Isolation and expression of cDNAs from rainbow trout (Oncorhynchus mykiss) that encode two novel basic helix-loop-Helix/PER-ARNT-SIM (bHLH/PAS) proteins with distinct functions in the presence of the aryl hydrocarbon receptor. Evidence for alternative mRNA splicing and dominant negative activity in the bHLH/PAS family.

cDNAs encoding two distinct basic helix-loop-helix/PER-ARNT-SIM (bHLH/PAS) proteins with similarity to the mammalian aryl hydrocarbon nuclear translocator (ARNT) protein were isolated from RTG-2 rainbow trout gonad cells. The deduced proteins, termed rtARNTa and rtARNTb, are identical over the first 533 amino acids and contain a basic helix-loop-helix domain that is 100% identical to human ARNT. rtARNTa and rtARNTb differ in their COOH-terminal domains due to the presence of an additional 373 base pairs of sequence that have the characteristics of an alternatively spliced exon. The presence of the 373-base pair region causes a shift in the reading frame. rtARNTa lacks the sequence and has a COOH-terminal domain of 104 residues rich in proline, serine, and threonine. rtARNTb contains the sequence and has a COOH-terminal domain of 190 residues rich in glutamine and asparagine. mRNAs for both rtARNT splice variants were detected in RTG-2 gonad cells, trout liver, and gonad tissue. rtARNTa and rtARNb protein were identified in cell lysates from RTG-2 cells. Transfection of rtARNT expression vectors into murine Hepa-1 cells that are defective in ARNT function (type II) result in rtARNT protein expression localized to the nucleus. Treatment of these cells with 2,3,7,8-tetrachlorodibenzo-p-dioxin results in a 20-fold greater induction of endogenous P4501A1 protein in cells expressing rtARNTb when compared with rtARNTa, even though both proteins effectively dimerize with the aryl hydrocarbon receptor. The decreased function of rtARNTa appears to be due to inefficient binding of rtARNTa·;AHR complexes to DNA. In addition, the presence of rtARNTa can reduce the aryl hydrocarbon receptor-dependent function of rtARNTb in vivo and in vitro.

The aryl hydrocarbon receptor (AHR) 1 and aryl hydrocarbon receptor nuclear translocator (ARNT) proteins are constitutively expressed basic helix-loop-helix (bHLH) transcription factors that are thought to mediate many of the biological effects of halogenated aromatic hydrocarbons typified by 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (1)(2)(3). These proteins are members of the bHLH/PAS family of transcription factors that also includes the Drosophila proteins SIM and PER (4 -7). The common characteristics of these proteins are bHLH regions and a domain of approximately 300 amino acids termed PAS (PER, AHR ARNT, SIM) that is distal to the bHLH motif. Deletion and mutational analysis has identified the basic region of these proteins as critical for DNA binding, a finding consistent with other bHLH and bHLH leucine zipper proteins (8 -12). The HLH and PAS domains appear to have multiple functions that include dimerization, hsp90 binding, and ligand binding in the AHR (8,11,13). Deletion analysis has defined regions in the COOH-terminal portion of both the AHR and ARNT as important in transactivation (8,(13)(14)(15). The hypoxia inducible factor 1a (Hif-1a) and a novel murine ARNT termed ARNT-2 have recently been added to the bHLH/PAS family (16,17).
The current model of AHR-mediated signal transduction in mammals hypothesizes that the AHR exists in association with hsp90 in a multiprotein complex primarily within the cytoplasmic compartment of cells (18 -21). Following ligand binding, the AHR becomes localized within the nuclear compartment in a form that presumably has dissociated from hsp90 (20 -21). Once in the nucleus, a fraction of the ligand-bound AHRs form heterodimers with the nuclear ARNT protein, whereas the majority of the AHRs are rapidly degraded (20,(22)(23)(24)(25). The AHR•ARNT heterodimers are then capable of binding the consensus xenobiotic response element sequence 5Ј-TNGCGTG-3Ј (XRE) to modulate gene expression (12,26,27). Presently, the complement of genes regulated by the AHR and ARNT is un-determined in most tissues, but the profound effects of TCDD exposure on early life stage development and differentiation in many mammalian and aquatic species support a role for these proteins in key cellular pathways (28 -30). Consistent with this hypothesis, gene-targeted mutation of the AHR locus in mice results in defects in immune system development and liver fibrosis (31). In addition, it has been reported that a reduction in AHR protein modulates the cell cycle and differentiated state of Hepa-1 cells (32). An animal model containing reduced ARNT protein has yet to be produced, but recent reports have significantly expanded the understanding of ARNT function (16,33,34). For example, ARNT can form heterodimers with SIM, PER, and Hif-1a, can form homodimers in vitro, and can bind to the E-box motif 5Ј-CACGTG-3Ј as a homodimer and drive transcription of reporter genes downstream of this sequence. These results suggest that ARNT has functions independent of the AHR and highlight the importance of identifying additional ARNT proteins and the processes they mediate.
In an effort to understand the function and evolution of ARNT, and to obtain the molecular tools necessary to study AHR-mediated signal transduction in aquatic species, a cDNA library was produced from the RTG-2 rainbow trout gonad cell line and screened with mouse ARNT cDNA. cDNAs encoding two novel proteins with similarity to ARNT were obtained. The deduced proteins, termed rtARNT a and rtARNT b , are identical over the first 533 amino acids but differ in their COOH-terminal domains due to the presence or absence of a 373-bp region that subsequently shifts the reading frame. The presence of distinct COOH-terminal domains does not appear to affect dimerization with the murine AHR but results in proteins that exhibit positive (rtARNT b ) or negative (rtARNT a ) function on AHR-mediated signal transduction in vivo and in vitro. These results highlight a previously uncharacterized mechanism for generating diversity within the bHLH/PAS family of proteins.

EXPERIMENTAL PROCEDURES
Materials-Specific antibodies against either the mouse AHR (A-1) or mouse ARNT protein (R-1) are identical to those described previously (20). All antibodies are affinity purified IgG fractions. For Western blot analysis goat anti-rabbit antibodies conjugated to horseradish peroxidase were utilized. For immunohistochemical studies, goat anti-rabbit IgG conjugated to Texas Red (GAR-TR) was used. Both of these reagents were purchased from Jackson Immunoresearch (West Grove, PA). Polyclonal rabbit ␤-actin antibodies were purchased from Sigma. Polyclonal rabbit antibodies against mammalian P4501A1 were a generous gift from Dr. Colin Jefcoate (University of Wisconsin of Pharmacology, Stanford University). The cells were propagated at 37°C in Dulbecco's minimum essential media containing 5% fetal bovine serum. Rainbow trout gonad cells (RTG-2) were obtained from American Type Culture Collection (ATCC, Rockville, MD). Cells were propagated at 20°C in Hepes-buffered minimum essential media supplemented with 10% fetal bovine serum. All cells were passaged at 1-week intervals and used in experiments during a 2-month period.
