Regulation of human class I alcohol dehydrogenases by bile acids.

Class I alcohol dehydrogenases (ADH1s) are the rate-limiting enzymes for ethanol and vitamin A (retinol) metabolism in the liver. Because previous studies have shown that human ADH1 enzymes may participate in bile acid metabolism, we investigated whether the bile acid-activated nuclear receptor farnesoid X receptor (FXR) regulates ADH1 genes. In human hepatocytes, both the endogenous FXR ligand chenodeoxycholic acid and synthetic FXR-specific agonist GW4064 increased ADH1 mRNA, protein, and activity. Moreover, overexpression of a constitutively active form of FXR induced ADH1A and ADH1B expression, whereas silencing of FXR abolished the effects of FXR agonists on ADH1 expression and activity. Transient transfection studies and electrophoretic mobility shift assays revealed functional FXR response elements in the ADH1A and ADH1B proximal promoters, thus indicating that both genes are direct targets of FXR. These findings provide the first evidence for direct connection of bile acid signaling and alcohol metabolism.

metabolic pathways including vitamin A (retinol) oxidation, which is the rate-limiting step in the conversion of retinol to retinoic acid ( 17 ) and, interestingly, bile acid metabolism. Thus, ADH1B has been shown to oxidize the bile alcohol 5 ␤ -cholestane-3 ␣ ,7 ␣ ,12 ␣ ,26-tetrol to the 3 ␣ ,7 ␣ ,12 ␣trihydroxy-5 ␤ -cholestanoic acid, an intermediate in bile acid synthesis ( 18 ), whereas ADH1C was identifi ed as the sole bile acid 3 ␤ -hydroxysteroid dehydrogenase present in human liver cytosol that promotes epimerization of iso bile acids to 3 ␣ -hydroxy bile acids, which are subsequently secreted by hepatocytes into bile ( 19,20 ). In this study, we investigated the effect of both the endogenous bile acid chenodeoxycholic acid (CDCA) and the synthetic FXR agonist GW4064 on ADH1 gene expression. The data presented nucleotide repeat separated by one nucleotide (IR1) and, in the presence of specifi c agonists, activates transcription of target genes involved in bile acid, cholesterol, lipoprotein, and glucose metabolism ( 10 ). FXR has been shown to have relevance in the attenuation of clinically important conditions such as gallstone disease ( 11,12 ), cholestasis ( 13 ), and fatty liver disease ( 14 ). Most of these hepatoprotective functions of FXR can be attributed to the induction of genes involved in bile acid detoxifi cation and xenobiotic metabolism, including phase I oxidation enzyme CYP3A4 ( 15 ) and a number of phase II conjugation enzymes ( 16 ) and phase III effl ux transporters ( 10 ).

Cell transfection and reporter assays
Huh7 cells were transiently transfected as previously described ( 23 ). After 6 h, cells were treated for 24 h with the vehicle (DMSO), 100 M CDCA, or 1 M GW4064 as described above. Luciferase activities were assayed as previously described ( 23 ). All transfections were performed in triplicate, and similar results were obtained in at least three independent experiments.

Western blot analysis
Whole protein cell extracts were obtained from Huh7 cells in Nonidet P-40 lysis buffer (150 mM NaCl, 50 mM Tris-HCl, 1% Nonidet P-40) supplemented with a mixture of protease inhibitors (Sigma-Aldrich) and 0.1 mM phenylmethylsulfonyl fl uoride (Sigma-Aldrich). Fifty micrograms of protein were resolved in 10% polyacrylamide gels, transferred onto Immobilon-P membranes (Millipore, Bedford, MA) and probed with antibodies: anti-pan herein identify the nuclear receptor FXR as a regulator of human ADH1 genes, thus linking bile acid signaling and alcohol metabolism.

