Role of the Per/Arnt/Sim Domains in Ligand-dependent Transformation of the Aryl Hydrocarbon Receptor*

The aryl hydrocarbon receptor (AhR) mediates the toxic and biological effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin and related compounds. In a process termed transformation, ligand binding converts the AhR into its high affinity DNA binding form that represents a dimer of the AhR and Arnt, a closely related nuclear protein. During transformation, protein chaperone Hsp90 is thought to be replaced by Arnt in overlapping binding sites in the basic helix loop helix and PASB domains of the AhR. Here, analysis of AhR variants containing a modified PASB domain and AhR PASA-PASB fragments of various lengths revealed (i) an inhibitory effect on transformation concomitant with Hsp90 binding in the PASB domain, (ii) an ability of the PASA-PASB fragment of the AhR to reproduce key steps in the transformation process, and (iii) a ligand-dependent conformational change in the PASA domain consistent with increased PASA exposure during AhR transformation. Based on these results, we propose a new mechanism of AhR transformation through initiation of Arnt dimerization and Hsp90 displacement in AhR PASA/B domains. This study provides insights into mechanisms of AhR transformation, dimerization of PAS domain proteins, and Hsp90 dissociation in activation of its client proteins.

The AhR is a ligand-dependent transcription factor that mediates the toxic and biological effects of a diverse spectrum of chemicals including the environmental contaminant 2,3,7,8tetrachlorodibenzo-p-dioxin (dioxin, TCDD) 2 (1,2). The AhR belongs to the family of basic helix loop helix (bHLH) and Per-Arnt-Sim (PAS) domain proteins that also includes mouse HIF1␣, SimI, and CLOCK proteins (3). In its native unliganded state, the AhR is cytosolic and exists in a complex with two molecules of Hsp90 and the proteins XAP2 and p23 (4). In the absence of ligand, Hsp90 binding is thought to maintain the AhR in an inactive state, possibly, through association in the regions that overlap AhR Arnt binding sites (bHLH and PASA/B domains) (5,6). Hsp90 appears to bind to the ligandbinding PASB and N-terminal DNA-binding bHLH domains and following ligand binding, remains associated with the AhR during nuclear translocation (6 -8). Once in the nucleus, Hsp90 dissociates or is more likely displaced by Arnt from their overlapping binding sites in the AhR leading to formation of an active transcriptional dimer of AhR/Arnt (9).
The conversion of the AhR into its high affinity DNA binding form is termed transformation and is presumably driven by a ligand binding-mediated conformational change in the AhR. Although the occurrence of conformational change(s) in the AhR has been suggested by limited proteolysis (10,11), mechanism and functional significance of the presumed conformational change(s) during AhR transformation remain mostly unknown. It has been previously suggested that ligand-dependent nuclear translocation of the AhR complex results from exposure of its nuclear localization sequence (NLS) in the bHLH domain, with the exposed bHLH domain providing a binding surface that allows initiation of Arnt dimerization (12,13). However, this mechanism still remains to be proven experimentally, and certain evidence such as the disparate effects of the Hsp90 stabilizing agent sodium molybdate on AhR transformation and nuclear translocation do not suggest a mechanistic connection between these steps of AhR activation (8).
The AhR contains two repeats of the PAS motif, the PASA and PASB domains. Although the PASA domain is essential for formation of the stable AhR:Arnt dimer, the PASB domain contains a ligand binding site and one of the Hsp90 binding sites (6,14,15). The PASB domain has been suggested to exert an inhibitory effect on AhR transformation and its deletion results in an AhR that is constitutively active with respect to its transformation/DNA binding and transcriptional activation functions (16 -18). Although the ability of Hsp90 to bind to the PASB deleted AhR was not determined, the constitutive activity of this AhR suggests deficient Hsp90 binding and/or exposure of the Arnt dimerization interface(s) on the AhR. Either of these events (loss of Hsp90 binding or exposure of the Arnt dimerization interface) could be initiating AhR transformation, and the PASB domain appears to play central role in ligand dependence of such initiation. Nonetheless, the overall mechanism of the AhR transformation and the role of various AhR domains remain to be elucidated. Accordingly, here we describe a series of studies examining the molecular events involved in the initiation of AhR transformation (loss of Hsp90 and Arnt dimerization). Our results are consistent with a mechanism by which AhR transformation is mediated through ligand-dependent initiation of Arnt dimerization within the AhR PAS domains and further our understanding of the molecular events in AhR signal transduction.