Isolation of cDNA Clones-Refer to Fig. 1 for orientation and length of all clones. A Uni-ZAP XR cDNA library was produced from RNA isolated from a rainbow trout gonad cell line (RTG-2) by Stratagene (La Jolla, CA). The unamplified library contained 5.8 ϫ 10 6 plaques, and 8 ϫ 10 5 plaques from the amplified library were screened with a full-length mouse ARNT cDNA. Four independent cDNAs (c2, c3, c4, and c5) of approximately 1600 bp were isolated and sequenced as detailed (35). All clones were overlapping, with three (c3, c4, and c5) containing a putative ATG start codon and open reading frame for the length of the fragment. To obtain clones with novel 3Ј sequence, 4 ϫ 10 5 recombinants were screened with the longest cDNA (c3). Two additional cDNAs internal to c3 were isolated (c6, c7) that did not extend the previous sequence. We then utilized a PCR-based method (3Ј-rapid amplification of cDNA ends; 36) to obtain sequence 3Ј to clone 3. Briefly, 1 g of RTG-2 total RNA was reverse-transcribed, and an oligonucleotide primer to nucleotides 1701-1717 containing a XhoI restriction site (primer 1) was used in combination with a poly(T) primer (primer 2) to amplify the reverse-transcribed RNA by PCR. Amplified cDNA was cut with XhoI and ligated into pBSK (Stratagene, La Jolla, CA). Colonies were hybridized with c3 cDNA, and three cDNAs were identified which contained an insert of 857 bp (pcr900). To confirm that the newly isolated pcr900 cDNA was the authentic 3Ј end of the original cDNAs, PCR was performed on reverse-transcribed RTG-2 RNA with a 5Јoligonucleotide corresponding to sequence present in clones c2, c3, c4, c5, and c7 (primer 3) and a 3Ј-oligonucleotide at 2467-2484 (primer 4). Amplified cDNA was cloned into pBKS and screened with c5 cDNA by colony hybridization (37). Fourteen colonies were isolated. Seven clones contained inserts with the predicted 1862-bp insert (pcrG41), and seven contained an insert of 2232 bp (pcrG79). 4 ϫ 10 5 plaques were then screened with the pcr900 cDNA, and two identical cDNAs (c16/11) were isolated that contained sequence identical to that in pcr900, pcrG41, and pcrG79.
Generation of Full-length ARNT cDNAs and Expression Vectors-Complete rtARNT cDNAs were generated as follows. An oligonucleotide (primer 5) with sequence corresponding to the putative ATG start site (nucleotide 1), containing both a Kozak sequence (38) and a HindIII restriction site, was used in combination with an oligonucleotide corresponding to nucleotides 1225-1243 (primer 6) to amplify a fragment from c5. Since a convenient restriction site was not present in the putative c5 sequence, it was necessary to change nucleotide 1243 from an adenine to a cytosine, thus creating a BamHI restriction site internal to the c5 sequence. The base substitution did not change the encoded amino acid since AGG and CGG both encode an arginine. The HindIII-BamHI fragment was ligated into pBKS (pKS-5). A primer set corresponding to nucleotides 1244 -1262 (primer 7) and 2243-2262 (primer 8) was then used to amplify a 3Ј fragment from either the pcrG41 or pcrG79 cDNA. The resulting fragments were ligated into the BamHI-XbaI site of pKS-5 to create full-length cDNAs corresponding to pcrG41 (pKSrtARNT a ) and pcrG79 (pKSrtARNT b ). To generate eukaryotic expression vectors, primer 5 and primer 8 were used to amplify full-length HindIII-XbaI fragments from pKSrtARNT a and pKSrtARNT b . The inserts were ligated into pRc/CMV (Invitrogen) to generate pMVrtARNT a and pMVrtARNT b . Primer 9 and primer 10 were used to amplify a full-length mARNT cDNA from plasmid pmARNT. The HindIII-XbaI fragment was then ligated into pRcCMV to create pMV m ARNT. A rtARNT a construct with a specific COOH-terminal domain was generated by amplifying a 1601-bp fragment from c3 with primers 5 and 13. This fragment was cloned into the HindIII-BamHI site of pBKS (pKS-U41). Primers 15 and 8 were used to amplify a 540-bp BamHI-XbaI from pKSrtARNT a containing the rtARNT a reading frame, and this was ligated into pKS-U41 (pKSrtARNT104). All clones were verified by sequencing.
Nucleotide and Amino Acid Sequence Analysis-Sequencing of DNA was accomplished using Sequenase Version 2.0 DNA Sequencing Kit (U. S. Biochemical Corp.). All sequences were read on both DNA strands. DNA sequence was compiled and analyzed by Lasergene software (DNASTAR Inc., Madison, WI). Amino acid comparisons were carried out in the MEGALIGN program using the method of Hein (39).
Generation of Polyclonal Antibodies against rtARNT-A 1078-bp fragment was amplified by PCR from clone c3 using primers 11 and 12. The fragment spans the 5Ј end of the rtARNT and contains sequence which encodes the bHLH region and part of the PAS domain (amino acids 2-353). The fragment was ligated into the bacterial expression vector pQE8 (Qiagen, Chatsworth, CA), and the recombinant protein was purified as previously detailed (20). SDS-PAGE analysis indicated that the expressed protein, termed rtBEARNT (rainbow trout bacterial expressed ARNT), migrated at approximately 40 kDa, consistent with the deduced amino acid sequence. Rabbit polyclonal antibodies were produced against rtBEARNT by Charles River Laboratories (Southbridge, MA) and subsequently affinity purified as detailed (20). The affinity purified IgGs with specificity toward rtARNT were termed rt-84 and rt-85.
Production of Total Cell Lysates and Cytosol-Total cell lysates for Western blot analysis were prepared by sonicating cell pellets in 1 ϫ lysis buffer and Nonidet P-40 as detailed previously (20,22). Cytosol was prepared from type II Hepa-1 cells by homogenization in MENG buffer supplemented with leupeptin (10 g/ml), aprotinin (20 g/ml), and phenylmethylsulfonyl fluoride (100 uM). Homogenates were centrifuged at 22,000 ϫ g in a refrigerated microcentrifuge, and aliquots of the supernatant were stored at Ϫ70°C prior to use in EMSA and immunoprecipitation studies. Protein concentrations were determined by the Coomassie Blue Plus assay (Pierce) using bovine serum albumin as the standard.
