Alpha-crystallin is a target gene of the farnesoid X-activated receptor in human livers.

Alpha-crystallins comprise 35% of soluble proteins in the ocular lens and possess chaperone-like functions. Furthermore, the alphaA subunit (alphaA-crystallin) of alpha crystallin is thought to be "lens-specific" as only very low levels of expression were detected in a few non-lenticular tissues. Here we report that human alphaA-crystallin is expressed in human livers and is regulated by farnesoid X-activated receptor (FXR) in response to FXR agonists. AlphaA-crystallin was identified in a microarray screen as one of the most highly induced genes after treatment of HepG2 cells with the synthetic FXR ligand GW4064. Northern blot and quantitative real-time PCR analyses confirmed that alphaA-crystallin expression was induced in HepG2-derived cell lines and human primary hepatocytes and hepatic stellate cells in response to either natural or synthetic FXR ligands. Transient transfection studies and electrophoretic mobility shift assays revealed a functional FXR response element located in intron 1 of the human alphaA-crystallin gene. Importantly, immunohistochemical staining of human liver sections showed increased alphaA-crystallin expression in cholangiocytes and hepatocytes. As a member of the small heat shock protein family possessing chaperone-like activity, alphaA-crystallin may be involved in protection of hepatocytes from the toxic effects of high concentrations of bile acids, as would occur in disease states such as cholestasis.

␣-Crystallins comprise 35% of soluble proteins in the ocular lens and possess chaperone-like functions. Furthermore, the ␣A subunit (␣A-crystallin) of ␣ crystallin is thought to be "lens-specific" as only very low levels of expression were detected in a few non-lenticular tissues. Here we report that human ␣A-crystallin is expressed in human livers and is regulated by farnesoid X-activated receptor (FXR) in response to FXR agonists. ␣A-Crystallin was identified in a microarray screen as one of the most highly induced genes after treatment of HepG2 cells with the synthetic FXR ligand GW4064. Northern blot and quantitative realtime PCR analyses confirmed that ␣A-crystallin expression was induced in HepG2-derived cell lines and human primary hepatocytes and hepatic stellate cells in response to either natural or synthetic FXR ligands. Transient transfection studies and electrophoretic mobility shift assays revealed a functional FXR response element located in intron 1 of the human ␣A-crystallin gene. Importantly, immunohistochemical staining of human liver sections showed increased ␣A-crystallin expression in cholangiocytes and hepatocytes. As a member of the small heat shock protein family possessing chaperone-like activity, ␣A-crystallin may be involved in protection of hepatocytes from the toxic effects of high concentrations of bile acids, as would occur in disease states such as cholestasis.
␣-Crystallins are members of the small heat shock protein family and constitute the major protein component of the vertebrate ocular lens. The 20-kDa proteins encoded by the two ␣-crystallin genes, ␣Aand ␣B-crystallin, multimerize with each other to form the Յ800-kDa native high molecular weight lenticular ␣-crystallins (for review, see Refs. [1][2][3][4]. ␣A-Crystallin expression is reported to be largely restricted to the lens, with some evidence of very low levels of expression in spleen, thymus, and retina (5,6). In contrast, non-ocular expression of ␣B-crystallin has been observed in a number of tissues includ-ing heart, skeletal muscle, brain, and lung (7,8).
Although the in vivo function and mechanism of action of ␣-crystallins are not well understood, in vitro studies suggest that they possess chaperone-like properties (9,10). The prevalent hypothesis is that ␣-crystallins function as molecular chaperones to sequester unfolded proteins in the lens to maintain the refractive power essential for normal lens function (2). Besides its chaperone activity, ␣-crystallins have also been shown to possess autokinase activity (11,12), but the significance of this activity is unknown. In vitro studies have also demonstrated that ␣-crystallins associate with cytoskeletal elements, consistent with a possible structural role for this protein in the lens (13)(14)(15). Mutations in both ␣Aand ␣B-crystallin have been linked to cataractogenesis in humans (16 -19). Consistent with this finding, ␣A-crystallin knock-out mice developed mild cataracts at as early as 7-weeks of age, which further progressed into severe lens opacification as a result of formation of inclusion bodies composed mainly of ␣B-crystallin (20). However, ␣B-crystallin knock-out mice surprisingly did not develop cataracts (21).