EXPERIMENTAL PROCEDURES
Antibodies-Monoclonal 3G3p90 anti-Hsp90 antibody (6,7) was produced at Antibodies Incorporated (Davis, CA) from hydridoma cells kindly provided by Dr. G. Perdew (Penn State University, PA). Monoclonal M2 anti-FLAG antibody and control murine IgG were purchased from Sigma and AffiniPure goat anti-mouse IgM from Jackson ImmunoResearch. The anti-AhR N-terminal end SE6 antibody was kindly provided by Dr. G. Clark (Xenobiotic Detection Systems) and its specificity toward in vitro expressed AhR and AhR variants, C57BL mouse cytosolic AhR (19), and in vitro expressed Arnt (no recognition) is demonstrated in supplemental Fig. S1.
In Vitro Protein Expression and Ligand-dependent Transformation-Wild type mAhR, mArnt, mAhR variants, GST fusion constructs, and GST were synthesized in vitro in the presence of L-[ 35 S]methionine (MP Biomedicals) or unlabeled L-methionine using the TNT Quick-coupled transcription/ translation rabbit reticulocyte lysate kit (Promega). The resulting AhR and Arnt translation mixtures were mixed in a 1:1 (v/v) ratio for DNA binding and dimerization experiments and in a 1:3 ratio for Hsp90 dissociation studies, and incubated with 20 nM TCDD or 2% (v/v) DMSO (the solvent control) for 1-2 h at room temperature. For FLAG co-immunoprecipitations, the in vitro synthesized 35 S-labeled AhR variants were transformed in the presence of Arnt-FLAG protein instead of Arnt. For gel retardation experiments, 3 l of AhR:Arnt translation mixture was diluted to 10 l with MEDG buffer (25 mM MOPS, pH 7.5, 10% (v/v) glycerol, 15 mM KCl, 1 mM EDTA, 1 mM dithiothreitol) and incubated in the presence of 20 nM TCDD or 2% (v/v) DMSO for 2 h at room temperature.
Co-immunoprecipitation Assays-Goat anti-mouse IgM was conjugated to Affi-Gel 10 (Bio-Rad) as described by the manufacturer and incubated with 5% (w/v) bovine serum albumin in MEDG buffer for 1 h at 4°C. Aliquots of the conjugated matrix were incubated with excess 3G3 antibody (2-3 g/reaction) for 1 h at 4°C with shaking and washed once with MEDGN (MEDG, 0.1% Nonidet P-40) buffer. Ten l of transformation mixture (or 35 S-labeled protein mixture) was incubated with a total of 10 l of the prepared 3G3-bound matrix for 1 h at 4°C with shaking. The beads were washed 3 times with 1 ml of MEDGN (supplemented with 20 mM sodium molybdate where indicated), analyzed by SDS-PAGE as previously described (22), and resolved by Phosphorimaging (GE Healthcare or Fujifilm). In FLAG co-immunoprecipitations, the transformation mixtures were incubated with 1 g of the anti-FLAG antibody or IgG control for 1 h at 4°C and then with 10 l of Protein G Plus-agarose (Santa Cruz) (pre-washed with 2% (w/v) bovine serum albumin in PBS) for 1 h at 4°C with shaking. The samples were washed 3 times with PBS and proteins were resolved by SDS-PAGE as previously described (22).
GST Pull-down and Limited Proteolysis Assays-Following AhR transformation reactions, 35 S-labeled GST-AhR1-3 and mutated GST-AhR3 constructs were incubated in the presence of 20 nM TCDD, 2% (v/v) DMSO for 1 h at room temperature, and incubated with 10 l of prewashed (with PBS) glutathioneagarose (Sigma) for 30 min at room temperature with shaking. Then, the samples were washed once with PBS, incubated with 1 g/ml trypsin for the indicated periods of time at room temperature, washed 3 times with PBS, and analyzed by SDS-PAGE as previously described (22).
Hydroxyapatite Ligand Binding Assays-[ 3 H]TCDD (13 Ci/mmol) was obtained from Dr. Safe (Texas A&M University). [ 3 H]TCDD specific binding to the in vitro synthesized proteins diluted in MEDG buffer to 8 mg/ml protein was conducted in the presence of 2 nM [ 3 H]TCDD for the indicated periods of time and measured following the previously published protocol for cytosolic AhR (22). Equivalent amounts of unprogrammed in vitro synthesized reactions were used as a nonspecific binding control as previously described (23).