Western Blot Analysis and Quantification of Protein-Protein samples were resolved by denaturing electrophoresis on discontinuous polyacrylamide slab gels (SDS-PAGE) and were electrophoretically transferred to nitrocellulose as described (20,37). Immunochemical staining was carried out with varying concentrations of primary antibody (see text and figure legends) in BLOTTO buffer supplemented with DLhistidine (20 mM) for 1-2 h at 22°C. Blots were washed with three changes of TTBSϩ for a total of 45 min. The blot was then incubated in BLOTTO buffer containing a 1:10,000 dilution of goat anti-rabbit horseradish peroxidase for 1 h at 22°C and washed in three changes of TTBSϩ as above. Prior to detection, the blots were washed in TBS for 5 min. Bands were visualized with the enhanced chemiluminescence (ECL) kit as specified by the manufacturer (Amersham Corp.). Multiple exposures of each set of samples were produced. The relative concentration of rtARNT, P4501A1, and actin was determined by computer analysis of the autoradiographs as detailed previously (22,40). All exposures analyzed by this method were within the linear range established for each antibody (22,40). In most instances, the mean and standard deviation of at least three independent samples are reported.
Southern Blot Analysis of PCR Products-PCR products were fractionated on 1% agarose gels and transferred to nitrocellulose by vacuum blotting. Blots were probed with 32 P-labeled cDNA fragments, washed, and exposed to film as detailed (37).
RNA Isolation and Northern Blot Analysis-Total RNA was isolated from culture cells with TriReagent as detailed by the manufacturer (MRC, Cincinnati, OH). RNA was resolved on formaldehyde gels and transferred to nitrocellulose by vacuum blotting. Blots were probed with 32 P-labeled cDNA fragments, washed, and exposed to film as detailed (36).
In Vitro Expression of rtARNT Protein-Recombinant rtARNT protein was produced from plasmids pKSrtARNT a , pKSrtARNT b , pMVr-tARNT a , pMVrtARNT b , and pMV m ARNT using the TNT Coupled Rabbit Reticulocyte Lysate Kit essentially as detailed by the manufacturer (Promega, Madison, WI). Upon completion of the 90-min reaction, samples were either combined with an equal volume of 2 ϫ gel sample buffer and boiled for 5 min or stored at Ϫ20°C for use in functional studies. The actual concentration of rtARNT protein expressed in each reaction was determined by Western blot analysis with the rt-84 antibody. Briefly, the linear range of a standard curve containing known amounts of the rtBEARNT antigen was established. Samples from TNT reactions were then resolved along with the standard curve, and the relative intensity of each band was determined by computer as detailed (22,39). Linear regression was then used to determine the concentra-tion of ARNT in each TNT sample. On average, the pMVrtARNT a and pMVrtARNT b constructs generated 300 ng of rtARNT protein per TNT reaction (6 ng/l). Both constructs routinely produced similar amounts of rtARNT.
In Vitro Activation of AHR⅐ARNT Complexes and Electrophoretic Mobility Shift Assay-AHR⅐ARNT complexes were produced by combining approximately 50 -80 ng of in vitro translated rtARNT a or rtARNT b protein with 100 g of type II cytosol in the presence of 10 nM TCDD or 0.5% Me 2 SO at 30°C for 2 h. The amount of rtARNT a and rtARNT b protein added to each sample was always identical as determined by Western blotting of the input protein (as detailed above). Samples were used immediately or stored at Ϫ70°C.
Oligonucleotides XRE-1 and XRE-2 were annealed and labeled with [ 32 P]dCTP by Klenow fill in (37). The double-stranded fragment corresponds to the consensus XRE-1 of the CYP1A1 promoter as described previously (41). 10 -20 g of wild type nuclear extract or in vitro activated type II cytosol was then incubated at 22°C for 15 min in 1 ϫ gel shift buffer supplemented with KCl (80 mM) and poly(dI-dC) (0.1 mg/ ml). In some instances, 0.5-1.0 g of affinity purified rt-84, A-1 or preimmune IgG was included in the sample. Approximately 4 ng of 32 P-labeled XRE was then added to each sample, and the incubation continued for an additional 15 min at 22°C. The samples were resolved on 5% acrylamide, 0.5% TBE gels, dried, and exposed to film.
Eukaryotic Transfections-Approximately 2 ϫ 10 6 type II cells were plated into 60-mm culture dishes and incubated at 37°C for 16 -24 h. 1-2 g of appropriate plasmid vectors were then transfected into cells with LipofectAMINE or Lipofectin reagent as detailed by the manufacturer (Life Technologies, Inc.). Following a 24-h recovery period, cells were incubated in the presence of 2 nM TCDD or Me 2 SO for an additional 20 -40 h. Cells were harvested from plates by trypsinization, and total cell lysates were prepared as detailed above. In some experiments the cells were also transfected with 1.0 g of pSV-␤-galactosidase to monitor transfection efficiency (Promega, Madison, WI). In these instances, harvested cells were split into two equal fractions. One fraction was used for analysis of ␤-galactosidase activity as detailed (Promega, Madison, WI), and the other fraction was used for the preparation of total cell lysates. Three, four, or five plates were transfected with each plasmid being evaluated, and experiments were completed three times. For immunological studies of transfected cells, 2 ϫ 10 6 type II cells were plated into 60-mm culture dishes containing poly-L-lysinecoated glass coverslips and then transfected and treated as detailed above.
Immunochemical Staining of Cells-All immunocytochemical procedures (fixation, staining, and photography) were carried out as described previously (20). The cells were observed on a Zeiss Axiophot microscope using the 568-nm filter. On average, 15-20 fields (5-20 cells each) were evaluated on each slip and three were photographed to generate the raw data. Experiments were repeated at least three times. The concentrations of antibodies used are detailed in the text.
