Characterization of a subset of the basic-helix-loop-helix-PAS superfamily that interacts with components of the dioxin signaling pathway.

In an effort to better understand the mechanism of toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin, we employed an iterative search of human expressed sequence tags to identify novel basic-helix-loop-helix-PAS (bHLH-PAS) proteins that interact with either the Ah receptor (AHR) or the Ah receptor nuclear translocator (ARNT). We characterized five new “embers f the AS superfamily,” or MOPs 1-5, that are similar in size and structural organization to the AHR and ARNT. MOPs 1-4 have N-terminal bHLH and PAS domains and C-terminal variable regions. MOP5 contained the characteristic PAS domain and a variable C terminus; it is possible that the cDNA contains a bHLH domain, but the entire open reading frame has yet to be completed. Coimmunoprecipitation studies, yeast two-hybrid analysis, and transient transfection experiments demonstrated that MOP1 and MOP2 dimerize with ARNT and that these complexes are transcriptionally active at defined DNA enhancer sequences in vivo. MOP3 was found to associate with the AHR in vitro but not in vivo. This observation, coupled with the fact that MOP3 formed tighter associations with the 90-kDa heat shock protein than the human AHR, suggests that MOP3 may be a conditionally active bHLH-PAS protein that requires activation by an unknown ligand. The expression profiles of the AHR, MOP1, and MOP2 mRNAs, coupled with the observation that they all share ARNT as a common dimeric partner, suggests that the cellular pathways mediated by MOP1 and MOP2 may influence or respond to the dioxin signaling pathway.

The AHR, 1 ARNT, SIM, and PER are the founding members of an emerging superfamily of regulatory proteins (1)(2)(3)(4). The AHR and ARNT are dimeric partners that transcriptionally up-regulate genes involved in the metabolism of xenobiotic compounds. The AHR is activated by a number of widespread environmental pollutants, such as the prototypical agonist, TCDD. In the absence of ligand, the AHR is primarily cytosolic and functionally repressed, presumably as the result of its tight association with HSP90 (5). Current models suggest that agonist binding initiates translocation of the receptor complex to the nucleus and concomitantly weakens the AHR⅐HSP90 association. Within the nucleus, HSP90 is displaced, and the AHR dimerizes with its partner ARNT, resulting in a bHLH-PAS dimer with binding specificity for enhancer elements upstream of gene products that metabolize foreign chemicals (6). In Drosophila, SIM is the master regulator of midline cell lineage in the embryonic nervous system (7). Genetic, in vitro, and in vivo studies suggest that SIM may also dimerize with an ARNT-like protein in Drosophila and regulate enhancer sequences present in the sim, slit, and Toll structural genes (7)(8)(9)(10). The Drosophila PER protein plays a role in the maintenance of circadian rhythms. PER has been shown to form heterotypic interactions with a second Drosophila protein TIM in vivo, and homotypic interactions with the ARNT molecule in vitro (11,12).
The distinguishing characteristic of these proteins is a 200 -300 stretch of amino acid sequence similarity known as the PAS domain. In the AHR, the PAS domain has been shown to encode sites for agonist binding, surfaces to support dimerization with other PAS domains, as well as surfaces that form tight interactions with HSP90 (1). In addition to the PAS domain, the AHR, ARNT, and SIM also harbor a bHLH motif that plays a primary role in dimer formation. The bHLH motif is found in a variety of transcription factors that utilize homotypic interactions to dimerize and regulate various aspects of cell growth and differentiation (1,(13)(14)(15). Dimerization specificity is conferred by sequences within both the bHLH and determinants within secondary interaction surfaces, such as the "leucine zipper" or "PAS" domains (1,12,16). Interestingly, these dimerization surfaces also appear to restrict pairing to within a given bHLH protein superfamily, thus minimizing cross-talk between important cellular pathways (17).
At the time we began this work, the AHR and ARNT were the only mammalian bHLH-PAS proteins that had been identified (see "Discussion"). Because other bHLH protein families utilize multiple homotypic interactions to provide fine control in the regulation of various gene batteries, we predicted that additional bHLH-PAS proteins existed in the mammalian genome and that a subset of these proteins would dimerize with either the AHR or ARNT. We propose that identification of such partners and determination of their pairing rules are the first steps in characterizing the potential points of cross-talk between different bHLH-PAS-mediated signaling pathways, as well as understanding the pleiotropic responses initiated by potent AHR agonists like TCDD. These observations led us to initiate a search for additional family members from libraries of human ESTs.

EXPERIMENTAL PROCEDURES
Search Strategy-The bHLH-PAS domains of the huAHR, huARNT, drSIM, and the PAS domain of drPER were used as query sequences in BLASTN searches of the GenBank™ data base between December of 1994 and October of 1995, using the following default values: data base ϭ NR (NR, non-redundant subset), expect ϭ 10, word length ϭ 12 (18). Preliminary experiments comparing AHR and PER led us to define candidate ESTs as those "hits" that yielded scores of 150 or higher. As a method to confirm the similarity of these EST sequences to known bHLH-PAS proteins, each candidate EST subsequently was compared with the NR subset of GenBank™ using the BLASTX program, matrix ϭ blosum 62, word length ϭ 3. Only ESTs that retrieved known bHLH-PAS proteins by this method of confirmation were further characterized.
Oligonucleotide Sequences-Sequences of oligonucleotides are given below. In cases where the oligonucleotide was used in gel shift assays, the 6-bp target sequence is underlined.
Cloning Strategy-In an effort to obtain extended open reading frames for each EST, an anchored-PCR strategy was employed to amplify additional flanking sequence from a variety of commercial cDNA libraries that were constructed in the phagemid Lambda Zap (tissues: HepG2, fetal brain, and skeletal muscle) (Stratagene, La Jolla, CA) ( Table I) (19). The resulting PCR products were subjected to agarose gel electrophoresis, transferred to a nylon membrane, and analyzed by hybridization with a 32 P-labeled probe generated from the corresponding parent EST plasmid (Table I). After autoradiography, the positive PCR products were purified by gel electrophoresis and cloned using the pGEM-T vector system (Promega, Madison, WI). Dideoxy sequencing was performed to characterize each positive clone (20).
Plasmid Construction for Expression in Vivo-Sequence information from each EST was used to design PCR primers for the amplification of cDNA from commercially available libraries. Expression plasmids were constructed by standard protocols (21). For a summary of clone designations, PCR primers, DNA templates, and GenBank™ accession numbers, refer to Table I. A brief description follows.
MOP1 Expression Vectors-Oligonucleotides OL404 and OL365 were used as primers in a PCR to amplify a 970-bp fragment from a HepG2 cell cDNA library. This fragment was cloned into the pGEM-T vector in the T7 orientation and designated PL439. To generate pGMOP1, the SalI/XhoI fragment of hbc025 was subcloned into SalI-digested PL439. To increase transcription efficiency of the MOP1 cDNA, pGMOP1 was digested with KpnI and SacI and this fragment subcloned into the corresponding sites of pSputk generating PL415 (Stratagene, La Jolla, CA) (22). The complete ORF of the MOP1 cDNA was amplified using the PCR and oligonucleotides OL425 and OL536. This fragment was digested with BamHI and ligated into the BamHI site in the pSport polylinker (Life Technologies, Inc.). This plasmid was designated PL611.
