were fixed in 4% paraformaldehyde in PBS and treated for 20 minutes with 5-bromo-4-chloro-3′-indolyphosphate/nitro-blue tetrazolium and (BCIP/NBT) Liquid Substrate System (Sigma-Aldrich). Cells were then washed with deionized water and air-dried. For Von Kossa staining, cells were fixed in 1% formaldehyde in PBS for 20 minutes and incubated with 5% silver nitrate solution for 30 minutes under a bright light, and then treated with stop solution. For Alizarin Red staining, cells were fixed in ice-cold 95% ethanol for 20 minutes, stained with 1% Alizarin Red at room temperature for 10 minutes, rinsed with distilled water, and observed microscopically (IX-70, Olympus, Tokyo, Japan). To quantitate Alizarin Red staining, 10% cetylpyridinium chloride was added to stained cells to extract the stain, which was assayed spectrophotometrically by absorption at 550 nm using a standard curve (mg/mL).
Dual-energy X-ray absorptiometry analysis
For DXA analysis, 2-month-old Aldh2*2 Tg and WT mice were euthanized, individual left femurs were dissected free of tissue and fixed in 70% ethanol, and bone mineral density (BMD) was analyzed.
For dynamic histomorphometry, mice were injected with 10 mg/kg body weight of calcein in 2% NaHCO3 5 days and 1 day before euthanasia. The right proximal tibia was removed from each mouse, fixed in 70% ethanol, and embedded in glycol methacrylate (Wako Pure Chemical Industries, Ltd., Osaka, Japan) without decalcification. Then, serial sections (3 µm in thickness) were cut longitudinally using a rotary microtome (RM2255; Leica Microsystems, Wetzlar, Germany) and sections were stained with toluidine-blue and tartrate resistance acid phosphatase (TRAP) as described.19, 20 The bone morphometric analysis in secondary spongiosa was performed in sections stained with toluidine-blue using OsteoMeasure software (OsteoMetrics, Inc. Decatur, GA, USA) as described,21–23 and cortical thickness was analyzed in hematoxylin and eosin (HE) staining sections by NIH Image (http://rsb.info.nih.gov/ij/). Stained sections were observed under a microscope (Keyence Co, Osaka, Japan).
Whole-cell lysates were obtained by solubilizing cells at 4°C in 100 µL lysis buffer containing 10 mM Tris (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1% Triton X-20, plus 1 mM phenylmethylsulfonylfluoride, 10 µg/mL aprotinin, 10 µg/mL leupeptin, 1 mM Na3VO4, 5 mM NaF, and 0.1% SDS. Lysates were cleared by centrifugation at 900g at 4°C for 15 minutes. Equivalent amounts of protein were separated by SDS-PAGE and transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, Darmstadt, Germany). Proteins were detected using anti-4HNE (Calbiochem), anti-ALDH218, anti-phospho Smad1,5,8 (Cell Signaling Technology), and anti-actin antibodies. Immunoreactive bands and biotinylated molecular weight standards were visualized by ECL (GE Healthcare, Little Chalfont, UK). 4-HNE bands were quantified using NIH image.
RNA isolation and quantitative real-time RT-PCR
Total RNA from cultured cells and tissues was isolated using the Qiagen reagent (Qiagen Inc., Hilden, Germany) or TRIzol reagent (Life Technologies Ltd), respectively. After denaturation of total RNA at 70°C for 5 minutes, cDNAs were synthesized from total RNA using an oligonucleotide dT primer by reverse transcription (Clontech/Takara Bio Inc., Shiga, Japan). Quantitative real-time RT-PCR was performed using SYBR Green PCR Master Mix (Takara Bio Inc.) with a Takara Prism 7000 sequence detection system, according to the manufacturer's instructions. Samples were matched to a standard curve generated by amplifying serially-diluted products using the same PCR reactions. β-actin expression served as internal control. Primer sequences were as follows:
Osteoblastic cells were washed twice with phosphate buffered saline without calcium and magnesium (PBS) (−), harvested, suspended in 0.25 M sucrose and 0.3 mM EDTA (pH 7.0), and homogenized. The sample was centrifuged at 900g for 10 minutes at 4°C to obtain whole-cell fractions. Total ALDH enzymatic activity was measured by detecting reduced NAD+ at 340 nm in 3.5 mg of the whole-cell fraction, except for cell membranes, as described, with minor modifications.20 The reaction mixture contained 50 mM sodium pyrophosphate buffer (pH 8.8), 0.5 mM NAD+, 0.1 mM pyrazole, 5 mM acetaldehyde, and 2 µM rotenone.