Animal experiments
Experimental protocols with mice were performed with the approval of the animal ethics committee of the University of Barcelona (Spain). Male 10-week-old C57BL/6 mice were injected intraperitoneally with either vehicle (corn oil 5% DMSO) or GW4064 (GlaxoSmithKline, Research Triangle Park , NC) dissolved in vehicle (10 mg/ml) at a dose of 50 mg/kg. After 8 h, mice were sacrifi ced, and livers were excised, snap-frozen in liquid nitrogen, and stored at Ϫ 80°C until analysis.

Cell culture and treatment conditions
Human hepatoma HepG2 and Huh7 cells and rat hepatoma FAO cells were cultured in DMEM supplemented with 10% FBS. Primary mouse hepatocytes were isolated by the collagenase method as previously described ( 21 ) and cultured for 6 h before treatments. Human primary hepatocytes were obtained commercially (Ready Heps TM Fresh Hepatocytes, Lonza, Basel, Switzerland) and maintained in Hepatocyte Complete Medium (HCM TM bulletkit, Lonza). Cells were treated with ligands in the same medium supplemented with 10% charcoal stripped FBS (Biological Industries).

Fig. 2.
Constitutive active FXR induces ADH1A and ADH1B expression. Huh7 cells were infected with increasing multiplicity of infection (MOI) of adenovirus expressing either VP16 (AdVP16) or a constitutively active chimera of VP16 and FXR (AdVP16FXR) for 48 h. Specifi c mRNA levels of ADH1A , ADH1B , and ADH1C were determined by real-time quantitative PCR, normalized to 18S content, and expressed relative to controls, which were set as 1 (mean ± SEM). * P < 0.05; ** P < 0.01; *** P < 0.001 versus 25 MOI AdVP16. The results are representative of two independent experiments from triplicate dishes.

Alcohol dehydrogenase activity assay
ADH enzymatic activity was determined using Alcohol Dehydrogenase Assay kit (catalog #K787, BioVision, Milpitas, CA) following the manufacturer's instructions, and normalized to protein content in each sample.

Statistical analysis
Data are expressed as mean ± SEM as determined by analysis of multiple independent samples, as indicated in fi gure legends. Signifi cant differences were assessed using a two-tailed Student's t -test. Values of P < 0.05 were considered to be signifi cant.

In silico analysis of FXR response elements
The analysis of genomic sequences for the identifi cation of putative FXR response elements (FXREs) was performed using the NUBIScan computer algorithm .

In vitro transcription/translation and electrophoretic mobility shift assay
Double-stranded oligonucleotides corresponding to the sequences spanning nt Ϫ 108 to Ϫ 82 of ADHA1 , nt Ϫ 131 to Ϫ 106 of ADH1B , or nt Ϫ 133 to Ϫ 107 of ADH1C were radiolabeled and used in an electrophoretic mobility shift assay (EMSA) as previously described ( 23 ). A probe containing the FXRE of the human ileal-bile acid-binding protein ( I-BABP ) gene promoter was used as a control. For competition experiments, increasing fold molar excess of unlabeled probes was included during a 15 min preincubation on ice. The probe mut ADH1A IR1 corresponds to the sequence spanning nt Ϫ 108 to Ϫ 82 of ADHA1 harboring mu- presence of vehicle (DMSO) or FXR agonists. Again, no signifi cant induction of ADH1C was observed in response to FXR ligands, whereas ADH1A and ADH1B mRNA levels were strongly increased ( Fig. 1B ). In addition, FXR activation failed to induce cytochrome P450 2E1 ( CYP2E1 ), cytosolic aldehyde dehydrogenase ( ALDH1A1 ), or mitochondrial aldehyde dehydrogenase ( ALDH2 ) mRNA levels. Next, Western blot analyses performed on cell lysates from Huh7 treated with FXR agonists revealed that the quantity of ADH1 protein was robustly increased compared with cells receiving vehicle ( Fig. 1C ).

