Identification of Functional Domains of the Aryl Hydrocarbon Receptor*

Functional domains of the mouse aryl hydrocarbon receptor (Ahr) were investigated by deletion analysis. Ligand binding was localized to a region encompass- ing the PAS B repeat. The ligand-mediated dissociation of Ahr from the 90-kDa heat shock protein (HSP90) does not require the aryl hydrocarbon receptor nuclear translocator (Arnt), but it is slightly enhanced by this protein. One HSP90 molecule appears to bind within the PAS region. The other molecule of HSP90 appears to require interaction at two sites: one over the basic helix-loop-helix region, and the other located within the PAS region. Each mutant was analyzed for dimerization with full-length mouse Arnt and subse- quent binding of the dimer to the xenobiotic responsive element (XRE). In order to minimize any artificial steric hindrances to dimerization and XRE binding, each Ahr mutant was also tested with an equivalently deleted Arnt mutant. The basic region of Ahr is required for XRE binding but not for dimerization. Both the first and second helices of the basic helix-loop-helix motif and the PAS region are required for dimer- ization. These last results are analogous to those previously obtained for Arnt (Reisz-Porszasz, S., Probst, M.R., Fukunaga, B. N., and Hankinson, O. (1994) Mol. Cell. Biol. 14, 6075–6086) compatible with the notion that equivalent regions of Ahr and Arnt associate with each other. Deletion of the carboxyl-terminal half of Ahr does not affect dimerization or XRE binding but,

The aryl hydrocarbon receptor (Ahr) 1 binds a variety of environmentally important carcinogens, including polycyclic aro-matic hydrocarbons, and certain halogenated aromatic hydrocarbons, such as TCDD. Following ligand treatment, induction of CYP1A1 and several other enzymes involved in xenobiotic metabolism occurs in most tissues. Certain of these enzymes (including CYP1A1) are involved in the metabolism of polycyclic aromatic hydrocarbons to active genotoxic metabolites, and Ahr therefore plays an important role in carcinogenesis by these compounds. Ahr also mediates most, if not all, of the carcinogenic and toxic effects of the halogenated aromatic hydrocarbons, although metabolism of these compounds does not appear to be involved (reviewed in Ref. 1).
The unliganded Ahr is located in the cytoplasm of the mouse hepatoma cell line, Hepa-1, as part of a complex that also contains two molecules of the 90-kDa heat shock protein (HSP90) and perhaps another protein of approximately 43 kDa (2). It is not known whether both, one, or neither of the HSP90 molecules bind Ahr directly. HSP90 appears to be required for ligand binding by Ahr (3). Binding of ligand leads to dissociation of Ahr from HSP90. In ligand-treated cells, Ahr is found in the nuclear fraction, from which it can be extracted in the form of a complex with the aryl hydrocarbon receptor nuclear translocator protein (Arnt) (4,5). This complex is probably a heterodimer of the two proteins, although the presence of additional small protein(s) has not been rigorously excluded. Arnt appears to be a nuclear protein in Hepa-1 cells (5,6). Some evidence indicates that dissociation of Ahr from HSP90 occurs in the nucleus and that HSP90 may play a direct role in translocating Ahr into this organelle (7,8). It has been proposed that Arnt promotes the dissociation of Ahr from HSP90 (9). Transcriptional activation of the cyp1a1 gene results from the binding of the Ahr/Arnt heterodimer to short DNA sequences, termed xenobiotic responsive elements (XREs), located in the 5Ј-flanking region of the gene (4, 10 -12). Both Ahr and Arnt bind directly to the XRE sequence (13).
Mouse Ahr and Arnt are 20% identical in amino acid sequence. They also show a striking resemblance in overall structure (14 -16). Both proteins contain bHLH motifs toward their amino termini. However, Ahr and Arnt represent a novel subclass of bHLH-containing transcription factors because they differ from most or all other such proteins in that (i) activation of the Ahr complex requires ligand, (ii) the XRE sequence differs from the E-box sequence, which is the recognition sequence for nearly all other bHLH-containing transcription factors (reviewed in Ref. 17), and (iii) both proteins contain an approximately 300-amino acid segment of sequence similarity, called the PAS domain. The PAS domains of each protein contain two copies of an approximately 50-amino acid degenerate direct repeat, referred to as the PAS A and PAS B repeats. The Drosophila proteins single-minded (Sim) and period (Per) also contain PAS domains, and this domain has been shown to mediate homodimerization of Per and heterodimerization of Per with Sim (18).