Assay of acetaldehyde concentration by gas-liquid chromatography
Samples were prepared according to a previous report.24, 25 Each blood sample was collected in a heparinized vial, and plasma fractions were analyzed by gas chromatography. In the final step, gas analysis was done using a Hitachi 164F flame ionization detector (FID) gas chromatograph Hitachi Ltd. (Tokyo, Japan). Columns packed with polyethylene glycol having an average molecular weight of 20 kDa (PEG-20M) (3 m × 2 mm) were used at a column temperature of 110°C and a helium flow rate of 30 mL/min. The relative concentration of acetaldehyde was determined by peak height compared with that of the internal control. Ethanol at 0.01% was included in every sample.
Detection of apoptosis
Apoptosis was measured using the Apo Alert Annexin V-FITC Apoptosis Detection Kit (Clontech/Takara Bio Inc.), and apoptotic cells were analyzed by FACS analysis.
Cell proliferation assay
For cell proliferation assay, 2.5 × 104 MC3T3-E1 cells were plated into 96-well culture plate with BMP2 (300 ng/mL) in the presence or absence of acetaldehyde or acetaldehyde + Trolox C for 4 days, and the cells counts were examined by Cell Counting Kit-F (Calcein-AM; Dojindo, Kumamoto, Japan).
Human study was approved by an Institutional Ethical Review Board (Keio Hospital #16-17-1), and written informed consent was obtained from all subjects. Human bone marrow cells were obtained by aspiration from patient donors suffering from hip osteoarthritis.
Statistical analysis was performed by Student's t test or one-way ANOVA, followed by Tukey-Kramer test to determine significance between groups. In this context, significant differences were defined as p ≤ 0.05.
The Aldh2*2 transgenic mouse exhibits osteoporosis
To evaluate the effect of the dominant-negative form of ALDH2 (ALDH2*2) on bone metabolism, we analyzed mice expressing Aldh2*2 (Aldh2*2 Tg) (Fig. 1). Aldh2*2 expression in transgenic mice was driven by the Actin promoter and detected in tissues such as skeletal muscle and heart, as described.18 We analyzed two independent lines of Aldh2*2 Tg mice that expressed the transgene at high and low levels, respectively (Supplementary Fig. S1). Soft X-ray and dual-energy X-ray absorptiometry (DXA) analysis demonstrated that Aldh2*2 Tg mice exhibited reduced radio-opacity and BMD relative to WT mice (Fig. 1A, B and Supplementary Fig. S1), indicative of osteoporosis. The reduction of bone mass seen in different Aldh2*2 Tg lines was positively correlated with Aldh2*2 expression levels. For further analysis we chose the high-expressing line, since those mice exhibited more apparent bone phenotypes. Trabecular bone and osteoclast formation, as shown by toluidine blue/Villanueva-Goldner and tartrate resistance acid phosphatase (TRAP) staining, respectively, were both inhibited in Aldh2*2 Tg compared with WT mice (Fig. 1C). In contrast, defects in growth plate chondrocytes were not seen in Aldh2*2 Tg (Supplementary Fig. S2). Bone morphometric analysis demonstrated that osteoclast parameters such as osteoclast number per bone perimeter (Oc.N/BP), osteoclast surface per bone surface (OS/BS), and eroded surface per bone surface (ES/BS) were all reduced in Aldh2*2 Tg compared to WT mice (Fig. 1D). Although reduced osteoclast activity could potentially promote increased bone mass, the significant reduction in bone mass evident in Aldh2*2 Tg mice suggested that the decrease was due to impaired osteoblastogenesis. Indeed, bone volume per tissue volume (BV/TV) was reduced in Aldh2*2 mice relative to WT mice (Fig. 1D), and trabecular bone and cortical bone thickness (Tb.Th and Co.Th) and bone formation rate per bone surface (BFR/BS) were significantly reduced in Aldh2*2 Tg compared to WT mice (Fig. 1D), supporting the idea that osteoblastogenesis was inhibited and suggesting an underlying cause for osteoporosis.