Immunoprecipitation-100 g of type II cytosol was activated in the presence of either 100 ng of [ 35 S]methionine-labeled rtARNT or 100 ng of "cold" rtARNT produced in vitro as detailed above. The sample was then split into aliquots and incubated with 5 g of affinity purified antibody in the presence of 1 ϫ immunoprecipitation buffer for 1 h at 22°C. The samples were put into fresh tubes, supplemented with 10 l of a 50% slurry containing protein A-Sepharose beads (Sigma), and incubated for 1 h at 22°C. Samples were then centrifuged at 500 ϫ g for 2 min, and the beads were removed to fresh tubes. The beads were washed for a total of 45 min with three changes of IP wash buffer, mixed with 25 l of 2 ϫ gel sample buffer, heated for 10 min at 95°C, and resolved by SDS-PAGE. The gels were incubated with Amplify (Amersham Corp.) for 30 min and exposed to film. Immunoprecipitations of cold samples were blotted to nitrocellulose and stained with appropriate antibodies as detailed above.

RESULTS AND DISCUSSION
Cloning of Rainbow Trout ARNT cDNAs-To obtain rainbow trout ARNT cDNAs a random primed cDNA library was prepared from the rainbow trout gonad cell lines (RTG-2). The choice of RTG-2 cells was due to the identification of the AHR in these cells (42), and exquisite induction of CYP1A following exposure to polychlorinated biphenyls, polychlorinated dibenzodioxins, and polychlorinated dibenzofurans (43). Based on the model of AHR-mediated signal transduction in mammals, these results implied that a functional ARNT-like protein was expressed in RTG-2 cells. A 2.4-kb cDNA containing the entire murine ARNT (mARNT) coding sequence was used as a probe to library, and four overlapping cDNA clones (c2, c3, c4, and c5) were isolated (Fig. 1A). The largest cDNA (c3) contained a 1.86-kb insert with a putative methionine translation initiation codon and an open reading frame of 1.73 kb that ran the length of the cDNA. The methionine was preceded by nucleotides that fall within the Kozak consensus sequence (38). Clones c4 and c5 contained sequence that matched c3, confirming the initiation codon and the reading frame. Since none of the cDNAs contained a complete open reading frame, the RTG library was screened with clone c3. Two additional cDNAs were identified that contained sequence identical to c2, c3, c4, and c5 ( Fig. 1) but did not extend the reading frame.
To obtain sequence that was distal to clone c3, an oligonucleotide corresponding to nucleotides 1701-1717 (primer 1) was used in combination with a poly(T) primer (primer 2) to amplify fragments from reverse-transcribed RTG-2 RNA as detailed under "Experimental Procedures." A set of 854-bp cDNAs (pcr900) were isolated that contained identical sequence found in the final 100 bp of clone c3 (Fig. 1A). These cDNAs extended the open reading frame an additional 110 bp before reaching an in-frame termination codon at nucleotide 1911. While the pcr900 cDNAs terminated in a (A) 17 stretch, a consensus polyadenylation signal could not be identified indicating that the sequence did not represent the true 3Ј end of the mRNA. To confirm that the newly isolated cDNA was continuous with the cDNAs obtained from the RTG-2 library, RT-PCR was performed with a second set of primers. The 5Ј primer corresponded to nucleotides 614 -632 and was sequence contained within c2, c3, c4, c5, and c7. The 3Ј primer represented sequence within the 3Ј-untranslated region of pcr900 (nucleotides 2467-2484). Surprisingly, the use of these primers resulted in the amplification of two distinct cDNA fragments. The cDNAs identified as pcrG41 contained 1.86-kb inserts that matched perfectly the sequence previously identified in the RTG-2 cDNAs and pcr900. These results confirmed the authenticity of pcr900. The cDNAs identified as pcrG79 contained 2.23-kb inserts that perfectly matched the sequence of pcrG41 but contained 373 bp of novel sequence inserted within the putative open reading frame at nucleotide 1601 ( Fig. 1A and Fig. 2). The new sequence contained an open reading frame for its entire length that was continuous with the previously identified reading frame of c3. However, the introduction of 373 bp shifted the COOH-terminal reading frame such that the amino acid sequence was different than that identified in the pcrG41 cDNAs (Fig. 2). The new open reading frame reached an in-frame termination codon (TGA) at nucleotide 2171. To confirm the authenticity of the new sequence, the RTG-2 library was screened with the pcr900 cDNA. Two identical clones were isolated (c16/11) that contained sequence present in pcr900, pcrG41, and pcrG79. These clones confirmed the location of the two termination codons and the existence of the 373-bp sequence since the first 38 nucleotides of c16/11 were identical to the last 38 nucleotides of the 373-bp sequence (Fig. 1A).
Two Distinct Forms of ARNT May Result from Alternative RNA Splicing- Fig. 2 shows the nucleotide and amino acid sequence constructed from the overlapping cDNAs. The complete cDNAs encode two bHLH/PAS proteins and are classified as rainbow trout ARNT (rtARNT) based on approximately 55% amino acid sequence identity to mammalian ARNT and only 19% identity to mammalian AHR, Drosophila SIM, or Drosophila PER (Fig. 1B). It is unclear whether the two isoforms of rtARNT arise from distinct genes or are generated as a result of alternative RNA splicing within the 3Ј portion of the putative open reading frame. In the human, genomic Southern blots suggested that only one ARNT gene is present (6). However, the recent identification of cDNAs for a novel murine ARNT termed ARNT2 suggests that genes of similar structure may exist (16). It is important to note, however, that the nucleotide and amino acid sequences of mARNT and mARNT2 are divergent. That is not the case within regard to the rainbow trout ARNT cDNAs described in this report. The nucleotide sequences are 100% conserved except for the presence of an additional 373-bp region in certain clones. Importantly, this sequence is flanked by consensus splice acceptor and donor sites (Ref. 44; Fig. 2, boxed sequences), and thus, it is likely that the two cDNAs originate from the same gene due to alternative RNA splicing. For this reason the encoded proteins have been named rtARNT a and rtARNT b . rtARNT a represents the protein that is missing the 373-bp sequence. The deduced amino acid sequence consists of 533 NH 2 -terminal residues that are identical to rtARNT b and a COOH-terminal region of 104 amino acids. The molecular mass of rtARNT a is 70 kDa. rtARNT b has a deduced molecular mass of 79 kDa due to the 123 residues encoded by the extra 373-bp region. The COOHterminal tail of rtARNT b contains 67 amino acids that are distinct from rtARNT a (Fig. 2).