MOP2 Expression Vectors-The PCR was employed using OL477 and OL450 to amplify a 931-bp MOP2 fragment from a HepG2 cDNA library. This fragment was cloned into pGEM-T in the SP6 orientation and designated PL424. Using OL560 and OL590, PCR amplification from this same library yielded a 3Ј fragment of the MOP2 cDNA. This fragment was cloned into pGEM-T in the SP6 orientation and was designated PL445. PL424 was digested with SalI and EcoRI and the fragment ligated into a SalI/EcoRI-digested PL445 to generate a full ORF MOP2 expression vector designated PL447. The complete ORF of the MOP2 cDNA was cloned into pSport as follows; PL447 was digested with SacII, treated with the Klenow fragment of DNA polymerase I in the presence of dNTPs, and subsequently digested with SalI. This fragment was purified and ligated into pSport, digested with HindIII, repaired with Klenow, then digested with SalI. This construct was designated PL477.
MOP3 Expression Vectors-Using the primers OL145 and OL489 and a human fetal brain cDNA library as template, the PCR was used to obtain a 1380-bp fragment. This fragment was isolated and cloned into pGEM-T as above, and this plasmid designated PL487. A fragment of MOP3 was obtained by the PCR using Pfu polymerase (Stratagene), primers OL657 and OL689, and PL487 as template. To obtain a fulllength MOP3 cDNA fragment, the megaprimer fragment obtained above was used in the PCR against oligonucleotide OL611 using IM-AGE clone 50519 as a template (23). This product was cloned into pGEM-T in the SP6 orientation as above and designated PL425.
MOP4 Expression Vectors-Using primers OL520 and OL145 and a HepG2 cDNA library as template, the PCR was performed to isolate a 5Ј fragment of the MOP4 cDNA. This fragment was cloned in the T7 orientation of pGEM-T and designated PL448. The cDNA insert of the phage clone F9047 (from C. C. Liew, University of Toronto, Toronto, Canada) was amplified by the PCR using oligonucleotides OL418 and OL419 and subcloned into the pGEM-T vector (24). This clone was designated PL420. An EcoRI fragment of PL448 was isolated and cloned into a partially EcoRI digested PL420. This clone was subjected to the PCR using oligonucleotides OL698 and OL146, the fragment cloned into the pGEM-T vector, and designated PL545.
MOP5 Expression Vectors-The PCR was used to obtain a 1260-bp fragment of the MOP5 gene using oligonucleotides OL685 and OL686 and IMAGE clone 42596 as template. This fragment was purified and subcloned into the pGEM-T vector as above in the SP6 orientation. This plasmid was designated PL528, and subsequently digested with SalI and partially digested with NcoI. This fragment was ligated into NcoI/ SalI-cut pSputk, and the resulting vector designated PL554.
Hypoxia-responsive Luciferase Reporters-The epo-luciferase plasmid, pGL2EPOEN, was constructed as follows. The hypoxia-responsive enhancer from the 3Ј region of the EPO gene was amplified by PCR using oligonucleotides OL499 and OL500 and human genomic DNA as template (amplified fragment corresponds to nucleotides 127-321 as reported in the EPO structural gene sequence found in GenBank™ accession no. GBL16588). This fragment was digested with KpnI and NheI and cloned into the corresponding sites of the plasmid pGL2-Promoter (Promega).
Antibody Production-Antisera against MOP1, MOP2, AHR, and ARNT were prepared in rabbits using immunization protocols that have been described previously (25,26). Crude antisera were chosen for use in all coimmunoprecipitation experiments, and the preimmune sera from the same rabbit served to preclear the samples. For MOP1, the plasmid hbc025 was digested with EcoRI and the 604-bp fragment was treated with the Klenow fragment of DNA polymerase 1 in the presence of dNTPs and cloned into the SmaI site of the histidine tag fusion vector pQE-32 (Qiagen, Chatsworth, CA). This clone, designated PL377, was transformed by electroporation into M15(REP4) cells for expression under IPTG induction. The expressed protein was purified from 8 M urea using nickel-nitrilotriacetic acidagarose, extensively dialyzed against 25 mM MOPS, pH 7.4, 100 mM KCl, and 10% glycerol before its use as an immunogen. Antiserum produced against this protein was designated R3752. For AHR, the human cDNA clone PL71 (27) was digested with BamHI and cloned into the corresponding site of the histidine fusion vector pQE31 (Qiagen). The AHR protein fragment was expressed and purified exactly as described for MOP1 (above). Antiserum produced against this protein was designated R2891. For MOP2, a SacI/PstI fragment of PL445 was cloned into SacI/PstI cut pQE-31 to generate PL456. This clone, designated PL456, was transformed into M15(REP4) cells and the protein expressed under IPTG induction. The histidinetagged fusion protein was first extracted in guanidine hydrochloride, dialyzed extensively, and purified on nickel-nitrilotriacetic acid-agarose as above. Antiserum produced against this protein was designated R4064. ARNT-specific antiserum was raised against huARNT protein purified from baculovirus as described previously (28).