Overexpression of constitutively active FXR increases ADH1A and ADH1B mRNA levels
To further confi rm the role of FXR in the regulation of the ADH1 gene cluster, we performed gain-of-function studies by ectopically expressing a constitutively active form of FXR (VP16FXR) ( 22 ) in Huh7 cells. The expression of ADH1A and ADH1B , but not ADH1C , was induced by FXR overexpression ( Fig. 2 ). Collectively, these data suggest that FXR induces specifi cally the expression of ADH1A and ADH1B isoenzymes.

The inductions of ADH1A and ADH1B by CDCA and GW4064 require FXR expression
Because bile acids may exert their signaling actions through FXR-independent pathways ( 9, 25 ), we silenced FXR by siRNA to ascertain whether the inductions observed upon treatment with CDCA were dependent on FXR. Huh7 cells were transfected with nontargeting siRNA or siRNA complexes directed against FXR (siFXR) before treatments . siFXR-mediated knockdown of endogenous FXR levels ( Fig. 3A ) almost completely eliminated the induction of ADH1A and ADH1B by bile acid CDCA ( Fig.  3B ). As a control, CDCA-dependent induction of small heterodimer partner ( SHP ), a well-known FXR target gene ( 10 ), was similarly attenuated in siFXR-transfected cells. Likewise, as depicted in Fig. 3C , we also confi rmed by FXR knockdown the requirement of FXR in the responses of these genes to GW4064 to rule out any effect of this synthetic ligand that could relate to weak agonistic effects on other receptors ( 26 ).

FXR regulation of ADH1 does not occur in rodents
To examine whether FXR regulation of ADH1 also occurs in rodents, C57BL/6 mice were treated with either vehicle or FXR agonist GW4064 and analyzed for hepatic transcript levels. Unexpectedly, no signifi cant change of Adh1 mRNA levels was observed following activation of FXR, in spite of the marked induction of the known FXR target Shp ( Fig. 4A ). To exclude the possibility that the dissimilar fi ndings in human hepatocytes and in mouse liver were attributable to methodological differences, mouse primary hepatocytes were incubated in the presence of vehicle or GW4064. Again, Shp was highly induced after FXR activation, whereas no signifi cant change was detected in Adh1 transcript levels ( Fig. 4B ). Likewise, analyses performed on rat hepatoma FAO cells treated with CDCA, GW4064, or vehicle showed induction of Shp in response and because ADH1 enzymes are reported to take part in bile acid metabolism (18)(19)(20), we evaluated whether this nuclear receptor modulates ADH1 gene expression. Human primary hepatocytes were exposed for 24 h to CDCA, a natural bile acid agonist of FXR, or GW4064, a synthetic and specifi c FXR agonist, and analyzed for the expression of the three ADH1 genes. As illustrated in Fig. 1A , both FXR ligands signifi cantly increased ADH1A and ADH1B mRNA levels. As expected, the well-characterized FXR target gene PLTP ( 24 ) was also induced when primary human hepatocytes were treated with FXR agonists. In contrast, ADH1C mRNA was not induced after activation of FXR ( Fig. 1A ). To confi rm these results in human hepatoma cell lines, HepG2 cells were also incubated in the Specifi c Shp and Adh1 mRNA levels normalized to 18S content are expressed relative to controls, which were set as 1 (mean ± SEM). ** P < 0.01; *** P < 0.001 versus untreated controls.

