A common East Asian-specic ALDH2 mutation causes obesity and insulin resistance: therapeutic effect of reducing toxic aldehydes by ALDH2 activation

Obesity and type 2 diabetes have reached pandemic proportion. In particular, the population with diabetes is expected to rise rapidly in East and South Asia. ALDH2 (acetaldehyde dehydrogenase 2, mitochondrial) is the key metabolizing enzyme of acetaldehyde and other toxic aldehydes, such as 4-hydroxynonenal (4-HNE). A missense mutation, Glu504Lys of ALDH2 (denoted as the ALDH2*2 allele) is prevalent in 560 million East Asians, resulting in reduced ALDH2 enzymatic activity. We found that Aldh2*2/*2 homozygous knock-in (KI) mice mimicking human Glu504Lys mutation were prone to develop diet-induced obesity, glucose intolerance, insulin resistance, and fatty liver on a high-fat high-sucrose diet compared with controls. The Aldh2 KI mice demonstrated reduced energy expenditure and thermogenesis. Proteomic analyses of the brown adipose tissue (BAT) of the Aldh2 KI mice identi�ed increased 4-HNE-adducted proteins involved in fatty acid oxidation and electron transport chain. Fatty acid oxidation rate and mitochondrial electron transport activity were reduced in the BAT of the Aldh2 KI mice, which explained the decrease in thermogenesis and energy expenditure. AD-9308 is


Introduction
Obesity and type 2 diabetes have reached pandemic levels.The 2019 International Diabetes Federation reported that there were 463 million people living with diabetes worldwide.This number is projected to increase rapidly to 578 million in 2030 1 .East Asia and South Asia, the most populous regions in the world, are expected to be the major contributors of new cases 1 .This pandemic is primarily caused by a high-calorie diet enriched in fat and sugar, sedentary lifestyle, and their interaction with genetic factors 1,2 .
Approximately 36% of East Asians (560 million) or nearly 8% of the global population, carry an inactivating missense mutation of the ALDH2 gene (denoted as the ALDH2*2 allele), which causes Glu504Lys substitution 3 .This mutation results in a reduction of the ALDH2 enzymatic activity by 60-80% in heterozygous carriers (ALDH2*1/*2) and ~ 90% in homozygous carriers (ALDH2*2/*2) 4 .ALDH2*2 heterozygotes and ALDH2*2 homozygotes exhibit sensitivity to alcohol, with presentations ranging from facial ushing, headache, and tachycardia due to a rapid increase in circulating acetaldehyde concentrations 5,6 .Epidemiological studies suggest a correlation between this inactivating mutation and several diseases including oral cancer, esophageal cancers, and blood and solid tumors associated with Fanconi anemia 7,8 .Importantly, several large-scale meta-analyses of genome-wide association studies revealed that this inactivating mutation is strongly associated with type 2 diabetes 9 , body mass index 10 , and serum lipids 11 in East Asians.A validation study further con rmed a close association of ALDH2*2 with visceral fat distribution in 2,958 Chinese subjects 12 .In addition, carriers of the inactivating ALDH2*2 variant had 2-3 times increased risk of non-alcoholic fatty liver disease among Japanese subjects 13 .
The ALDH2 gene encodes mitochondrial aldehyde dehydrogenase 2 (ALDH2), a major acetaldehydemetabolizing enzyme responsible for ~ 95% acetaldehyde metabolism due to its low Km for acetaldehyde 14 .Acetaldehyde is produced by alcohol dehydrogenase (ADH) following alcohol ingestion.Acetaldehyde can also be generated endogenously from the intermediate metabolisms or by gut microbial ora 15 .Air pollution, thermal degradation of plastics, and spoiled food are also sources of toxic aldehyde [16][17][18] .
Acetaldehyde is listed as group I carcinogen by the International Agency for Research on Cancer 19 .In addition to acetaldehyde, ALDH2 also metabolizes various bioactive toxic aldehydes, including acrolein, malondialdehyde and 4-hydroxynonenal (4-HNE).Among these bioactive aldehydes, the most intensively studied has been 4-HNE, a lipid peroxidation product that forms covalent adduct to macromolecules such as protein, lipid and DNA, causing cellular damage [20][21][22][23][24] .
With the advent of high-throughput screening, a small molecule, Alda-1, was identi ed as an activator for both the wild-type ALDH2and mutant ALDH2 enzymes 25 .Based on X-ray crystallography and enzyme kinetics, the binding of Alda-1 partially blocks the exit tunnel of substrates, thus increasing the likelihood of productive encounters between reaction intermediate and the catalytic site 26 .Administration of Alda-1 has been effective to treat myocardial infarction, aortic aneurysm, atrial brillation, Alzheimer's disease and nociceptive pain in mice model by enhancing the clearance of 4-HNE and in rodent model 27,27−29 .The discovery of ALDH2 activators offers an opportunity to test the therapeutic potential of reducing toxic aldehydes for treating a variety of diseases 30 .AD-9308 is a highly water soluble and orally bioavailable prodrug of a potent and highly selective ALDH2 activator AD-5591, a new generation ALDH2 activator that has the improved biological activities and pharmacological properties compared to Alda-1 31 .
In this study, we used aldh2*2/*2 homozygous knock-in (KI) mice, which mimic the East Asian-speci c Glu504Lysmutation, to evaluate the effect of this mutation on diet-induced obesity, glucose homeostasis, fatty liver, and serum lipids and tested the therapeutic effect of a novel ALDH2 activator AD-5591 by dosing its prodrug AD-9308 for treating metabolic disorders.

