Identification and Functional Characterization of Two Highly Divergent Aryl Hydrocarbon Receptors (AHR1 and AHR2) in the TeleostFundulus heteroclitus

The aryl hydrocarbon receptor (AHR) is a ligand-activated transcription factor through which 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds cause altered gene expression and toxicity. The AHR belongs to an emerging multigene family of transcription factors possessing basic helix loop helix (bHLH) and Per-ARNT-Sim (PAS) domains. Most bHLH-PAS proteins occur as duplicates or “paralog groups” in mammals, but only a single mammalian AHR has been identified. Here we report the cDNA cloning of two distinct AHRs, designated FhAHR1 and FhAHR2, from a single vertebrate species, the teleost Fundulus heteroclitus (Atlantic killifish). Both Fundulus AHR proteins possess bHLH and PAS domains that are closely related to those of the mammalian AHR. FhAHR1 and FhAHR2 are highly divergent (40% overall amino acid identity; 61% identity in the N-terminal half), suggesting that they arose from a gene duplication predating the divergence of mammals and fish. Photoaffinity labeling with 2-azido-3-[125I]iodo-7,8-dibromodibenzo-p-dioxin and velocity sedimentation analysis using 2,3,7,8-[1,6-3H]TCDD showed that both FhAHR1 and FhAHR2 exhibit specific, high-affinity binding of dioxins. Both AHRs also showed specific, TCDD- and ARNT-dependent interactions with a mammalian xenobiotic response element. The two Fundulus AHR genes displayed different tissue-specific patterns of expression; FhAHR1 transcripts were primarily expressed in brain, heart, ovary, and testis, while FhAHR2 transcripts were equally abundant in many tissues. Phylogenetic analysis demonstrated that Fundulus AHR1 is an ortholog of mammalian AHRs, while AHR2 forms in Fundulus and other fish are paralogous to Fundulus AHR1 and the mammalian AHRs and thus represent a novel vertebrate subfamily of ligand-binding AHRs.

The aryl hydrocarbon receptor (AHR) 1 is a ligand-activated transcription factor through which 2,3,7,8-tetrachlorodibenzop-dioxin (TCDD) and other polyhalogenated and polycyclic aromatic hydrocarbons cause altered gene expression and toxicity (1)(2)(3). The AHR and its dimerization partner ARNT (AHR nuclear translocator) belong to an emerging family of transcription factors that contain basic helix loop helix (bHLH) and Per-ARNT-Sim (PAS) domains (4,5). Many of the mammalian bHLH-PAS family members occur as "paralog groups," each of which contains duplicated genes with related functions (two to three genes per group, reviewed in Refs. 5 and 6). In contrast, a mammalian AHR paralog group like those of other bHLH-PAS proteins has not been reported.
Members of the bHLH-PAS family play key roles in development, hypoxia-induced gene expression, the control of circadian rhythmicity, and phototransduction (reviewed in Refs. 4 and 5). Unlike most other bHLH-PAS family proteins, the AHR is ligand-activated, although its endogenous ligand is not known. The AHR acts in concert with ARNT to regulate target genes such as cytochrome P450 -1A1 (CYP1A1) via binding to consensus xenobiotic responsive enhancer elements (XREs) (3,7). In addition to its role in regulating drug-metabolizing enzymes, possible endogenous functions for the AHR include cell cycle control (8,9), developmental signaling (10 -12), and modulation of growth factor signal transduction pathways (13,14).
We are using the estuarine fish Fundulus heteroclitus (mummichog or Atlantic killifish) as an early vertebrate model for studying the evolution and adaptive significance of AHR signaling pathways. This species has several features that make it a valuable experimental system (15). F. heteroclitus is abundant in both pristine and polluted environments; it is sensitive to dioxins but is capable of developing heritable dioxin resistance after multigenerational exposure (16 -18). Unlike some nonmammalian vertebrate models, such as rainbow trout and Xenopus laevis, F. heteroclitus is diploid (19), facilitating the analysis of gene function and diversity. The developmental biology and population genetics of F. heteroclitus are well known (20,21). Finally, this species possesses an inducible CYP1A1 (22), indicating the existence of a functional AHR signaling system.
Recently, we reported the presence of AHR (23,24) and ARNT (25) homologs in F. heteroclitus. In the course of these studies, we obtained evidence for a second F. heteroclitus AHR gene. Partial sequences of these two Fundulus AHRs (PAS domains only) were reported as part of a phylogenetic analysis of the bHLH-PAS family (6). Here we report the full cDNA sequence and functional characterization of both F. heteroclitus AHRs (FhAHR1 and FhAHR2). These AHR paralogs are highly divergent (40% overall amino acid identity; 61% identity in the more conserved N-terminal half), yet both proteins exhibit specific, high affinity binding of dioxins as well as ligand-and ARNT-dependent binding to XRE sequences. Differences in the structural features, ligand-binding specificity, and patterns of expression of FhAHR1 and FhAHR2 suggest distinct functions.