The farnesoid X-activated receptor (FXR, also NR1H4) 1 is a member of the nuclear hormone receptor superfamily (22,23). Nuclear hormone receptors are ligand-activated transcription factors that regulate transcription of target genes to control key physiological processes including reproduction, development, and metabolism. FXR expression is specific to liver, kidney, adrenal gland, and intestine, with low levels of expression in fat and the heart (24,25). Recent studies identified four FXR isoforms that are produced from the single human or murine gene as a result of use of alternative promoters and alternative RNA splicing (25,26). The four isoforms (FXR␣1-␣4) are expressed in different ratios in the above-mentioned tissues. More importantly, they have been shown to regulate transcription of some target genes in an isoform-specific manner (25,27).
One important hepatic function of FXR is the maintenance of bile acid homeostasis through coordinated regulation of hepatic bile acid transporters and catabolism of cholesterol into bile acids (33)(34)(35). In response to elevated levels of bile acids in the liver, FXR is activated and induces a number of ATP binding cassette (ABC) transporters, including the bile salt export pump (BSEP, ABCB11), multidrug resistant-associated protein 2 (MRP2, ABCC2), and the human multidrug resistant protein 3 (MDR3, murine mdr2, ABCB4) (36 -38). BSEP and MRP2 efflux bile salts from hepatocytes into the bile duct, whereas MDR3/mdr2 flips phospholipids across the bile canalicular membrane before their dissociation into the bile, thus aiding in the desorption and solubilization of bile salts and cholesterol in the bile (39 -41). In addition, activation of FXR leads to the repression of cholesterol 7␣-hydroxylase (CYP7A1), the ratelimiting enzyme in the bile acid synthesis pathway, via complex feedback inhibition mechanisms involving the small heterodimer partner (SHP) protein (42) and/or the human fibroblast growth factor 19 (43). Moreover, bile acids also repress their own biosynthesis through FXR-independent pathways (44).
Under normal conditions the regulatory events mentioned above are sufficient to protect the liver from accumulation of bile acids. Perturbation of this intricate regulatory cascade either as a result of certain disease states or genetic mutations leads to accumulation of bile acids in the liver and subsequent liver damage, a condition known as cholestasis (45,46). Recent studies have demonstrated that FXR activation may protect the liver from cholestasis-induced damage (47). Here we provide evidence that human ␣A-crystallin is expressed in the human livers and is a direct target gene of FXR. We propose that the induction of ␣A-crystallin expression in response to bile acid-activated FXR contributes to cellular defense against bile acid-induced hepatotoxicity.

EXPERIMENTAL PROCEDURES
Materials-The synthetic FXR-specific ligand GW4064 was a gift from Dr. Patrick Maloney (GlaxoSmithKline) (48). Mammalian expression vectors for rat FXR (pCMX-rFXR) and human RXR␣ (pCMX-hRXR␣) were gifts from Dr. Ron Evans (Salk Institute, La Jolla, CA), and that for human FXR␣1 was a gift from Dr. Bryan Goodwin (Glaxo-SmithKline). Primary cultures of human hepatocytes and hepatic stellate cells were obtained from ADMET Technologies (Durham, NC). Dexamethasone was from the Sigma. Insulin-transferrin-selenium and all other tissue culture reagents were from Invitrogen. Mammalian expression vectors for individual human FXR isoforms (pcDNA3.1-hFXR␣1, -␣2, -␣3, and -␣4) were generated using standard procedures. Polyclonal antibodies to recombinant bovine ␣A-crystallin, the amino-terminal peptide of ␣A-crystallin, and the carboxyl-terminal peptide of ␣B-crystallin were generous gifts from Dr. Joseph Horwitz (UCLA Jules Stein Eye Institute) (20,49). The sources of other reagents have been described elsewhere. Dimethyl sulfoxide (Me 2 SO) and chenodeoxycholic acid (CDCA) were purchased from Sigma.
Cell Culture and Stable Cell Lines-The generation and maintenance of stably infected HepG2-Neo and HepG2-FXR cells have been described (50). CV-1 cells were cultured in modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C and 5% CO 2 . Human hepatocytes were obtained as described (37).
Primary Culture of Human Hepatocytes-Primary human hepatocytes were cultured on Matrigel-coated 6-well plates at a density of 1.5 ϫ 10 6 cells/well. Culture media consisted of serum-free Williams' E medium supplemented with 100 nM dexamethasone, 100 units/ml penicillin G, 100 g/ml streptomycin, and 1ϫ insulin-transferrin-selenium. After overnight culture cells were treated with GW4064, added to the culture medium from a 1000ϫ stock in Me 2 SO. Control cultures received vehicle (0.1% Me 2 SO) alone. Twenty-four to forty-eight hours after treatment, total RNA was isolated using Trizol reagent (Invitrogen) according to the manufacturer's instructions. Real-time PCR was performed using a 7900HT sequence detection system from Applied Biosystems.