Results
Aldh2 KI mice carrying the East Asian-speci c Glu504Lys mutation were prone to develop diet-induced obesity and fatty liver To assess the effect of the East Asian-speci c ALDH2 Glu504Lys mutation on diet-induced obesity and related metabolic traits, Aldh2 KI and WT mice littermates were placed on either regular chow or high-fat high-sucrose diet (HFHSD) for 24 weeks since the age of 4 weeks.At the end of study, the body weights of Aldh2 KI and WT mice were 42.45 ± 6.59 and 37.78 ± 5.54 g respectively (Fig. 1a, P < 0.0001).However, no difference of body weight gain was observed between Aldh2 KI and WT mice fed a chow diet (Fig. 1b).Aldh2 KI mice fed on HFHSD also had more white fat, including inguinal (1.33 ± 0.1 vs. 1.02 ± 0.09 g, P = 0.04), mesenteric (0.68 ± 0.091 vs. 0.51 ± 0.051 g, P = 0.077), and perigonadal fat (2.29 ± 0.27 vs. 1.48 ± 0.13 g, P = 0.004) and increased liver weight (1.89 ± 0.07 vs. 1.64 ± 0.06 g, P < 0.01), but less brown adipose tissue (BAT) weight (0.10 ± 0.01 vs. 0.14 ± 0.10 g, P = 0.016) compared with WT littermates (Fig. 1c).No such differences were found between Aldh2 KI and WT mice on chow diet except for the smaller BAT found in Aldh2 KI mice.Body composition analysis revealed signi cantly increased fat mass (12.46 ± 0.3 vs. 8.85 ± 1.23 g, P = 0.04), and to a lesser extent, increased lean mass and total water conten in Aldh2 KI mice compared with the WT mice fed on HFHSD (Fig. 1d). Figure 1e showed the representative gross appearance of mice, BAT, perigonadal fat, and liver.H&E stain using of perigonadal fat showed hypertrophic adipocytes with more crown-like necrosis in Aldh2 KI mice compared with controls (Fig. 1f, 1 h).There was no difference in the number of adipocytes between the two groups (Fig. 1f-h), indicating hypertrophy rather than hyperplasia of white adipose tissue.Aldh2 KI mice also had signi cantly higher hepatic triglycerides contents (0.249 ± 0.055 vs. 0.124 ± 0.027 mg/mg liver tissue, P = 0.005), and more severe hepatic steatosis than the WT mice on HFHSD (Fig. 1i, 1j).However, there was no difference in muscle triglycerides content (Fig. 1k).These results suggest that the East Asian-speci c ALDH2 Glu504Lys mutation promoted HFHSD-induced obesity and fatty liver in mice.
Aldh2 KI mice had reduced energy expenditure and impaired adaptive thermogenesis Energy expenditure measured by indirect calorimetry showed that Aldh2 KI mice had lower energy expenditure (Fig. 2a) than WT mice.There was no difference in food intake between Aldh2 KI and WT mice (Fig, 2b).These data indicated Aldh2 KI mice were prone to diet-induced obesity due to reduced energy expenditure but not intake.
Energy expenditure is composed of basal metabolic rate, physical activity, and adaptive thermogenesis including cold-induced and diet-induced thermogenesis.Indirect calorimetry showed no difference in basal metabolic rate as indicated by the energy expenditure in resting (light) phase (Fig. 2a).We also measured energy expenditure by monitoring physical activities of the mice.Daily physical activity, including wheel rotations (Fig. 2c) and travel distances (Fig. 2d) were similar between the two groups.
HomeCage monitoring systems did not detect signi cant differences in various mouse behaviors including awakening, feeding, hanging, rearing up, resting, twisting, and walking except for slightly increased grooming behavior in Aldh2 KI mice (Fig. 2e).
For cold tolerance test, Aldh2 KI mice had lower rectal temperature after 12 hours of prolonged cold exposure at 4 °C (Fig. 2g), indicating cold intolerance.These data suggest that Aldh2 KI mice may develop obesity due to reduced energy expenditure resulting from impaired adaptive thermogenesis.
Aldh2 KI mice displayed reduced insulin sensitivity and impaired glucose tolerance Since the ALDH2 Glu504Lys mutation is reported to be associated to type 2 diabetes in genome-wide association studies in East-Asians (3) , we further examined glucose homeostasis in mice.On HFHSD, insulin tolerance test (ITT) showed signi cantly higher blood glucose levels with 76.5% reduction of the inverse area under curve (AUC) of glucose levels in Aldh2 KI mice, indicating increased insulin resistance (Fig. 2h).Intraperitoneal glucose tolerance (i.p.GTT) showed signi cantly higher blood glucose levels with AUC of glucose levels increased by 109% in Aldh2 KI mice (Fig. 2i).Oral glucose tolerance test (OGTT) also showed signi cantly higher blood glucose levels in Aldh2 KI mice; AUC of glucose levels during the test were increased by 111%in KI mice, indicating worsened glucose intolerance (Fig. 2j).Aldh2 KI mice displayed a compensatory increase in insulin secretion after oral glucose load (Fig. 2k).No differences in glucose homeostasis, measured by glucose and insulin tolerance tests, were found between Aldh2 KI and WT mice fed a chow diet (Supplementary Fig. 1).
Aldh2 KI mice had reduced mitochondrial fatty acid oxidation rate and lower respiratory transport chain activity in brown adipose tissue due to 4-HNE adduction In view of the reduced adaptive thermogenesis observed in the Aldh2 KI mice, we compared the expression levels of Ucp1, the major thermogenic protein in BAT and white fat including inguinal and perigonadal fat between the Aldh2 KI and WT mice by real-time quantitative PCR (RT-qPCR).There was no difference in Ucp1 expression in the BAT or white fat between the Aldh2 KI and WT mice fed on HFHSD (Fig. 3a).Immunoblots further con rmed that there was no difference in the level of Ucp1 in the BAT, the tissue where Ucp1 was normally expressed (Supplementary Fig. 2).