Deletion analysis indicates that both ␣-helices of the bHLH region of Arnt are required for dimerization with Ahr and that the basic region is required for XRE binding but not for dimerization (16). The XRE sequence is asymmetrical. Arnt binds to the side of the XRE that is identical in sequence to a half-site of an E-box, while Ahr binds to the side of the XRE not resembling an E-box half site (19). This is consistent with the observation that the basic region of Arnt conforms well to the consensus sequence for other bHLH proteins, while the basic region of Ahr conforms only poorly. Deletion of either the A or B segments of the PAS region of Arnt slightly reduces dimerization with Ahr, while deletion of the complete PAS region severely affects dimerization (16). Thus Arnt possesses multiple domains required for maximal heterodimerization with Ahr.
We describe here a mutational analysis of Ahr that complements our previous mutational analysis of Arnt. We have used Ahr and Arnt proteins that are derived from the same species (mouse); have designed the deletion mutants such that individual functional domains are deleted precisely, so that we could accurately assess their putative roles in Ahr function; and have performed a comprehensive in vitro and in vivo phenotypic analysis of the mutants.

MATERIALS AND METHODS
Constructs-pSportAhr, the mouse Ahr cDNA (15) in the pSV.Sport1 expression vector (Life Technologies, Inc.) was kindly provided by Dr. Christopher Bradfield (Northwestern University Medical School, Chicago, IL). Expression constructs for the mouse Ahr (pcDNAI/Neo/Ahr) and Arnt (pcDNAI/Neo/mArnt) cDNAs and generation of the Arnt mutants were described previously (16).
The ⌬C mutant of Ahr was generated by polymerase chain reaction using pSportAhr as template. The 5Ј primer for ⌬C was the same 5Ј polymerase chain reaction primer used to generate the Ahr expression construct pcDNAI/Neo/Ahr (16), while the 3Ј primer contained a fournucleotide random sequence, a XhoI restriction site, a stop codon, and the complement of Ahr bases 1181-1191. After digestion with the appropriate restriction enzymes, the polymerase chain reaction product was ligated into the similarly digested pcDNAI/Neo vector (Invitrogen, San Diego, CA), for in vitro and in vivo expression. Internal deletion mutants of Ahr were generated by the oligonucleotide directed mutagenesis system of Nakamaye and Eckstein (20) using the Oligonucleotide-Directed In Vitro Mutagenesis System, version 2.1. (Amersham Corp.). Mutagenesis of the Ahr cDNA was performed on a 1.2-kilobase SalI/BamHI fragment (1-1254 bp) from pSportAhr cloned into M13 mp18. After confirmation of each mutation by sequencing, the mutagenized fragment was transferred to pcDNAI/Neo for expression. This was accomplished by a three-way ligation of a 1.039-kilobase SacII/ NarI fragment of pcDNAI/Neo/Ahr (containing pcDNAI/Neo sequence 1183-2196 bp and Ahr cDNA sequence Ϫ15 to 11 bp), a NarI/BamHImutagenized fragment of the Ahr cDNA (bp 11-1254 less the deletion), and a 8.715-kilobase SacII/BamHI fragment of pcDNAI/Neo/Ahr (containing pcDNAI/Neo sequences 0 -1182 bp and 2258 -6969 bp and Ahr cDNA sequence 1255-3077 bp). This ligation altered the 5Ј region of the pSportAhr fragment to that of pcDNAI/Neo/Ahr, which resulted in a 4-fold increase in in vitro expression (16). Plasmids were prepared by the Qiagen maxiprep procedure according to the supplier's protocols (Qiagen, Chatsworth, CA).
Cells and Cell Culture-The mouse hepatoma cell line, Hepa-1, and monkey kidney tumor cell line, CV-1, were cultured in nucleoside-free ␣-minimal essential medium (Irvine Scientific, Santa Ana, CA) supplemented with 10% heat-inactivated and dextran/charcoal-treated fetal calf serum as described previously (21).
In Vitro Transcription and Translation-The Ahr and Arnt cDNAs and their mutant derivatives were all contained in the pcDNAI/Neo vector in the appropriate orientation for in vitro expression from the T7 polymerase promoter. The constructs were expressed in the TNT T7coupled reticulocyte lysate system in the presence or absence of L-[ 35 S]methionine (final concentration, 1 mCi/ml; specific activity, Ͼ1,000 Ci/mmol; Amersham) according to the protocol from the supplier (Promega, Madison, WI). All reactions were incubated for 90 min at 30°C. The degree of expression of each construct was assayed by subjecting an aliquot from the incubation performed in the presence of L-[ 35 S]methionine to SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and subsequent quantitation on an AMBIS radioanalytic imaging system (AMBIS Inc., San Diego, CA) (henceforth referred to as ␤-scanning).
