Ligand-dependent interaction of the aryl hydrocarbon receptor with a novel immunophilin homolog in vivo.

In an effort to identify regulators of aryl hydrocarbon receptor (AHR) signaling, we have employed the yeast two-hybrid system to screen for human proteins that interact in a ligand-dependent manner with the AHR. After screening 1.4 × 106 clones from a human B cell library, two distinct clones were identified that associated specifically with the liganded receptor. No clones were identified that interacted preferentially with the unliganded AHR. One of the ligand-dependent clones, ARA9, encodes a novel 330-amino acid protein with regions of amino acid sequence similarity to the 52-kDa FK506-binding protein known to be associated with the glucocorticoid receptor. Yeast two-hybrid experiments with ARA9 demonstrated a strong interaction with the AHR that is enhanced 11-fold in the presence of the ligand β-naphthoflavone. In vitro experiments using proteins generated in reticulocyte lysates confirmed this interaction and indicated that ARA9 can be co-immunoprecipitated with the AHR using antisera raised specifically for either the AHR or the 90-kDa heat shock protein. The observation that ARA9 has a high affinity for both the 90-kDa heat shock protein-associated and ligand-activated forms of the AHR suggests that ARA9 is a component of the AHR-signaling pathway in vivo.

The AHR 1 is a ligand-activated transcription factor that regulates the expression of xenobiotic metabolizing enzymes in response to binding environmental pollutants such as 2,3,7,8tetrachlorodibenzo-p-dioxin (1). The AHR is a member of the Per-ARNT-Sim homology domain superfamily of regulatory proteins that also includes the AHR's dimer partner, ARNT, and such proteins as HIF1␣ and Sim (reviewed in Ref. 2). Members of this family are distinguished by a region of similarity of approximately 200 amino acids termed PAS (3). In the AHR this domain is involved in dimerization with other PAScontaining proteins, ligand binding, and association with hsp90 (4 -6). Most members of this superfamily also contain a basic helix-loop-helix domain immediately N-terminal to the PAS region (2). The basic domain mediates the recognition and binding of these factors to specific DNA sequences in enhancer elements that regulate transcription of target genes (6,7). The helix-loop-helix domain functions as a primary dimerization surface that directs interactions with appropriate dimeric partners (8,9).
Although no obvious structural relationship is apparent, the AHR and certain members of the steroid receptor superfamily exhibit similarities in their signaling mechanism (10,11). Biochemical studies have indicated that the unliganded AHR and the GR are located in the cytosol in a complex with a dimer of the molecular chaperone hsp90 and other cellular proteins (12)(13)(14). hsp90 has been shown to be an important regulator of receptor activity. Genetic studies in yeast systems deficient in hsp90 have demonstrated an absolute requirement for hsp90 in both glucocorticoid and AHR signaling (10,11,15). Biochemical studies have correlated the association of hsp90 with an increase in ligand binding capacity, and dissociation of hsp90 has been correlated with an increase in DNA binding capacity (16 -19). Finally, current models of signaling for the GR and AHR are quite similar. In response to ligand binding in the cytosol, receptor-hsp90 complexes translocate to the nucleus where hsp90 dissociates and the activated GR homodimerizes or the AHR heterodimerizes with ARNT. The resulting complexes attain binding specificity for their cognate enhancer elements to regulate transcription of specific batteries of genes.
In addition to hsp90, a number of other intracellular proteins appear to play a role in GR signaling. The GR has been shown to exist in a complex with two other heat-inducible proteins, hsp70 and FKBP52 (20). hsp70 binds to the hormone binding domain of the GR and may be required for hsp90 to bind to the receptor (21). FKBP52 is an immunophilin of the FK506 binding class that binds directly to hsp90 in the non-liganded receptor heat shock protein complex and may play a role in targeted protein movement (20 -24). Other proteins that have been recovered in native steroid receptor complexes bound to hsp90 include a highly acidic 23-kDa protein and an immunophilin of the cyclosporin A binding class known as Cyp-40 of unknown function (20).
In contrast to GR, less is known about the composition of the cytosolic and ligand-activated forms of the AHR. Chemical cross-linking experiments have revealed the presence of a 46-kDa protein of unknown function in the AHR cytosolic complex (25)(26)(27). Purification of ligand-activated AHR complexes using DNA affinity chromatography has indicated that the DNAbound form of the AHR is found associated with ARNT and a number of unknown proteins including those with apparent molecular masses of 110, 57, and 54 kDa (28,29). In an effort to identify additional proteins that are part of the AHR complex or that may regulate its activity, we used the yeast two-hybrid system to screen for proteins that associate with the AHR in a ligand-dependent manner in vivo and then confirmed these interactions by co-immunoprecipitation assays in vitro. We report here the isolation of a cDNA clone encoding a protein whose interaction with the AHR in vivo is detected only in the presence of ligand and with sequence similarity to the FKBP52 protein found associated with the GR.