Primary Culture of Human Hepatic Stellate Cells-Primary human hepatic stellate cells were cultured on 6-well plates at a density of 1 ϫ 10 6 cells/well. Culture media consisted of Dulbecco's modified Eagle's medium supplemented with 15% fetal bovine serum, 2 mM glutamine, 100 units/ml penicillin G, and 100 g/ml streptomycin. After overnight culture, cells were either collected for RNA isolation or treated with 300 nM GW4064 or vehicle (0.1% Me 2 SO) for the indicated times. After treatment, total RNA was isolated using Trizol reagent, and real-time PCR was performed as described above.
RNA Isolation and Northern Blot Assay-Unless otherwise indicated, HepG2 cells were cultured in medium containing super-stripped fetal bovine serum for 24 h before the addition of vehicle (Me 2 SO) or ligand. After an additional 24 -48 h, total RNA was isolated. Northern blot assays were carried out as described previously (37). All membranes were hybridized to a control probe (18 S or 36B4) to correct for small differences in the amount of RNA loaded in each lane. The RNA levels were quantitated using a PhosphorImager (ImageQuant software; Amersham Biosciences).
Western Blot Analysis-HepG2-FXR cells were either untransfected or transfected with expression plasmids encoding FXR and RXR and treated with Me 2 SO or GW4064 for 24 h before protein isolation. Cells were lysed in buffer (50 mM Tris⅐Cl, pH 7.4, 25 mM KCl, 5 mM MgCl 2 , 1 mM EDTA) containing 1 mM phenylmethylsulfonyl fluoride and Complete EDTA-free protease inhibitor mixture (Roche Applied Science) by 3 freeze-thaw cycles. Protein concentrations of the lysates were determined with the Bradford-based assay (Bio-Rad) using bovine serum albumin as the standard. Fifty micrograms of total protein were loaded onto a 15% SDS-PAGE mini-gel followed by electrophoresis and transfer to a polyvinylidene difluoride membrane (Millipore, MA). Anti-␣Acrystallin polyclonal antibodies were used at 1:500 dilution followed with horseradish peroxidase-conjugated anti-rabbit IgG (Amersham Biosciences) at 1:10,000 dilution. Signal was detected using ECL Plus reagent (Amersham Biosciences).
Electrophoretic Mobility Shift Assays (EMSAs)-EMSAs were performed essentially as described (51,37). Human RXR␣ and human FXR␣1 were synthesized from CMX-hRXR␣ and pSG5-hFXR␣1 expression vectors, respectively, using the TNT T7-Coupled Reticulocyte System (Promega). Unprogrammed lysate was prepared using the empty pCMX vector. Binding reactions contained 10 mM HEPES, pH 7.6, 40 mM NaCl, 2.5 mM MgCl 2 , 0.5% Nonidet P-40, 10% glycerol, 0.5 mM dithiothreitol, 2 g of poly(dI-dC), 25 g of bovine serum albumin, and 1-4 l of each receptor protein. Control incubation received unprogrammed lysate alone. Oligonucleotide probes were end-labeled with [␥-32 P]ATP using polynucleotide kinase (New England Biolabs). Specific activities of end-labeled oligonucleotide probes were determined by scintillation counting. Volumes of 32 P-labeled oligonucleotide probes equivalent to 25,000 cpm were used in each reaction. Reactions were incubated on ice for 15 min before the addition of radioactively labeled oligonucleotide probes and where indicated in the presence of antibodies against RXR␣ (Santa Cruz, CA). In competition assays, competitor oligonucleotides were added at 50-, 100-, and 250-fold molar excess. Samples were held on ice for another 5 min, and the protein-DNA complexes were resolved on a pre-electrophoresed 5% polyacrylamide gel in 1ϫ TBE (45 mM Tris borate, 1 mM EDTA) at 4°C. Gels were dried and autoradiographed at Ϫ70°C for 8 h. The double-stranded oligonucleotides were annealed from the following single-stranded oligonucleotide and an oligonucleotide corresponding to the complementary sequence; FXRE1 WT, 5Ј-gagcaggtgggggtcatagtcctgaaagccaga-3Ј; FXRE2 WT, 5Ј-catgacaccaagggcagtgacctcattccacag-3Ј; FXRE2 mut, 5Ј-catgacaccaaAAgcagtgaTTtcattccacag-3Ј. IR-1 elements are indicated in boldface, and mutations are capitalized.