Yeast Two-hybrid Analysis-A modified yeast interaction trap was employed to identify those MOPs that could interact with the AHR or ARNT in vivo. LexAMOP chimeras were constructed to fuse the bHLH-PAS domains of the MOP proteins with the DNA binding domain of the bacterial protein LexA amino acids 1-202) (29). To amplify the region corresponding to the bHLH-PAS domains of MOP1, OL715 and OL716 were employed in the PCR using PL415 as template. To amplify the region corresponding to the bHLH-PAS domains of MOP2, OL717 and OL718 were employed in the PCR using PL447 as template. To amplify the region corresponding to the bHLH-PAS domains of MOP3, OL719 and OL720 were employed in the PCR using PL486 as template. To amplify the region corresponding to the bHLH-PAS domains of MOP4, OL721 and OL722 were employed in the PCR using PL545. Since a more detailed domain map existed for the AHR, a construct was made with a fine deletion of the transactivation domain. The N-terminal portion of the AHR was amplified by the PCR using oligonucleotides OL180 and OL124 and pmuAHR as template (30). This product was digested with KpnI and SalI, and cloned into the corresponding sites of pSG424 (31). This clone was designated PL187. The 3Ј end of the AHR cDNA was amplified by PCR using oligonucleotides OL201 and OL202 and pmuAHR as template. This product was digested with NotI and cloned into the corresponding site of pSGAhN⌬166 (30). This clone was designated PL188. PL188 was digested with KpnI and XbaI, and this fragment cloned into the corresponding sites of PL187. This clone was designated PL204. A cDNA fragment of the AHR was amplified by the PCR using OL392 and OL393 and PL204 as template. This product was cloned using the pGEM-T system and designated pGTAHR⌬TAD. This construct was digested with XhoI and this fragment ligated into SalIcut pBTM116 (32). This construct was designated pBTMAHR. Lex-AARNT was constructed by PCR using oligonucleotides OL226 and OL176 and PL87 as template. The PCR product was cloned into pGEM-T as above, and the BamHI fragment cloned into a BamHIdigested pGBT9 vector (Clontech). This construct was cut with BamHI and subcloned into a BamHI-digested pBTM116. This construct was designated LexAARNT. Following amplification these products were purified and cloned into the pGEM-T vector. These clones were designated PL537 (MOP1), PL538 (MOP2), PL539 (MO3), and PL540 (MOP4). These plasmids were digested with BamHI/PstI (PL537), BamHI/SalI (PL538 and PL540), and EcoRI/SalI (PL539), and these fragments ligated into the appropriately digested pBTM116. These clones were designated LexAMOP1, LexAMOP2, LexAMOP3, and Lex-AMOP4, respectively. Full-length expression plasmids harboring the AHR and ARNT were constructed as follows: PL104 (pSporthuAHR) was cut with SmaI, the insert purified and subcloned into SmaI site of pCW10, and this plasmid was designated PL317. This clone was digested with SmaI and subcloned into a SmaI-cut pRS305, and this clone designated pRSAHR. The ARNT cDNA (PL101) was digested with NotI and XhoI and cloned into the corresponding sites of pSGBMX1. This plasmid was designated PL371, and subsequently was digested with NotI and XhoI, and this fragment cloned into the corresponding sites of pSGBCU11. This clone was designated PL574. The LexAMOP fusion protein constructs were cotransformed with a yeast expression vector containing the full-length AHR or ARNT into L40, a yeast strain containing integrated lacZ and HIS3 reporter genes. As controls, Lex-AAHR and LexAARNT were cotransformed with AHR and ARNT. The strength of interaction was visually characterized by 5-bromo-4-chloro-3-indolyl ␤-D-galactoside plate assays, performed after 3 days of growth on selective medium (33). To provide quantitation of the interaction strength, multiple colonies from yeast harboring each bHLH-PAS combination colonies were grown overnight in liquid medium. Liquid cultures were grown for 5 h and assayed for lacZ activity using the Galacto-Light chemiluminescence reporter system (Tropix, Bedford, MA). To determine the effect of AHR agonists on these interactions, yeast were also grown on plates or in liquid culture with and without 1 M ␤NF (34).
To ensure expression of each bHLH-PAS construct, Western blot analysis was performed using antibodies raised against the LexA DNA binding domain. Yeast extracts were prepared from 15-ml overnight cultures derived from multiple colonies of yeast expressing each Lex-AMOP fusion protein. Cultures were subjected to centrifugation at 1200 ϫ g for 5 min, and the pellet was resuspended in 500 l of 6 M guanidinium HCl, 0.1 M sodium phosphate buffer, 0.01 M Tris, pH 8.0. This suspension was transferred to a fresh Eppendorf tube containing 500 l of acid washed glass beads (Sigma), and mixed on the max setting in a Bead-Beater (BioSpec, Bartlesville, OK) for 3 min at 4°C. The samples were cleared by centrifugation at 14,000 ϫ g, and 400 l of supernatant was precipitated with 400 l of 10% trichloroacetic acid on ice. After clearing by centrifugation at 14,000 ϫ g for 20 min at 4°C, the extracts were resuspended in SDS loading buffer and subjected to SDS-PAGE analysis. Following electrophoresis, proteins were transferred to nitrocellulose membrane and detected with LexA antiserum and secondary antibodies linked to alkaline phosphatase by standard protocols (35).
Transient Transfection of Hep3b Cells-pGL2EPOEN was cotransfected with pSport, PL611 (pSportMOP1), or PL477 (pSportMOP2) using the Lipofectin protocol (Life Technologies, Inc.). Briefly, the expression vector was mixed with the Epo reporter and the ␤-galactosidase control plasmid pCH110 (Clontech) and incubated with cells in a six-well plate for 5 h at 37°C in serum-free medium. Following incubation, fresh medium was added, and the cells were incubated with or without 75 M cobalt chloride at 37°C overnight before the cells were harvested. Cell extraction and luciferase assays were performed according to manufacturer's protocols (Promega), and ␤-galactosidase assays were performed using the Galacto-Light assay according to manufacturer's protocols (Tropix).
Coimmunoprecipitation with HSP90 -Each MOP construct was in vitro translated in the presence of [ 35 S]methionine in a TNT-coupled transcription/translation system (Promega). HSP90 immunoprecipitation assays were performed with monoclonal antibody 3G3p90 or a control IgM antibody, TEPC 183 (Sigma) essentially as described (36). Each immunoprecipitation was subjected to SDS-PAGE, and the resulting gel was dried. The level of radioactivity in each coprecipitated protein band was quantified on a Bio-Rad GS-363 Molecular Imager System. The amount of bHLH-PAS protein immunoprecipitated with the HSP90 antibody is presented as a percentage of the amount of murine AHR immunoprecipitated in parallel assays.

EST Search
Our initial BLAST searches in December 1994 were performed with the bHLH-PAS or PAS domains of all family members known at that time (AHR, ARNT, SIM, and PER). In these searches we identified an EST clone, hbc025, derived from human pancreatic islets (Table I) (37). To confirm this similarity, we performed a BLASTX search, comparing hbc025 to the GenBank™ data base and found that this sequence was most homologous to Drosophila SIM (Fig. 1). This EST clone was designated MOP1. The bHLH-PAS domains of all family members, including MOP1, were again searched from May to October of 1995. Human ESTs that recorded BLASTN scores above 150 were again retrieved and confirmed using the BLASTX algorithm. This routine resulted in the discovery of six ESTs with significant homology to the bHLH-PAS domains of known members (Fig. 1, Table I).