Several features of the rtARNT proteins are worth noting. These proteins share the highest degree of amino acid sequence similarity with the human ARNT (Fig. 1B) and show 73% identity to this protein over the first 457 amino acids. Within this region, rtARNT shows 100% identity to hARNT in the basic region, 100% identity in the HLH, and 77% identity over the PAS domain. There are no amino acid additions or deletions within the 325 residues of the bHLH/PAS region of rtARNT as compared with the hARNT. Distal to the PAS region, the similarity between the rtARNT proteins and mammalian ARNT becomes significantly diminished. The amino acid sequence encoded by the 373-bp region of rtARNT b and the COOH- terminal 67 residues show 25 and 44% identity to hARNT, respectively. However, there are "pockets" of glutamine and asparagine residues in conserved locations that have been shown to be important for transactivation function in the mammalian AHR and ARNT proteins (13)(14)(15). For example, the COOH-terminal tail of rtARNT b contains the sequence VW-PQWQGQ that is also found in the hARNT. In contrast, the COOH-terminal portion of the rtARNT a shows essentially no similarity to any mammalian ARNT (Ͻ13%). The striking feature of this region is that 49% of the 104 amino acids are either proline, serine, or threonine residues. Unlike the rtARNT b sequence, there are no pockets of conserved residues that correlate to mammalian ARNT because the sequence is essentially devoid of glutamine and asparagine residues (6% QN). However, PST residues have been identified in transactivation domains of other transcription factors (45,46) and have also been shown to play a role in the turnover of nuclear regulatory proteins (47).
In Vitro Expression of rtARNT a and rtARNT b Protein and Evidence for Distinct COOH-terminal Ends-The sequence data presented in Fig. 2 show that the major differences between rtARNT a and rtARNT b are the lack of an exon that encodes 123 amino acids and the presence of distinct COOHterminal domains. To confirm these results and begin the functional analysis of these proteins, complete cDNAs for each rtARNT were constructed and ligated into pBKS or pRcCMV expression vectors as detailed under "Experimental Procedures." rtARNT protein was then expressed in vitro, resolved by SDS-PAGE, and analyzed by Western blotting with the rt-84 antibody. In vitro expression of pMVrtARNT a resulted in a protein of approximately 70 kDa, whereas pMVrtARNT b produced a larger protein migrating at approximately 79 kDa (Fig.  3). Both forms of rtARNT appeared to be expressed at similar levels by this in vitro system and were smaller than the 87-kDa mARNT (Fig. 3, lane 1). When the rtARNT proteins were stained with an antibody produced against the 460-amino acid COOH-terminal end of mARNT (R-1; Ref. 20), rtARNT b showed moderate reactivity while the rtARNT a was not stained. 2 These results are consistent with the deduced amino acid sequences and support the hypothesis that rtARNT b and rtARNT a express distinct COOH-terminal domains. To further establish 2 R. S. Pollenz, unpublished observations. that rtARNT a contained the 104-amino acid PST-rich domain distal to the splice site, clone c3 was used as a template to amplify a fragment containing a BamHI site inserted at the putative splice site (nucleotide 1601). A 540-bp fragment from the 3Ј end of rtARNT a was then ligated at the BamHI site so that the encoded protein was in the PST-rich rtARNT a reading frame (pKSrtARNT104). A Western blot of the in vitro expressed protein is shown in Fig. 3. The protein produced from pKSrtARNT104 migrated at the same molecular mass as rtARNT a but was larger than the protein produced from clone c3. Clone c3 contains only the first 1801 bp of rtARNT a ; therefore, it encodes a truncated protein of 600 amino acids that would be the expected size of rtARNT a if it contained the QN-rich rtARNT b COOH-terminal end. Collectively, these results provide strong evidence that the rtARNT a protein contains a COOH-terminal sequence that is highly divergent from mammalian ARNT proteins and is also distinct from rtARNT b .
Identification of rtARNT a and rtARNT b mRNAs and Protein in Rainbow Trout-To evaluate the expression of rtARNT a and rtARNT b in vivo, total RNA was isolated from RTG-2 cells, rainbow trout gonad tissue, and rainbow trout liver. RT-PCR was performed on these samples with a set of oligonucleotides (primers 16 and 17) that spanned the putative splice site at nucleotide 1601, and Southern blots of the PCR products were then probed with 32 P-labeled c3 cDNA as detailed under "Experimental Procedures." Correct amplification from this primer set should result in a 810-bp fragment from rtARNT a and an 1183-bp fragment from rtARNT b . The results presented in Fig.  4 show that an 810-and 1183-bp fragment are amplified from all RNA samples (Fig. 4A). 3 A longer exposure of the lane containing the liver sample is included to highlight the expression of both RNAs in this tissue (Fig. 4A, Liver-2). Overall, the expression of the rtARNT messages appear to be highest in gonad tissue and RTG-2 cells, and this trend was also observed on Northern blots of the identical RNA samples. 2 The significance of this finding is not known at the present time. Indeed, the tissue-specific regulation and expression of ARNT message or protein has not been thoroughly investigated in most species (49) and remains an important goal of future research.
To determine whether the proteins corresponding to the different rtARNT RNAs were expressed, total cell lysates were prepared from RTG-2 cells and analyzed by Western blots with the rt-84 antibody as detailed under "Experimental Procedures." Fig. 4B shows a representative blot. Several immunoreactive bands were observed in RTG-2 cells. The most highly reactive band had a molecular mass of approximately 79 kDa and co-migrated with the in vitro expressed rtARNT b . A band that co-migrated with rtARNT a was also observed, but its concentration was significantly lower than rtARNT b . Additional bands migrating with a molecular mass in the range of mammalian ARNT were also observed on some gels (6,15,16), as were lower molecular mass species. These bands may represent additional rtARNT isoforms or may simply be nonspecific reactivity of the rt-84 antibody. These results suggest that both rtARNT messages and protein are expressed in RTG-2 cells. Importantly, the results presented in this section indicate that RNA species corresponding to rtARNT a and rtARNT b can be detected in gonad and liver tissues and thus, are not an artifact of cell culture.
rtARNT Protein Expressed in Murine Hepa-1 Cells Is Nuclear-Type II and Group C Hepa-1c1 cell variants are characterized by high expression of a functional AHR protein, essentially undetectable levels of ARNT protein and nonresponsiveness to TCDD as evaluated by CYP1A1 induction and P4501A1 activity (20, 50 -52). These cells have been used as a model system to evaluate the ability of various mammalian ARNT expression constructs to restore AHR-mediated signal trans-

FIG. 4. Analysis of rtARNT expression in RTG-2 cells and rainbow trout tissues.