Immunohistochemical Studies-Fresh-frozen normal, cirrhotic, and cholestatic human liver sections were obtained from the Tissue Procurement Core Laboratory of the Department of Pathology and Laboratory Medicine at UCLA. The protocol was approved by the Institutional Review Board of UCLA. Hematoxylin and Eosin staining were performed by the core laboratory using standard protocols. Immunohistochemical studies using anti-␣A-crystallin polyclonal antibodies against recombinant bovine protein, amino-or carboxyl-terminal peptide, were performed as follows; frozen sections were thawed at room temperature for 30 min, fixed in 100% methanol for 10 min at Ϫ20°C, and then air-dried at room temperature for 20 min. Sections were then blocked with hydrogen peroxide (0.3%) to quench endogenous peroxidase activities. After rinsing with PBS, the sections were in turn blocked with 10% normal goat serum (20 min), avidin (15 min), and biotin (15 min) (Avidin/Biotin Blocking Kit, Vector laboratories, CA) at room temperature. Sections were then incubated with one of the anti-␣A-crystallin antibodies for 1.5 h at room temperature in a humidity chamber and washed thoroughly with phosphate-buffered saline. Secondary antibodies were added for 30 min, and the sections were washed with phosphate-buffered saline. The sections were developed using the VECTORSTAIN ABC peroxidase system and either the 3,3Ј-diaminobenzidine or the 3-amino-9-ethylcarbazole substrates from Vector Laboratories, counterstained with hematoxylin, and dehydrated, and the slides were mounted in xylene-based mounting medium.

␣A-Crystallin Expression Is Induced in HepG2 Cells and
Primary Human Hepatocytes in Response to FXR Ligands-In an attempt to identify target genes that are regulated by FXR, a microarray screen was carried out with RNAs isolated from HepG2 cells that stably express the neomycin resistance gene (HepG2-Neo) or the neomycin resistance gene and rat FXR␣2 (HepG2-FXR) (50) that had been treated for 24 h with either vehicle (Me 2 SO), the natural FXR ligand CDCA (100 M), or the synthetic FXR-specific ligand GW4064 (1 M). Analysis of the microarray data revealed a number of genes whose mRNA levels were induced upon treatment with either the natural or synthetic FXR ligand. One such induced mRNA corresponded to ␣A-crystallin (CRYAA).
This result was intriguing since CRYAA expression was reported to be largely restricted to the lens of the eye and absent from the liver. Consistent with earlier studies (5, 6), Northern blot analysis indicates that CRYAA mRNA was barely detectable in untreated HepG2 cells (Fig. 1, A and B) or primary human hepatocytes (Fig. 1C). However, CRYAA mRNA levels exhibited a dramatic 3.1-7.1-fold induction when HepG2-Neo or HepG2-FXR cells were treated with 100 M CDCA, respectively (Fig. 1, A and B). Incubation of the cells with higher levels of CDCA (250 M) resulted in a less robust induction of CRYAA mRNA (Fig. 1A). We have noted previously that the expression of other FXR target genes, such as apolipoprotein C-II (apoC-II), also decreased in the presence of CDCA concentrations of 250 M (52). The data of Fig. 1C shows that CRYAA mRNA levels were also induced 7.2-fold after a 48-h treatment of primary human hepatocytes with GW4064. A similar increase in CRYAA mRNA levels was noted in FXR-agonisttreated HuH7 human hepatoma cell line (data not shown). As expected, the well characterized FXR target gene SHP was induced when either HepG2 cells or primary human hepatocytes were treated with FXR agonists (Fig. 1). These data confirmed the prediction derived from analysis of the microarray data and demonstrated that CRYAA mRNA levels are highly induced in response to FXR ligands in both HepG2 cells and primary human hepatocytes.
To rule out the possibility that the induction of CRYAA mRNA was due to a general effect of FXR ligands on the expression of heat shock proteins, HepG2-FXR cells were treated with either vehicle (Me 2 SO) or the synthetic FXRspecific ligand GW4064 (1 M) for 24 h before RNA isolation. The data of Fig. 1B illustrate the specificity of CRYAA mRNA induction in response to FXR ligand treatment. In contrast to CRYAA, neither heat shock protein 27 (HSP27) nor heat shock protein 70 (HSP70) mRNA levels were induced after activation of FXR (Fig. 1B). These data suggested that FXR ligand treatment specifically increased CRYAA mRNA levels but not that of other heat shock proteins.