Gel Retardation Assays-Annealed double-stranded oligonucleotides containing the AhR:Arnt DNA binding site (DRE3) from the murine CYP1A1 upstream regulatory sequence were 32 P-labeled and gel retardation analysis was conducted as detailed previously (22,24).
Western Blot Analysis-SDS-PAGE and the electrophoretic transfer of the protein samples were performed as previously described (22). The primary anti-AhR antibody SE-6 was added at 0.2 mg/ml and detection performed using the Amersham Biosciences ECL Western blotting analysis system (GE Healthcare).

RESULTS
Hsp90 Binding to the bHLH and PASB Sites of the AhR-It has been previously suggested that Hsp90 maintains the AhR in an inactive state through its binding to at least two separate regions, the bHLH and PASB domains, although the exact sequences involved remain to be determined (6,7,25). Deletion of the ligand binding PASB domain (amino acids 287-422) generates an active (ligand-independent) AhR that constitutively dimerizes with Arnt, binds to DNA, is capable of nuclear translocation, and activates gene transcription (17) despite the fact that it retains one Hsp90 binding site (in the bHLH domain). These observations raise the question whether the Hsp90-mediated inhibitory effect on AhR transformation is binding site dependent (e.g. is mediated by Hsp90 bound to the PASB domain). However, Hsp90 binding to this PASB-deleted AhR (AhR⌬PASB) was not previously determined. To analyze Hsp90 binding to this deletion variant, an identical AhR⌬PASB expression construct (Fig. 1a) was generated and the AhR⌬PASB protein synthesized in vitro. Although the level of protein expression of the AhR⌬PASB protein was comparable with that of the wild type AhR (wtAhR, compare input lanes in Fig. 1b), we detected little Hsp90 binding to AhR⌬PASB (5.6% of input compared with 54% of input for the wtAhR (Fig. 1b)). The observed deficient Hsp90 binding to the AhR⌬PASB likely explains the previously observed constitutive activity of this construct in AhR transformation and nuclear translocation (17).
Deletion of the PASB domain was sufficient to destabilize Hsp90 binding to the bHLH domain of the AhR⌬PASB protein, although it has been previously reported that C-terminal deletion of the AhR containing the bHLH but not the PASB domains could bind Hsp90 (6,7). This observation suggested that a change in the overall structure of the AhR resulting from the internal deletion may destabilize Hsp90 binding in the bHLH domain of AhR⌬PASB. To minimize destabilizing structural effects of internal deletion, we generated a chimeric AhR, in which the AhR PASB domain (amino acids 263-409) was substituted with the comparable PASB domain of Arnt (amino acids 326 -470) (Fig. 1a). The exchanged region in this AhR/ PASB-Arnt chimera was slightly different from that of the original deletion in the AhR⌬PASB to allow swapping of structurally homologous regions between the two proteins. Interestingly, whereas Arnt does not bind Hsp90 (7) and the AhR⌬PASB is deficient in this binding, insertion of the Arnt PASB domain partially restored Hsp90 binding to the chimeric AhR as compared with the AhR⌬PASB (compare 29% of input for AhR/PASB-Arnt to 5.6% of input for AhR⌬PASB and 54% of input for wtAhR, Fig. 1b). These results suggest that some loss in Hsp90 binding in AhR⌬PASB is likely due to a change in the overall structure of the AhR and the remainder due to loss of the Hsp90 binding within PASB. We assume that the restored Hsp90 binding occurs within the AhR bHLH domain binding site of AhR/PASB-Arnt because Arnt (and presumably the Arnt PASB) does not bind Hsp90 (7).