The cloning of two distinct AHRs from F. heteroclitus, a teleost fish, provides the first definitive evidence for a vertebrate AHR paralog group. A single AHR form has been cloned from each of several mammalian species (26 -30). AHR fulllength cDNA sequences have been reported from three other species of fish: tomcod (31), zebrafish (32), and trout (33). In trout, two AHR forms have been identified. However, these share 97% amino acid identity, indicating that they are derived from a restricted gene duplication related to the recent tetraploid ancestry of salmonid fish. Here we show that F. heteroclitus AHR2 and all of the full-length AHRs reported from other fish species (including both trout AHRs) belong to a distinct AHR2 clade that is paralogous to, and separate from, FhAHR1 and the mammalian AHRs reported to date. The AHR2 clade thus represents a novel subfamily of ligand-binding bHLH-PAS proteins within the vertebrate AHR family. EXPERIMENTAL (35). Degenerate primers for ␤-actin (36) were a gift from Dr. M. Sogin (Marine Biological Laboratory, Woods Hole, MA). These were used to amplify an 1100-bp partial ␤-actin cDNA from Fundulus liver by RT-PCR with annealing at 52°C. Fundulus-specific actin primers (actin F and actin R) were designed based on the sequence of this fragment.

Fish Collection and RNA Isolation
F. heteroclitus (mummichog) were collected in salt marshes on Cape Cod, MA. The dissected organs were immediately frozen in liquid nitrogen. Total RNA was isolated using RNA STAT-60 (Tel-Test B, Inc.). Poly(A) ϩ RNA was purified with oligo(dT) spin columns (5 Prime 3 3 Prime, Inc.).

RT-PCR
One g of poly(A) ϩ RNA was reverse transcribed with random hexamers using the Gene-Amp RNA-PCR kit (Perkin-Elmer) following the manufacturer's directions. The anneal/extend temperature varied (50 -60°C) based on the primers used. In some cases, an additional extension period of 1 min at 72°C was added to the cycling.

Cloning and DNA Sequencing
The PCR products were cloned into pCNTR (5 Prime 3 3 Prime, Inc.), pT7Blue(R) (Novagen), or pGEM-T Easy (Promega). Large inserts were digested with various restriction enzymes and subcloned into pBluescript KS Ϫ (Stratagene) for sequencing using SequiTherm Excel II long-read cycle sequencing kits (Epicentre Technologies, Madison, WI) and an automated DNA sequencer (LI-COR 4000L). Both strands from multiple clones were sequenced to ensure accuracy.

Genomic Library Screening
The first Fundulus FixII genomic library (Powers) was screened using mouse AHR cDNA fragments of 423 and 700 bp (spanning amino acid residues numbers 26 -420) that were labeled with [␣-32 P]dCTP by random priming. Hybridization was in 20% formamide, 6 ϫ SSPE, 5 ϫ Denhardt's, and 0.1% SDS at 37°C. Filters were washed in 2 ϫ SSC, 0.1% SDS at 48°C. Approximately 3 ϫ 10 5 plaques were screened. One positive plaque was identified and DNA was isolated. Multiple restriction enzymes were used to digest the DNA and several fragments were subcloned into pBluescript (Stratagene) and sequenced. Approximately 4 ϫ 10 5 plaques from the second Fundulus genomic library (Crawford) were screened using a 690-bp partial FhAHR2 cDNA fragment (amino acid residues number 117-331) (24) as described above. Seven positive plaques were identified and DNA was isolated. Restriction fragments from these genomic clones were subcloned and sequenced.
cDNA Library Construction and Screening F. heteroclitus liver and heart cDNA libraries were constructed using ZAP-cDNA Gigapack II Gold cloning kit (Stratagene). Approximately 1.9 ϫ 10 6 plaques were screened from each library using 32 P-labeled partial FhAHR1 (amino acid residues numbers 252-440) and FhAHR2 (amino acid residues numbers 117-331) cDNAs. Filters were prehybridized in 50% formamide, 6 ϫ SSPE, 0.1% SDS, and 0.05 ϫ blotto at 42°C. Hybridization was carried out in the same solution at 42°C overnight. Filters were washed at 65°C in 2 ϫ SSC, 0.1% SDS prior to autoradiography. One FhAHR2 positive plaque from the heart cDNA library was isolated and the pBluescript phagemid containing the insert was excised and sequenced.

5Ј-3Ј RACE
Adaptor-ligated, oligo(dT)-primed, double-stranded liver or ovary cDNA was synthesized using a Marathon cDNA Amplification kit (CLONTECH). Gene specific primers were coupled with adaptor primers in the PCR reactions and the products were cloned and sequenced. For FhAHR1 5Ј-RACE, primer 1-256R and nested primer 1-239R were used. For FhAHR1 3Ј-RACE, primer 1-780F and nested primer 1-846F were used. For FhAHR2 3Ј-RACE, primer 2-410F and nested primer 2-475F were used. In a modified FhAHR1 5Ј-RACE procedure, cDNA synthesis was performed at a higher temperature to overcome secondary structure in RNA. The first-strand cDNA was synthesized with the Thermoscript RT System (Life Technologies) at 70°C with a genespecific primer, 1-317R. Following second-strand synthesis and adaptor ligation, primers 1-109R and nested 1-76R were used for the PCR.
pDPFhAHR2-The M2f/2-1090R primer pair was used to amplify a 3250-bp FhAHR2 cDNA from liver. The Kozak consensus sequence was incorporated into the 5Ј primer M2f. The PCR product was digested with XhoI and KpnI (primer sites) and ligated into the SalI/KpnI sites of pDP18 (Ambion, Inc.). Advantage cDNA polymerase mixture (CLON-TECH) was used in the PCR reactions to maximize fidelity of amplification. All FhAHR1␣ and FhAHR2 cDNA constructs were verified by complete sequencing.