We next examined the expression of mitochondrial respiratory complex I to V component in BAT.These protein components are essential for the maintenance of the proton gradient in mitochondrial electron transport chain (ETC) and are required for UCP1-mediated thermogenesis.There was no difference in mitochondrial respiratory complex I to V protein components expression in the BAT of Aldh2 KI and WT mice (Supplementary Fig. 3).We also observed no difference in microscopic morphology of BAT, the major thermogenic organ (Supplementary Fig. 4).
Since the ETC protein expression studied showed no difference between Aldh2 KI and WT mice, it implied that the decreased energy expenditure and thermogenesis were likely due to impaired protein functions.
We found increased protein carbonylation (Fig. 3b) and 4-HNE-adducted mitochondrial proteins (Fig. 3c) in the BAT of the KI mice compared with WT mice.Using liquid chromatography tandem mass spectrometry (LC MS/MS) analysis, we identi ed 20 4-HNE-adducted BAT mitochondrial proteins in Aldh2 KI mice and 10 4-HNE adducted mitochondrial proteins in WT mice, with 8 proteins which are present in both Aldh2 KI and WT mice (Fig. 3d, 3e).
Three identi ed adducted sites were present in proteins involved in fatty acid oxidation (FAO) including 3ketoacyl-CoA thiolase and propionyl-CoA carboxylase.Eleven sites were present in proteins involved in the maintenance of mitochondrial electron transport chain (ETC) including NADPH dehydrogenase (complex I), succinate dehydrogenase (complex II), cyochrome b-1 complex (complex III), ATP synthase, and glycerol-3-phoshphate dehydrogenase.Three adducted sites were present in the MICOS complex, which are essential for the maintenance of inner mitochondrial membrane structure.Hence, we further evaluated the capacity of mitochondrial FAO and ETC to determine the effect of 4-HNE modi cation of these proteins.The ex vivo FAO rate of the whole BAT tissue isolated from Aldh2 KI mice was signi cantly reduced by 16% (P = 0.03) compared with WT mice (Fig. 3f).Consistent with the whole BAT tissue, the FAO rate of the cultured primary brown adipocytes isolated from the Aldh2 KI mice was also signi cantly reduced by 22% (P = 0.03, Fig. 3g).Furthermore, addition of 4-HNE decreased FAO rate of induced primary brown adipocytes isolated from WT mice in a dose-dependent manner; FAO rates were reduced by 36% (P < 0.01) and 60% at 5 and10 µM4-HNE, respectively (P < 0.01)(Fig.3h).In line with the LC-MS/MS nding, we found a signi cant reduction in the enzymatic activity of mitochondrial respiratory complexes I, II, and III, but not complex IV, in Aldh2KI mice compared with WT mice (Fig. 3i).Oxygen consumption rate was also signi cantly decreased in induced primary brown adipocytes isolated from Aldh2 KI mice compared with WT mice (Fig. 3j).Since FAO is the major energy source of thermogenesis and the mitochondrial ETC is essential for maintaining proton gradient required for thermogenesis, the increased protein carbonylations may explain the impaired thermogenesis of Aldh2 KI mice.
ALDH2 activator AD-9308/AD-5591 activated wild-type and mutant human ALDH2 enzymatic activity AD-9308 is a valine ester prodrug of a potent and selective small molecule ALDH2 activator AD-5591.When administered in vivo, AD-9308 is rapidly converted to AD-5591 by esterase hydrolysis (Fig. 4a).In vitro, AD-5591 treatment increases the catalytic activity of recombinant wild-type and mutant human ALDH2 (Fig. 4b).In vivo, AD-9308 administration showed a good pharmacokinetic pro le in mice when administered orally or intravenously with good bioavailability in rodents and dogs (Supplementary Table 2).
Yet, in our modeling analyses, AD-5591 forms one additional hydrogen bond with Ala461(Fig.4e), which may explain its higher a nity to ALDH2 than for Alda-1.
ALDH2 activator AD-9308/AD-5591 lowered 4-HNE and attenuated diet-induced obesity, fatty liver, insulin resistance, and glucose intolerance in both WT and Aldh2 KI mice In view of the marked reduction in Aldh2 expression in diet-induced obese mice (Supplementary Fig. 6), we next explored whether Aldh2 activation can rescue these obesity-associated phenotypes.From the age of 10 weeks when HFHSD was started, Aldh2 KI and WT mice were treated with vehicle, 20 mg/kg/day or 60 mg/kg/day of AD-9308 by oral gavage for 20 weeks (Fig. 5a).As shown in Fig. 5B and 5C, AD-9308 treatment effectively reduced the diet-induced weight gain in both Aldh2 KI and WT mice dosedependently.Weight of perigonadal fat, inguinal fat, omental fat, and liver decreased with AD-9308 treatment dose-dependently (Fig. 5d-5 h).AD-9308 treatment reduced the extent of hepatic steatosis (Fig. 5i, 5j) and the levels of hepatic triglycerides contents in both Aldh2 KI and WT mice in a dosedependent manner (Fig. 5k).
Pathological examination reveled no abnormalities in liver or kidney in both Aldh2 WT and KI mice treated with AD-9308 for 20 weeks (Supplementary Table 3).Serum alanine aminotransferase (ALT) and creatinine levels were also not different between groups after AD-9308 treatment for 20 weeks (Supplementary Fig. 7).