Transformations and Library Screening-Transformation of plasmids into the strain L40 was performed by a modified lithium acetate protocol as described (10,34). For library screening, the transformants were replica-plated after 4 days to fresh plates of the same media and incubated for 2 days. The colonies were again replica-plated to fresh plates that contained either no agonist or 1 M ␤NF, then incubated at 30°C for 2 days, and assayed for lacZ expression as described (35).
Northern Blot Analysis-Fifty nanograms of ARA9 and 25 ng of actin cDNA were random-primed and used as a probe to hybridize human multiple tissue and human immune system Northern blots (CLON-TECH, Palo Alto, CA) as described (33).
Quantitative ␤-Galactosidase Assays-S. cerevisiae strain L40 was transformed with the appropriate plasmids and plated on selective media either treated by spreading 100 l of 1 mM ␤NF over the surface (78.5 cm 2 ) of the medium or left untreated. The plates were incubated for 4 days at 30°C. For each assay, three colonies were suspended in 100 l of buffer Z (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 35 mM 2-mercaptoethanol). Five microliters of the cell suspension were diluted in 170 l of buffer Z, and the A 600 was measured to estimate cell density. The remaining cell suspension was pelleted by centrifugation, and ␤-galactosidase activity was determined by a chemiluminescent reporter assay protocol (Galacto-light, Tropix, Bedford, MA). Light units were divided by the A 600 to normalize values to cell density.
Co-immunoprecipitation Assay-AHR, ARNT, and ARA9 were expressed in vitro using the TNT-coupled rabbit reticulocyte lysate (Promega) using the plasmids pSportAHR, pSportARNT, and PL580 as templates, respectively (4). In our hands, this expression system typically produces about 10 fmol of 35 S-labeled translated protein per 50 l of reaction mixture (4). Associations were formed by mixing 5 l of the appropriate in vitro translated proteins in a 1.5-ml microcentrifuge tube, and co-immunoprecipitations were performed as described previously (33). For co-immunoprecipitations using the AHR, 0.1 l of 1 mM ␤NF or Me 2 SO was first added to the reaction and incubated for 2 h at 30°C.

RESULTS AND DISCUSSION
Two-hybrid Assay Screening Strategy-To identify novel factors involved in AHR signaling, we designed a two-hybrid assay that would detect proteins that interact with the AHR in a ligand-dependent manner. A plasmid containing a LexA-AHR fusion that had a deletion in the transcriptional activation domain (LexA-AHR⌬TAD (36)) was used to screen a human B lymphocyte cDNA library that was fused to the GAL4 transcriptional activation domain (6,37). The transformants were plated on selective media containing 1 M AHR agonist ␤NF or onto untreated plates. The surviving colonies from each group were twice replica-plated onto agonist and nonagonist plates and assayed for both histidine prototrophy and lacZ expression exclusively under either agonist or nonagonist conditions. 700,000 clones were screened under each condition. No clones were identified that interacted preferentially with the unliganded AHR. However, we isolated 13 positive colonies that interacted with AHR in a ligand-dependent manner. Ten of these clones retained ligand-dependent interactions with the LexA-AHR⌬TAD construct upon secondary screening against the unrelated protein, lamin (38). Sequence analysis revealed that five clones represented a distinct cDNA of 2.2 kilobase pairs, designated ARA3 (for Ah receptor-activated), and the remaining five clones represented a second cDNA of 1.03 kilobase pairs that was designated ARA9. In this paper, we describe the characterization of the ARA9 cDNA.
Analysis of the ARA9 cDNA-The original ARA9 clone, pbARA9, contained an open reading frame of 325 amino acids and lacked a Kozak consensus methionine (39). To obtain additional upstream sequence information, expressed sequence tags were identified by a BLAST search of the GenBank data base (33). Three of the corresponding cDNAs were obtained from the IMAGE consortium at Washington University and subjected to sequence analysis. One of the expressed sequence tag clones, PL580 (GenBank accession number R50134), contained the entire ARA9 sequence as well as additional 3Јuntranslated sequences, a poly(A) tail, and an additional 131 nucleotides at the 5Ј end for a total of 1244 nucleotides (Fig.  1A). This clone also contained a Kozak methionine with an upstream in-frame stop codon at position Ϫ99 (39). The ARA9 mRNA is approximately 1.35 kilobases in length and is expressed in all tissues examined with higher levels of expression in heart, placenta, and skeletal muscle (Fig. 2).