In Vitro Protein Synthesis
TNT Quick Coupled Reticulocyte Lysate Systems (Promega) were used to synthesize unlabeled or 35 S-labeled proteins following manufacturer's directions. Five l of the TNT reactions were subjected to SDS-polyacrylamide gel electrophoresis, followed by fluorography and autoradiography. The 35 S-labeled proteins were quantified by scintillation counting of excised gel fragments.

Velocity Sedimentation Analysis
AHR proteins were synthesized by in vitro transcription and translation and analyzed by velocity sedimentation on sucrose gradients in a vertical tube rotor (41). For each AHR, two identical TNT reactions (100 l total) were combined, diluted 1:1 with MEEDMG buffer (25 mM MOPS, pH 7.5, 20°C, containing 1 mM dithiothreitol, 1 mM EDTA, 5 mM EGTA, 0.02% NaN 3 , 20 mM Na 2 MoO 4 , 10% (v:v) glycerol (23) plus a mixture of protease inhibitors (23)), split into two 100-l aliquots, and incubated with [ 3 H]TCDD (2 nM) Ϯ TCDF (400 nM) for 1-2 h at 4°C. The [ 3 H]TCDD concentration was verified by sampling each tube for total counts. After incubation, 90 l from each tube were applied to 10 -30% sucrose gradients that were prepared in MEEDMG buffer using the method of Coombs and Watts (42). The gradients were spun for 140 min at 60,000 rpm at 4°C in a VTi 65.2 rotor, after which 150-l fractions were counted using a Beckman LS5000TD scintillation counter. Specific binding is defined as the difference between total binding (incubations containing [ 3

Electrophoretic Mobility Shift Assay
Electrophoretic mobility shift assays were performed essentially as described (25), using combinations of proteins synthesized in TNT lysates (2.5 l each) as detailed in the legend to Fig. 7. Following incubation of the proteins for 2 h in the presence of dimethyl sulfoxide or 25 nM TCDD in dimethyl sulfoxide, 1 g of poly[d(I⅐C)] (Roche Molecular Biochemicals) was added, NaCl concentration adjusted to 100 mM, and reactions incubated an additional 15 min with or without 100 ng of unlabeled mouse XRE probe or mutated XRE probe (mutated core sequence is 5Ј-TTGCTTG-3Ј; mutated base underlined). Finally, 1 ng of 32 P-end labeled mouse XRE probe (mouse DRE3 (44) approximately 10 5 cpm) was added and the reactions incubated for 15 min more. Samples were electrophoresed on 4% polyacrylamide (29:1 acrylamide:bis) gels in 0.5 ϫ TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA). The dried gel was subjected to autoradiography.
Prior to use in the DNA binding assay, unprogrammed and ARNTcontaining TNT lysates were extracted once with a quantity of dextrancoated charcoal equal to the mass of total protein in the sample. Because Fundulus AHR proteins are inactivated by charcoal treatment, 2 TNT lysates were extracted once with dextran-coated charcoal prior to AHR synthesis (0.1 mg/mg protein). Charcoal treatments reduced sequence-specific, constitutive background XRE binding.

Tissue-specific Expression
Poly(A) ϩ RNA (1 g) was reverse transcribed with random primers (Gene-Amp RNA-PCR kit, Perkin-Elmer) and an equal aliquot of cDNA was used in each of three PCR reactions with AmpliTaq Gold DNA polymerase (Perkin-Elmer). The primer pairs used were 1F/1Ra, 2F/ 2Ra, and actin F/actin R for FhAHR1, FhAHR2, and ␤-actin, respectively. The primer pairs for FhAHR1 and FhAHR2 were designed to overlap exon-exon junctions, thereby avoiding amplification of any contaminating genomic DNA. All three sets of primers were designed to amplify similar length fragments (372, 346, and 349 bp for FhAHR1, FhAHR2, and ␤-actin, respectively) to minimize differences in the efficiency of amplification. The linear range for the PCR was determined by varying the number of cycles from 20 to 35 with 3-cycle increments and using 2, 4, 6, and 8 l of template cDNA (data not shown). Five l of template cDNA and 28 cycles were chosen as the amplification parameters. The cycling conditions were: 95°C/10 min (94°C/15 s, 60°C/30 s) for 28 cycles; 72°C/7 min. Under these conditions, the amount of PCR products amplified from gonad cDNA was linearly related to cycle number and amount of template. Ten-l aliquots of each reaction (volume verified to be in the linear range for imaging) were subjected to agarose gel electrophoresis and subsequent ethidium bromide staining. The gels were photographed using a Kodak DCS 200 digital camera and Adobe Photoshop software. The integrated density of each amplified fragment was determined from the digital image by NIH Image software, version 1.60.