Discussion
The prevalence of obesity and diabetes mellitus has surged in the past decades and is predicted to continue to rise, especially in East and South Asia 1 .This trend is largely caused by high-calorie diet enriched in fat and sugar, sedentary lifestyle, and their interaction with genetic predisposition 1, 2 .Genome-wide association studies have con rmed many genetic loci associated type 2 diabetes and obesity.Speci cally, several genetic loci are East Asian-speci c. Genetic variants in or near the CDKAL1, KLF9, GP2, ALDH2, and ITIH4 genes are associated with obesity 4,32,33 and genetic variants in or near the GDAP1, PTF1A, SIX3, ALDH2, and PAX4 genes 3,34 are speci cally associated with type 2 diabetes in East Asians.Among them, the ALDH2 Glu504Lys mutation that affects 560 million East Asians or nearly 8% of global population is associated with body mass index, type 2 diabetes, and serum lipids in large metaanalyses of genome-wide association studies.
We demonstrated that Aldh2 knock-in mice carrying the East Asian-speci c Glu504Lys mutation were more prone to develop diet-induced obesity, fatty liver, insulin resistance and glucose intolerance than WT mice on HFHSD.Importantly, the ALDH2 activator AD-9308 increased both the catalytic activity of WT and mutant enzyme, reduce serum 4-HNE levels, and effectively alleviated diet-induced obesity, fatty liver, insulin resistance, and glucose intolerance in both Aldh2 KI and WT mice in a dose-dependent manner.
BAT is a highly specialized organ enriched in Ucp1 for adaptive thermogenesis.Although Aldh2 KI mice exhibited impaired thermogenesis, unexpectedly, they did not show reduced Ucp1 expression.Instead, we found that several key mitochondrial proteins involved in mitochondrial FAO and ETC were modi ed by 4-HNE adduction.Consistently, previous studies have shown that 4-HNE is mainly generated from oxidation of mitochondrial membranes, with 30% of 4-HNE-adducted proteins located within mitochondria 35,36 .Our data further support that the 4-HNE adduction to mitochondrial proteins can be an underlying mechanism of ALDH22 inactivation that leads to the reduction of mitochondrial FAO and ETC function.
Mitochondrial FAO and ETC are required for the maintenance of the proton gradient in the intermembranous space, which is essential for Ucp1-mediated adaptive thermogenesis.In our study, we found that the thermogenic capacity of Aldh2 KI mice was reduced.Fatty acids serve as the main fuel suppliers for thermogenesis 37 .It has been estimated that intracellular fatty acids in the BAT contribute 74-84% of the fuel for thermogenesis upon cold challenge 37 .Cpt1 is the rate-limiting enzyme for the translocation of fatty acids into mitochondria for β-oxidation.Cpt1b +/− mice developed fatal hypothermia following cold challenge 38 .Adipose-speci c Cpt2-knockout mice presented a hypothermic phenotype when exposed to cold 39 .Mice de cient in fatty acid β-oxidation enzymes, including very-long-chain acyl-CoA dehydrogenase (VLCAD), long-chain acyl CoA dehydrogenase (LCAD), and short-chain acyl CoA dehydrogenase (SCAD) also displayed cold intolerance [40][41][42] .These data indicate that mitochondrial FAO is critical for adaptive thermogenesis.Furthermore, BAT-speci c Lkb1-knockout mice, which have reduced expression of ETC complex proteins, also developed impaired thermogenesis 43 , indicating that the integrity of mitochondrial ETC machinery is essential for adaptive thermogenesis.These data strongly support our ndings that 4-HNE adduction to mitochondrial proteins involved in mitochondrial FAO and ETC could lead to impaired adaptive thermogenesis.
ALDH2 metabolizes bioactive toxic aldehydes by oxidation.In addition to ALDH2, one of the major pathways to detoxify 4-HNE is mediated through glutathione transferases (GSTs) by conjugation to glutathione.Gsta4 is one of the isoforms of GST with the highest conjugation activity for 4-HNE.