The increase in CRYAA mRNA expression in response to FXR ligand treatment in HepG2 cells resulted in a concomitant increase in CRYAA protein level. Western blot analysis using polyclonal antibodies toward recombinant bovine ␣A-crystallin demonstrated that ␣A-crystallin protein levels also increased in response to an FXR-specific ligand (Fig. 1D, compare lane 2  versus lane 1). Furthermore, this induction is FXR-dependent since transient transfection of plasmids encoding FXR and RXR further augmented the level of ␣A-crystallin protein and the subsequent response to FXR ligand treatment (Fig. 1D,  compare lanes 3 and 4 versus lanes 1 and 2). We also noted that the apparent molecular weight of ␣A-crystallin in HepG2 cells is higher than that of bacterially expressed purified recombinant bovine ␣A-crystallin (Fig. 1D, lane 1-4 versus lane 5). To rule out the possibility that the bands in lanes 1-4 of Fig. 1D correspond to ␣B-crystallin, a duplicate blot was hybridized with anti-␣B-crystallin antibodies. No signal was detected, although the antibody was shown to react strongly with recombinant ␣B-crystallin (data not shown). Therefore, we hypothesize that the difference in molecular weight may be due to previously uncharacterized post-translational modifications of ␣A-crystallin in HepG2 cells.
To investigate whether CRYAA mRNA was expressed and regulated in rodent livers, we isolated RNAs from livers of wild-type mice gavaged with either vehicle or GW4064 (30 mg/kg twice a day for 4 days). Northern blot analysis indicated that CRYAA mRNA was present at very low levels that were unchanged after the GW4064 treatment (data not shown). In contrast, the hepatic levels of SHP mRNA were highly induced (data not shown). Taken together, these findings suggested that regulation of CRYAA mRNA by FXR is human-specific since it was only observed in human-derived hepatocytes but not in murine livers.
The First Intron of the ␣A-Crystallin Gene Confers FXR-dependent Transcription Activation in Transient Transfection Assay-With one reported exception (53), FXR activation of known target genes requires that FXR binds to an FXR response element (FXRE) as an FXR/RXR heterodimer (24,54). Most known FXREs consist of two half-sites with the consensus sequence AGGTCA arranged as an inverted repeat separated by one nucleotide (IR-1) (22). Nevertheless, recent studies have identified a few FXR target genes that possess FXREs other than the classic IR-1 motif (37,27,55). Analysis of the published nucleotide sequence of the proximal promoter and first intron of the CRYAA gene revealed a number of putative IR-1 motifs that are similar to previously identified FXR-response elements. To identify which if any of these putative IR-1 sequences are functional, we initially generated two luciferase reporter gene constructs. A 557-bp region of the proximal promoter that includes the transcription start site was cloned into the pGL3 luciferase reporter vector (pGL3-500). A second reporter construct was generated by cloning the first intron of the CRYAA gene upstream of the minimal pTK promoter-luciferase reporter vector (pTK-CryAAint1) (Fig. 2).
CV-1 cells were transiently transfected with one of these reporter gene constructs in the presence or absence of expression plasmids encoding human RXR␣ and rat FXR (corresponding to the FXR␣2 isoform) (24,25). Transfected cells were subsequently treated with either vehicle (Me 2 SO) or the synthetic FXR ligand GW4064 for 48 h. Both the pGL3-500 construct and the empty pGL3 vector exhibited low and unregulated transcriptional activities ( Fig. 2A). In contrast, the pTK-CryAAint1 reporter gene exhibited specific and potent transcriptional activation after treatment with GW4064 and co-expression of FXR and RXR (Fig. 2B). Notably, the luciferase activity and fold induction were far greater than that seen with the positive control reporter gene driven by two copies of a well characterized IR-1 element (pTK-2X IR-1) from the hepatic control region of the human apolipoprotein E/C-I/C-IV/C-II gene cluster (Fig. 2B) (52). These results suggested that a potent and functional FXRE is present in the first intron of the CRYAA gene.
Identification of Two Putative FXREs in the First Intron of the CRYAA Gene-The transient transfection assays described in Fig. 2B suggest that the first intron of the CRYAA gene contains one or more functional FXRE. Analysis of the genomic sequence of this region identified two putative FXREs (Fig. 3A,  FXRE1 and FXRE2).
To determine whether FXR/RXR heterodimers bound to these putative FXREs, we performed EMSAs using in vitro transcribed and translated rat FXR and human RXR␣ proteins and radioactively labeled oligonucleotide probes corresponding to either FXRE1 or FXRE2. As is evident in Fig. 3B, FXR/RXR heterodimers bound to both FXREs in vitro (lanes 4 and 10). However, FXRE2 led to the formation of a more robust shifted complex than FXRE1, suggesting that FXR/RXR heterodimers may bind to FXRE2 with higher affinity. An antibody specific to human RXR␣ supershifted the complexes (lanes 5 and 11), whereas an antibody to YY1 (Yin Yang 1) has no effect (lanes 6 and 12). The well characterized IR-1 from the phospholipid transfer protein (PLTP) served as a positive control (Fig. 3B,  lanes 13 and 14) (50, 56). Competition EMSAs were also carried out using non-radioactively labeled oligonucleotides corresponding to FXRE2 from the CRYAA gene, a mutated FXRE2 containing four transversions, or the IR-1 from the PLTP promoter (Fig. 4). FXRE2 was chosen as the competitor because the data in Fig. 3 suggest that it binds FXR/RXR heterodimer with higher affinity than FXRE1. As seen in Fig. 4, increasing concentrations of oligonucleotides corresponding to either unlabeled FXRE2 or IR-1 from the PLTP promoter reduced the formation of the FXR/ RXR heterodimer with the radioactive probe (lanes 5-7 and lanes [11][12][13]. In contrast, no competition was observed when the mutant FXRE2 was used as a competitor (Fig. 4, lanes  8 -10). These data demonstrated that the interaction between FXR/RXR and FXRE2 is specific.