Aldh2*2 Tg mice show inhibition of osteoblast differentiation
To determine the effects of Aldh2*2 on bone formation, osteoblastogenesis was analyzed in osteoblastic cells derived from Aldh2*2 Tg or WT mice (Fig. 2). Osteoblast differentiation, as evaluated by ALP, Alizarin Red, and Von Kossa staining (Fig. 2A), was strongly inhibited in osteoblastic cells derived from Aldh2*2 Tg compared to WT mice. Calcium deposition, as analyzed by quantitative Alizarin Red staining, was also significantly decreased in Aldh2*2 Tg osteoblastic cells (Fig. 2B). Furthermore, expression of ALP as analyzed by RT-PCR and real-time PCR, was significantly decreased in Aldh2*2 Tg osteoblasts compared with WT cells (Fig. 2C, D). These results indicate that the Aldh2*2 mutation promotes impaired osteoblast differentiation.
Acetaldehyde accumulates in Aldh2*2 Tg mice
Aldh2*2 Tg osteoblastic cells showed inhibited Aldh2 enzymatic activity compared to WT and even Aldh2-deficient osteoblastic cells (Supplementary Fig. S3), supportive of the idea that Aldh2*2 has a competitive-inhibition or dominant-negative function. Indeed, reduced bone mass seen in Aldh2*2 Tg mice was more severe than that seen in Aldh2-deficient mice (Supplementary Fig. S4). Aldh2-deficient mice reportedly exhibited reduced bone mass with alcohol consumption16; however, Aldh2*2 Tg mice exhibited reduced bone mass without alcohol consumption, suggesting that Aldh2*2 regulates physiological bone homeostasis. ALDH2 is known to play a pivotal role in an aldehyde metabolic pathway. Gas-liquid chromatography analysis showed increased acetaldehyde levels in blood plasma of Aldh2*2 Tg compared to WT mice (Supplementary Fig. S5). Thus, we next examined the effects of acetaldehyde on osteoblastogenesis. Osteoblastic MC3T3-E1 cells were cultured with BMP2 to induce osteoblast differentiation in the presence or absence of acetaldehyde, and osteoblastogenesis was analyzed by ALP protein and mRNA expression (Fig. 3). Differentiation evaluated by ALP staining (Fig. 3A) and detection of ALP expression by RT-PCR (Fig. 3B) and real-time PCR (Fig. 3C) was significantly inhibited by acetaldehyde in WT osteoblasts. Furthermore, expression of osteocalcin (OCN) and osteopontin (OPN), both markers of osteoblast differentiation, was also significantly inhibited by acetaldehyde in WT osteoblasts (Fig. 3C).
Antioxidant restores the inhibited osteoblastogenesis by acetaldehyde
Increased levels of aldehyde adducts are implicated in various disease states and are detected in oxidized lipoproteins in osteoarthritic articular tissues, atherosclerotic lesions, cardiac tissue from patients with coronary artery disease, and brains of individuals with Alzheimer's disease.26–28 Peroxidated lipid-protein, which occurred by accumulation of acetaldehyde or reactive oxygen spices, reportedly accumulates in cells with ALDH2 dysfunction.11–13, 18 Indeed, Western analysis indicated significant accumulation of 4HNE, a major peroxidated lipid-protein in osteoblasts derived from Aldh2*2 Tg compared to WT mice (Fig. 4A), suggesting that Aldh family enzymes inhibit 4HNE formation, while Aldh2*2 antagonizes that activity and thereby promotes 4-HNE accumulation. 4HNE accumulation was also induced by acetaldehyde treatment of C3H/10T1/2 mesenchymal stem cells (Fig. 4B), suggesting that defects in osteoblastogenesis resulting from Aldh2*2 expression or acetaldehyde treatment are likely due to increased levels of oxidative stress induced by peroxidated lipid-protein. Indeed, osteoblast differentiation shown by ALP and OCN expression was significantly inhibited by acetaldehyde, and such inhibition was significantly rescued by treatment with the vitamin E analogue Trolox C (Fig. 4B–D), an antioxidant.29, 30 This observation suggests that osteoblastogenesis inhibited by acetaldehyde is due, at least in part, to increased oxidant levels.