The DNA and deduced amino acid sequence of ARA9 is shown in Fig. 1A. It contains an open reading frame encoding a novel protein of 330 amino acids followed by a short 3Ј-untranslated region ending in a polyadenylation sequence. The deduced amino acid sequence has a predicted molecular weight of 38, which is consistent with the estimated molecular mass of 37 kDa as determined from the electrophoretic mobility of in vitro translated ARA9 on SDS-polyacrylamide gel electrophoresis (Fig. 4). The N-terminal portion of ARA9 contains a region of 79 amino acids, which is 30% identical to a region in FKBP52 and 28% identical to FKBP12. A second region of 11 amino acids is 63% identical to a region in FKBP12 (Fig. 1B). The C terminus harbors three regions among residues 182-215, 234 -267, and 268 -301 that conform to the consensus TPR domain and are 21, 24, and 47%, respectively, identical to TPR domains found within FKBP52. TPR domains are highly degenerate 34-amino acid sequences containing two regions able to form short amphipathic ␣-helices that are thought to mediate protein-protein interactions in a number of proteins (40,41). These domains have been identified in many proteins involved in such diverse functions as cell division in yeast, protein import, and Drosophila development (42).
Interaction of the AHR and ARNT with ARA9 -To obtain quantitative data describing the ligand-dependent interaction of the AHR with ARA9 and to determine whether ARA9 interacts with the structurally related AHR dimeric partner, ARNT, LexA fusions of AHR⌬TAD or ARNTC⌬325 were co-transformed with pbARA9 into the strain L40, plated on selective media, and ␤-galactosidase activity measured as an indication of the relative strength of the interactions in the presence or absence of ␤NF (for AHR⌬TAD). In the presence of ARA9, induction of ␤-galactosidase expression was increased 25-fold over control in the absence of agonist, and this interaction is enhanced approximately 11-fold in the presence of ␤NF. These experiments demonstrate that ARA9 interacts with the AHR in a conditional manner (Fig. 3). Co-transformation of Lex-AARNTC⌬325 with ARA9 only weakly induced ␤-galactosidase activity (2.7-fold over control) suggesting that ARNT may not associate with ARA9 in vivo or that this interaction is much weaker than that with AHR (Fig. 3).
To further characterize the interactions of ARA9 with the AHR and ARNT, co-immunoprecipitation assays were performed using full-length in vitro translated AHR, ARNT, and 35 S-labeled full-length ARA9. Using AHR antisera, 35 S-labeled ARA9 could be co-precipitated with the AHR both in the presence and absence of ligand (Fig. 4, lanes 2-5). We also asked whether ARA9 was a component of the AHR⅐hsp90 complex. The AHR⅐ARA9 complex was detected using anti-hsp90 antisera, indicating that ARA9 is present in AHR⅐hsp90 complexes (Fig. 4, lanes 6 -9). However, ARA9 could not be co-precipitated with hsp90 in the absence of AHR (Fig. 4, lanes 10 and 11). These results suggest that either ARA9 binds to the AHR directly or that stable or high affinity association with hsp90 is AHR-dependent.
The observation that ARA9-AHR interactions are ligand-dependent in vivo and ligand-independent in vitro suggests that ARA9 constitutively associates with the AHR regardless of the ligand binding status of the receptor, and that the apparent ligand dependence of the complex in vivo reflects the change in subcellular localization from the cytosol to the nucleus where it is able to activate transcription. This observation also suggests that the ARA9-AHR interaction occurs at multiple stages of the signaling pathway and that the ligand dependence of the twohybrid assay is a reflection of the fact that the interaction can be maintained during nuclear translocation, hsp90 association/ dissociation, dimerization, and DNA binding of the LexA-AHR⌬TAD chimera. The negative co-immunoprecipitation data with ARA9 and ARNT (Fig. 4, lanes 12 and 13) supports the in vivo experiments which demonstrated a weak interaction between ARA9 and ARNT and indicate that ARA9 may have marked restrictions in the number of proteins with which it can interact. With this in mind, it is interesting to point out that our preliminary two-hybrid data suggest that ARA9 does not interact with GR (data not shown).