TABLE I MOP cDNA clone information Row 1, clones (in parentheses) containing the candidate ESTs were requested from their laboratory of origin. Row 2, the GenBank accession number for each original EST is indicated. Row 3, oligonucleotides used in library screening. Sequence information generated from this original clone was used to design oligonucleotides for use in an anchored-PCR strategy, whereby gene-specific and vector-specific primers were used to amplify 5Ј and 3Ј portions of the cDNA. Vector specific primers corresponded to modified T3 (5Ј, OL145) or T7 (3Ј, OL146) primers. A matrix of gene-specific primers against annealing temperature (50 -65°C) was attempted for each clone, generally leading to at least one successful reaction. Row 4, the cDNA libraries from which additional sequence of positive clones was identified. Row 5, size of ORFs. We define a complete ORF by the presence of an in-frame stop codon 5Ј to a methionine codon that lies within a Kozak consensus sequence for translational initiation. The 3Ј end of open reading frames are defined by the presence of an in-frame termination codon. An asterisk (*) denotes a clone which does not meet these criteria (see text). Row 6, GenBank ™ accession numbers of the final MOP cDNAs are given. MOP1 cDNA Cloning To more completely characterize the similarities and domain structures of the candidate clones, an anchored-PCR strategy was employed to obtain additional flanking cDNA sequence using phagemid libraries as a template. Comparison of amino acid sequences of these bHLH-PAS proteins is displayed in Fig.  2. Upon characterization of the open reading frames, it was learned that two of these ESTs (F06906 and T77200) corresponded to the same gene product (Table I, Fig. 1). Thus, we designated these remaining five unique cDNAs as "members of PAS superfamily" or MOPs 1-5. The PCR strategy provided what appeared to be the complete ORFs of MOP1, MOP2, and MOP3 based upon the following criteria. 1) At their 5Ј ends these clones contain an initiation methionine codon (AUG) downstream of an in-frame stop codon, and 2) at their 3Ј ends these clones contain an in-frame stop codon followed by no obvious open reading frames. In addition, the nucleotide sequences flanking of the MOP1 and MOP2 most 5Ј AUG codons (see GenBank™ accession nos. U29165 and U51626) are in reasonable agreement with the proposed optimal context for translational initiation, i.e. CCACCAUGG (38,39). Using the same anchored-PCR technique, we were unable to obtain the complete open reading frames of MOP4 or MOP5 (19). We did identify a potential start methionine for MOP4 and the 3Ј stop codon for MOP5 (Fig. 2). Our preliminary designation of the MOP4 start methionine is tentative and is based on its proximity to the start methionines of MOP1, MOP2, MOP3, AHR, and SIM (Fig. 2) (1, 3).

Tissue-specific Expression
To characterize the tissue-specific expression patterns of the MOP mRNAs, Northern blots of poly(A) ϩ RNA from eight human tissues were probed with random primed cDNA restriction fragments (Fig. 3). Single transcripts of 3.6 kb (MOP1), 6.6 kb (MOP2), and 3.2 kb (MOP3) were detected. Expression levels of each mRNA varied significantly between tissues, with MOP1 being highest in kidney and heart, MOP2 highly expressed in placenta, lung, and heart, and MOP3 highly expressed in skeletal muscle and brain. No detectable message was detected for MOP4 or MOP5 by our Northern blot protocol.

Identification of Novel AHR or ARNT Partners
Interaction of MOPs with the AHR and ARNT in Vitro: Coimmunoprecipitation experiments-We first performed coimmunoprecipitation experiments to determine if MOPs 1-4 had the capacity to interact with either the AHR or ARNT in vitro.  Table I). Bottom, results of BLASTX searches are shown with the most homologous PAS members. The numbers adjacent to the query sequence (top) are relative nucleic acid positions of the homology. The numbers adjacent to the subject sequence (bottom) are relative amino acid positions of the most homologous bHLH-PAS protein.
These proteins were expressed in a reticulocyte lysate system in the presence of [ 35 S]methionine and then incubated in the presence or absence of the AHR or ARNT. Complex formation was assayed by coimmunoprecipitation with AHR-or ARNTspecific antisera, followed by quantitation of coimmunoprecipi-tated 35 S-labeled MOP by phosphoimage analysis (Fig. 4). Interactions were identified by a reproducible increase in an AHR-or ARNT-dependent precipitation of MOP protein. Because we have observed considerable variability in this coimmunoprecipitation assay, each experiment was performed at least three times. A representative result is presented in Fig. 4.
In the AHR interaction studies (Fig. 4, top), we observed that MOP3 was coimmunoprecipitated with AHR. The positive control, ARNT⅐AHR interaction, was also reproducible, but weaker (Fig. 4, top). Neither MOP1, MOP2 or MOP4 could be shown to interact with the AHR by this protocol. The ARNT protein displayed a broad range of interactions and was shown to coimmunoprecipitate with AHR (positive control), MOP1 and MOP2, but not MOP3 or MOP4 (Fig. 4, bottom).
Interaction of MOPs with the AHR and ARNT in Vivo: Yeast Two-hybrid Experiments-To determine if MOP⅐AHR or MOP⅐ARNT complexes could form in vivo, a modified interaction trap was employed (40,41) (Fig. 5). Fusion proteins were constructed in which the DNA binding domain of the bacterial repressor, LexA, was fused to the bHLH-PAS domains of the MOP proteins (Fig. 5, panel A). The bHLH-PAS domains were chosen because they harbor both the primary and secondary dimerization surfaces of this family of proteins and they typically do not harbor transcriptional activity that would interfere with this assay (35). Interactions were tested by cotransformation of each LexAMOP construct with either the full-length AHR or ARNT into the L40 yeast strain, which harbors an integrated lacZ reporter gene driven by multiple LexA operator  Each lane consists of 2 g of twice selected poly(A) ϩ mRNA prepared from human heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas (Clontech). The blots from top to bottom were probed for: MOP1, MOP2, MOP3, and GAPDH; exposure times were 5 days, 1 day, 3 days, and 8 h, respectively. Random primed cDNA fragments were as follows: MOP1, a 604-bp EcoRI fragment of hbc025; MOP2, a 649-bp EcoRI/HindIII fragment of IMAGE clone 67043; MOP3, a 1121-bp HindIII fragment of IMAGE clone 50519; MOP4, a 696-bp SacI fragment of clone PL420 (see Table I); MOP5, a 539-bp BamHI/HindIII fragment of IMAGE clone 42596; and GAPDH (rat), a 1.3-kb PstI fragment of PL89 (pGAPDH). IMAGE clone designations are given in Table I. sites (32). In this system, LexAMOP fusions that interact with AHR or ARNT drive expression of the lacZ reporter gene.
We assessed the relative strength of these interactions by both a direct lacZ plate assay and by quantitation of the reporter activity in a liquid culture (Fig. 5B). In all cases, these two methods of detection were equivalent (data not shown). To test the validity of this model system as a method to detect bHLH-PAS interactions, LexAAHR and LexAARNT constructs were cotransformed with either the full-length ARNT or AHR. In these control experiments, we were able to demonstrate the specificity of AHR⅐ARNT interaction and its dependence on the presence of the agonist ␤NF. The LexAAHR⅐ARNT interaction in the presence of ␤NF was 913-fold above background, while the LexAARNT⅐AHR interaction in the presence of ␤NF was 14-fold above background. Both combinations showed ligand inducibility. The LexAAHR⅐ARNT interaction in the presence of ␤NF was 6.4-fold greater than LexAAHR⅐ARNT in the absence of ligand, while the LexAARNT⅐AHR interaction in the presence of ␤NF was 2.0-fold over LexAARNT⅐AHR in the absence of ligand. Despite our ability to readily detect the agonist-induced LexAARNT⅐AHR interaction in the two-hybrid system, we were unable to detect any LexAMOP that could interact with the AHR (Fig. 5B). That is, none of the LexAMOP fusion proteins appeared to interact with cotransformed AHR and drive lacZ expression in the absence or presence of ligand (Fig. 5B).