Although removal/reduction of Hsp90 inhibitory association may explain previously reported ligand-independent transformation and nuclear translocation by the AhR⌬PASB (17), it was of interest to analyze these activation steps for AhR/PASB-Arnt because this AhR variant retained some Hsp90 binding. AhR transformation of the wtAhR, AhR⌬PASB, and the AhR/PASB-Arnt chimeras was examined indirectly by measuring AhR DNA binding using gel retardation analysis (22). As expected, a ligand-dependent increase in complex formation was only observed with the wtAhR, whereas AhR⌬PASB exhibited constitutive DNA binding activity (Fig. 2a). The significantly higher amount of DNA binding by the AhR⌬PASB (Fig. 2, a and  b) is consistent with decreased association with Hsp90 in this AhR variant. Despite retaining some Hsp90 binding, the AhR/ PASB-Arnt construct also demonstrated ligand-independent dimerization and DNA binding (Fig. 2a) suggesting that the presumed lack of Hsp90 binding to the PASB domain of this chimera is sufficient to allow constitutive Arnt dimerization, AhR transformation, and DNA binding. Interestingly, the levels of constitutive DNA binding by AhR/PASB-Arnt were similar to that of wtAhR in the presence of TCDD (Fig. 2a). Moreover, the similarity in kinetics of the time course of transformation/ DNA binding of these two proteins (Fig. 2b) suggests similarities in the mechanism of dimerization/DNA binding between b FIGURE 1. Hsp90 binding to wtAhR and AhR variants containing a modified PASB domain. a, wild type (wt) and variant AhRs used in these studies. Protein sequences are not to scale. b, Hsp90 binding to wtAhR, AhR⌬PASB (⌬PASB), and AhR/PASB-Arnt (PASB-Arnt). 35 S-Labeled wtAhR and AhR variants were co-immunoprecipitated with anti-Hsp90 antibody (3G3). The washing buffer contained 20 mM sodium molybdate. Input aliquots (10% input) and co-immunoprecipitated proteins were analyzed by SDS-PAGE, visualized, and quantitated by Phosphorimager analysis. Positions of protein markers are indicated. The results are representative of three independent experiments. the ligand-bound wtAhR and AhR/PASB-Arnt. These findings are consistent with the previously suggested inhibitory role of the AhR PASB domain in AhR transformation (16,18) and suggest that although the AhR PASB is important in ligand-dependent control of transformation, it contributes little to maximum levels and kinetics of AhR:Arnt dimerization and DNA binding.
To analyze cellular localization of the constructs of interest, wtAhR, AhR⌬PASB, and AhR/PASB-Arnt were tagged at the C-terminal end with yellow fluorescent protein (YFP). The resulting constructs demonstrated unchanged levels of protein expression, Hsp90 binding, and DNA binding compared with their untagged counterparts (supplemental Fig. S3). COS-1 cells were transiently transfected with YPF-tagged AhR constructs, incubated in the presence of TCDD (or solvent control DMSO) for 1-2 h, and AhR localization determined by fluorescence microscopy. Cells transfected with the control vector pcDNA3 did not produce a fluorescent signal (data not shown). Although incubation of the transiently transfected COS-1 cells with TCDD resulted in a shift of fluorescent signal of the wtAhR-YFP from cytosolic and cytosolic/nuclear localization to predominantly nuclear localization (indicating nuclear import of the AhR complex), AhR⌬PASB-YFP fluorescence was primarily localized within the nuclei independent of ligand, as expected (Fig. 3a). In contrast, (AhR/PASB-Arnt)-YFP exhibited predominantly cytosolic and cytosolic/nuclear localization (Fig. 3a) and it was expressed at levels comparable with that of AhR⌬PASB as determined by Western blotting assay (Fig. 3b). It is unclear why AhR/PASB-Arnt was less efficient in nuclear translocation than AhR⌬PASB or wtAhR in the presence of TCDD, although presumed Hsp90 binding to the bHLH domain of this AhR chimera may mask the NLS. Nonetheless, it is significant that the same AhR protein was fully active in transformation in vitro (as evidenced by DNA binding, Fig. 2) and yet inactive in nuclear translocation. Although the previously proposed mechanism of initiation of AhR transformation through the bHLH domain relied on presumed ligand-dependent exposure of this domain (which contains the NLS), the example of the AhR/PASB-Arnt demonstrates that the AhR protein may not need to be able to translocate into the nucleus to be active in AhR transformation (conversion into its DNA binding form).