Fig. 5.
Characterization of a functional IR1 in the proximal promoter of ADH1A and ADH1B . A: Schematic representation of human ADH1 gene cluster and localization of the IR1 elements identifi ed by NUBIScan in each proximal promoter. Alignment of the three IR1 and the FXRE consensus is shown below. The RGGTCA half-sites are indicated by arrows. B: The promoters of ADH1A and ADH1B , but not of ADH1C , respond to activated FXR. Huh7 cells were transfected with a plasmid containing luciferase reporter constructs driven by ‫ف‬ 2-2.9 kb fragments corresponding to ADH1 gene promoters (pGL3-ADH1A , pGL3-ADH1B , and pGL3-ADH1C , respectively), or the empty pGL3-basic vector (bv) as negative control, along with a plasmid expressing FXR, or the empty expression vector pSG5 as control, and then treated for 24 h with vehicle (Control) or 1 M GW4064 and luciferase activities were measured. C: IR1 elements in ADH1A and ADH1B , but not ADH1C , confer FXR responsiveness to a heterologous promoter. Experiments were performed as in (B) with reporter constructs containing four copies of the IR1 site identifi ed in the proximal promoter of ADH1 genes cloned in front of a heterologous thymidine kinase (TK) promoter-driven luciferase gene. pGL3-TK reporter vector was used as negative control. D: Disruption of ADH1A and ADH1B IR1 elements by site-directed mutagenesis abrogates the response to FXR. Experiments were performed as in (B) with the indicated reporter constructs containing wild-type or IR1 mutated sequence. E: Conversion of ADH1C IR1 to ADH1A IR1 element by site-directed mutagenesis confers FXR responsiveness to ADH1C promoter. Experiments were promoter. In contrast to the wild type, the activity of the construct bearing the mutated IR1 (pGL3-ADH1C mut 1A ) was markedly increased by FXR ( Fig. 5E ). Collectively, these findings indicate that ADH1A and ADH1B are direct target genes of FXR and that the IR1 elements located at their respective proximal promoters are essential for the response to FXR.

Binding analysis of FXR to the IR1s in the ADH1A and ADH1B promoters
To determine whether FXR-RXR heterodimers directly bind the IR1s identifi ed in the proximal promoters of ADH1 genes, EMSAs were performed. Incubation of in vitro translated FXR and RXR together with radiolabeled double-stranded oligonucleotides containing the IR1 located at the proximal promoter of ADH1A or ADH1B resulted in a specifi c retarded complex ( Fig. 6A , lanes 4 and  6, respectively). In contrast, the probe containing the IR1 in the ADH1C promoter failed to form a specifi c DNA protein complex ( Fig. 6A , lane 8). The well-characterized IR1 from I-BABP served as a positive control ( Fig. 6A , lane 2) ( 30 ). As expected, more detailed gel-shift analyses demonstrated that FXR indeed binds as a heterodimer with RXR to these IR1s because the appearance of a robust retarded complex with ADH1A IR1 ( Fig. 6B , lane 4) or ADH1B IR1 (data not shown) required the presence of both FXR and RXR proteins. The specifi city of these complexes was demonstrated by competition with increasing concentrations of either the unlabeled oligonucleotide dimers corresponding to the wild-type IR1 or I-BABP FXRE probes ( Fig. 6B ,  lanes 5-7 and lanes 11-13), whereas no disappearance of the retarded complexes was appreciated when a mutated IR1 was used as a competitor ( Fig. 6B , lanes 8-10).

Activation of FXR increases ADH1 enzymatic activity
Having shown that activation of FXR increases ADH1 mRNA and protein levels, we next investigated whether the activity of the enzyme was also increased. Exposure of Huh7 cells to CDCA or GW4064 resulted in an increase of alcohol dehydrogenase activity ( Fig. 7A ). To confi rm that such increase was dependent on FXR, we transfected Huh7 cells with FXR-specifi c or nontargeting siRNA prior to treatments with GW4064. As depicted in Fig. 7B , when FXR expression was silenced, the stimulatory effect of GW4064 on ADH activity was completely abolished. Thus, we conclude that FXR regulates not only ADH1 expression but also the overall cellular ADH activity.

DISCUSSION
The present study identifi es the human ADH1A and ADH1B genes as direct targets of the bile acid-activated nuclear receptor FXR. The evidence for this fi nding is the to FXR ligands, but no change in rat Adh1 transcript levels ( Fig. 4C ). Taken together, these data indicate that induction of ADH1 genes by activation of FXR is species specifi c because it was observed in human but not in rodentderived hepatocytes.