Ligand Binding Assay-The photoaffinity ligand 2-azido-3-[ 125 I]iodo-7,8-dibromodibenzo-p-dioxin (specific activity 2176 Ci/mmol) was generously provided by Dr. Gary Perdew (Purdue University, West Lafayette, IN). TNT incubation mixtures containing equimolar amounts of each protein were diluted in HEDG buffer (25 mM HEPES, 1.5 mM EDTA, 1 mM dithiothreitol, 10% glycerol) and incubated with 0.15 Ci of ligand for 30 min at room temperature. After cooling the mixtures on ice, dextran/charcoal was added to remove unbound ligand. After centrifugation, the supernatants were irradiated at 366 nm by means of a Blak-Ray ML-49 UV lamp (UVP Inc., San Gabriel, CA) at a distance of 8 cm for 10 min. After acetone precipitation, pellets were dissolved in Laemmli sample buffer (22), boiled, and subjected to SDS-PAGE and subsequent autoradiography. Quantitation of labeled proteins was performed on a LKB Ultroscan enhanced laser densitometer (Pharmacia Biotech Inc.).
Binding to HSP90 -Goat anti-mouse IgM ( chain-specific; Pierce) was coupled to CNBr-activated Sepharose (Pharmacia) according to the manufacturer's instructions. Ahr was synthesized in vitro in the presence of [ 35 S]methionine. The TNT reaction mixtures were incubated with 10 nM TCDD (in Me 2 SO, to a final concentration of 0.2% Me 2 SO) or solvent alone for 1.5 h at room temperature in the presence or absence of equimolar amounts of in vitro synthesized, unlabeled Arnt and/or 7.5 nM double-stranded synthetic oligonucleotide containing mouse XRE-1. They were then diluted 100-fold with HSP immunoprecipitation buffer (25 mM MOPS, 50 mM NaCl, 10% glycerol, 6 mg/ml bovine serum albumin, 1 mM EDTA, 0.5% Tween 20, and 0.02% NaN 3 ) and incubated on ice for 2 h. Aliquots were then incubated for two more hours at 4°C with anti HSP90 monoclonal antibody 3G3 (Affinity Bioreagents, Neshanic Station, NJ) or control IgM (TEPC 183, Sigma), each preadsorbed to anti-mouse IgM-Sepharose. Pellets were washed five times with the same buffer, heat-denatured in SDS sample buffer, and subjected to SDS-PAGE on a 7.5% gel. The dried gel was exposed to x-ray film and subsequently quantitated by ␤-scanning. For co-immunoprecipitation of the mutant derivatives of Ahr, the same procedure was used, except that the TNT reactions mixtures were directly diluted 200-fold in HSP immunoprecipitation buffer, in the absence of TCDD, Arnt, or XRE-1.
Heterodimerization Assay-Arnt (or its mutant derivatives) was synthesized in vitro in the presence of [ 35 S]methionine, while Ahr (or its mutant derivatives) was synthesized in the absence of isotope. The in vitro transcription and translation reactions containing equimolar amounts of Ahr (or its mutant derivatives) were mixed with 3 parts Arnt (or its mutant derivatives). The mixture was incubated with 10 nM TCDD (in Me 2 SO, to a final concentration of 0.2% Me 2 SO) or solvent alone, for 2 h at room temperature. The mixture was then adjusted to 25 mM HEPES, 1.2 mM EDTA, 10% glycerol, 200 mM NaCl, 0.1% Nonidet P-40, pH 7.4 (immunoprecipitation buffer), in a final volume of 100 l. Excess affinity-purified polyclonal antibody to Ahr (13) or the corresponding preimmune immunoglobulin G fraction was then added, and the incubation continued for an additional 2 h. The antigen-antibody complex was precipitated with protein A-Sepharose CL-4B beads (Pharmacia) for 1 h at room temperature. The pellets were washed four times with immunoprecipitation buffer, boiled in SDS sample buffer (22), and then subjected to SDS-PAGE on a 7.5 or 15% gel. In addition to exposure to x-ray film, the dried gels were analyzed by ␤-scanning to quantitate the amount of radioactivity in each immunoprecipitate. The low signal generated by the preimmune IgG in the presence of TCDD was subtracted from the signal generated by the Ahr antibodies in calculating the values for the amount of co-immunoprecipitated Arnt protein in the presence of TCDD, while values obtained in the absence of TCDD were adjusted by subtracting the (low) signal generated by the preimmune IgG in the absence of TCDD.
XRE Binding-In vitro synthesized, unlabeled Ahr or its mutant derivatives were mixed with unlabeled Arnt or its mutant derivatives (1:1 molar ratio) and incubated in the presence of 10 nM TCDD (in Me 2 SO, to a final concentration of 0.2% Me 2 SO) or solvent alone at room temperature for 1.5 h. The mixture was then adjusted to 25 mM HEPES (pH 7.5), 200 mM KCl, 10 mM dithiothreitol, 10% glycerol, 5 mM EDTA, 50 g of poly(dI-dC)⅐(dI-dC)/ml, and the incubation continued for 20 min at room temperature. The mixture was then incubated for an additional 20 min in the presence of a 32 P-labeled double-stranded synthetic oligonucleotide containing mouse XRE1 (16). Samples were analyzed on a 4.5% nondenaturing polyacrylamide gel in 1 ϫ HTE buffer (200 mM HEPES, 100 mM Tris, 5 mM EDTA, pH 8.0). After exposure to x-ray film, the dried gels were used to quantitate the XRE⅐Ahr⅐Arnt complex by ␤-scanning.