A, Southern blots. Total RNA was reverse-transcribed and then subjected to PCR along with pKSrtARNT a and pKSr-tARNT b as detailed under "Experimental Procedures." The amplified products were resolved on a TBE agar gel, blotted to nitrocellulose, and hybridized with 32 P-labeled c3 cDNA. Solid arrow indicates 1183-bp fragment amplified from rtARNT b . Open arrow indicates the 810-bp fragment corresponding to rtARNT a . The liver-1 and liver-2 lanes represent two exposures of the same sample. The band present between the 1183 and 810 bands represents a PCR artifact. 2 B, Western blots. Total cell lysates from RTG-2 were resolved by SDS-PAGE, blotted to nitrocellulose, and stained with 1 g/ml anti-rtARNT (rt-84), followed by GAR-HRP (1:10,000). The blots were visualized by ECL. rtARNT a and rtARNT b proteins were produced in a coupled TNT reaction as detailed under "Experimental Procedures." The molecular mass of standard proteins is indicated on the left (kDa). duction (6,11,13,15,16). Since a similar cell line has yet to be developed from any fish species, type II cells were utilized to assess rtARNT function. The first set of experiments were designed to establish whether rtARNT proteins were expressed appropriately in the type II cells. pMVrtARNT a or pMVr-tARNT b were transfected into type II cells grown on glass coverslips, fixed, and then stained with rt-84 as detailed under "Experimental Procedures." The majority of cells in each field lacked a fluorescent signal, but a few cells showing intense nuclear fluorescence were detected (Fig. 5). Cells transfected with pMVrtARNT a or pMVrtARNT b were essentially indistinguishable in (i) the subcellular location of the fluorescence, (ii) the average number of stained cells, and (iii) the relative intensity of the fluorescent signal (Fig. 5). The specificity of the antibody is demonstrated by the lack of nuclear staining in untransfected cells (Fig. 5, E and F). These results are consistent with previous immunological studies that have identified mARNT and rtARNT as a nuclear proteins (20,22,53). 4 Indeed, the rtARNT contains the sequence RXXKRR in its NH 2terminal region that is precisely conserved in all mammalian ARNT proteins and may represent the nuclear localization signal (54). 5 These results indicate that the rtARNT a and rtARNT b proteins are expressed appropriately in mammalian cells.
rtARNT a Does Not Complement Endogenous P4501A1 Protein Expression in Type II Hepa-1 Cells-Having established that both rtARNT proteins were highly expressed in the proper context in type II cells, the next set of experiments focused on whether rtARNT a or rtARNT b could complement AHR-mediated signal transduction in this model. The parameter chosen to monitor the complementation of this pathway was TCDDmediated induction of endogenous P4501A1 protein. This end point was utilized instead of traditional XRE-driven reporter gene expression due to the finding that TCDD-mediated changes in reporter gene activity in type II cells are only 3-5-fold and do not approach the high levels of change normally observed in other cell types (11,13,15,16). These results are problematic because the TCDD-mediated changes in reporter gene activity are small due to very high levels of background reporter gene expression. 6 This indicates that the reporter gene is being activated in a TCDD-independent manner in type II cells, and thus, the assessment of ARNT-specific function may not be accurate. The evaluation of TCDD-mediated changes in the endogenous expression of P4501A1 overcomes this problem since transfection of ARNT expression vectors into type II cells does not result in induction of P4501A1 without TCDD exposure (Fig. 6A).
Type II cells were transfected with pMVrtARNT a , pMVr-tARNT b , or a 1:1 mixture of both plasmids (based on weight). The parent vector (pRc/CMV) was also included in appropriate samples so that each sample was transfected with an identical amount of DNA. Following a 24-h recovery period, cells were treated with 2 nM TCDD or Me 2 SO for 20 h and total cell lysates prepared. Equal amounts of total protein were resolved by SDS-PAGE, transferred to nitrocellulose, and evaluated for rtARNT, actin, and P4501A1 protein expression as detailed under "Experimental Procedures." Fig. 6A shows Western blots from a representative experiment with each lane representing a single dish of transfected cells that were treated with TCDD. It can be observed that the concentration of rtARNT is essentially constant whether cells are transfected with pMVr-tARNT a , pMVrtARNT b , or both vectors. Identical results were observed in samples that had not been exposed to TCDD, and a representative sample is shown in Fig. 6A (lane 11). In contrast, when identical blots were stained for P4501A1, this protein was detected in lysates derived from pMVrtARNT b -transfected cells and was essentially absent in lysates derived from pMVrtARNT a -transfected cells. Importantly, P4501A1 protein was not detected in lysates derived from untreated cells (Fig.  6A, lane 11).
To quantify the level of P4501A1 protein in these samples, the relative concentration of rtARNT, ␤-actin, and P4501A1 was determined by computer analysis of the scanned bands as previously detailed (22,40). The relative concentration of P4501A1 protein was then divided by the relative concentration of rtARNT and plotted. In three independent experiments, the level of P4501A1 detected in cells expressing rtARNT a was 7 Ϯ 6% of the amount detected in cells expressing rtARNT b . These results were confirmed by staining transfected cells with antibodies against P4501A1 following TCDD exposure. Fig. 6B shows that high levels of intense perinuclear staining are predominantly observed in cells transfected with pMVrtARNT b . Cells transfected with pMVrtARNT a show low levels of P4501A1 staining, consistent with the reduced level of P4501A1 detected by Western blotting (Fig. 6A). Duplicate 4 RTG-2 cells stained with rt84 show strong nuclear reactivity (R. S. Pollenz, manuscript in preparation). 5 Preliminary deletion studies with mARNT, suggest that removal of the RXXKRR domain results in a protein with a predominantly cytoplasmic location in type II cells (R. S. Pollenz, manuscript in preparation). 6 Very high levels of basal luciferase activity were detected in type II cells co-transfected with pMVrtARNT a or pMVrtARNT b and an XREdriven luciferase reporter plasmid. However, in all experiments the level of luciferase activity associated with cells expressing rtARNT b was 3-10-fold higher than the luciferase activity associated with cells expressing rtARNT a . (H. L. Root and R. S. Pollenz, unpublished observations.) coverslips stained with rt-84 showed similar patterns of rtARNT a and rtARNT b expression as presented in Fig. 5. 2 These results support the hypothesis that rtARNT b , but not rtARNT a , can complement TCDD-induced induction of P4501A1 in vivo.