Identifi cation and functional characterization of putative FXREs in the promoters of ADH1 genes
In order to determine whether FXR directly controls ADH1 expression and also with the aim to understand the molecular mechanisms involved in the differences observed in the FXR-dependent regulation of the distinct human ADH1 genes, the promoters of ADH1A , ADH1B , and ADH1C were in silico analyzed for putative FXREs using the NUBIScan computer algorithm. Most known FXREs consist of an IR of the RGGTCA hexad with minor variants ( 27,28 ). Accordingly, an IR1 motif was identifi ed as a putative FXRE in the proximal promoter of each of the three genes. As illustrated in Fig. 5A , IR1 elements in ADH1A , ADH1B , and ADH1C promoters differ from 2, 3, and 4 nucleotides, respectively, from the consensus FXRE. The presence of a T at position 3 and a C at position 10 of the IR1 element are permissive (e.g., FXREs of FGF19 and ALAS1 , respectively) ( 10,29 ). In contrast, A into G conversion at the sixth position, as it occurs in the ADH1C element, is detrimental to receptor binding ( 27,28 ). To determine if the ADH1 proximal promoters were able to confer a response to FXR, we performed transient transfection assays in Huh7 cells with luciferase reporter constructs under the control of ‫ف‬ 2-2.9 kb fragments corresponding to ADH1 gene promoters in the presence or absence of a plasmid encoding FXR and the agonist GW4064. The data in Fig. 5B show that reporter activity of ADH1A and ADH1B promoter constructs was increased by activated FXR, whereas no signifi cant effects were observed on the ADH1C promoter construct or the promoterless pGL3 basic vector. These results are consistent with the higher similarity of ADH1A and ADH1B IR1 elements to the consensus FXRE. To further confi rm this observation, the IR1 element for each promoter was cloned upstream of the thymidine kinase promoter-driven luciferase reporter gene. As expected, transient transfection assays showed that the reporter constructs containing the IR sequences of ADH1A or ADH1B , but not of ADH1C , were strongly transactivated by FXR ( Fig. 5C ).
To assess the importance of the identifi ed IR1 elements in FXR-dependent activation, we introduced single nucleotide mutations to disrupt the IR1 in the context of ADH1 promoters to generate pGL3-ADH1A mut and pGL3-ADH1Bmut constructs, respectively. As depicted in Fig. 5D , these point mutations completely abrogated the response to FXR. We next investigated the effects of performing site-directed mutations to convert the ADH1C IR1 element into the ADH1A IR1 element in the context of the of ‫ف‬ 2.6 kb ADH1C performed as in (B) with the constructs containing wild-type ADH1C promoter or a modifi ed version containing the IR1 converted to ADH1A IR1 element. Results are expressed as -fold induction over control. *** P < 0.001 versus untreated controls. The results are representative of three independent experiments from triplicate dishes. following: a ) the expression of ADH1A and ADH1B is induced in HepG2 cells, Huh7 cells, and primary human hepatocytes in response to natural and synthetic FXR ligands; b ) overexpression of constitutively active FXR augments ADH1A and ADH1B mRNA levels; c ) IR1 elements identified at the proximal promoters of ADH1A and ADH1B are able to confer FXR response and bind FXR-RXR heterodimers; d ) alcohol dehydrogenase activity is increased upon FXR activation; and e ) silencing of FXR abolished the effects of FXR agonists on ADH1 expression and activity.
At the present time, the reason why FXR regulates ADH1 genes is not immediately obvious. Previous studies have shown that 3 ␤ -dehydrogenation of iso bile acids is catalyzed by ADH1C isoenzyme ( 19,20 ), yet we have failed to observe signifi cant changes in ADH1C expression in response to FXR agonists. Moreover, although ADH1B has been shown to catalyze a step of bile acid synthesis ( 18 ), it is counterintuitive to presume that this is the explanation for the FXR-dependent regulation of ADH1B given the inhibitory effects that FXR exerts on the biosynthesis of bile acids ( 10 ). In fact, more recent reports have concluded that ADH1 activity may be dispensable for bile