In Vivo Functionality-The Ahr cDNA constructs, with or without the Arnt cDNA, were cotransfected along with the CAT reporter plasmid, pMC6.3k (23) (6 g of each plasmid construct), into 5 ϫ 10 5 CV-1 cells/60-mm dish by the method of Chen and Okayama (24). 16 h after transfection, cells were subjected to a 15% glycerol shock for 1.5 min and then refed media with or without 10 nM TCDD. They were harvested 24 h later. In an attempt to eliminate potential inducers of CYP1A1 in the medium used to grow the cells, the fetal calf serum was treated with dextran-coated charcoal, and the medium was stored in the absence of light. Cell lysates were produced by three cycles of freezing (dry ice/ethanol) and thawing (37°C water bath). Endogenous acetyl transferases were inactivated by incubation at 65°C for 10 min. CAT assays were performed as described previously (25). Protein concentrations were determined by Bradford assay. Results were calculated as nmol of CAT activity/mg of protein/30 min.

RESULTS
Ahr Mutant Constructs-The mutant constructs are presented in Fig. 1. Ahr mutant constructs ⌬bHLH, ⌬b, ⌬HLH, ⌬H1, ⌬H2, ⌬AB, ⌬A, and ⌬B had the indicated segments excised precisely by oligonucleotide-directed mutagenesis. Since the boundaries for the Ahr basic region have not been fully defined, due to the lack of consensus with other bHLH proteins, we designed boundaries for this region according to alignments made with other bHLH proteins and also as indicated by previous investigators (14,15) Sequence alignment of the PAS regions of Arnt, Ahr, Per, and Sim shows that these regions are most similar over the A and B repeats and immediately flanking sequences (15,26). The boundaries for the PAS A and PAS B deletions were chosen so as to eliminate these most conserved regions. ⌬C lacks C-terminal amino acids 398 -805. All constructs were inserted into the expression vector pcDNAI/Neo for in vitro and in vivo expression. Two independent clones were isolated and analyzed for each mutation.
Ligand Binding-The mutant constructs were synthesized in vitro in the transcription-translation system and then analyzed for binding to the photoaffinity ligand 2-azido-3-[ 125 I]iodo-7,8-dibromodibenzo--dioxin. A representative ligand binding assay is presented in Fig. 2. Hepa-1 cytosolic extract (first lane) and in vitro synthesized Ahr (second lane) both generated strongly labeled bands migrating in the correct position (ϳ95 kDa) for Ahr. Furthermore, a 100-fold molar excess of unlabeled TCDD eliminated this band in the in vitro extract, confirming that it corresponds to Ahr (third lane). The deletion of the entire bHLH domain or segments thereof (⌬b,⌬H1,⌬H2) did not affect ligand binding. Similarly, deletion of amino acids 398 -805 (⌬C) resulted in a protein that maintained full ligand binding capacity. Thus, deletions outside the PAS domains had no effect on ligand binding. Deletion of the PAS A domain reduced ligand binding to 30% of that of full-length Ahr. The PAS B deletion completely abolished binding, as did deletion of the complete PAS AB region (⌬AB). The results of two independent experiments are summarized in Table I.
Binding to HSP90 -Treatment with TCDD in vivo causes dissociation of Ahr from HSP90. McGuire and co-workers (9) demonstrated that Ahr synthesized in reticulocyte lysate associates with HSP90 that is present in the lysate. However, they did not find any effect of TCDD on the amount of Ahr that complexed with HSP90. Additional studies suggested that Arnt is required for TCDD-dependent dissociation of Ahr from HSP90. We synthesized Ahr in the presence of [ 35 S]methionine in a coupled T7 transcription-reticulocyte lysate translation system. Binding of HSP90 to Ahr was then assayed by coimmunoprecipitating the 35 S-labeled Ahr with a monoclonal antibody to HSP90 under conditions in which the antibody was limiting. We observed association of Ahr with HSP90, but in contrast to the observations of McGuire and co-workers, found that TCDD treatment reduced the amount of Ahr associated with HSP90. We also investigated whether the addition of in vitro synthesized Arnt and/or an XRE-containing doublestranded oligonucleotide enhanced the effect of TCDD. A representative experiment is presented in Fig. 3A. A slight effect of Arnt in increasing the dissociation of HSP90 from Ahr was observed. (The ratios of the amount of HSP90 co-immunoprecipitated in the TCDD-treated mixture relative to the corresponding nontreated mixture were 0.59 Ϯ 0.03, 0.46 Ϯ 0.15, 0.57 Ϯ 0.14, and 0.43 Ϯ 0.08 for the mixtures with no additions (equivalent to lanes 1 and 2 in Fig. 3A), with added Arnt, with added XRE, and with added Arnt plus XRE, respectively. (The means and standard deviations are given for four separate experiments. The mean value for the ratio obtained with Arnt plus XRE is significantly different (p Ͻ 0.05) from that of the mixture with no additions.) The mean and standard deviation for the pooled values for the control mixtures and the mixtures to which the XRE was added are 0.58 Ϯ 0.09 and differ significantly (p Ͻ 0.05) from the mean for the pooled values of the mixtures containing Arnt and the mixtures containing Arnt plus the XRE (0.44 Ϯ 0.11). We were unable to detect Arnt in the reticulocyte lysate by Western blot analysis (data not shown) and estimate that the concentration of Ahr in the lysate after in vitro translation of the corresponding cDNA is at least 100-fold greater than the concentration of Arnt. Our experiments therefore indicate that TCDD-dependent dissociation of Ahr from HSP90 does not require Arnt and are therefore at variance with the conclusions of McGuire and co-workers. However our experiments do indicate that Arnt may enhance the dissociation of Ahr from HSP90. This effect could conceivably result from Arnt sequestering Ahr and preventing the latter's reassociation with HSP90.