Two of four MOP proteins tested were found to interact with ARNT in the two-hybrid assay. Both the LexAMOP1 and Lex-AMOP2 interactions with full-length ARNT were extremely robust, 36-and 28-fold above background, respectively (Fig.  5B). When compared with the LexAAHR⅐ARNT interaction in the presence of ␤NF, the LexAMOP1-ARNT and LexAMOP2-ARNT interactions were 24% and 69% as intense. These differences in LexAMOP1⅐ARNT and LexAMOP2⅐ARNT interaction could be attributed to differences in expression levels or to subtle differences in chimera construction. To control for relative expression of the LexAMOP fusions, protein extracts were prepared and Western blot analysis was performed with LexAspecific antiserum (Fig. 5C). We observed expression for each LexAMOP fusion proteins, indicating that negative results with LexAMOP3 and LexAMOP4 are not due to lack of expression.
DNA Binding and Specificity-Prompted by the observation that MOP1 and ARNT and MOP2 and ARNT specifically interact, we next examined the ability of MOP⅐ARNT dimeric complexes to bind those DNA response elements recognized by other bHLH-PAS protein complexes. Reports from a number of laboratories have demonstrated that bHLH-PAS dimers can bind to a variety of DNA elements: "DRE," TNGCGTG (42); "CME," ACGTG (8); "SAE," GTGCGTG (10); and "E-box," CANNTG (9, 10). Using a gel-shift assay, we observed that MOP1⅐ARNT complexes specifically bound CACGTG and TACGTG (Fig. 6A, lanes 6 and 7), while the complex failed to bind GTGCGTG, TTGCGTG, and a non-palindromic E-box, CATGTG (lanes 8 -10). Previous reports have demonstrated that ARNT homodimers are capable of binding the CACGTG sequence in vitro and that this complex can drive reporter gene expression in vivo (9,10). Our results suggest that the MOP1⅐ARNT dimeric complex binds the CACGTG oligonucleotide with a higher affinity than either MOP1 or ARNT alone (Fig. 6A, lanes 11-14). MOP1 failed to form a productive DNA binding complex with the AHR with any of the bHLH-PAS family response elements (Fig. 6A, lanes 2-5). As a comparison of MOP1⅐ARNT and MOP2⅐ARNT DNA binding, we provide results from gel shift assays using a double-stranded oligonucleotide containing a core TACGTG hexad binding site (Fig.  6B). We observed that both MOP1⅐ARNT and MOP2⅐ARNT bound the TACGTG-containing oligonucleotide with approximately equal capacity and that neither ARNT, nor MOP1, nor MOP2 could bind this DNA sequence alone. As additional controls, we confirmed the presence of the MOP proteins in the complex by showing that antisera raised against these proteins retarded the mobility of the complex (Fig. 6B, lanes 6 -11).
Interaction of MOPs with HSP90 -In an effort to assess the ability of MOPs to interact with HSP90, we performed coimmunoprecipitation assays with HSP90-specific antibodies. Given the remarkable stability of the HSP90 complex with the

FIG. 6. Gel-shift analysis of the MOP1-MOP2 and ARNT interaction in vitro.
A, 32 P-labeled oligonucleotides corresponding to the known bHLH-PAS response elements were used in each analysis. 32 Poligos indicates the core sequences of the given radiolabeled, doublestranded oligonucleotide used in that reaction. All mixing experiments were performed with equimolar amounts of each protein unless otherwise indicated. The proteins were coincubated at 30°C for 30 min, followed by the addition of poly(dI-dC) (100 ng, 10 min at room temperature) and 5 ϫ 10 4 cpm of the 32 P-labeled and annealed oligonucleotides. All oligonucleotides share identical sequences flanking these variable core sequences. Arrows indicate the position of the MOP1⅐ARNT and AHRC⌬516⅐ARNT. Lane 1, the migration of the AHRC⌬516 ⅐ARNT⅐ 32 P-TTGCGTG complexes as a positive control. Lanes 2-5, gel shifts performed with various radiolabeled oligonucleotides and equimolar mixtures of MOP1 and AHRC⌬516. Lanes 11-14, gel-shifted band observed in the presence of MOP1⅐ARNT, MOP1 alone, ARNT alone, and 2-fold excess of ARNT with the 32 P-TCACGTG radiolabeled oligonucleotide. B, MOP1 and MOP2 were expressed and quantitated in reticulocyte lysates as described under "Experimental Procedures." ARNT was expressed and purified from a baculovirus expression system as described under "Experimental Procedures." Oligonucleotides OL568 and OL569 containing a TACGTG core sequence were labeled and annealed as described under "Experimental Procedures." Each DNA binding reaction contained 0.6 fmol of MOP1 or MOP2 and 20 fmol of ARNT. The proteins were coincubated at 30°C for 30 min, followed by the addition of poly(dI-dC) (100 ng, 10 min at room temperature) and 5 ϫ 10 4 cpm of the 32 P-labeled and annealed oligonucleotides.  6, 8, 9, and 11). Preimmune serum (PI) from the same rabbits was used in lanes 7 and 10 as a control for nonspecific inhibition of the binding by serum. Antisera were added after the oligonucleotide incubation step and an additional incubation of 10 min at room temperature was added. Samples were loaded onto a 0.3 ϫ Tris borate-EDTA buffer nondenaturing polyacrylamide gel (5%), dried, and subjected to autoradiography.
AHR from the C57BL/6J mouse, we used this receptor species as a reference and compared all interactions relative to it. As additional controls, we immunoprecipitated ARNT and the human AHR as negative and positive controls, respectively. Despite our ability to readily detect huAHR⅐HSP90 interactions, we were unable to detect ARNT, MOP2, or MOP5 interactions with HSP90 (Fig. 7). In contrast, huAHR, MOP1, MOP3, and MOP4 all immunoprecipitated with HSP90-specific antisera. MOP3 formed the tightest interaction with HSP90, followed by the huAHR, MOP4, and MOP1 (71%, 53%, 31%, and 17%, respectively.

DISCUSSION
Hypothesis-Our hypothesis was that additional bHLH-PAS proteins are encoded in the mammalian genome and that some of these proteins are involved in mediating the pleiotropic response to potent AHR agonists like TCDD. This idea arose from the observation that other bHLH superfamilies employ multiple dimeric partnerships to control complex biological processes, such as myogenesis (MyoD/myogenin), cellular proliferation (Myc, Max, Mad), and neurogenesis (achaete-scute/ daughterless) (43)(44)(45). The observation that bHLH proteins often restrict their dimerization to within members of the same gene family (i.e. "homotypic interactions") and that this restriction may occur as the result of constraints imposed by both primary (e.g. bHLH) and secondary dimerization surfaces (e.g. leucine zippers and PAS), prompted us to screen for additional bHLH-PAS proteins and test each protein for its capacity to interact with either the AHR or ARNT. The ultimate objective of our search was to identify MOPs that were physiologically relevant partners of either the AHR or ARNT in vivo. Our prediction was that such proteins might respond to or modulate the AHR signaling pathway and thus provide mechanistic insights into TCDD toxicity.