acid biosynthesis, because the bile acid pool is unchanged when ADH1 is absent ( 31 ), and because mitochondrial CYP27 performs all steps in the formation of 3 ␣ ,7 ␣ ,12 ␣ -trihydroxy-5 ␤ -cholestanoic acid from the corresponding 3 ␣ ,7 ␣ ,12 ␣triol ( 32,33 ). Nevertheless, given the bile acid detoxifying role of FXR ( 16 ), we cannot exclude that the regulation of ADH1 genes by FXR responds to protective pathways whereby ADH1A and ADH1B isoenzymes metabolize yet unknown bile alcohols and/or bile acids to less toxic products. In agreement with this hypothesis, several bile alcohols and bile acids that are intermediates in the bile acid synthetic pathway have been identifi ed as highly effi cacious ligands for FXR ( 34 ). Also in this regard, it is interesting to  suggests a potential effect on ethanol metabolism that should be taken into account in future clinical trials with FXR modulators.
consider the sequential turn-on of ADH1 genes during liver development. Whereas ADH1A isoenzyme is found during early stages of fetal liver development and ADH1B appears at later fetal stages, ADH1C is only detected several months postnatally ( 4 ). Therefore, it is tempting to speculate that the isoenzyme-specifi c regulation by FXR that we observe may correspond to a mechanism of protection during fetal liver development. In this context, it is worth noting that fetal bile acid synthesis differs markedly from that of the adult, and "atypical" bile acids are found in human fetal bile. As an example, C-4 hydroxylated bile acids, which account for 5-15% of the total biliary bile acids of the fetus, are exclusive to early human development ( 35 ). Hence, it will be of considerable interest for future liver development studies to determine whether human ADH1 isoenzymes metabolize fetal specifi c bile acids.
Strikingly, the FXR-dependent induction of ADH1 genes is species specifi c, because rodent hepatic Adh1 mRNA levels were not increased after activation of FXR ( Fig. 4 ). Such a species-dependent regulation has also been observed for other human FXR target genes, including syndecan-1 ( 36 ), fi brinogen ( 37 ), ␣ A-crystallin ( 38 ), peroxisome proliferator-activated receptor ␣ ( 39 ), hepatic lipase ( 40 ), ALAS1 ( 29 ), fetuin-b ( 41 ), and PCSK9 ( 42 ). The reason for such expansion in the repertoire of FXR targets in humans, compared with rodents, remains obscure. Probably, it is a refl ection of species differences in bile acid composition as well as the novel roles that bile acids have acquired as signaling molecules in humans. Interestingly, a number of studies have focused on the evolutionary history of ADH1 in order to understand specifi c processes regarding primate adaptation to dietary alcohols ( 43 ). The three human ADH1 paralogs mainly originated from sequential duplications of an ancestral ADH1 gene after the divergence between rodents and primates. Several lines of evidence indicate that the fi rst split was between ADH1C and the gene that gave rise ADH1B and ADH1A ( 43 ). The current fi ndings that FXR induces ADH1A and ADH1B but not ADH1C correlates with this evolutionary mechanism that places ADH1C as the outgroup. Consequently, it is tempting to postulate that the FXR response appeared after the fi rst split and before the split between ADH1A and ADH1B .
Inasmuch as ADH1 enzymes catalyze the rate-limiting steps of retinol and ethanol metabolism in humans ( 3,17 ), the current data suggest that the activation of FXR is likely to enhance the metabolism of retinol and ethanol. On the other hand, we did not observe a signifi cant effect of FXR activators in the mRNA levels of CYP2E1 , an enzyme that also contributes to ethanol metabolism in chronic alcohol ingestion. In addition, incubation of human cells with FXR agonists failed to increase ALDH1A1 or ALDH2 mRNA levels. Therefore, additional studies will be required to defi ne whether treatments with FXR ligands might effectively modify ethanol metabolism in vivo.
In conclusion, our data reveal a direct stimulatory role for FXR on ADH1, which points to a possible involvement of ADH1 in bile alcohol and/or bile acid catabolism. Furthermore, the FXR-dependent activation of ADH1 also