We also analyzed the binding of certain Ahr mutants with HSP90. The results of a representative experiment are presented in Fig. 3B. The degree of HSP90 binding to each Ahr derivative was determined by quantitative radioanalytic imaging (referred to as ␤-scanning). The value for each mutant relative to that for full-length Ahr in the same experiment was calculated, and the mean values for four experiments are presented in Table I. ⌬A bound HSP90 at undiminished efficiency. ⌬bHLH and ⌬B each bound HSP90 at approximately 50% efficiency. ⌬AB did not bind HSP90. HSP90 was not precipitated by the control monoclonal antibody from any incubation.
Dimerization of Ahr Mutants with Full-length Arnt-Equimolar amounts of in vitro synthesized and unlabeled fulllength or mutant Ahr proteins were mixed with a 3-fold excess of [ 35 S]methionine-labeled full-length Arnt protein in the presence or absence of 10 nM TCDD and incubated at room temperature for 2 h to allow for heterodimerization. The mixtures were then incubated with an excess of affinity-purified antibodies prepared against amino acids 12-31 of Ahr (13). (Although the deletions in ⌬b and ⌬bHLH overlap the segment to which the antibodies were raised, these two proteins were immunoprecipitated at the same efficiency as full-length Ahr (data not shown).) The resulting immunoprecipitates were subjected to SDS-PAGE, and the amount, in each case, of the co-immunoprecipitated Arnt protein was determined by ␤-scanning. The value for each construct was calculated as a percentage of the amount of Arnt co-immunoprecipitated by full-length Ahr in the same experiment. The average values for all experiments are presented in Table I. The results of a representative experiment are shown in Fig. 4.
The first six lanes of Fig. 4 represent the controls for the co-immunoprecipitation assay and utilized full-length Ahr and Arnt incubated in the absence or presence of TCDD, and treated with preimmune IgG or Ahr antibodies, as indicated. The data demonstrate that TCDD treatment increased the amount of Arnt co-immunoprecipitated with Ahr and that very little Arnt was precipitated from the co-incubation mixture upon treatment with the preimmune IgG preparation.
Deletion of the complete helix region or individual helices (⌬bHLH, ⌬HLH, ⌬H1, and ⌬H2) resulted in proteins with greatly reduced or nondetectable ability to co-immunoprecipitate with Arnt, demonstrating a requirement for both H1 and H2 for dimerization. Deletion of the complete PAS domain (⌬AB) or either PAS A (⌬A) or PAS B (⌬B), also resulted in proteins lacking the ability to dimerize with Arnt. The dimerization activity of ⌬A was much lower than its ligand binding activity, suggesting that the PAS A region may contribute toward dimerization. However, an alternative explanation is that the PAS A region is not directly involved in dimerization a Results of two independent experiments. ϩ indicates ligand binding activity at least as great as that manifested by full-length Ahr; ϩϪ indicates activity 30% that of Ahr; Ϫ indicates no detectable ligand binding activity.
b Each value is the mean of 4 to 6 independent determinations (Ϯ standard errors); each determination being expressed as a percentage of the same parameter measured with full-length Ahr (ϩTCDD) on the same occasion.
c Signal obtained using preimmune IgG with full-length Ahr and Arnt not treated with TCDD subtracted from each independent determination of each construct. d Signal obtained using preimmune IgG with full-length Ahr and Arnt treated with TCDD subtracted from each independent determination of each construct. e Significantly different from cells transfected with pcDNAI/NEO lacking the Ahr insert (i.e. p Ͻ 0.05). Otherwise all other CAT activities were not significantly different from cells transfected with pcDNAI/NEO lacking the Ahr insert, correspondingly untreated or treated with TCDD (i.e. p Ͼ 0.05).