Role of the PASA and PASB Domains in the AhR Transformation-Constitutive transformation of the AhR/ PASB-Arnt in vitro suggests that this process may be initiated through AhR PAS domains because exposure of the bHLH domain of this chimeric AhR (where the NLS is located) appears to be restricted due to its lack of constitutive nuclear localization in transfected cells (Fig. 3a). Because in the inactive state, Hsp90 is associated with the Arnt dimerization interface in the AhR, the hypothetical mechanism of initiation of AhR transformation through the PASB may involve ligand-dependent dissociation of Hsp90 from the binding site in this domain. To test this possibility, we examined various steps of AhR transformation (ligand binding, Hsp90 displacement and dimerization with Arnt) using AhR-GST fusions containing various lengths of AhR PASA and PASB domains as well as GST fulllength AhR. A GST tag was added to stabilize and rapidly isolate the expressed protein fragments for analysis. GST-AhR 1, 2, and 3 span amino acids 230 -421, 167-421, and 83-421 of the wtAhR, respectively (Fig. 4a), and were each expressed in vitro at levels comparable with that of wtAhR and free GST (Fig. 4b and Table 1), although protein expression levels of GST-AhR-FL were significantly lower. Co-immunoprecipitation experiments revealed that all constructs exhibited high levels of Hsp90 binding (37-45% of input), comparable with that of wtAhR (46% of input, Fig. 4b), whereas the GST negative control protein alone did not bind Hsp90 (Fig. 4b). Thus, whereas GST-AhR-FL was expressed at much lower levels, it had similar Hsp90 binding suggesting that its protein stability was not due to deficient association with Hsp90. However, we chose to not perform functional studies on this construct due to the low levels of expression and functional activity in transformation studies (data not shown). In contrast, we observed that the GST-AhR 1-3 constructs were fully active in AhR transforma- tion (as detected by dimerization with Arnt and Hsp90 displacement, Fig. 5). Thus, whereas these protein fragments differ from the full-length AhR and have an added tag, their ability to undergo transformation suggested that activities of the PASA and PASB domains in the AhR transformation are retained in isolation from the other AhR domains.
[ 3 H]TCDD specific binding to GST-AhR 1-3 was determined to be 34 Ϯ 6, 39 Ϯ 4, and 47 Ϯ 5%, respectively, of that of wtAhR (Table 1) demonstrating that each GST-AhR fusion retains the ability to bind ligand, albeit less than that of a comparable amount of in vitro synthesized wtAhR. A comparable reduction in ligand binding has been previously reported for an AhR fragment spanning the same region as GST-AhR1 (15). Whether this reduction in [ 3 H]TCDD specific binding results from a reduced binding affinity and/or capacity remains to be determined.
The small difference in Hsp90 binding to the wtAhR in experiments demonstrated in Figs. 1b and 4b (54 versus 46% of input) could be attributed to the presence of 20 mM sodium molybdate in the washing buffer. We did not include this AhR: Hsp90 binding stabilizing agent in the washing buffer in the latter experiment because at the next step, the process of liganddependent Hsp90 dissociation was examined using this co-immunoprecipitation technique.
Addition of ligand and/or Arnt did not alter levels of Hsp90 binding to GST-AhR1 and GST-AhR2 as determined by coimmunoprecipitation with anti-Hsp90 antibody (Fig. 5a) suggesting that ligand binding to the PASB domain of this AhR fragment (Table 1) does not result in Hsp90 dissociation from its binding site(s) within this domain. However, with GST-AhR3, which includes the PASA and PASB domains, Hsp90 binding was reduced in the presence of both ligand and Arnt to a level comparable with that of wtAhR (Fig. 5b). Hsp90 binding b

FIGURE 3. AhR/PASB-Arnt does not translocate into the nucleus in transient transfections in COS-1 cells. a, wtAhR-YFP, AhR⌬PASB-YFP (⌬PASB), or (AhR/PASB-Arnt)-YPF (PASB-Arnt) were transfected in COS-1 cells, incubated
in the presence of 10 nM TCDD (ϩ) or 0.1% (v/v) DMSO control (Ϫ) for 2 h, and YFP fluorescence was detected using an Olympus IX71 fluorescent microscope. Images were colorized using Wasabi software. In each incubation, 300 cells were examined for cytosolic (and cytosolic/nuclear) or predominantly nuclear localization of the AhR fluorescent signal, and the resulting percentages are shown. b, COS-1 cells were transfected with the indicated plasmid and incubated with TCDD as in panel a, lysed, and analyzed for protein expression by Western blotting assay using anti-AhR SE6 antibody. A gel representative of three independent experiments is shown. The arrows indicate the positions of the wtAhR and variants.   NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47

Role of PAS Domains in AhR Transformation
to either wtAhR or GST-AhR3 in the presence of TCDD and Arnt (ϩ/ϩ) was statistically lower than that in the presence of any other TCDD/Arnt combination (Fig. 5b). Thus, whereas ligand binding did not result in Hsp90 dissociation from the PASB domain, the presence of the PASA domain was sufficient to initiate Hsp90 dissociation from the PASA-PASB fragment, but only in the presence of TCDD and Arnt. Co-immunoprecipitation experiments with anti-FLAG antibody revealed that ligand-dependent dimerization of GST-AhR3 with Arnt-FLAG was comparable with that of wtAhR, whereas GST-AhR1 and GST-AhR2 demonstrated significantly lower levels of dimerization, which was, however, liganddependent (Fig. 5c). This finding suggests that although the PASB domain by itself is capable of initiating dimerization with Arnt, it cannot support the full measure of dimer formation.