Expressed Sequence Tag Approach-In an effort to rapidly identify expressed genes, Venter and colleagues (46) developed the EST approach, whereby a cDNA library is constructed and randomly selected clones are sequenced from both vector arms. These partial sequences, generally 200 -400 bp, are deposited in a number of computer data bases that can be readily analyzed using a variety of search algorithms. In the past year, the IMAGE (Integrated Molecular Analysis of Genomes and their Expression) Consortium has deposited over 300,000 human ESTs, generated from different tissues and developmental time periods into publicly accessible data bases, identifying approx-imately 40,000 unique cDNA clones 2 (47). The availability of these sequences and plasmids harboring their corresponding cDNA clones provided the impetus to identify novel members of the bHLH-PAS family by nucleotide homology screening of available EST data bases.
When we began these studies, the human AHR and ARNT and the Drosophila SIM and PER were the only PAS proteins that had been described. Therefore, we used the nucleotide sequences encoding their PAS domains as query sequences in BLASTN searches of the available EST data bases. Using this strategy in an iterative fashion and confirming each hit with a BLASTX search (nucleotide against protein search), we identified five MOP cDNAs (Fig. 1). Using PCR, we were able to obtain the complete ORF of MOPs 1-3, and extensive but incomplete ORFs of MOP4 and 5. The inability to obtain the complete ORFs of MOP4 and MOP5 appears to be related to the low copy number of their mRNAs in the tissues examined. Interestingly, MOPs 1-4 displayed adjacent bHLH-PAS domains and variable C termini. This domain structure is identical to that observed in the AHR, ARNT, and SIM. MOP5 contained a consensus PAS domain, but our inability to amplify its 5Ј end did not allow us to determine if it is also a bHLH protein.
While this work was in progress, Wang and colleagues identified two factors involved in cellular response to hypoxia, HIF1␣ and HIF1␤. These proteins are identical to MOP1 and ARNT, respectively (48). Thus, of the five MOPs we have cloned, four have not been previously characterized. For consistency in this report, we use the term MOP1 for HIF1␣. Because of historical precedent and since the term MOP1 was meant to be temporary nomenclature, we will refer to this clone as HIF-1␣ in future reports.
Since cDNAs encoding the complete open reading frames for MOPs 1-3 were available, our initial studies focused on these proteins. Although these data should be considered preliminary, MOP4 and MOP5 were included in some studies. Since our MOP4 clone contained the sequences analogous to those involved in the dimerization, transcriptional activation, and DNA binding of other bHLH-PAS proteins, this cDNA was included in most interaction analyses (30,35). Since our MOP5 clone did not harbor a bHLH domain (required for dimeric interactions with other MOPs), but did encode a complete PAS domain (analogous to the AHR region required for HSP90 interaction), this clone was included in the HSP90 coimmunoprecipitation experiment.
Tissue-specific Expression-To provide a preliminary indication of the biological relevance of each MOP cDNA to the AHR signaling pathway, we performed Northern blot analyses to determine which MOPs displayed overlapping tissue-specific expression profiles with either the AHR or ARNT mRNAs. We observed that each MOP mRNA displayed a unique tissuespecific distribution, with MOP1 being highest in kidney and heart, MOP2 highly expressed in placenta, lung, and heart, and MOP3 highly expressed in skeletal muscle and brain. Previous studies conducted in our laboratory indicated that the human ARNT is ubiquitously expressed with highest levels in skeletal muscle and placenta, while the human AHR is most prevalent in placenta, lung, and heart and lower levels in brain, liver, and skeletal muscle (27,49). The observation that these bHLH-PAS proteins are coexpressed in a variety of tissues supports the idea that cross-talk between these signaling pathways may be occurring in vivo and that multiple tissue-specific interactions may be taking place. In particular, the observation that AHR FIG. 7. Interaction of MOPs with HSP90. In vitro expressed radiolabeled muAHR, huAHR, and MOPs1-5 were immunoprecipitated with HSP90-specific antisera. Following precipitation, SDS-PAGE analysis was performed, the gel dried, and exposed to a phosphor imaging plate for analysis. The values presented are the average of two separate experiments, and compared with the muAHR (set at 100%). Values from these two assays differed by less than 5%. and MOP2 have coincident expression profiles in human tissues suggested to us that these proteins should be among the first candidates to be screened for signaling pathway interaction. An additional and equally important interpretation of these unique MOP expression profiles is that unidentified partners exist for these bHLH-PAS proteins and that they regulate a number of undescribed biological pathways.
Interaction Strategy-Our interaction screening strategy was based on functional data and the detailed domain maps available for the AHR and ARNT. An important assumption used in the design and interpretation of our studies is that some of the MOPs may be constitutively active in vivo (like ARNT) and others may be conditionally active, possibly requiring ligand-activation to dimerize in vivo (like AHR). We chose to employ coimmunoprecipitation as an initial interaction screen for a number of reasons. First, AHR and ARNT-specific antibodies are available that have been shown to precipitate AHR⅐ARNT complexes. This suggests that if MOP⅐AHR or MOP⅐ARNT interactions occurred in vitro, that these same antibodies would recognize and precipitate such complexes. Second, data from a number of laboratories, using independently derived antibodies, indicate that coimmunoprecipitation of AHR⅐ARNT complexes is independent of AHR-ligand (50,51). This observation suggests that AHR or ARNT interactions with conditional MOP proteins might still be identified by coimmunoprecipitation even in the absence of knowledge about how to activate a conditional MOP (e.g. identification of its ligand).
As a secondary screen to characterize interacting MOPs, we employed a yeast interaction trap commonly referred to as the "two-hybrid assay" (40). Support for use of this system comes from our previous observation that LexAAHR chimeras are functional in yeast and provide a good model of AHR signaling and ARNT interaction (34). In addition, this method provides an independent confirmation of those interactions identified by coimmunoprecipitation and also provides a demonstration that interactions can occur in vivo. One advantage of comparing two-hybrid results with coimmunoprecipitation results is that it may reveal conditional MOPs that require activation prior to dimerization. An example of this can be seen with the AHR and ARNT. In the absence of ligand, the AHR appears to reside primarily in the cytosol and ARNT appears to be primarily nuclear (26,35). This compartmentalization appears to be part of a cellular mechanism to prevent inappropriate interaction of these proteins and minimize constitutive activity of the complex. As a LexA chimera in yeast, the AHR is repressed until addition of ligand (34). Therefore, in vivo systems such as the two-hybrid assay may yield negative results for conditional MOP proteins in the absence of the factors required for their activation (e.g. ligands).
In light of the above considerations, our interpretation of the coimmunoprecipitation and two-hybrid interaction results are as follows. First, since the MOP1⅐ARNT and the MOP2⅐ARNT interactions were confirmed in two independent systems, these interactions appear to involve MOPs that have constitutive activity. Second, the observation that MOP3 interacts with the AHR in vitro, but not in vivo, suggests that MOP3 may be a conditional MOP that has the capacity to interact with the AHR or other MOPs in vivo upon its activation by a ligand (this idea gained support from HSP90 interaction studies below). The suspicion that MOP3 is a conditional bHLH-PAS protein, coupled with the observation that MOP3 and AHR have disparate expression profiles, led us to delay study of this interaction until we learn how to activate MOP3 or until we have evidence that these two proteins are expressed in the same cell type. Finally, our observation that ARNT can form dimers with two out of four MOPs examined suggests that ARNT is a promiscuous bHLH-PAS partner that may be a focus of crosstalk between different MOP signaling pathways. The multiplicity of ARNT partnerships is supported by recent observations from a number of laboratories (9,10,48).