f ND, not done. g All values but these are significantly different (p Ͻ 0.05) from the corresponding value obtained with wild-type Ahr in the presence of TCDD. h Transfected with equivalent amount of the expression vector pcDNAI/NEO lacking the Ahr insert. and that ⌬A cannot dimerize because the HLH and PAS B regions of the ⌬A derivative of Ahr and full-length Arnt cannot properly align (as discussed below). The basic domain deletion mutant (⌬b) retained full ability to co-immunoprecipitate with Arnt, demonstrating that this region is not involved in dimerization. In addition, deletion of amino acids 398 -805 (⌬C) also produced a protein with undiminished capacity to associate with Arnt, indicating that no domains required for dimerization exist in this region. XRE Binding-We previously showed that when equimolar amounts of in vitro synthesized Arnt and Ahr were mixed in the absence or presence of 10 nM TCDD, incubated with 32 Plabeled XRE 1, and then analyzed by gel mobility shift assay, a particular gel-shifted band was produced. This band was identified as the Ahr⅐Arnt⅐XRE complex because it was not formed with either protein on its own and because its intensity was increased by TCDD treatment, greatly reduced in the presence of a 100-fold excess of unlabeled XRE, and unaffected by a 100-fold excess of a mutant XRE (16). In the present experiments, each mutant Ahr protein was mixed with an equimolar amount of full-length Arnt, and the amount of Ahr•Arnt•XRE complex that was generated was quantitated by ␤-scanning. The value for each mutant was calculated as a percentage of that produced by full-length Ahr in the same experiment. The average results of five different experiments are presented in Table I, and a representative autoradiogram is presented in Fig. 5. Consistent with their greatly reduced ability to heterodimerize with the Arnt protein, ⌬bHLH, ⌬HLH, ⌬H1, ⌬H2, ⌬AB, ⌬A, and ⌬B were each unable to generate an Ahr⅐Arnt⅐XRE complex. ⌬b, which lacks the basic domain, although it is fully capable of dimerizing with Arnt, was unable to bind the XRE. ⌬C formed an XRE complex at approximately the same efficiency (75%) as it dimerized with Arnt.
Interaction of Ahr Mutants with Arnt Mutants-It is possible that the lack of dimerization or XRE binding activity of a particular mutant protein (Ahr or Arnt) with its appropriate wild-type heterodimeric partner could be due to inability of the appropriate interfaces of the two proteins to interact. In an attempt to address this possibility, we performed dimerization and XRE binding assays on pairs of equivalently deleted Ahr and Arnt mutant constructs. The Arnt mutants are illustrated in Ref. 16. Consistent with previous single mutant assays for both Arnt (16) and Ahr (this paper), ⌬bHLH, ⌬HLH, ⌬H1, ⌬H2, ⌬AB, and ⌬B (Ahr ⌬B was dimerized with Arnt bHLHA) pairings expressed no significant TCDD-induced dimerization or TCDD-induced XRE binding activities. Results of representative experiments are presented in Figs. 6 and 7, and the average values for replicate experiments are presented in Table II tively), confirming that all domains necessary for high level in vitro functionality reside within these two constructs.
In the presence of TCDD, Ahr ⌬A dimerized with Arnt ⌬A at 35% of the efficiency with which full-length Ahr and full-length Arnt dimerized with each other. This is approximately the same efficiency with which Ahr ⌬A binds ligand and suggests that, like the PAS A region of Arnt (16), the PAS A region of Ahr plays little if any role in dimerization. The poor dimerization efficiency of Ahr ⌬A with full-length Arnt is therefore probably due to inappropriate association of the two proteins. In contrast to their partial ability to dimerize, the two ⌬A constructs, when mixed together, were completely incapable of binding the XRE. Thus either PAS A is directly involved in binding to the XRE (which would appear to be unlikely), or the conformation of the heterodimer of Ahr ⌬A and Arnt ⌬A is still aberrant to such a degree that DNA binding is precluded.
In Vivo Functionality of the Ahr Mutants-Ahr was cotransfected with the plasmid pMC6.3k, both with and without Arnt, into monkey CV-1 cells. (pMC6.3k contains the region from about nucleotide Ϫ6300 to nucleotide ϩ2566 of the rat cyp1a1 gene fused to the CAT reporter gene). The cells were then assayed for CAT activity in the presence and absence of TCDD (Table I). CV-1 cells were found by Western blot analysis to lack detectable Ahr, and TCDD-inducible CAT activity was highly expressed in these cells only when they were also cotransfected with Ahr. Despite the fact that Western blot analysis demonstrated that CV-1 cells have levels of Arnt protein comparable with those in Hepa-1 cells (data not shown), Arnt stimulated the TCDD-inducible CAT activity of CV-1 cells containing Ahr and pMC6.3k about 6-fold. However, in the absence of Ahr, Arnt had little if any effect on CAT activity, as demonstrated by the observation that the activity in CV-1 cells cotransfected with full-length Arnt and pMC6.3k in the presence of TCDD was not significantly different from the activity in cells cotransfected with pMC6.3k and the vector pcDNAI/Neo lacking the Arnt cDNA (p Ͼ 0.05). As we found earlier with Arnt (16) and others have found with Ahr (27,28), CAT activities in the absence of TCDD were nearly equal to those generated in its presence. (Uninduced CAT activities for Ahr in the absence and presence of Arnt were 56 and 91% of their respective induced activities (data not shown).) This may be related to overexpression of the proteins in transfected cells.