Osteoblastic cell differentiation was induced by Smad 1,5,8 phosphorylation under the stimulation of BMP2.31 However, the phosphorylation of Smad 1,5,8 induced by BMP2 was strongly inhibited by acetaldehyde treatment and such inhibition was restored by Trolox C in osteoblastic cells (Fig. 5A). Total Smad1 expression was reduced by acetaldehyde treatment and was recovered by Trolox C in osteoblastic cells (Fig. 5A). Furthermore, expression of runx2 and osterix, both of which are essential transcriptional factors for osteoblastogenesis,32, 33 was also inhibited by acetaldehyde treatment and was restored by Trolox C in osteoblastic cells (Fig. 5B).
Osteoblast proliferation was inhibited by acetaldehyde treatment but was restored by Trolox C treatment (Fig. 5C). Similarly, apoptosis was induced by acetaldehyde treatment but was restored by Trolox C treatment in osteoblasts (Fig. 5D).
Induction of adipogenesis in Aldh2*2 Tg mice or acetaldehyde-treated osteoblastic cells
Peroxisome proliferator-activated receptor gamma (PPARγ) is an essential transcription factor regulating adipogenesis.34, 35 Activation of PPARγ inhibits the runx2 function in osteoblasts.36 Interestingly, we found that PPARγ2 expression was significantly elevated in osteoblastic cells from Aldh2*2 Tg mice or in acetaldehyde-treated osteoblasts compared to cells from WT mice or nontreated osteoblasts, respectively (Fig. 6A and data not shown). Similarly, PPARγ1 expression was induced in Aldh2*2 Tg osteoblasts and in acetaldehyde-treated osteoblasts (Supplementary Fig. S6). Adipogenesis was indeed stimulated in Aldh2*2 Tg compared to WT littermate mice, and increased adipogenic droplets were evident in Aldh2*2 bone sections (Fig. 6B). Expression of the adipogenic markers PPARγ2 and adipose fatty acid binding protein 2 (aP2) was significantly upregulated in bone from Aldh2*2 Tg compared to WT mice, further suggesting that adipogenesis was stimulated in transgenic mice (Fig. 6B). Increased PPARγ2 expression promoted by acetaldehyde was significantly downregulated by antioxidant treatment (Fig. 6C). PPARγ reportedly inhibits osteoblastogenesis,34–36 suggesting that inhibited osteoblastogenesis seen in Aldh2*2 Tg mice or in acetaldehyde-treated cells is due to increased PPARγ expression. Indeed, osteoblastogenesis inhibited by acetaldehyde was significantly rescued following treatment with the PPARγ inhibitor GW9662 (Fig. 6D). Although not statistically significant, inhibition of ALP expression as well as elevation of PPARγ2 expression were also detected in osteoblasts derived from human subjects carrying the ALDH2*2 gene compared to individuals carrying the WT allele (Fig. 6E). Overall, our data provides new insight into mechanisms underlying osteoporosis and suggests novel therapeutic strategies.
Osteoporosis is a complex disease promoted by various events, such as estrogen deficiency and gene mutations.1, 2 In this study, we demonstrate that Aldh2*2, a dominant-negative form of Aldh2 protein, induces accumulation of acetaldehyde, which in turn leads to osteoporosis due to impaired osteoblastogenesis. Acetaldehyde treatment induced accumulation of 4HNE, peroxidated lipid-protein in osteoblasts, which in turn induced apoptosis and inhibited differentiation and proliferation in osteoblasts. The presence of Aldh2*2 or treatment with acetaldehyde stimulated PPARγ expression in osteoblastic cells, an effect antagonized by antioxidant treatment. Antioxidant treatment restored inhibited osteoblast differentiation by acetaldehyde. Because PPARγ inhibitors restored osteoblastogenesis inhibited by acetaldehyde, PPARγ induction is likely a key event in inhibiting osteoblastogenesis by Aldh2*2. ALDH2*2 inhibits osteoblast differentiation by inducing accumulation of acetaldehyde followed by accumulation of 4HNE adducts, oxidative stress, and in turn induces osteoblast apoptosis, inhibits proliferation and differentiation by elevating PPARγ expression. Thus, therapies with antioxidants or PPARγ-inhibitors could potentially antagonize osteoporosis induced by ALDH2*2 or excess alcohol intake.