Consistent with the present nding, disruption of Gsta4 in mice increased 4-HNE levels and caused obesity in mice 44 .Disruption of the Gst-10 gene, which causes a 50% increase in 4-HNE adducts in C. elegans also resulted in fat accumulation and direct treatment with 4-HNE increases lipid storage in C. elegans 45 .Conversely, over-expression of Gst-10 led to 4-HNE reduction and a lean phenotype 45 .Glutathione peroxidase 4 (Gpx4), which resides in the inner mitochondrial membrane, is a scavenger that reduces lipid peroxides.De ciency of Gpx4 in mice increased the number of 4-HNE adducts and exacerbated glucose intolerance, dyslipidemia, and fatty liver 46 .Collectively, these data indicate that increased aldehydic load due to excessive 4-HNE cause metabolic disorders.Indeed, overexpression of ALDH2 has been shown to decrease both the heart and liver weight and mitochondrial injury in mice fed on a high-fat diet 47 .Interestingly, diet-induced obesity was not observed in these ALDH2 over-expressing mice 47 .
Our study also showed that the Aldh2 KI mice had more severe hepatic steatosis than the WT mice when fed on HFHSD, which can be reversed by AD-9308 treatment.In line with the present results, Alda-1, a prototype of the ALDH2 activator has been shown to alleviate nonalcoholic hepatic steatosis in apolipoprotein E-knockout mice 48 and reversed alcohol-induced hepatic steatosis in animals 49 , supporting activation of ALDH2 also prevent both alcohol and non-alcoholic hepatic steatosis.
Our study has several limitations.First, obesity and related metabolic phenotypes in the Aldh2 KI mice were observed only when fed on HFHSD and not when placed on chow diet, the metabolic phenotypes were not different between the Aldh2 KI and WT mice.Therefore, our ndings may not be generalized to the entire population.The HFHSD for mice consists of 58% calories from fat and 12% calories from sucrose; while the chow diet is composed of 13% calories from fat and 3% calorie from sucrose.According to the Nutrition and Health Surveys in Taiwan (NAHSIT) in 2008, the average calorie intake from fat and sucrose is 33% and 8% in adults in Taiwan 50 .In the National Health and Nutrition Examination Survey of U.S adults., the average calorie intake from fat and sucrose is 35% and 14% in 2012 51,52 .The sucrose content is comparable between HFHSD for mice and modern human diet but the fat content of HFHSD is higher than human diet.Even if this concern is relevant, the ndings of this study are still substantial given the large number of East Asians (560 million people) carrying this inactivating mutation (Glu504Lys) in the ALDH2 gene.Second, our experiments did not include heterozygous knock-in mice.Third, although the metabolic disorders were normalized in mice treated with AD-9308 for 5 months without pathological changes in the liver and kidney, long-term safety should be formally determined with GLP standard.Last, humans and other mammals have 19 different aldehyde dehydrogenases (ALDH) and at least six are found in the mitochondria 30 .Although AD-9308 does not activate ALDH1A1, ALDH3A1, ALDH4A1, ALDH5A1 and ALDH7A1 members of the ALDH family (data not shown), it is possible that AD-9308 may still have non-speci c effect on other ALDHs.
Our study has important clinical implications.We showed that reduced activity of the mitochondrial enzyme, ALDH2, exacerbates obesity-associated pathologies.These pathologies correlate with increased aldehydic load and inactivation of critical mitochondrial proteins involved in FAO and ETC.Signi cantly, we showed that treatment with an activator of ALDH2, such as AD-9308 prevented these pathologies.These data provide strong evidence for a critical role of toxic aldehydes accumulation and defective ALDH2 activity in the pathogenesis of obesity, diabetes, and fatty liver disease.Our study provide a new strategy by targeting ALDH2 for the treatment of obesity-associated metabolic disorders which is rising rapidly in human populations, particularly in the East and South Asia.