FXRE2 Is a Functional Enhancer Element in the Natural Context of the CRYAA Gene-To determine whether the FXREs present in the first intron of the human CRYAA gene are functional in their natural context, a luciferase reporter gene construct was generated that was under the control of a genomic fragment containing the proximal promoter region plus exon 1, intron 1, and part of exon 2 (Ϫ1320 to ϩ1581) of the CRYAA gene (Fig. 5A). To ensure the production of a functional luciferase protein, the start codon located in exon 1 of the CRYAA gene was cloned in-frame with the ATG of the luciferase gene in the vector (pGL3-pE1E2 in Fig. 5A). Translation of the resulting transcript should produce a fusion protein containing 94 amino acids of the amino-terminal CRYAA sequence (derived from exons 1 and 2) in-frame with luciferase.
CV-1 cells were transiently transfected with the pGL3-pE1E2 promoter-reporter gene construct in the presence or absence of plasmids expressing human RXR␣ and human FXR␣1 (Fig. 5A). Transfected cells were subsequently treated with either vehicle (Me 2 SO) or the synthetic FXR ligand (GW4064) for 48 h. The data of Fig. 5A demonstrated that pGL3-pE1E2 is highly induced by GW4064 after co-transfection of FXR-encoding expression plasmids. We conclude that the putative FXRE(s) in intron 1 retained its enhancer activity in the natural context of the gene. Induction of luciferase activity was unaffected by a mutation in FXRE1 (compare Fig. 5B to 5A). However, mutation of FXRE2 completely abolished induction of reporter gene activity (Fig. 5C), and luciferase levels declined to those of the empty vector (Fig.  5D). Taken together, these data clearly demonstrate that FXRE2 in intron 1 of the human CRYAA gene functions as a bona fide FXR response element.
Because we have previously demonstrated that certain FXR target genes are equally activated by each of the four FXR isoforms, whereas other genes are responsive to specific FXR isoforms (25,27,57), we performed transient transfection using the wild-type pGL3-pE1E2 reporter gene in the presence or absence of co-transfected plasmids encoding hRXR␣, rFXR␣2, or individual human FXR isoforms. Each human FXR isoform was expressed from the same vector (pcDNA3.1) to ensure similar expression and allow for a direct comparison. The data of Fig. 6A show that under these conditions all four human FXR isoforms are able to induce ␣Acrystallin reporter gene transcription activity. However, hFXR␣2 and hFXR␣4, the isoforms lacking the MYTG motif adjacent to the DNA binding domain (25), induce a far more robust transcriptional activation than the two isoforms that contain this motif (Fig. 6A).
No induction was noted with the empty vector (Fig. 6B).
Immunohistochemical Staining of Human Liver Sections Reveals ␣A-Crystallin Expression in Hepatocytes and Cholangiocytes-Because we have shown that CRYAA is an FXR target gene and its expression in human hepatocytes is induced in response to FXR ligands, we performed immunohistochemical staining of human liver sections using antibodies raised against a degenerate peptide corresponding to the amino ter- The IR-1 elements in FXRE1 and FXRE2 are in boldface, whereas the arrows indicate the orientation of each half-site. B, electrophoretic mobility shift assays were performed as described under "Experimental Procedures." 32 P-Labeled oligonucleotide probes containing sequences of FXRE1, FXRE2, or the previously characterized IR-1 element from the promoter of the human phospholipid transfer protein (hPLTP) gene (50,56) were incubated with in vitro transcribed and translated hFXR␣1 and/or hRXR␣, as indicated. Antibodies against RXR␣ were also included in the reaction in lanes 5, 11, and 14. The FXR/RXR␣/ 32 Plabeled oligonucleotide probe complex and the supershifted anti-RXRa/ FXR/RXR␣/ 32 P-labeled oligonucleotides probe complex are indicated.