The relative abundance of FhAHR1␣ and FhAHR1␤ transcripts in different tissues was assessed via semi-quantitative RT-PCR as described above, except that cDNA synthesis was primed with oligo(dT). The upstream primers (1S-888F for ␣ and 1L-888F for ␤) spanned the splice junction at which the sequences diverge. The downstream primers (1S-1015R for ␣ and 1L-960R for ␤) hybridized to the 3Ј-UTR in each transcript. The amplified fragments were 379 and 216 bp for FhAHR1␣ and FhAHR1␤, respectively.

RESULTS
Identification of Two AHR Genes in F. heteroclitus-We initially used two approaches to identify and clone a Fundulus AHR. In the first approach, RT-PCR using liver RNA and degenerate primers AhR-A1/AhR-B1 produced a 690-bp partial AHR cDNA, the sequence of which we reported previously (24).
For the second approach, mouse AHR cDNA fragments (26) were used as probes to screen a Fundulus genomic library at low stringency. One positive plaque was isolated and several contiguous fragments spanning 9 kilobases from this genomic clone were subcloned and sequenced. Coding regions were identified based on sequence similarity to mammalian AHR exons 7, 8, 9, and 10 (26, 45). Surprisingly, the deduced amino acid sequence of the genomic clone shared only 62% identity with the RT-PCR product, suggesting the presence of an additional AHR gene in F. heteroclitus. To confirm the presence of two AHR genes in this species, a second genomic library was screened at low stringency using the 690-bp RT-PCR product as probe. The seven positive clones belonged to two groups, corresponding to the two sequences obtained previously. These two genes were designated FhAHR1 (first identified from the genomic clone) and FhAHR2 (first identified as the RT-PCR product), for reasons explained under "Discussion." The complete coding sequences of FhAHR1 and FhAHR2 cDNAs were obtained via additional RT-PCR, screening of cDNA libraries, and RACE procedures, as illustrated in Fig. 1 and described below. The full-length sequences reported here confirm and extend our earlier suggestion of two AHRs in this species, based on partial sequences (6,24).
FhAHR1 cDNA-Using primers 1-252F and 1-439R, designed based on the FhAHR1 genomic DNA sequence, a 567-bp fragment was amplified from heart cDNA. Degenerate primers AhR-A1 and Qf were coupled with specific primers 1-239R and 1-191R in RT-PCR reactions that yielded an additional 593 bp of coding sequence. The remainder of the coding and UTR sequence for FhAHR1 was obtained by 5Ј-and 3Ј-RACE (Fig.  1). Two distinct 3Ј-RACE products (1␣ and 1␤) were identified (Fig. 2). Sequence of genomic clones in this region indicated that the 1␣ and 1␤ cDNAs are generated by differential exon splicing at a point corresponding to codon 876 (Fig. 2). As assessed by semi-quantitative RT-PCR, FhAHR1␣ is the predominant or only form expressed in all tissues; a very low level of FhAHR1␤ expression was detectable only in ovary (data not shown). Based on these results, we chose to concentrate our subsequent characterization efforts on the FhAHR1␣ cDNA.
FhAHR1␣ has an open reading frame of 942 amino acid residues that encodes a protein with a predicted molecular mass of 103.7 kDa.
FhAHR2 cDNA-Screening of Fundulus heart and liver cDNA libraries with the 690-bp RT-PCR product resulted in isolation of a partial (1.84 kilobases) heart cDNA. The N-terminal sequence, including 45 bp of 5Ј-UTR for FhAHR2, was obtained by cRACE. A 1.9-kilobase 3Ј-RACE product provided the rest of the coding sequence and 560 bp of 3Ј-UTR (Fig. 1). The full-length FhAHR2 contains 905 amino acids with a predicted molecular mass of 99.9 kDa (Fig. 3).
Sequence Comparisons-Alignment of fish and mammalian AHRs (Fig. 4) reveal a striking dichotomy between the well conserved N-terminal half and the poorly conserved C-terminal half (termed "hypervariable" by others (28)). The sharp boundary between the two regions (residue ϳ406 -414; see Fig. 4) corresponds closely with the C-terminal boundary of the ligand-binding domain (26, 46 -48). Because of the low sequence identity among AHRs within the distal half of the protein and the resulting uncertainties in the C-terminal alignments, comparisons of the N-terminal halves of the proteins are most informative. FhAHR1 is most closely related to the mammalian AHRs in this region (70 -71% identity) whereas FhAHR2 is most closely related to the tomcod AHR (74% identity) and shares less sequence identity with mammalian AHRs (60 -62%) (Table I). Importantly, FhAHR1 and FhAHR2 are as different from each other in this region (61% identity) as each is from other AHR sequences (Table I).
Within the conserved N-terminal region, there are three segments of high sequence identity separated by short stretches of highly variable residues (Fig. 4). The first conserved segment encompasses the basic and helix loop helix motifs identified previously in mammalian AHRs. Fundulus AHR1 and Fundulus AHR2 share 83 and 73% identity, respectively, with the human AHR in this region. Notably, both Fundulus AHRs possess the basic 1 and basic 2 domains found in mammalian AHRs (26,49). In addition, specific basic region residues identified as essential for XRE binding of the mammalian AHR (mouse residues P 34 SKRHR 39 (50)) and for nuclear localization (Arg 12 -His 38 (51)) are retained in both FhAHR1 and FhAHR2. The nuclear export sequence (Lys 62 -Ser 72 (51)) is also well conserved in both Fundulus AHRs.