Animal experiments, administration of diets and drug treatment
All animal experiments were performed according to institutional ethical guidelines and were approved by the Institutional Animal Care and Use Committee (IACUC), which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC).The mice were housed at 22-24 °C with light: dark cycles of 12:12 hours.The mice were fed either a HFHSD (cat.no.D12331, Research Diets) which provided 58% calorie from fat and 12.5% calorie from sucrose or chow diet (cat no.5001, Lab Diet).For the animal experiments, AD-9308 was dissolved in water and was delivered daily to mice by oral gavage.Only male mice were used in this study.

Generation of Aldh2*2/*2knock-in mice
Aldh2 KI mice carrying the human ALDH2 Glu504Lys mutation were generated by introducing the Glu504Lys mutation within the mouse gene 29 .Both Aldh2 WT controls and Aldh2 KI mice were littermates from mated heterozygous mice.
Energy expenditure, food intake and physical activity Metabolic measurements (food and water intake, locomotor activity, VO 2 consumption and VCO 2 production) were obtained using the Promethion metabolic phenotyping system (Sable Systems).Monitoring was performed for 5 days after mice have been acclimatized to the cages for 2 days.

Hepatic triglycerides content measurement
Approximately 80 mg of liver tissue was homogenized in 1800 µl of chloroform/methanol (2:1).Then, 360 µl of H 2 O was added.The homogenates were centrifuged and the lower 200-µl layer was added with 100 µl of chloroform with 4% Triton X-100 and then dried in a chemical hood.The dried pellet was redissolved with 200 µl of H 2 O for determination of triglyceride concentrations with Wako TG LabAssay kit (cat.no.290-63701, Wako) Cold tolerance test and diet-induced thermogenesis test For the cold tolerance test, 24-week-old mice with matched average body weight from the two groups were placed individually on HFHSD in a 4°C chamber.The rectal temperature of the mice was measured after 0, 1, 2, 3, 4, 5, 6, 12, and 18 hours.For the diet-induced thermogenesis test, 24-week-old mice were fasted overnight for 18 hours.Then, their rectal temperature was measured at 0, 15, 30, 60, 120, 180, and 240 min after HFHSD refeeding.