FIG. 4.
Competition studies identify FXRE2 from the ␣A-crystallin gene as a high affinity binding site for FXR/RXR. EMSAs were carried out as described under "Experimental Procedures." Unlabeled oligonucleotide probes containing sequences of FXRE2, mutated FXRE2, or IR-1 from the hPLTP promoter (a previously characterized high affinity binding site for FXR/RXR) were added at 50-, 100-, and 250-fold molar excess to binding reaction mixtures containing 32 Plabeled FXRE2 oligonucleotide probes and in vitro translated and transcribed hFXR␣1 and/or hRXR␣ as indicated.
minus of bovine, rodent, and human ␣A-crystallin (49). Hematoxylin and eosin staining was performed on human liver samples that were either normal (Fig. 7A), cirrhotic (Fig. 7B), or cholestatic (Fig. 7C). As evident in Fig. 7D, specific staining of ␣A-crystallin is seen in cholangiocytes, the epithelial cells that line the bile ducts, of the normal human liver. In sections of the human cirrhotic liver, specific staining of ␣A-crystallin is seen in cells located at the periphery of regenerative modules encircled by fibrotic tissues (Fig. 7, E and F, arrows). There is also strong staining of cells located within the fibrotic tissues (Fig.  7, E and F, arrowheads). These latter cells likely correspond to differentiating liver progenitor cells that form reactive bile ductules in response to hepatic damage (58) or to myofibroblasts derived from activated hepatic stellate cells (HSCs) (59,60). In addition, in cholestatic liver sections taken from a patient suffering from primary biliary atresia, hyperproliferation of bile ducts is evident, and ␣A-crystallin expression is seen in cholangiocytes that line these bile ducts (Fig. 7G, arrows). Strikingly, the staining is specifically found on the lumen-facing side of these cholangiocytes. Bile droplets can be seen within some of these latter bile ducts (Fig. 7H, arrow). Although the antibody used in these stainings has been shown previously to recognize both ␣Aand ␣B-crystallin (49), as expected, immunohistochemical stainings of these same liver sections using antibodies specific to ␣B-crystallin failed to detect any specific signal (data not shown). Taken together, these data suggest that ␣A-crystallin is specifically expressed in hepatocytes, cholangiocytes, and possibly activated HSCs in human livers by FXR under pathological conditions that are associated with increased hepatic levels of bile acids.

Real-time Quantitative PCR Confirms Expression and Regulation of ␣A-Crystallin by an FXR Agonist in Primary Human
Hepatic Cell Cultures-Because immunohistochemical staining of human liver sections revealed the presence of ␣Acrystallin in multiple hepatic cell types, we carried out realtime quantitative PCR analysis using RNAs isolated from primary cultures of human hepatocytes, cholangiocytes, and HSCs. The data confirm that ␣A-crystallin is both expressed in primary human hepatocytes (Fig. 8B) and HSCs (Fig. 8A) and is induced after treatment with GW4064. Moreover, the results in Fig. 8B demonstrated that ␣B-crystallin is not responsive to FXR ligand treatment. Repression of CYP7A1 mRNA in the primary human hepatocytes was used as a positive control for FXR-ligand responsiveness. To date, we have been unable to demonstrate ␣A-crystallin expression in primary human cholangiocyte cultures, although FXR is present both in these cells (data not shown) (61) and in were shown in each panel after normalization to ␤-galactosidase activities to control for differences in transfection efficiency. Data are given as mean Ϯ S.D.; p values were determined by unpaired two-tailed Student t test; *, p Ͻ 0.05. murine cholangiocytes (62). This discrepancy may result from de-differentiation and de-polarization of cholangiocytes in culture and/or difference in growth environment. DISCUSSION The present study identifies the human CRYAA gene as a direct target of FXR and shows that ␣A-crystallin protein is expressed in various cell types of cirrhotic and cholestatic human livers. This finding is intriguing because ␣A-crystallin was generally considered to be a "lens-specific" protein. The murine ␣A-crystallin gene, in contrast to its human counterpart, is not regulated by FXR. Consistent with this observation, analysis of the murine ␣A-crystallin genomic sequence reveals that the functional FXRE identified in the first intron of the human gene was not conserved in the murine gene. Although nuclear receptor response elements are usually located in the proximal promoters of target genes, there have been several examples in which the response elements are found in the distal enhancers (52) or in the intron of the target genes (63,43). In addition, other differences have been noted between the human and rodent ␣A-crystallin genes; for example, the murine, but not the human, gene encodes alternative transcripts as a result of alternative mRNA splicing of an exon located in the first intron of the rodent ␣A-crystallin gene (64,65). Other human-specific FXR target genes have also been identified; these include syndecan-1 (27) and fibrinogen (57). In addition, activation of the closely related nuclear receptor liver X receptor (LXR) is known to activate the murine but not the human genes encoding CYP7A1 (66,67) and LXR␣ itself (68). Detailed studies demonstrated that the liver X receptor response element in the promoter of the murine Cyp7a1 gene is not conserved in the human gene. This provides a mechanism by which rodents can eliminate excess dietary cholesterol through the activation of Cyp7a1 expression and the resulting enhanced conversion of cholesterol to bile acids. This renders rodents less susceptible to diet-induced hypercholesterolemia than humans, who lack this compensatory response (69 -72).