The observed correlation of Arnt dimerization and Arnt-dependent Hsp90 dissociation is consistent with the previous sug-gestion of Hsp90 displacement by Arnt during AhR transformation (6,13,26,27) and, being contingent on the presence of the PASA, implicates this domain in stabilizing the AhR:Arnt dimer in transformation. Taken together, these data suggest a mechanism of ligand-dependent transformation of the AhR PASA-PASB fragment, in which Hsp90 is displaced by Arnt from the ligand-bound PASB domain binding site due to stabilization of the AhR/Arnt interaction by the AhR PASA domain.
Although GST-AhR1 exhibited ligand-dependent dimerization with Arnt (at approximately half of that observed with GST-AhR3 (Fig. 5c)), there was no detectable Hsp90 dissociation in contrast to that observed with GST-AhR3 or wtAhR (Fig. 5b). It is possible that a lower level of dimerization of this construct with Arnt masks corresponding changes in its association with Hsp90 or that an alternative dimerization interface is used for Arnt dimer formation in this AhR PAS domain fragment.
In contrast to a previous study (6), we did not observe a significant decrease in bound Hsp90 following ligand binding to the wtAhR in the absence of Arnt (Fig. 5b), although a small decrease in bound Hsp90 was observed for the wtAhR and GST-AhR3 (which contains PASA and PASB domains) in the presence of ligand (compare ϩ/Ϫ to ϩ/ϩ reactions in Fig. 5b). Although we hypothesized that following ligand binding Hsp90 association is weakened with the wtAhR (or the PASA-PASB fragment), a higher stringency washing buffer did not result in ligand-dependent differences in binding in the absence of Arnt (data not shown). It is unlikely that Hsp90 fully dissociates from the AhR simply as a result of ligand binding and this is consistent with previous findings that Hsp90 remained bound to the AhR complex through ligand-induced nuclear translocation, and that the ligand-bound nuclear AhR complex contained Hsp90 (8,28). It is possible that ligand binding results in weakened AhR/Hsp90 interactions, however, this remains to be confirmed.
Ligand Binding Results in a Conformational Change in the AhR PASA Domain-Previously, a conformational change in the AhR was reported using limited proteolysis (10). In these experiments, a protected AhR fragment was only formed in the presence of both Arnt and TCDD, which approximately spanned the AhR Arnt dimerization interface (bHLH through PASB domains) (10). Thus the apparent conformational change(s) could have represented a difference between the AhR-Hsp90 and AhR-Arnt complexes. Seeking to demonstrate a conformational change in the absence of Arnt, we modified this approach and performed limited proteolysis on 35 S-labeled GST-AhR3 bound to glutathione beads. The assumption of this strategy was that first, a smaller length of the analyzed fragment would generate fewer proteolytic fragments and second, only proteolytic fragments containing the GST moiety would be isolated (on glutathione beads) following limited proteolysis. Such an approach would allow approximate mapping of the cleavage site(s).
Glutathione precipitation/limited proteolysis of 35 S-labeled GST-AhR3 revealed a fragment of ϳ43 kDa, which was moderately increased in the presence of ligand (Fig. 6a). Assuming the GST moiety at the N-terminal end of this fragment, the cleavage site was tentatively mapped to amino acid 225 of AhR (the GST-AhR3 contains amino acids 83-419 of the AhR). There is a close clustering of trypsin cleavage sites in this arginine-cysteine-rich region of the AhR with arginine residues at 212, 215, 217, and 219. To confirm the size of the fragment, a truncated GST-AhR3 was generated spanning amino acids 83-209 of the AhR (truncated right before the RC-rich region). The resulting truncated GST-AhR3 fragment was indeed very close to the detected 43-kDa proteolytic fragment on the SDS gel (Fig. 6b). We subjected this construct to trypsin degradation for clear comparison in size between the 43-kDa proteolytic fragment and the undigested truncated GST-AhR3 (Fig. 6b). Thus, the cleavage site is located close to amino acid 209 of AhR, or within the PASA domain according to the domain alignment of the bHLH-PAS proteins from a recent study (29).