ARA9 is a candidate for the previously described 43-kDa subunit of the AHR-cytosolic complex (26). Chemical cross-linking experiments using murine Hepa1c1c7 cytosol have shown that the AHR-cytosolic complex exists as a heterotetramer (25,26). One of these subunits is the AHR, two others have been identified as a hsp90 dimer, and the fourth is an unidentified 43-kDa subunit (26). In a previously reported study, a heteromeric intermediate of 146 kDa was isolated using immunoabsorbed anti-AHR antibody and was postulated to contain the 97-kDa AHR plus the 43-kDa-associated protein suggesting that this factor may be directly associated with the AHR (25,26). Additionally, the 43-kDa subunit was shown not to be stably associated with cytosolic-hsp90 complexes in Hepa1c1c7 extracts (43). These findings are in agreement with the data presented here in which ARA9, a human protein of similar size to p43, appears to directly interact with the AHR, but not with cytosolic hsp90 in the absence of receptor.
In overall structure, ARA9 is related to FKBP52, also known in the literature as p59 and hsp56 (reviewed in Ref. 20) (Fig.  1B). FKBP52 is one of several chaperone proteins found in the untransformed GR heterocomplex and has an overall length of 459 amino acids (20). It binds directly to hsp90 in the heat shock complex through three TPR units located within the C-terminal half of the protein (44,45). FKBP52 also contains two FKBP12-like domains termed I and II. Although the function of domain II is unknown, domain I has high identity (49%) with FKBP12 and has been shown to exhibit peptidylprolyl isomerase activity and bind to the immunosuppressants FK506 and rapamycin (44,46). The biological significance of this activity is unclear since blocking the peptidylprolyl isomerase activity with FK506 does not affect the cytosolic heterocomplex assembly or proper folding of the GR into a high affinity ligand binding conformation (47). Furthermore, FKBP52 does not appear to be necessary for GR⅐hsp90 heterocomplex assembly (48). Recent experiments, however, have indicated that it may play a role in receptor translocation to the nucleus upon activation by ligand (23,24).
In contrast to FKBP52, ARA9 is a smaller protein (330 amino acids) and has only one FKBP12-like domain with weaker identity to FKBP12 (28%) and three TPR domains in the C terminus. It has yet to be established if ARA9 exhibits FK506 FIG. 2. Northern blot analysis of ARA9. Each lane contains 2 g of poly(A) ϩ mRNA from the indicated human tissue. The blot was probed with a random-primed probe to ARA9. The ARA9-hybridized blots were exposed to film at Ϫ80°C with intensifier screens for either 2 days (left panel) or 5 days (right panel, immune system tissues). The blots were reprobed with actin to control for loading (data not shown).

FIG. 3. In vivo interaction of ARA9 with AHR and ARNT.
pbARA9 was co-transformed into S. cerevisiae with a LexA fusion of either AHR⌬TAD or ARNTC⌬325 and assayed for interaction with ARA9 in the presence or absence of ligand. As a control, the AHR⌬TAD and ARNTC⌬325 LexA fusions were transformed into the strain L40 with an empty GAL4 activation domain vector to measure base-line ␤-galactosidase activity in the absence of ARA9 expression. Each bar represents the average of three independent experiments. ␤-Galactosidase expression was measured and normalized to the number of cells assayed. binding and peptidylprolyl isomerase activity and how this may affect receptor function. One important difference between ARA9 and FKBP52 is that while FKBP52 has been found to be associated with hsp90 through chemical cross-linking and coimmunoprecipitation experiments, ARA9 does not appear to stably interact with hsp90 in the absence of AHR. Possible explanations for this difference in hsp90 binding affinity in the absence of AHR may be that 1) all TPR domains do not form equally tight interactions with hsp90, 2) the affinity of the ARA9-hsp90 association may be below the limit of detection in our assay, and 3) the hsp90 antibodies used in our experiments may recognize hsp90 complexes different from those used to co-precipitate hsp90-FKBP52 complexes (i.e. our hsp90 antibodies may not be able to recognize hsp90⅐ARA9 complexes). In any case, while ARA9 appears to be in the same class structurally as FKBP52 and although both proteins associate with nuclear receptors, their roles in receptor signaling may be distinct.
Because of its stable, ligand-dependent interaction with AHR, ARA9 is likely to have an important role in receptor function. If ARA9 and p43 are the same protein and ARA9 does indeed form part of the AHR cytosolic complex it may associate with the AHR complex and function along with hsp90 in the folding and stabilization of the receptor in an inactive form. Alternatively, since ARA9 also appears to interact with the AHR as part of a DNA-binding complex, it is possible that ARA9 functions as a co-activator forming a bridge between the receptor and basal transcription factors similar to the way that steroid receptor co-activators such as SRC-1 and Trip1 are thought to function (49,50). Given its structural similarity with FKBP52, it is possible that ARA9 may have a similar role in AHR signaling possibly acting as a targeting molecule to direct receptor trafficking to the nucleus. Finally, we cannot rule out the possibility that ARA9 has a function unrelated to AHR signaling and that this function is dependent upon association with AHR.