␣-Crystallins are stress-related and cell-protective proteins. Their ability to bind partially denatured proteins to prevent further denaturation or aggregation has been well established (1)(2)(3). ␣-Crystallins have also been shown to aid in the refolding of partially denatured proteins in vitro (73,74). Notably, Hatters et al. (75) reported on the ability of ␣-crystallins to inhibit amyloid formation by lipid-poor apoC-II, one of the known target genes of FXR (52). In the absence of lipids, apoC-II was shown to assume a less stable conformation in vitro, contributing to a higher probability of self aggregation and fiber formation (75).
In vitro studies with ␣Aand ␣B-crystallin led to the suggestion that each protein has independent chaperone-like activity and functions (76 -78). For example, ␣B-crystallin is normally expressed at relatively high levels in the heart, and the lack of ␣B-crystallin expression has been implicated in desmin-related myopathy (79 -81). More recently, ␣B-crystallin was shown to associate with the perinuclear Golgi in a human glioblastoma cell line, suggesting a possible role for ␣B-crystallin in the Golgi reorganization during cell division (82).
The induction of ␣A-crystallin expression in the liver by ligand-activated FXR, reported herein, likely implies a novel role for ␣A-crystallin in bile acid homeostasis and/or protection of hepatocytes and cholangiocytes from excess bile acids. Immunohistochemical staining demonstrated that ␣A-crystallin is expressed in cells that border the regenerative nodule of hepatocytes and the fibrotic portal tract in human cirrhotic liver (Fig. 7). These cells are thought to be derived from liver progenitor cells (also known as oval cells), which are able to differentiate into hepatocytes or cholangiocytes upon liver injury as part of the liver regeneration mechanism (58,83). The fact that ␣A-crystallin is induced in isolated primary hepatic stellate cell cultures also suggests a role in injury response, as stellate cells are involved in repair of damaged liver (59,60). Taken together, these data suggest a new role for ␣A-crystallin as a stress-induced hepatoprotective protein.
Excess hepatic bile acid levels are highly detrimental to normal liver function (84). A number of FXR-dependent pathways have been identified that ensure hepatocytes are protected from excess bile acids. For example, activated FXR down-regulates the rate-limiting enzyme in bile acid synthesis, CYP7A1, by mechanisms that include activation of SHP and/or fibroblast growth factor 19 (42,43) and induces numerous transporter proteins in the liver, including BSEP, MRP2, and MDR2/3, to facilitate the transport of bile acids and phospholipids into the bile. Mutations in these latter transporters have been linked to various forms of cholestatic liver disease associated with increased hepatic levels of bile acids (85)(86)(87).
Although the precise mechanism mediating the cytotoxic effect of bile acids remains obscure, it has been generally attributed to their hydrophobicity which likely disrupts membrane integrity and denatures proteins (88). As a result, hepatocytes have developed a number of mechanisms to protect themselves from the toxic effects of bile acids. For example, toxic secondary bile acids such as lithocholic acid also activate the pregnane X-receptor, a nuclear receptor known to be important in regulating hepatic xenobiotic metabolism and detoxification (89 -91).
The results presented in this study demonstrate that a specific heat shock protein, ␣A-crystallin, is a direct target of FXR in human liver. These data suggest that ␣A-crystallin induction represents yet a third mode of defense for cells exposed to excess bile acids. To our knowledge, ␣A-crystallin expression has not been shown previously to respond to stress-induced stimuli outside of the lens. Immunohistochemical staining of human liver sections demonstrate that ␣A-crystallin is expressed in hepatocytes and/or differentiating liver progenitor cells of cholestatic and cirrhotic human livers in addition to cholangiocytes of both normal and pathological human livers (Fig. 7). Consequently, we hypothesize that ␣A-crystallin functions to prevent protein denaturation and cell damage in the face of excess intracellular bile acids. The data presented here suggest that activation of FXR might also protect human livers from the deleterious effects that result from excessive intracellular bile acid levels.