Quantitation of the limited proteolysis results is presented in Fig. 6c and was carried out as follows. The specific bands (of the GST-AhR3 and 43-kDa proteolytic fragment) from the DMSO and TCDD reactions were quantitated using the phosphorimager technique, and normalized to corresponding GST-AhR3 intensities from DMSO and TCDD-treated "no trypsin" control reactions (to correct for differences in the input of [ 35 S]GST-AhR3). In addition, the percentages of the 43-kDa band were adjusted using a factor of 16/10, which reflects the fact that this fragment would contain 10 methionines (when its C terminus is mapped around AhR amino acid 209) compared with 16 methionines in the GST-AhR3 fragment.
Comparison of the digestion time profiles revealed that the amount of the 43-kDa fragment was ϳ2-fold higher in the presence of TCDD as compared with control (DMSO) reactions at all incubation times (Fig. 6c). The ligand-dependent difference in formation of the 43-kDa fragment can result from either higher accumulation in the presence of TCDD or faster degradation in the presence of DMSO, or a combination of both. At 5 and 10 min the amount of undigested GST-AhR3 is lower than that of the 43-kDa fragment indicating higher stability of the 43-kDa fragment compared with the undigested form. Moreover, the amounts of the 43-kDa fragment in the DMSO and TCDD reactions decreased proportionally (Fig. 6c) suggesting that the 43-kDa fragment has similar degradation rates in the DMSO and TCDD reactions. Hence, the ligand-dependent difference in its formation could be due to differences in accumulation rather than degradation of the fragment.
Given that no higher molecular mass product(s) was present in amounts close to that of the 43-kDa fragment at 2 min (about 25% of bound input, Fig. 6c), the 43-kDa fragment and GST appear to be the primary products of proteolytic digestion suggesting that the 43-kDa fragment is formed directly from GST-AhR3. The increase in ligand-dependent accumulation of the 43-kDa fragment indicates that ligand binding results in increased cleavage of the GST-AhR3 in the PASA domain. Such preferential cleavage would be consistent with increased liganddependent exposure of the PASA domain.
Next, we attempted to identify the specific residue(s) responsible for producing the 43-kDa fragment. Predicted trypsin cleavage sites spanning amino acids 106 to 354 were mutated to alanine, and the resulting GST-AhR3 variants were subjected to limited proteolysis (Fig. 7). Although no single mutation resulted in complete disappearance of the 43-kDa band, mutation of arginine 217 to alanine (R217A) resulted in a substantial reduction of the 43-kDa band (Fig. 7). Moreover, this cleavage site was close to the predicted position (amino acids 209 -225). For these reasons, further analysis involving this mutation was performed.
The lack of complete disappearance of the 43-kDa protein band by a single mutation could suggest that this protein band actually results from cleavage at multiple trypsin sites that are close to each other. In fact, there are four putative tryptic sites in the RC-rich stretch where Arg 217 is located. The apparent slightly shorter size of the band corresponding to the 43-kDa fragment with the R217A mutant (Fig. 8c,  the corresponding peak iii in the wt GST-AhR3 and its R217A mutation) suggested that cleavage sites at Arg 212 and Arg 215 could be alternatively utilized. Accordingly, we generated double mutations R212A/R217A and R215A/R217A and analyzed proteolytic patterns of these GST-AhR3 constructs as well as those of wt GST-AhR3 single mutations R212A, R215A, and R217A (Fig. 8).
Proteolytic cleavage of single mutations R212A and R215A revealed proteolytic patterns that were very close to that of the wild type GST-AhR3 (Fig. 8b). Among the minor differences, fragments ii and v were present in these mutations and not in the wt GST-AhR3. All mutations containing R217A demonstrated proteolytic patterns dramatically different from the wt GST-AhR3 and similar among each other. Specifically, fragments i, ii, and v were increased, and iii (which corresponds to the 43-kDa complex) and iv were decreased compared with the wt GST-AhR3 (Fig. 8c). Moreover, fragments i-iv appeared slightly shifted toward smaller molecular weight and this was more apparent on longer proteins gels (data not shown). Decrease in the 43-kDa proteolytic fragment and the concomitant increase of proteolytic fragments at surrounding cleavage sites suggest that Arg 217 is involved in cleavage either as one of the cleavage sites, or directly affecting conformation of the cleavage site. Because none of the mutations of predicted trypsin sites nearby resulted in the loss of the 43-kDa fragment, we hypothesize that the cleavage event generating this fragment may involve low specificity trypsin activity. Consistent with this idea, preparations of trypsin incubated at room temperature, which potentially results in autoproteolysis and generation of pseudotrypsin, demonstrated improved cleavage of the Arg 217 mutation (data not shown). There are multiple hydrophobic residues in and around the RC-rich region that could be involved, and future analysis will target these residues.