The second conserved segment is the ϳ67 amino acid region containing the PAS-A box (residues 105-171 of FhAHR1). The third segment is the ϳ210 amino acid region (residues 196 -409 of FhAHR1) that includes the PAS-B box as well as large flanking regions. This portion encompasses the putative ligand-binding domain of the mammalian AHRs (mouse aa 230 -397 (26, 46 -48)). Within this ligand-binding domain, FhAHR1 and FhAHR2 share 79 and 61% amino acid identity, respectively, with the human AHR. Previously, residue 375 of the murine AHR has been identified as a site where a A 3 V change results in a 4-fold decrease in binding affinity (52)(53)(54). The equivalent residues in Fundulus AHR1 and AHR2 are alanine (Ala 376 and Ala 373 , respectively), as in the high affinity mouse AHR b-1 allele. Another residue shown to be critical for AHR function, Cys 216 (55), is also conserved in FhAHR1 and FhAHR2.
Although overall sequence relatedness, conserved bHLH and PAS domains, and functional characteristics (see below) identify Fundulus AHR1 and AHR2 both as Ah receptors, there are features of the primary sequence that distinguish these two proteins. 1) An "LXCXE" motif, which has been suggested to be important for interaction of the mammalian AHR with the retinoblastoma (Rb) protein (56), is also present in Fundulus AHR1, whereas in Fundulus AHR2 the sequence "MYCAD" occurs at this location (Fig. 4). 2) The C-terminal half of the Fundulus AHR1 includes a glutamine (Gln)-rich region (Fig. 2); a similar domain has been identified in mammalian AHRs (26) and has been implicated in their transactivation function (47,(57)(58)(59)(60). In contrast, the Gln-rich domain is reduced in Fundulus AHR2. However, other potential transactivation domains, including acidic and P/S/T-rich domains, occur in the distal halves of both FhAHR1 and FhAHR2 (see Figs. 2 and 3).
Phylogenetic Analyses-Earlier, we inferred the relationship of these two Fundulus AHRs to other AHR sequences using partial amino acid sequences encompassing only the PAS domain (6). Phylogenetic analysis of the full-length Fundulus AHR1 and AHR2 sequences with other published full-length AHRs (Fig. 5) shows that the vertebrate AHRs fall into two distinct groups, consistent with our earlier hypothesis. In both distance and maximum parsimony analyses, Fundulus AHR1 and all mammalian AHRs form a monophyletic group (clade), while Fundulus AHR2 clusters with other recently identified fish AHRs. These distinct AHR1 and AHR2 clades are strongly supported in both analyses. F. heteroclitus is the first verte-brate species shown to possess both AHR types. The recently identified AHR homologs in Caenorhabditis elegans and Drosophila fall outside of the vertebrate AHR clusters. The topology of these and other trees resulting from phylogenetic analysis of AHR sequences (6) are consistent with a duplication of an ancestral AHR gene early in vertebrate evolution, prior to the divergence of tetrapods and ray-finned fishes.
In Vitro Expression and Functional Characterization of FhAHR1 and FhAHR2-The functional characteristics of the two Fundulus AHRs, including their capacity for ligand binding, was assessed following their expression in a coupled in vitro transcription and translation system. Labeling with [ 35 S]methionine demonstrated that proteins of the predicted size were synthesized in this system (Fig. 6A).
The ligand binding activities of the in vitro synthesized FhAHR1 and FhAHR2 were evaluated by photoaffinity labeling and velocity sedimentation analysis. The photoaffinity ligand 2-azido-3-[ 125 I]iodo-7,8-dibromodibenzo-p-dioxin specifically labeled both AHR proteins (Fig. 6B), as observed for mammalian AHRs expressed in vitro (26,28) and in vivo (61). Ligand binding was also measured using the reversible ligand [ 3 H]TCDD. Velocity sedimentation on sucrose gradients showed binding of [ 3 H]TCDD (2 nM) that was displaced by a 200-fold excess of unlabeled ligand (Fig. 6C). Such specific binding was not observed with unprogrammed lysate (Fig. 6C) or in lysate programmed with an empty expression vector (not shown). The sedimentation coefficients of FhAHR1 (9.2 S) and FhAHR2 (10.6 S) were similar to those obtained with the in vitro expressed human AHR (10.7 S; Fig. 6C) and mouse AHR (9.8 S; not shown) as well as those reported for mammalian AHRs in cytosol (62,63). The amount of specific binding obtained with the Fundulus AHRs (FhAHR1: 8 fmol/reaction; FhAHR2: 10 fmol/reaction; means of three experiments at 2 nM [ 3 H]TCDD) was less than that obtained with mouse and human AHRs under these conditions (64 and 37 fmol/TNT reaction, respectively).
DNA Binding-The strong sequence conservation of the DNA-binding regions of both Fundulus AHRs suggests that both of these proteins may bind canonical XRE sequences. We assessed the DNA binding ability of these proteins using electrophoretic mobility shift assays with TNT-synthesized proteins (Fig. 7). Both proteins exhibited TCDD-stimulated DNA binding in conjunction with human ARNT1. DNA binding was sequence-specific; the shifted band could be displaced by a 100-fold excess of unlabeled XRE, but not by an XRE with a point mutation in the core binding sequence (Fig. 7). Unprogrammed reticulocyte lysate did not support specific XRE binding (Fig. 7).