MOP1 and MOP2 Interactions with ARNT-The concordance of the coimmunoprecipitation and two-hybrid data led us to pursue the MOP1⅐ARNT and MOP2⅐ARNT interactions further. Given the pairing rules deduced from the interaction studies described above, we next attempted to determine if the MOP1⅐ARNT and MOP2⅐ARNT complexes bound specific DNA sequences in vitro. Earlier reports indicated that the basic region of each bHLH partner generates specificity for a distinct DNA half-site of at least 3 bp (10,16). Data from a number of laboratories have indicated that the ARNT protein displays specificity for the 3Ј-GTG half-site of the hexad target sequence, 5Ј-NNCGTG-3Ј, where 5Ј-NNC is the half-site of the ARNT partner (10,52). To determine the half-site specificity of the MOP1 protein when complexed with ARNT, we used gel shift analysis with oligonucleotides representing the response elements of all known bHLH-PAS proteins. These preliminary experiments indicated that MOP1⅐ARNT complex had greatest affinity for the 5Ј-CACGTG and 5Ј-TACGTG sites (Fig. 6A).
At the time these DNA binding specificity studies were being completed, a report appeared that defined the biological role of MOP1. Studies by Semenza and colleagues demonstrated that a protein complex, termed HIF1, regulates hypoxia-responsive genes such as EPO and VEGF, and is composed of a dimer of HIF1␣ and HIF1␤ subunits (48). HIF1␣ and HIF1␤ RNAs were both shown to be up-regulated in response to hypoxia or certain agents like cobalt chloride or desferrioxamine that stimulate an upstream "heme sensor" (48). Moreover, these studies demonstrated that the HIF1 complex bound to TACGTG-containing enhancer elements that regulated the expression of hypoxiaresponsive genes. These results were important for a number of reasons. First, these results provided independent support for our screening approach by demonstrating that the MOP1⅐ARNT complex we described did have biological relevance. Second, they confirmed our independently derived DNA binding site for the MOP1⅐ARNT complex. Third, they provided a logical approach to design a physiologically relevant reporter construct in our attempts to compare the interactions of MOP1 and MOP2 with ARNT in vivo (see below).
Because the MOP1 and MOP2 basic regions differed by only one amino acid residue and since this residue is not thought to be in a DNA contact position (53,54), we hypothesized that MOP2 would bind the same DNA half-site sequences as MOP1.
To confirm this, we performed MOP2⅐ARNT gel shift assays using a double-stranded oligonucleotide containing a core TACGTG hexad binding site (Fig. 6B). We observed that both MOP1⅐ARNT and MOP2⅐ARNT bound the TACGTG-containing oligonucleotide and that neither MOP1 nor MOP2 could bind this sequence in the absence of ARNT. As additional controls, we confirmed the presence of the MOP1 and MOP2 proteins in the complex by showing that antisera raised against these proteins retarded the mobility of the complex (Fig. 6B, lanes  6 -11).
To assay MOP1⅐ARNT and MOP2⅐ARNT interactions in vivo, we constructed a luciferase reporter driven by the hypoxiaresponsive TACGTG-containing enhancer from the human EPO gene (55). Our transient expression experiments in Hep3B cells consisted of cotransfection of this reporter with vector control, MOP1, or MOP2 in the presence or absence of cobalt chloride to stimulate the hypoxia heme sensor (Fig. 8) (56). ARNT has been shown previously to be expressed in HEP3B cells (55). This experiment confirmed that the TACGTG-containing enhancer sequence is responsive to cobalt and cotransfected MOP1 or MOP2 under normal oxygen tension (Fig. 8). The transfected MOP1 construct appeared to be responsive to hypoxia (3.5-fold over control), while the MOP2 construct was only slightly responsive (1.2-fold) (Fig. 8). MOP2 was more potent than MOP1 in driving expression of this reporter gene both in the presence and absence of cobalt chloride. This difference in efficacy of the MOP1 and MOP2 constructs in driving expression of the reporter plasmid in Hep3B cells could be explained by three possibilities. 1) The relative potency of the MOP2 transactivation domain may be much greater than MOP1, 2) the relative expression of MOP2 my be greater in this transient expression system than MOP1, or 3) the MOP1 may be partially repressed in vivo by HSP90 while MOP2 is not (see HSP90 discussion below). Given that our MOP2 antisera are not useful in Western blots, we could not assess the relative expression of the MOP1 and MOP2 clones in this system. Thus, we cannot rule out more efficient expression of MOP2 relative to MOP1.
MOP3 Is a Conditionally Active bHLH-PAS Protein-Data from a number of laboratories suggest that HSP90 is involved in AHR signaling in vivo (34,57). In vitro experiments suggest that HSP90 is required for high affinity ligand binding and that HSP90 "caps" the DNA binding domain, preventing the unliganded receptor from constitutively interacting with the ARNT protein and binding DNA (58,59). Moreover, a minimal region shown to repress transactivation of constitutive AHR deletion chimeras has been shown to mediate HSP90 binding (60). These observations suggest that HSP90 represses AHR activity by inhibiting constitutive dimerization and by anchoring the receptor in the cytosol away from its nuclear dimeric partner ARNT. Upon ligand binding, the AHR⅐HSP90 complex translocates to the nucleus where HSP90 dissociates from the complex and the AHR dimerizes with ARNT and binds DNA (6). Two lines of evidence suggest that MOP3, like the AHR, may be a conditionally active bHLH-PAS protein and that in the absence of an unidentified cognate ligand, might be repressed and unable to dimerize in vivo. First, MOP3 interacts with HSP90 even more efficiently than the human AHR, suggesting that MOP3 may be functionally repressed or anchored in the cytosol like the AHR (Fig. 7). Second, MOP3 interacts with AHR in the coimmunoprecipitation assay, but not in the yeast interaction trap (Fig. 4, top, and Fig. 5B). Similarly, the AHR interacts with ARNT in the coimmunoprecipitation assay, but interacts weakly, if at all, in the absence of ligand activation (Fig. 4, bottom) (34).