To further evaluate the functionality of the rtARNT proteins, P4501A1 expression was determined in type II cells transfected with a 1:1 mixture of pMVrtARNT a and pMVrtARNT b . Fig. 6A shows that the amount of rtARNT protein detected in lysates derived from co-transfected cells is similar to that detected in lysates of cells transfected with each individual vector. The ratio of rtARNT a to rtARNT b was approximately 0.7:1 in the experiment shown. However, the amount of P4501A1 protein present in lysates of co-transfected cells was reduced when compared with cells transfected with pMVrtARNT b alone (Fig.  6A, lanes 8 -10). The relative concentration of P4501A1 was determined by computer analysis of the scanned bands and divided by the relative concentration of rtARNT b . The results from two separate experiments are shown in Fig. 6C. Each bar represents the mean and standard deviation of three independent samples (the shaded bars represent the data presented in Fig. 6A). In both experiments, the relative concentration of P4501A1 was reduced by at least 50% when cells were cotransfected with pMVrtARNT a and pMVrtARNT b as compared with pMVrtARNT b alone. Because P4501A1 is the product of an endogenous gene, any change in its concentration between cells expressing similar levels of rtARNT b must be related to the co-transfected construct (i.e. co-expression of rtARNT a ). Indeed, the data presented in Fig. 6B did not change when normalized to the activity expressed from a ␤-galactosidase vector co-transfected in each sample (transfection efficiency). 7 These results indicate that the expression of rtARNT a can reduce the ability of rtARNT b to complement AHR-mediated signal transduction in vivo. Therefore, rtARNT a does function in the context of AHR-mediated signal transduction, but the predominant effect appears to be negative.
rtARNT a and rtARNT b Form Heterodimers with Murine AHR-The reduction in TCDD-induced expression of P4501A1 in type II cells expressing rtARNT a could be the result of inefficient formation of rtARNT a ⅐AHR complexes, decreased affinity of the rtARNT a ⅐AHR complex for DNA, or inefficient transactivation. To evaluate rtARNT binding with the AHR, type II cytosol was supplemented with identical concentrations of 35 S-labeled rtARNT protein, activated with TCDD, and immunoprecipitated as detailed under "Experimental Procedures." Fig. 7 shows the results from a representative experiment. An antibody specific to the murine AHR precipitated both rtARNT a and rtARNT b following activation of type II FIG. 6. Analysis of P4501A1 expression in TCDD-treated type II cells expressing rtARNT protein. A, 3-4 plates of type II cells were transfected with the combinations of constructs indicated in A (3.0 g of total DNA). Following a 24-h recovery, cells were treated with 2 nM TCDD for 20 h. Total cell lysates were prepared as detailed under "Experimental Procedures," and 18 g of each sample were subjected to denaturing gel electrophoresis on duplicate gels. Gels were blotted to nitrocellulose and stained with 1.2 g/ml rt-84 IgG or anti-P450 IgG (1:500) and anti ␤-actin IgG (1:1200) followed by GAR-HRP (1:10,00). The blots were visualized by ECL. B, type II Hepa-1 cells grown on glass coverslips were transfected with 1.5 g of pMVrtARNT a or pMVr-tARNT b . Following a 24-h recovery, cells were treated with 2 nM TCDD for 20 h. Coverslips were incubated with anti-P4501A1 IgG (1:500) followed by GAR-TR (1:750). C, P4501A1 expression in transfected cells. The relative level of P4501A1 and rtARNT b protein were determined by computer analysis of the autoradiograms as previously detailed (22). The relative concentration of P4501A1 was divided by the relative concentration of rtARNT b . Units are arbitrary. Each bar represents the average and S.D. of at least three samples. The solid bars represent the data from lanes 5 to 10 in A. cytosol with TCDD. Specificity is demonstrated by the lack of precipitation with preimmune IgG and the increase in rtARNT precipitated following activation with TCDD. Interestingly, the amount of rtARNT a precipitated by the AHR antibody was consistently 2-4-fold greater than rtARNT b . This trend was observed in three separate experiments and is highlighted in lane 11 of Fig. 7. In this sample, an equal amount of both rtARNT a and rtARNT b were incubated with type II cytosol and TCDD prior to immunoprecipitation. Although Western blot analysis showed that the initial sample contained equal amounts of both rtARNT proteins, rtARNT a was the predominant protein precipitated by the AHR antibody. Experiments carried out with unlabeled rtARNT a and rtARNT b produced similar results (10).
These results indicate that both rtARNT proteins can associate with the mouse AHR and are consistent with the finding that they contain identical bHLH/PAS regions that have been implicated in dimerization (8 -11). The results also indicate that the inability of the rtARNT a protein to complement AHRmediated signal transduction in type II cells is not related to reduced formation of rtARNT a ⅐AHR complexes. Indeed, the results suggest that rtARNT a may form heterodimers with greater efficiency than rtARNT b even if the complexes do not function in the context of the CYP1A1 promoter.
rtARNT a Has Reduced Affinity for Core XRE Sequences and Can Disrupt the Binding of rtARNT b ⅐AHR Complexes-To evaluate the interaction of rtARNT a ⅐AHR and rtARNT b ⅐AHR complexes with DNA, identical amounts of rtARNT a or rtARNT b protein were combined with type II cytosol and activated with TCDD as detailed under "Experimental Procedures." The formation of functional rtARNT⅐AHR complexes was then evaluated by electrophoretic mobility shift assays (EMSA) using a oligonucleotide containing the putative XRE sequence 5Ј-TT-GCGTG-3Ј (41). Fig. 8A shows the results of a typical assay. Samples containing rtARNT b and type II cytosol were able to shift the putative XRE following TCDD activation. The presence of both rtARNT and mAHR is demonstrated by the lack of a shifted band when antibodies against mAHR or rtARNT are added to the activated sample prior to the XRE. Addition of preimmune IgG did not affect the assay. These results are consistent with the positive function of rtARNT b in AHR-mediated signal transduction shown in Fig. 6. In contrast, type II cytosol supplemented with rtARNT a and activated with TCDD produced a very weak response in the gel shift assay. In four independent experiments, the amount of XRE shifted by rtARNT a was 5 Ϯ 5% of the amount shifted by an identical amount of rtARNT b . These results suggest that rtARNT a does not complement AHR-mediated signal transduction due to reduced rtARNT a •AHR binding at the XRE. Because of this finding, it was pertinent to investigate whether rtARNT a could negatively influence the association of rtARNT b •AHR dimers with the XRE.