FhAHR1 exhibited some sequence-specific DNA binding in the absence of exogenous TCDD. This interaction could be reduced by pretreating the TNT lysates with dextran-coated charcoal, suggesting that this XRE binding is not ligand-independent but rather due to a ligand intrinsic to the lysate. Similar TCDD-independent DNA binding was observed with FhAHR2, but this interaction was completely abolished with charcoal treatment; thus, the putative intrinsic ligand may have a higher affinity for FhAHR1. Seemingly constitutive XRE binding by AHR⅐ARNT complexes in electrophoretic mobility shift assays has also been observed with mammalian AHR proteins synthesized in vitro (Fig. 7 and Refs. 25, 28, and 64) as well as with other fish AHRs (32,33).
Tissue Distribution of Fundulus AHR1 and AHR2 mRNA-The tissue-specific expression of FhAHR1 and FhAHR2 transcripts was measured using a semi-quantitative RT-PCR method. FhAHR1 was expressed predominantly in brain, heart, ovary, and testis; expression of this transcript was detectable but very low in liver, kidney, and gill tissues (Fig. 8). In contrast, uniform levels of FhAHR2 expression were detected in heart, liver, ovary, testis, brain, kidney, and gill of adult Fundulus. DISCUSSION We describe here the presence of two highly divergent forms of AHR in a single vertebrate species, the teleost F. heteroclitus. These data provide the first molecular evidence for an AHR paralog group like those known for other bHLH-PAS proteins, and define a novel subfamily of ligand-binding bHLH-PAS proteins.
Fundulus AHR1 and AHR2 Are Both Homologs of the Mammalian AHR-Fundulus AHR1 and AHR2 share only 40% amino acid sequence identity overall (61% in the N-terminal half), but they each possess the essential features that define the original AHR (1,26). Our conclusion that both of these killifish genes are homologs of the mammalian AHR, rather than of other members of the PAS family, is based on several criteria.
(a) Both Fundulus AHRs share significant sequence identity with mammalian AHRs and are more closely related to these AHRs than to any other known PAS protein. In phylogenetic analyses of all PAS protein sequences, the fish and mammalian AHR sequences cluster together in a monophyletic group of vertebrate AHRs that is strongly supported by statistical analysis (6). Importantly, this clade containing FhAHR1 and FhAHR2 falls within the larger AHR gene family that also includes the invertebrate AHR homologs recently identified in C. elegans and Drosophila (6,65,66). This placement within the larger AHR family supports our designation of both Fundulus proteins as AHRs.
(b) Like mammalian AHRs, both killifish AHRs possess a PAS domain and a bHLH motif (as defined by Burbach et al. (26)). Among bHLH proteins, the bHLH domain of the mammalian AHR is distinctive (49,67). The bHLH domains of the Fundulus AHR proteins share high amino acid identity with those of mammalian AHRs (Fundulus AHR1, 83% identity; Fundulus AHR2, 73% identity; Fig. 4). Moreover, the bHLH domains of both FhAHR1 and FhAHR2 exhibit 100% conser- vation of the specific residues that have been shown to be essential for XRE binding of the mammalian AHR (P 34 SKRHR 39 ) (50).
(c) Both Fundulus AHR1 and Fundulus AHR2 exhibit specific, high-affinity binding of dioxins as assessed by photoaffinity labeling and velocity sedimentation analysis (Fig. 6). Mammalian AHRs exhibit a similar dioxin binding ability (26,28). In contrast, an invertebrate (C. elegans) AHR homolog does not appear to bind dioxins (65). Thus, the ability to bind halogenated aromatic hydrocarbons may be a unique characteristic of vertebrate AHRs.
(d) Both Fundulus AHRs exhibit sequence-specific binding to mouse XRE oligonucleotides in a gel shift assay (Fig. 7). For both proteins, XRE binding is ARNT-dependent and is enhanced by ligand (TCDD), as for mammalian AHRs.
(e) Several splice junctions have been identified from analysis of genomic clones containing portions of the FhAHR1 and FhAHR2 genes (Figs. 2 and 3). These data indicate that the gene structures of FhAHR1 and FhAHR2 are similar to each other and to that of the mouse AHR gene (45). 3 Fundulus AHR1 and AHR2 Exhibit Differences Suggesting Distinct Functions-Although both Fundulus AHR1 and AHR2 are bona fide AHRs by the criteria noted above, these two proteins differ in several ways. At the level of primary sequence, they share only a modest degree of amino acid identity. They also differ in the presence of specific motifs. The presence of a Gln-rich region in FhAHR1 is consistent with our designation of this form as the direct ortholog of mammalian AHRs, in which Gln-rich regions occur. The reduced Gln-rich region in FhAHR2 is a characteristic shared by other fish AHRs cloned to date (31)(32)(33). However, FhAHR2 contains other motifs (acidic and P/S/T-rich domains) known to participate in transactivation. The presence of an LXCXE motif also unites Fundulus AHR1 with the mammalian AHRs (56). Interestingly, the MYCAD motif in this location is shared by Fundulus AHR2 3 S. Karchner and M. Hahn, unpublished data.