Alternative explanations for the different MOP3⅐AHR interaction results obtained from our in vivo or in vitro systems must also be considered. For example, the structure of MOP3 may be different than the AHR and ARNT, such that positioning of the LexA domain adjacent to the bHLH-PAS domain may sterically hinder dimerization surfaces within this protein or lead to improper subcellular localization or instability of the chimera. One example of the potential negative impact of context sensitivity in the two-hybrid system can be observed in  (145-471). The scale at the bottom indicates number of amino acid residue substitutions. PAS family members that interact with HSP90, interact with the AHR, and interact with ARNT by the coimmunoprecipitation method are indicated by a ϩ, whereas members that do not interact are indicated by a Ϫ (see "Discussion"). An a denotes a bHLH-PAS member whose cDNA is not complete. Note that these interactions occur in vitro and may or may not be physiologically relevant. Where appropriate, the reference is included in parentheses. robust than the LexAARNT⅐AHR interaction. In addition, the LexAAHR⅐ARNT interaction is more responsive to the AHR ligand ␤NF than the LexAARNT⅐AHR combination (6.4-fold and 2.0-fold, respectively). This difference cannot be explained by the relative transactivation potencies of the transactivation domains of AHR and ARNT in yeast, and therefore must be a result of context sensitivity. A final consideration is that coimmunoprecipitations may be capable of detecting weak interactions that cannot be maintained at the low cellular concentrations of the various MOPs. Thus, the MOP3⅐AHR dimerization may be too weak to occur in vivo. In this regard, we have previously reported ARNT⅐ARNT homodimers that bind specific DNA enhancer sequences in vitro, but they are weakly active, if at all, in vivo (10).
It is also important to note that MOP1 and MOP4 also interact with HSP90 in the coimmunoprecipitation assay, albeit less strongly than MOP3 or human AHR (Fig. 7). The relatively weak interaction of MOP1 with HSP90 may be an indication that this protein is partially repressed in vivo and that it may have both constitutive and conditional activity. Such a phenomenon might explain why MOP1 has less transcriptional activity in our in vivo systems than MOP2, which does not interact with HSP90 ( Figs. 7 and 8). Finally, MOP4 did not interact with the AHR or ARNT in either the coimmunoprecipitation assay or the interaction trap. Although our experience with AHR indicates that interactions with conditional bHLH-PAS proteins can be observed by coimmunoprecipitation assays, MOP4's interaction with HSP90 may also indicate a requirement for activation and may inhibit the sensitivity of detecting interactions in vivo.
The bHLH-PAS Superfamily-Our experimental approach has significantly expanded the number of known members of this emerging superfamily of transcriptional regulators and has identified additional potential targets for TCDD signaling. During the preparation of this manuscript, five additional mammalian bHLH-PAS proteins have been identified, HIF1␣, SIM1, SIM2, ARNT2, and SRC-1 (48,51,(61)(62)(63)(64). To compare amino acid sequences of these proteins, we performed a CLUSTAL alignment with the bHLH-PAS domains of all the known family members using a PAM250 residue weight table (65). The two most related members were MOP1/HIF1␣ and MOP2, which shared 66% identity in the PAS domain. A comparison of these two proteins reveals only a single amino acid difference in the basic region and 83% identity in the HLH region. This sequence similarity is in agreement with our contention that MOP1/HIF1␣ and MOP2 function analogously, interacting with the same dimeric partners and binding similar enhancer sequences in vivo. A comparison of MOP3 and ARNT and a comparison of MOP5 and SIM reveal 40% and 38% identity in the PAS domain, respectively. The basic regions of MOP3 and ARNT have only three substitutions, while the HLH domains share 66% identity, again suggesting that the two proteins may regulate similar or identical enhancer sequences (half-sites).
A CLUSTAL alignment of the C termini of these MOP proteins and the previously identified PAS members demonstrated that these regions are not well conserved (data not shown) (1). This lack of conservation may indicate that the C termini of these genes have divergent functions, or that the functions harbored in the C termini can be accomplished by a variety of different sequences. For example, the C termini of the AHR, ARNT, HIF1␣, and SIM all harbor potent transactivation domains, yet display little sequence homology (35,66,67).
In an effort to characterize the evolutionary and functional relationships of these proteins, we performed a parsimony analysis to identify related subsets. A dendrogram represent-ing the primary amino acid relationship between the PAS domains of these proteins is illustrated in Fig. 9. This figure suggests that four major groups exist for eukaryotic PAS family members. The AHR, drSIMILAR, MOP1/HIF1␣, MOP2, drTRACHEALESS, MOP5, and SIM exist in one group; ARNT, muARNT2, MOP3, and MOP4 exist in another; and PER and huSRC-1 exist in their own groups. Interestingly, this pattern reflects some of what is known functionally about the existing PAS members. Members from the AHR group and ARNT, but not the PER group, have been shown to interact with HSP90 ( Fig. 7) (68). Only members from the ARNT group have been shown to interact with the AHR (Fig. 4, top) (51). Finally, members from all groups have been shown to interact with ARNT (Fig. 4, bottom) (10,12,48). The observation that ARNT has been shown to be capable of forming DNA binding homodimers as well as heterodimers with a number of previously identified members of the bHLH-PAS family (at least in vitro), suggests that it plays a role in a number of biological processes (9,10). Based on their similarity with ARNT, MOP3 and MOP4 may be candidates for binding DNA as homodimers, or for interacting with multiple bHLH-PAS members, possibly from the AHR group. PER is the only well characterized eukaryotic PAS protein so far identified that lacks a bHLH domain. huSRC-1 is the only PAS protein so far identified to interact with members of the steroid receptor family. Future investigation will focus on confirming the functional relationships of these MOPs, their pairing rules and DNA binding specificities.
Conclusion-In an effort to understand the molecular consequences to TCDD exposure, we have developed a protocol to identify novel PAS proteins as well as to define their pairing rules. These studies bring the total number of the mammalian PAS family to 11. Using a number of assays for protein-protein interactions and DNA binding specificity (10, 40), we have been able to determine that both MOP1 and MOP2 are productive dimerization partners of the ARNT protein and that these dimers recognize a TACGTG-containing enhancer element in vivo. The validity of our system is supported by the characterization of MOP1/HIF1␣ by an independent group that reached the same conclusions regarding dimerization and DNA binding specificity (48). These observations demonstrate the power of the approach and support its use in characterizing the emerging superfamily of bHLH-PAS proteins that will be revealed by EST technology in the coming years.
In addition to the relevance of these data to TCDD signaling, it also appears to be revealing additional factors important to cellular responses to hypoxic stress. Our analysis indicated that HIF1␣/MOP1 and MOP2 share a common dimeric partner, ARNT, and are capable of regulating a common battery of genes. This idea is supported by three lines of evidence; 1) both MOP1 and MOP2 interact with ARNT as defined by coimmunoprecipitation or two-hybrid assay, 2) they have similar DNA half-site specificities when complexed with ARNT, and 3) they are both transcriptionally active from TACGTG enhancers in vivo. The observation that HIF1␣/MOP1 and MOP2 have markedly different tissue distributions suggests that these two proteins may be regulating similar batteries of genes in response to different environmental or developmental stimuli. Alternatively, these proteins may be involved in restricting expression of certain groups of genes regulated by TACGTGdependent enhancers. Finally, it is possible that MOP2 is a subunit of a "HIF1-like" complex (i.e. a "HIF2␣") that regulates hypoxia-responsive genes in a distinct set of tissues.