Real-time quantitative PCR (RT-qPCR)
RT-qPCR was performed using SYBR green reagent (cat.no.11203ES08, YEASEN).The primer sequences for mouse Ucp1 and Ppia is list in Table S1.RT-qPCR reactions were performed using an ABI 7900HT FAST (Applied Biosystems).All qPCR reactions were run in duplicate for each sample.

Isolation of mitochondria from brown adipose tissue
BAT was removed from Aldh2 IWT and KI mice fed a HFHSD for 5 weeks.Isolation of mitochondria was performed according to published protocols 53 .Brie y, BAT was homogenized in 10% w/v of ice-cold mitochondrial isolation buffer by a Dounce homogenizer.The pellets were separated by centrifugation, and then resuspended in SET buffer (0.25 M sucrose, 1 mM EDTA, and 10 mM Tris-HCl, pH 7.4) for the fatty acid oxidation assay and LC-MS/MS analysis or in BES-sucrose buffer (0.25 M sucrose, 0.1 M KCl, and 20 mM BSE, pH 7.2) for the mitochondrial complex activity assay.

Fatty acid oxidation (FAO) assay
FAO measurements were performed using labeled 3 H-palmitic acid and 3 H 2 O production was assessed as previously described 54 .For FAO rate of differentiated primary cells, the capture of 3 H 2 O was measured after a 2-hour incubation with 5 mM palmitate/BSA buffer including 0.5 µCi [ 3 H]-palmitate (cat.no.PK-NET043001MC, PerkinElmer) in the presence of 1 mM carnitine.For FAO rate of whole BAT, the isolated mitochondria were placed in a 24-well plate and added incubated with reaction buffer (100 mM sucrose, 10 mM Tris-HCl, 5 mM KH 2 PO 4 , 0.2 mM EDTA, 80 mM KCl, 1 mM MgCl 2 , 2 mM L-carnitine, 0.1 mM malate, 0.05 mM coenzyme A, 2 mM ATP, 1 mM dithiothreitol, and 7% BSA/5 mM palmitate/0.01µCi/µl [ 3 H]-palmitate) at 37°C for 60 min. 3H 2 O was isolated by oil-water separation with chloroform, methanol and KCl/HCl.The average counts per minute (CPM) were measured using a liquid scintillation counter.

Biotin hydrazide staining for carbonylated protein
For detection of carbonylated protein, samples were chemically reduced by NaBH 4 and then incubated with 0.5 mM EZ-link biotin hydrazide (ca.no.21339, Pierce) for 1 hour.After coupling, the samples were separated by 10% SDS-PAGE gel and transferred to PVDF membrane.The membrane was blocked with 10% skim milk in PBS containing 0.05% Tween-20 (PBST) at 4°C overnight and then incubated with Streptavidin-HRP (cat no.890803, BD Biosciences) for 1 hour at room temperature.Chemiluminescence signals were developed with HRP substrate (cat.no.WBLUR0500, Millipore).
Plasmid construction, expression and puri cation of human ALDH2 in Escherichia coli E. coli BL21 (DE3) was transformed with pTrcHi-WT ALDH2 and pTrcHi -KI ALDH2 29 .The transformants were cultured and then induced to express recombinant proteins using 0.4 mM IPTG for 16-18 hours at 25 °C.The cells were harvested by centrifugation and broken in lysis buffer by sonication on ice.The recombinant ALDH2 protein was puri ed with NI-NTA resin (cat.no.88222, ThermoFisher) following the user manual.

ALDH2 enzymatic activity
ALDH2 activity was measured by monitoring the reduction rate of NAD + /min at 340 nm and 25 °C.Enzyme activity was assayed in 100 µl of reaction mixtures containing 50 mM Na 4 P 2 O 7 (pH 9.5), 0.01% BSA, 10 mM NAD + , 50 µM acetaldehyde and recombinant ALDH2 with or without 100 µM AD-5591 or Alda-1.One enzymatic activity unit was de ned as the reduction of 1 µmol NAD + per min by 1 mg of human ALDH2 protein.