FIG. 6. Ligand binding by Fundulus AHRs. A, in vitro transcription/translation of mouse (M), Fundulus AHR1 (F1), and Fundulus AHR2 (F2) cDNAs. pSPORTMAHR, pSPFhAHR1 18 -942 , and pDPF-hAHR2 constructs were expressed in TNT SP6-and T7-Quick Coupled Reticulocyte Lysate Systems to synthesize 35 S-labeled proteins. The TNT reactions were subjected to SDS-polyacrylamide gel electrophoresis, followed by fluorography and autoradiography. The second, lower molecular mass band in lane F1 is likely due to either translational initiation from a downstream initiation site or transcriptional pauses during the in vitro transcription/translation reaction, as noted by others for expressed AHRs (99). B, photoaffinity labeling of Fundulus AHR1 and AHR2 proteins (pSPFhAHR1 18 -942  and the other fish AHRs ( Fig. 4; others not shown). This difference between AHR1 and AHR2 suggests that the two AHR forms may differ in their ability to interact with the Rb gene product or related proteins.
Within the putative ligand-binding domain (corresponding to amino acids 230 -397 of the mouse AHR), FhAHR1 and FhAHR2 share 61% amino acid identity; they share 79 and 61% amino acid identity, respectively, with the human AHR. Although amino acids outside of this region may also contribute to the ligand binding specificity, the degree of difference in sequences within the ligand-binding domain suggests that these AHRs, and especially FhAHR2, may bind distinct ligands. This is supported by preliminary data from a yeast expression system, where FhAHR1 and FhAHR2 were differentially activated by BNF and other agonists of the mammalian AHR. 4 The observed differences in tissue-specific expression further suggest that FhAHR1 and FhAHR2 have distinct functions. The ubiquitous expression pattern of Fundulus AHR2 is similar to that of mammalian AHRs (68 -70). In contrast, the expression of Fundulus AHR1 transcripts was restricted to heart, brain, ovary, and testis, suggesting specialized roles in these tissues. These could include roles in cardiovascular or reproductive function, processes known to be altered by AHR ligands (71,72).
Multiple AHRs in Vertebrate Animals-Multiple forms (paralogs) are known for most other vertebrate bHLH-PAS proteins, including ARNT (64), SIM (73,74), HIF (75,76), Per (77), and others (reviewed in Ref. 5). It is somewhat surprising, therefore, that additional AHR forms have not been found previously in mammals or other vertebrate animals. Our results show that additional AHR forms (AHR paralogs) do occur in vertebrates. In naming the two Fundulus AHRs, we have followed the precedent set for these other paralog groups. Fundulus AHR1 shows greater similarity to the original mammalian AHR, both in terms of overall amino acid identity as well as the presence of specific motifs. Fundulus AHR2 is more divergent from the mammalian AHRs.
Interestingly, all of the full-length fish AHR sequences reported to date are significantly more similar to Fundulus AHR2 than to either Fundulus AHR1 or the mammalian AHRs (Fig. 5). Thus, the tomcod AHR (31), zebrafish AHR (32), and two recently duplicated trout AHRs (␣ and ␤) (33) should all be classified as AHR2 forms. In addition, partial AHR sequences from European flounder (78) and Poeciliopsis (5) also belong to the AHR2 clade. The two trout AHRs share 97% amino acid sequence identity, reflecting the recent tetraploid ancestry of salmonid fishes (79) and clearly distinguishing them from the highly divergent Fundulus AHRs described here. In trout and zebrafish, there is preliminary evidence for an additional AHR form that may correspond to Fundulus AHR1 (33, 80). Thus, many fish species possess two or more distinct AHRs, further , testis (not shown), brain (B), kidney (K), and gill (G) were reverse transcribed with random primers. An equal aliquot of cDNA was used in each of three PCR reactions for amplification of Fundulus AHR1, AHR2, and actin fragments. The PCR conditions were adjusted so that the amount of PCR product obtained was linearly related to the amount of template and the number of PCR cycles. Specificity for the target gene was demonstrated in control experiments (not shown). An aliquot of each PCR sample was subjected to agarose gel electrophoresis followed by ethidium bromide staining. B, the integrated density of each amplified fragment was determined from the digital image of the gels in A and plotted as a percentage of that of actin in the same tissue. Similar results were obtained when this experiment was repeated with a second independent reverse transcription reaction for each tissue (data not shown).
supporting the existence of an AHR paralog group.
We envision two possible scenarios for the origin of the AHR1 and AHR2 clades (Fig. 9). In scenario 1, the gene duplication that produced the direct ancestors of the two Fundulus AHRs occurred after the divergence of lineages leading to present day bony fish and mammals. If this scenario is correct, one would expect (a) that the two fish AHRs would be more closely related to each other and more distantly (and similarly) related to the mammalian AHR, and (b) that AHR2 orthologs would not exist outside of bony fish (or a subset of bony fish). In scenario 2, the AHR gene duplication occurred prior to the divergence of fish and mammalian lineages. In this case, two AHRs would be expected in other vertebrate species, and the second mammalian AHR either has been lost or has yet to be discovered. This scenario predicts that one of the two fish AHRs would be more closely related than the other to the known mammalian AHR. Our current and previous (6) findings contradict both predictions of scenario 1, but are entirely consistent with those of scenario 2. Both pairwise sequence analysis and more rigorous phylogenetic analyses indicate that (i) Fundulus AHR1 is substantially more closely related to the mammalian AHRs than is AHR2, and (ii) FhAHR2 is as different from FhAHR1 as it is from the mammalian AHRs ( Fig. 5 and Table I). Furthermore, we have identified both AHR1 and AHR2 orthologs in a cartilaginous fish, the dogfish Mustelus canis (6). If cartilaginous fish diverged from the main vertebrate lineage prior to the divergence of bony fish and mammalian lines, as commonly accepted (81), 5 both AHR1 and AHR2 must have been present in a mammalian ancestor (scenario 2).