We specifically showed that acetaldehyde accumulates in Aldh2*2 Tg mice blood plasma and that exogenous acetaldehyde induces 4HNE accumulation in osteoblasts. We also detected 4HNE accumulation in osteoblasts derived from Aldh2*2 mice and acetaldehyde-treated osteoblasts, suggesting that in osteoblasts Aldh2 likely functions in physiological detoxification of acetaldehyde. ALDH2 plays a major role in detoxification of aldehyde, which is produced by ADH during alcohol detoxification and, if accumulated, causes several pathologies associated with alcoholism3, 7. Individuals carrying Aldh2*2 are reportedly susceptible to “alcohol flush” syndrome due to altered alcoholic detoxification.37 Reduced bone mass was seen in Aldh2-deficient mice with alcohol consumption.16 However, Aldh2*2 Tg mice exhibited significantly reduced bone mass without alcohol consumption, indicating that Aldh2*2 plays a pivotal role in physiological osteoblastogenesis and bone homeostasis. Indeed, we showed that osteoblastogenesis in Aldh2*2 cells was significantly inhibited compared with WT cells even in the absence of alcohol in vitro.
Aldh2*2 could play two different roles in regulating osteoblast differentiation. Because ALDH family proteins form a hetero- or homotetramer,38 Aldh2*2 may act as a competitive inhibitor of non-mutant ALDH proteins in tetramer formation. ALDH2*2 could also exhibit dominant-negative activity and interfere with the activity of not only normal ALDH2 but other ALDH family members.4, 39 As a likely consequence, ALDH2*2 Tg exhibited severely reduced ALDH enzymatic activity and bone mass compared not only with WT but with Aldh2-deficient mice (Supplementary Figs. S2 and S3). Therefore, controlling aldehyde detoxification is likely crucial for individuals carrying ALDH2*2.
Osteoblasts are derived from mesenchymal stem cells, which are the common progenitors of adipocytes, and there is a delicate balance in lineage commitment between osteoblasts and adipocytes. PPARγ transcription factors are essential for adipogenesis, but they inhibit osteoblastogenesis. We showed that PPARγ expression is upregulated in Aldh2*2 Tg or acetaldehyde-treated, WT osteoblasts and 3T3-E1 cells, resulting in a lineage shift toward adipocytes and away from osteoblasts.
Aldh2*2 is implicated in diseases such as late-onset Alzheimer's disease11–13; however, the Aldh2*2 mutation reportedly induces resistance to ischemia/reperfusion in heart tissue.18 In bone tissue, homeostasis is regulated by bone resorption by osteoclasts and bone formation by osteoblasts, and we demonstrate that Aldh*2 inhibits osteoblastogenesis. Taken together, our study brings new insight to the molecular mechanisms underlying osteoporosis due to the Aldh2*2 mutation in bone homeostasis. Administration of antioxidant reagents and PPARγ inhibitors may provide potential therapeutic strategies for treating Aldh2*2 or alcohol-induced osteoporosis.
All authors state that they have no conflicts of interest.
We thank Prof. H. Matsuhashi at Hokkaido University of Education for technical support of gas-liquid chromatography analysis. T. Miyamoto was supported by a grant-in-aid for Young Scientists (A) and by Precursory Research for Embryonic Science and Technology (PREST), the Takeda Science Foundation, the Inamori Foundation and the Keio Kanrinmaru project, Japan. H. Hoshi was supported by a grant-in-aid for Young Scientists and by the Asahi Breweries Foundation.
Authors' roles: HH and WH performed experiments. YF and AF collected human samples. YM, KH, KM, RI, HM, SY, TM, HK, EK and AF performed some experiments. YS and TK maintained and genotyped mice. KK, KN and TK prepared Aldh2 KO mice. MS, KF, HM, KC and YT discussed results. IO and SO prepared Aldh2*2 Tg mice. TM wrote the manuscript.