Molecular docking
To visualize the interaction between the ALDH2 with activators Alda-1 and AD-5591, the X-ray structure of human ALDH2 (PDB ID: 1NZX) and NAD + were used.Ligand energy was minimized using the PyRx program before docking 55 .Three-dimensional models are generated by using the PyMOL program 56 and the 2D protein-ligand interaction diagrams were generated by the LigPlot + program 57 .

LC − MS/MS analysis of 4-HNE-modi ed proteins
The protein samples were prepared with in-solution trypsin digestion.The peptides were desalted with C18 Zip Tip (Millipore, USA) and dry by vacuum centrifugation.The desalted and dried peptides were resuspended in 0.1% formic acid.LC-MS/MS analysis was performed on a NanoACQUITY UPLC system (Waters, USA) coupled to a highresolution mass spectrometer (QE HF-X, Thermo Fisher Scienti c).The peptides were injected into a trap column (Symmetry C18, 2 cm × 75 µm i.d.) and then separated in a 25 cm × 75 µm i.d.BEH130 C18 column (Waters, USA) on a gradient from 0-85% buffer B (buffer A, 0.1% FA H 2 O; buffer B, 0.1% formic acid in acetonitrile).The mass spectrometer was operated in data-dependent mode with the following acquisition cycle: an MS scan (m/z 350-1600) recorded at resolution R = 60,000 and MS/MS scans recorded at resolution R = 15,000, which were acquired by HCD fragmentation with collision energy of 28.
MS/MS spectra were searched with the Mascot engine (v2.6,Matrix Science) against the UniProtKB mouse protein database using the following parameters: precursor peptide mass tolerance of 20 ppm and an MS/MS fragment tolerance of 0.02 Da with a maximum of two missed cleavage sites.The following modi cations were made to the peptides: static carbamidomethylation on cysteine, variable oxidation on methionine, variable deamidation of asparagine or glutamine, and various 4-HNE modi cations on cysteine, histidine and lysine.The cut-off threshold for acquiring signi cant peptide-tospectrum matches was P < 0.05.

Mitochondrial respiratory chain complex activity assay
The mitochondrial complex spectrophotometric assays were carried out using published protocols 58,59 .Complex I and complex II activities were measured spectrophotometrically by examining the decrease in absorbance due to the reduction of 2,6-dichlorophenolindophenol (DCPIP) at 600 nm.Activity was expressed in nanomoles of DCPIP reduction/min/mg protein (E = 19.1 mmol − 1 •cm − 1 ).Complex IIIspeci c activity and complex IV-speci c activity were measured by monitoring the reduction of oxidized cytochrome C(III) and oxidation of reduced cytochrome C(II) at 550 nm, respectively.The activity is expressed as a nanomole of reduced cytochrome of oxidized cytochrome C /min/mg protein (E = 18.5 mmol − 1 •cm − 1 ).

Oxygen consumption rate (OCR)
Stromal vascular fraction was isolated from BAT of 4-week-old Aldh2 WT and KI mice and seeded to Seahorse XF24 v7 cell culture plates (Agilent) at approximate density of 20,000 cells per well and was then differentiated to primary brown adipocyte as described above.On day7 of differentiation, cells were washed once with Seahorse medium and maintained in Seahorse medium in CO 2 -less incubators.Basal OCR were determined by three measurements, followed by three measurements after each injection drug: oligomycin (2 µM), FCCP (2 µM), Rotenone/Antimycin A (0.5 µM) using Seahorse XFe24 Analyzer (Agilent).OCR measurements of each stage were normalized to the protein concentration.Fatty acid oxidation (FAO) rate of the whole BAT tissue isolated from the Aldh2 WT and KI mice fed a HFHSD (n=24:20).(g) FAO of primary brown adipocytes isolated from the Aldh2 WT and KI mice fed a HFHSD (n=3 per group).(h) FAO rate of induced primary brown adipocytes isolated from the WT mice treated with 0, 5, and 10 μM 4-HNE (n=3 per group).(i) Mitochondrial electron transfer chain (ETC) complex I-IV enzymatic activity isolated from the BAT of Aldh2 WT and KI mice on HFHSD (n=4:4) (j) Oxygen consumption rate (OCR) of induced primary brown adipocytes isolated from Aldh2 WT and KI mice fed on HFHSD.Cells were treated with oligomycin, FCCP, and Rotenone/Antimycin A. OCR measurements were normalized to the protein concentration.(n=3 per group) *P<0.05,**P< 0.01.

Figure 1 Body
Figure 1