Samples containing a constant amount of rtARNT b protein, type II cytosol, and 32 P-labeled XRE were supplemented with increasing amounts of rtARNT a and then activated with TCDD or Me 2 SO as detailed under "Experimental Procedures." The samples were then evaluated by EMSA. Consistent with the data presented in Fig. 8A, type II cytosol supplemented with only rtARNT a did not show a strong shift even when the amount of rtARNT a was increased 6-fold (Fig. 8D, lanes 8 and  9). However, when both rtARNT a and rtARNT b were incubated with type II cytosol, the concentration of shifted XRE became dramatically reduced as the ratio of rtARNT a to rtARNT b increased (Fig. 8D, lanes 10 -14). When the ratio of rtARNT a to rtARNT b was 4:1, the specifically shifted XRE showed a faster electrophoretic mobility and was reduced by at least 60% when compared with the concentration of shifted XRE in type II cytosol supplemented with only rtARNT b . At a 6:1 ratio of rtARNT a to rtARNT b , the concentration of shifted XRE was reduced by 86% and showed a similar mobility to type II samples supplemented with only rtARNT a (Fig. 8D, compare lanes  9 and 14). Indeed, the reduced concentration of shifted XRE and the faster mobility are consistent with the presence of rtARNT a •AHR dimers (Fig. 8A). To confirm that these results were directly related to the presence of rtARNT a protein, and not the result of inhibition of the assay by high concentrations of TNT lysate, type II cytosol was supplemented with increasing amounts of rtARNT b and evaluated by EMSA. Fig. 8C shows that there is an increase in the amount of shifted XRE as the concentration of rtARNT b is increased and that the shift only occurs if TCDD is present.
These results indicate that the negative effect produced by rtARNT a on rtARNT b function is a reduction in rtARNT b ⅐AHR complexes associated with the XRE. The mechanism underlying this result is likely related to the formation of FIG. 8. Electrophoretic mobility shift analysis of rtARNT protein. A, 100 g of type II cytosol was combined with 50 ng of rtARNT a or rtARNT b and activated with 10 nM TCDD or 0.5% Me 2 SO for 2 h at 30°C. EMSA was then performed as detailed under "Experimental Procedures." Closed arrows show the specifically shifted band associated with rtARNT b ⅐AHR⅐XRE complex. Open arrows show the specifically shifted band associated with rtARNT a ⅐AHR⅐XRE complex. rtARNT a or rtARNT b did not shift the XRE in the absence of type II cytosol (lanes 12, 13). B, Western blot of the TNT lysates containing rtARNT a and rtARNT b . Both proteins are present at a similar concentration. C, 8 g of type II cytosol was combined with the indicated concentration of rtARNT b and activated with 10 nM TCDD or 0.5% Me 2 SO for 2 h at 30°C. EMSA was then performed as detailed under "Experimental Procedures." D, 8 g of type II cytosol was combined with the indicated concentration of rtARNT a and rtARNT b and activated with 10 nM TCDD or 0.5% Me 2 SO for 2 h at 30°C. EMSA was then performed as detailed under "Experimental Procedures." Closed arrow shows the specifically shifted band associated with rtARNT b ⅐AHR⅐XRE complexes. Open arrow shows the specifically shifted band associated with rtARNT a ⅐AHR⅐XRE complexes. rtARNT a ⅐AHR complexes that have low affinity for the XRE. Since the AHR is not a highly expressed protein (55,56), and appears to be rapidly depleted following ligand binding (22), the formation of nonfunctional rtARNT a ⅐AHR complexes would reduce the number of AHRs capable of associating with rtARNT b and thus abrogate TCDD-mediated changes in gene regulation. This hypothesis is consistent with the mechanism whereby other dominant negative regulators function. For example, the Id family of proteins are thought to reduce function of various bHLH proteins by forming heterodimers that are not capable of binding E-box sequences (57,58). In addition, ectopic expression of PER disrupts TCDD-mediated signaling in mammalian cells by reducing the number of functional ARNT⅐AHR complexes associated with the XRE (11). The negative activity of Id and PER is related to the absence of DNA binding basic regions in these proteins; therefore, they cannot regulate the transcription of genes. In contrast, both rtARNT proteins contain identical bHLH and PAS domains that have been shown to function in dimerization and DNA binding of ARNT⅐AHR complexes. Therefore, the results presented in this report suggest that the COOH-terminal region of ARNT can influence the ability of ARNT⅐AHR complexes to associate with XRE sequences. This function was likely missed in previous experiments that evaluated COOH-terminal domains of the AHR and ARNT because function was exclusively evaluated by deletion of the COOH-terminal end (13)(14)(15).
Conclusions and Implications-Dominant negative activity in AHR-mediated signal transduction pathway has been described in variants of the Hepa-1 cell lines (59). The molecular basis of the dominant negative phenotype is undefined. The results presented in this report suggest that (i) the basic, helixloop-helix or PAS domains do not have to be modified or deleted to produce a bHLH/PAS protein that can interact negatively on AHR-mediated signal transduction, and (ii) that changes in the amino acid composition of the COOH-terminal region of ARNT can influence the function of domains in the NH 2 -terminal region. These results are intriguing due to the observation that most of the heterogeneity within the different alleles of the AHR and among the various ARNT proteins occurs in the COOH-terminal regions (16,61,62). The structural modifications caused by the changes in COOH-terminal amino acid sequence will likely require x-ray crystallographic data to determine whether COOH-and NH 2 -terminal regions are in close proximity to each other or whether structural changes in the COOH-terminal region affect the folding of the NH 2 terminus.
The finding that the changes in COOH-terminal regions of the rtARNT proteins are due to the presence or absence of nucleotide sequence that is flanked by consensus splice acceptor and donor sites suggests that the two RNA species may be produced by alternative RNA splicing. There is precedent for alternative RNA splicing in the bHLH-PAS family of proteins. The human ARNT (6) and rat AHR (49) both appear to have splice variants, although the functional significance on AHRmediated events is unknown. Alternative RNA splicing has been identified as a mechanism for generating functional diversity within numerous families of genes (41,62,63). For example, the two isoforms of the prostaglandin EP 3 receptor are formed because of an alternatively spliced exon in the 3Ј portion of the gene that results in a change in reading frame and two distinct COOH-terminal ends (63). The similarity of EP 3 mechanism to the results described in this report suggest that alternative RNA splicing may be of similar importance in generating functional diversity within the bHLH/PAS family of proteins. Since the RTG-2 cells are inducible by TCDD (43), and the ratio of rtARNT b to rtARNT a is high, it is unclear how rtARNT a functions in AHR-mediated events of RTG-2 cells at this time. However, alternative splicing can occur in a temporal, tissue-specific, or developmental fashion (64). Therefore, the analysis of this process, the physiological function of rtARNT a in fish, the tissue-specific regulation of the rtARNT a and rtARNT b proteins, and the identification of structural homologues in mammals remain a focus of future research.