Certain Ahr mutant constructs were contransfected with pMC6.3k, with and without Arnt into CV-1 cells. The cells were treated with TCDD and then assayed for CAT activity. The CAT activity of each construct (with or without Arnt) is presented as a percentage of the activity obtained in the same experiment with cells contransfected with full-length Ahr and Arnt and treated with TCDD. As expected, ⌬bHLH, ⌬b, and ⌬A, which were shown not to bind the XRE in the presence of full-length Arnt, did not produce appreciable levels of CAT activity. ⌬C, although possessing nearly full in vitro dimerization and XRE-binding activities, produced no significant CAT activity, even with Arnt cotransfection. which is necessary if the roles of these regions are to be assessed accurately. We have previously reported on analogous mutational studies of Arnt (16). The various functional domains of Ahr deduced, as discussed below, from the current work and that of other investigators (29 -33) are illustrated in Fig. 8.
Whitelaw and co-workers (30) showed that a fusion protein containing only amino acids 230 -421 of Ahr possessed nearly full ligand binding activity, while Poland and co-workers (34) demonstrated that an Ahr mutant containing amino acids 1-403 bound ligand with normal efficiency. Since deletion of amino acids 398 -805 in ⌬C had no effect on ligand binding, our data allow us to narrow the boundaries of a domain absolutely required for ligand binding to amino acids 230 -397. Deletion of the PAS A region (amino acids 121-182) reduced ligand binding by 70%. This is consistent with previous results obtained with amino-terminal deletion mutants of Ahr (29,34). Since HSP90 may be required for binding of ligand to Ahr (3), ligand binding assays probably define a minimal region required for binding both HSP90 and ligand rather than a region that contacts ligand alone.
Our Ahr-HSP90 co-immunoprecipitation experiments were carried out under conditions in which increasing the amount of the antibodies to HSP90 led to a proportionate increase in the amount of HSP90 precipitated. Therefore the amount of each Ahr derivative coprecipitated with HSP90 should reflect the degree to which the derivative binds HSP90. Since ⌬bHLH binds only 50% of the amount of HSP90 as the full-length Ahr protein, the bHLH region appears to contain a necessary binding site for one of the HSP90 molecules. Interestingly, HSP90 has been shown to transiently interact with several bHLH proteins, and a binding site on MyoD has been mapped to a small region encompassing the bHLH region of this protein (35,36). The results from ⌬B suggest that a necessary binding site for the other HSP90 molecule is contained within amino acids 259 -374. Since ⌬AB is totally deficient in HSP90 binding, but ⌬A is unaffected, the segment from amino acid 182 to 374 appears to contain a region (or regions) required for binding both HSP90 molecules. The HSP90 molecule binding in the region between amino acids 259 and 374 may bind to an additional site contained between amino acids 182 and 259. The results of Poellinger and co-workers (30,31,37) also suggest that HSP90 binds over the bHLH region and over a region encompassing the PAS B repeat. However, these workers did not quantitate their results, and they could not deduce that different HSP90 molecules bind over the two regions. Furthermore many of their results were obtained with chimeric proteins containing Ahr fragments fused with a portion of the glucocorticoid receptor, and interpretation of these results is potentially confounded by the fact that the glucocorticoid receptor also binds HSP90, since it is possible that the segment of the glucocorticoid receptor contained in the chimeras contributes toward HSP90 binding.
In contrast to the results of McGuire and co-workers (9), we observed that TCDD treatment reduced the binding of the in vitro translated Ahr to HSP90, thus reflecting the in vivo situation and contradicting the proposal of McGuire and coworkers (9) that Arnt is required for dissociation of Ahr from HSP90. It should be noted that a requirement for Arnt in the latter regard is difficult to reconcile with the observation that the free Ahr monomer appears to be an intermediate in the dissociation process (38) and with evidence that TCDD can trigger dissociation of HSP90 from Ahr in Arnt-defective mutants of Hepa-1 cells (30). We did, nevertheless, obtain evidence that Arnt can enhance the dissociation of Ahr from HSP90.