Two rounds of genome duplications are proposed to have occurred early in vertebrate evolution, prior to the emergence of jawed vertebrates (82). Several recent studies have provided evidence that fish (especially zebrafish) contain additional members of some multigene families as compared with mammals (83). Based on these findings, it has been suggested that another genome duplication occurred in many or all fish lineages following the divergence of fish and mammalian lines (reviewed in Refs. 84 and 85). Evidence for four clusters of Hox genes in Fundulus (19), as in mammals (82), does not support such an additional duplication in an ancestor of this species. Furthermore, the existence of AHR2 orthologs in cartilaginous fish (see above) is inconsistent with AHR1 and AHR2 being the result of a duplication unique to bony fishes. Thus, we conclude that the gene or genome duplication that gave rise to AHR1 and AHR2 occurred prior to the emergence of the extant classes of jawed vertebrates.
AHR2 Ortholog in Mammals?-If AHR1 and AHR2 arose early in vertebrate evolution, is there any evidence for a mammalian AHR2? A single AHR gene has been cloned from each of several mammalian species (human, mouse (several alleles), rat, and rabbit) (26 -30). Data from AHR-null mice indicating that this AHR gene is necessary for CYP1A induction and TCDD toxicity (10,11,86) do not preclude the existence of a second AHR that could be involved in other aspects of dioxin signaling or in modulating AHR1 function. There is biochemical evidence for multiple forms of the AHR in mammals (87)(88)(89)(90); these may or may not represent products of multiple AHR genes. Thus, the presence of a second AHR in mammals remains uncertain.
A novel mouse AHR-like sequence originally designated "AHR2" (91) has been characterized and renamed "AHR repressor" (92). Although most closely related to mammalian AHRs, this protein contains a highly divergent PAS-B domain and may not bind dioxins or other AHR ligands. In preliminary phylogenetic analyses, AHR repressor falls outside of the AHR1 and AHR2 clades, but within the larger vertebrate/ invertebrate AHR family. 6 We conclude from this that AHR repressor is not the mammalian ortholog of fish AHR2.
In preliminary attempts to identify a human AHR2, we have screened a panel of cDNAs from adult human tissues using degenerate primers that amplify both AHR1 and AHR2 in fish. 2 Evaluation of multiple clones (10 -12 for each tissue) derived from RT-PCR products from human liver, lung, heart, small intestine, and placenta failed to reveal any sequences other than the human AHR identified earlier (28). These results suggest that a human AHR2, if present, is expressed at low levels or in a cell-specific or developmentally restricted manner.
Implications of a Second AHR Form in Fish-Further investigation of the AHR2 subfamily will contribute to our understanding of AHR-dependent signaling in vertebrates, including mammals. The multiple functions ascribed to the mammalian "AHR" (2, 3) could involve multiple AHR forms or pleiotropic effects of a single AHR. If humans (mammals) possess a single AHR, it is conceivable that its multiple roles, in adaptive response pathways versus endogenous pathways (2), for example, may be partitioned among the two fish AHRs. If so, future studies in Fundulus and other teleosts could reveal AHR functions not tractable in mammals. Alternatively, fish AHRs could have evolved novel physiological functions.
Characterization of fish AHRs may illuminate the evolution of AHR ligand-binding properties. An invertebrate (C. elegans) AHR homolog lacks dioxin binding activity (65) and a similar property has been suggested for the AHR in an early diverging fish, the lamprey Petromyzon (6,93). Other data suggest that AHR1 and AHR2 possess distinct, although overlapping, binding specificities. 4 Further analysis of structure binding characteristics for FhAHR1 and FhAHR2 in comparison to those of other vertebrate AHRs will enhance our understanding of the evolution of dioxin signaling.
Finally, the presence of two AHRs in Fundulus and other teleosts is important for the use of fish as models in biology and toxicology. Some fish species, including Fundulus, 7 are among the most sensitive vertebrates to dioxin toxicity (94). In addition, populations of Fundulus and other fish are capable of developing heritable dioxin resistance (16 -18). These and other features have led to the proposed use of fish as vertebrate models for studying mechanisms of dioxin action (5,32,95). The presence of two AHRs could contribute to the sensitivity of fish to AHR agonists, and alterations in one or both of these could underlie the acquired resistance. More generally, differences in dioxin sensitivity and structure-activity relationships among taxa (96, 97) could involve differences in AHR structure, function, and multiplicity. The identification of a second AHR 5 The basal position of cartilaginous fish in the evolution of jawed vertebrates has been questioned recently (98), but most morphological and molecular evidence remains in favor of the traditional view. subfamily provides a foundation on which to investigate mechanisms of differential dioxin sensitivity among populations, species, and classes of vertebrates.