DISCUSSION
Ligand-dependent AhR transformation is a complex event and is thought to be triggered by a ligand binding-driven conformational change(s) in the AhR protein. Previous studies have demonstrated that inactive unliganded AhR is associated with Hsp90, and that during transformation ligand-bound AhR dimerizes with Arnt in the nucleus, releasing Hsp90 at some step prior to dimer formation (9,30). Because Hsp90 and Arnt have overlapping binding sites in the AhR (in the bHLH and PASB domains) and do not appear to bind to the AhR at the same time, it was hypothesized that Hsp90 is displaced by Arnt during transformation and that in the absence of ligand Hsp90 maintains the AhR in a dimerization-incompetent state by masking the Arnt dimerization interface(s) on the receptor (6,13,25,27). In agreement with this hypothesis, we found that AhR⌬PASB, which is constitutively active in transformation, was deficient in Hsp90 binding. Moreover, AhR/PASB-Arnt was constitutively active in transformation/DNA binding while retaining ϳ50% of Hsp90 binding compared with wtAhR indicating that loss of Hsp90 association specifically in the PASB domain of the AhR is likely sufficient for constitutive activation of transformation. Interestingly, loss of some Hsp90 binding to the AhR/PASB-Arnt protein did not appear to affect the kinetics of AhR transformation that was similar to that of wtAhR in the presence of TCDD (Fig. 2b) suggesting similarities in the mechanisms of transformation/DNA binding between these two proteins. Further analysis of transformation events in the PASA-PASB fragments of the AhR revealed that this region is sufficient to reproduce transformation events and that following ligand binding, a conformational change can be detected in the PASA that is consistent with increased exposure of this domain. Such increased exposure of the PASA domain could be coincidental with an increase in dimerization potential of the PASA-PASB fragment following ligand binding and would explain the observed stabilizing role of the PASA on AhR:Arnt dimerization resulting in more efficient Hsp90 displacement. Together, these data are consistent with a mechanism of AhR transformation that is initiated through the PAS domains.
In contrast, the previously proposed mechanism suggested that initiation of AhR transformation occurred at the bHLH domain (12,13). This mechanism was based on predicted exposure of the NLS contained within the bHLH domain following ligand binding due to the observed nuclear import of the AhR complex (12). However, a more recent study suggested that nuclear import of the AhR complex may depend on a change in the NLS conformation rather than its exposure and that XAP2 could be involved in this process (31). Moreover, the prediction of ligand-dependent dissociation of Hsp90 from the bHLH domain (to account for presumed NLS exposure) was also questioned by the findings that the Hsp90 binding stabilizing agent sodium molybdate did not inhibit nuclear translocation of the AhR complex (8), and by the observation that deletion of the C-terminal transactivation domain of the AhR, which does not affect Hsp90 binding, resulted in constitutive ligand-independent AhR nuclear localization (32,33).
In this study, we found that the chimeric AhR variant, AhR/ PASB-Arnt, was capable of transformation but did not translo- cate into the nucleus (Fig. 3). These results further support the idea of a certain degree of independence between the effect of ligand binding on the transformation potential of the AhR and its ability to undergo nuclear localization. We suggest that the inability of the AhR/PASB-Arnt to undergo nuclear translocation is due to the loss of the ability of ligand to affect Hsp90 bound in the bHLH region. Although ligand binding should clearly affect conformation or exposure of the wtAhR NLS, the mechanism of this effect remains unclear. However, together with previous studies, our data suggest that the bHLH exposure is not responsible for initiation of dissociation of Hsp90 and dimerization with Arnt in AhR transformation.
Our studies also rule out another possible mechanism of initiation of transformation, namely through ligand-dependent a b c FIGURE 8. R217A mutation results in an altered proteolytic pattern. a, 35 S-labeled GST-AhR3 and its indicated point mutation variants were incubated in the presence of TCDD and analyzed by pull-down limited proteolysis analysis as described in the legend to Fig. 7. The gel was visualized and quantitated by Phosphorimager analysis, and density lanes profiles are presented (b and c)). Input controls (10% input) are shown. The 43-kDa fragment is indicated with an arrow. Results are representative of three independent experiments. b, density lane profiles for the wild type GST-AhR3 (blue), R212A (green), and R215A (cyan) variants. PSL, photostimulated luminescence units. c, density lane profiles for the wild type GST-AhR3 (blue), R217A (pink), R212A/R217A (peach), and R215A/ R217A (orange) variants.