As with Arnt, deletion of either ␣-helix of the bHLH domain of Ahr eliminated dimerization and XRE binding, while dele-  tion of the basic region eliminated XRE binding but not dimerization. Thus, like Arnt, the basic region of Ahr is required for DNA binding but not for dimerization. The observation that equivalent helix 1 deletion mutants of Arnt and Ahr failed to dimerize and that equivalent helix 2 deletion mutants also failed to dimerize reinforces the conclusion that both helices are required for dimerization. Although the PAS A-deleted mutant (⌬A) of Ahr was unable to dimerize with Arnt, this mutant dimerized with Arnt ⌬A at the same efficiency as it bound ligand. This suggests that the inability of Ahr ⌬A to dimerize with full-length Arnt is due to steric hindrance to dimer formation between these two proteins, and suggests that PAS A of Ahr is dispensable for dimerization. Deletion of PAS A of Arnt was previously shown to have little effect on its dimerization with Ahr (16). It has previously been shown that a mutant (C⌬516), containing amino acids 1-289, and deleted for most of the PAS B repeat, binds the XRE at normal efficiency, although it does not bind ligand (29). Thus the presence of ligand bound to Ahr per se, is not required for dimerization and subsequent XRE binding. Instead, the role of ligand in these processes appears to be to trigger release of HSP90 from Ahr, thereby allowing the latter to dimerize with Arnt. Ahr ⌬B is deficient in dimerization. However, we can make no conclusion as to whether the segment missing from ⌬B is required for dimerization, because the inability of this mutant to dimerize could be fully ascribable to the fact that it retains HSP90 binding. (The results obtained with C⌬516 demonstrate that if the segment missing from ⌬B is in fact required for dimerization, the relevant portion must be between amino acids 259 and 289). Interestingly, the equivalent mutant of Arnt (bHLHA) is only mildly deficient in dimerization. ⌬AB does not bind ligand or HSP90 and does not dimerize with Arnt, indicating that although the PAS A region (and also perhaps the PAS B region) is dispensable for dimerization, the presence of either the PAS A or the PAS B region is required and that both cannot be deleted without eliminating dimerization potential; or alternatively, the region that is deleted in ⌬AB, but not in either ⌬A or ⌬B, is necessary for dimerization. Like Ahr, deletion of the complete PAS domain of Arnt also prevents heterodimerization (16). Poellinger (37) and co-workers reported that a mouse Ahr mutant deleted for most of the PAS region (deleted for amino acids 84 -340 and therefore not identical to our ⌬AB mutant) could dimerize with human Arnt but did not bind the XRE. However, they did not quantitate their results, and importantly did not compare dimerization of the mutant with that of full-length Ahr (37).
Our results therefore indicate that helix 1, helix 2, and the PAS region of Ahr are all involved in dimerization with Arnt. The most plausible model for dimerization is that the various dimerization domains of Ahr associate with their corresponding counterparts in Arnt. All our results are consistent with this model.
Deletion of amino acids 398 -805 in ⌬C did not affect dimerization and reduced XRE binding to 75% of that of full-length Ahr. This contrasts with results of Dolwick and co-workers (29). They found that a mutant containing amino acids 1-492 only possessed 36% of wild-type binding activity. However, they assessed binding using the mouse Ahr mutant in conjunction with human Arnt, whereas our Ahr and Arnt constructs (and the mutant derivatives) were both of mouse origin. Our results are therefore of more biological relevance. The carboxylterminal halves of both Ahr and Arnt have been shown to have transcriptional activation potential when fused to heterologous DNA binding domains. The transcriptional activation domains encompass the glutamine-rich regions of the proteins and, at least in the case of Ahr, includes flanking regions as well. The amino-terminal halves of the proteins are devoid of transcriptional activation potential in these assays. (31,33,39). We found that Ahr ⌬C was inactive in the in vivo CAT assay. However, we previously found that an equivalently deleted mutant of Arnt (bHLHAB) suffered only a 50% reduction of activity when analyzed in a similar fashion (16). Thus the transcriptional activation domain of Ahr appears to play a more important role than that of Arnt with regard to activation of the cyp1a1 gene. Whitelaw and co-workers (31) came to the completely opposite conclusion. Our results are more likely to reflect the true in vivo situation, however, because our reporter plasmid for the CAT assays contained the normal architecture of the cyp1a1 enhancer-promoter, whereas Whitelaw and coworkers' reporter plasmid was an artificial construct, containing a single XRE element linked to the mouse mammary tumor virus promoter. Transcriptional activation of the cyp1a1 gene results from cooperative interaction of Ahr⅐Arnt dimers at several XREs and their coordinated interaction with proteins at the promoter, including components of the general transcription machinery (40). In Whitelaw's construct, the proteins at the heterologous promoter will probably not correspond to those at the cyp1a1 promoter, and transcriptional activation of this construct may occur in an aberrant fashion relative to the normal cyp1a1 promoter.