Resistance to 2,3,7,8-Tetrachlorodibenzo-p-dioxin Toxicity and Abnormal Liver Development in Mice Carrying a Mutation in the Nuclear Localization Sequence of the Aryl Hydrocarbon Receptor*

The Ah receptor (AHR) mediates the metabolic adaptation to a number of planar aromatic chemicals. Essential steps in this adaptive mechanism include AHR binding of ligand in the cytosol, translocation of the receptor to the nucleus, dimerization with the Ah receptor nuclear translocator, and binding of this heterodimeric transcription factor to dioxin-responsive elements (DREs) upstream of promoters that regulate the expression of genes involved in xenobiotic metabolism. The AHR is also involved in other aspects of mammalian biology, such as the toxicity of molecules like 2,3,7,8-tetrachlorodibenzo-p-dioxin as well as regulation of normal liver development. In an effort to test whether these additional AHR-mediated processes require a nuclear event, such as DRE binding, we used homologous recombination to generate mice with a mutation in the AHR nuclear localization/DRE binding domain. These Ahr nls mice were found to be resistant to all 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced toxic responses that we examined, including hepatomegaly, thymic involution, and cleft palate formation. Moreover, aberrations in liver development observed in these mice were identical to that observed in mice harboring a null allele at the Ahr locus. Taken in sum, these data support a model where most, if not all, of AHR-regulated biology requires nuclear localization.

The aryl hydrocarbon receptor (AHR) 1 regulates an adaptive metabolic response to a variety of planar aromatic chemicals that are widely dispersed in the environment. Over the last 20 years, the mechanistic details of this adaptive signaling pathway have been well characterized (1)(2)(3)(4). The AHR is a basic helix-loop-helix-PAS (bHLH-PAS) transcription factor. Upon binding agonists, the AHR translocates from the cytoplasm to the nucleus, where it forms a heterodimer with another bHLH-PAS protein known as the aryl hydrocarbon nuclear translocator (ARNT). This heterodimeric complex interacts with dioxin-responsive elements (DREs) within the genome and upregulates the transcription of a battery of xenobiotic metabolizing enzymes (XMEs). These regulated XMEs include the cytochrome P450s Cyp1a1, Cyp1b1, and Cyp1a2 and the phase II enzymes Gst-a1 and Ugt1-06 (reviewed in Refs. 2 and 3).
In addition to regulating an adaptive metabolic response, the AHR also mediates toxic responses to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and plays an important role in normal development. Early genetic and pharmacological experiments provided evidence that the AHR mediates toxic responses to TCDD and related pollutants (5). Highly reproducible toxic endpoints in rodent species include thymic involution, hepatomegaly, epithelial hyperplasia, and teratogenesis. More recently, generation of null alleles at the Ahr locus in mice revealed that the AHR also plays an important role in normal mammalian development (6 -9). Across laboratories, the most reproducible phenotype associated with the homozygous null allele is a smaller liver. We have proposed that smaller liver size is the result of the persistence of a fetal vascular structure known as the ductus venosus (DV) (10). Our hypothesis is that smaller liver size is due to hepatocyte atrophy, resulting from partial shunting of portal blood flow directly to the vena cava. Importantly, the mechanistic role of AHR in the persistence of the DV as well as the appearance of other vascular aberrations is unknown.
Our laboratory is interested in determining if the adaptive, toxic, and developmental pathways of AHR differ in any mechanistic aspect. In particular, we are interested in understanding whether cytosolic events, nuclear events, or DRE-mediated transcription play a role in TCDD-induced toxicity or in the establishment of a normal liver size. Although it may be assumed that all actions of AHR are dependent upon DRE-mediated transcription, a considerable body of evidence has been reported to the contrary. In this regard, signal transduction through cellular factors such as the cSrc kinase, the retinoblastoma protein (Rb), ceramide, steroid receptors, NF-B, and HIF1␣ has been reported to be modulated by AHR agonists in a manner that is independent of AHR DRE binding and transcriptional activation roles (11)(12)(13)(14)(15)(16)(17).
In an effort to formally examine the possibility that aspects of AHR biology are mediated through cytosolic events that do not require the AHR to translocate to the nuclear compart-ment, we undertook the approach of using gene targeting in mouse embryonic stem (ES) cells to generate a novel allele at Ahr that harbors a mutation that blocks both nuclear localization and DRE binding (designated as Ahr nls ). By examining the responses of these mutant animals to TCDD and through a comparison with other Ahr alleles, we are able to define the relative importance of cytosolic versus nuclear events in various aspects of AHR biology.

EXPERIMENTAL PROCEDURES
Strategy-The design of the AHR nls allele (and therefore the Ahr nls protein) is based on published results from several laboratories as well as practical considerations. Foremost is the observation that the residues required for DRE binding overlap with those required for nuclear localization. Previous mutational analysis reveals that the AHR nuclear localization sequence (NLS) is bipartite and resides within the basic region residues 12-17 and 37-42 (18 -20). These regions significantly overlap with those shown to be critical for DRE binding, including residues 9 -14 and 36 -39 (21-23) (Fig. 1A). Because of this domain overlap, we chose to introduce a series of nonconservative mutations to change Arg 37 , His 38 , and Arg 39 to Ala 37 , Gly 38 , and Ser 39 , respectively. Our prediction was that this would generate an AHR protein that is deficient in both nuclear localization and DRE binding. Based upon the work cited above and our previous domain mapping work, we predicted that these mutations would not influence ligand binding, chaperone binding, or dimerization with ARNT (24).
Expression Constructs-The construct pSV-AHR (PL 65) is previously described (25,26). To construct pTgTAHRT7 (PL1550), PL65 was used as a template for 22 rounds of PCR amplification using OL1503 (forward) and OL2639 (reverse) as primers. The resulting amplicon was cloned into the pTarget TM mammalian expression vector (Promega, Madison, WI). OL1503 contains a consensus Kozak start codon and a conservative mutation creating an SrfI site to match the targeting construct sequence (see below). The oligonucleotide OL2639 contains the region of AHR cDNA preceding the stop codon, followed by sequences for the T7 epitope and for a translational stop codon. The plasmid pSVAHR nls (PL1108) was made by three-step PCR using PL65 as template. In the first step, the forward oligonucleotide, OL1500, containing a 9-base substitution that changed codons Arg 37 to Ala 37 , His 38 to Gly 38 , and Arg 39 to Ser 39 , respectively, was used in a PCRamplification with the reverse oligonucleotide OL72. The product was then used as template for 20 rounds of amplification using OL1503 and OL72 to add the 5Ј amino acids. This product was then amplified using OL975 and OL72 to add a consensus Kozak start codon for increased expression. An SpeI/NaeI fragment was excised and used to replace the SpeI/NaeI fragment from PL65 to create pSVAHR nls (PL1108). Sequencing of the full-length constructs ensured that no other mutations were present.
Protein Analysis-Gel shifts and coimmunoprecipitations were performed as described previously (16,27,28). In vitro protein expression was carried out using the TNT coupled transcription/translation reticulocyte lysate system (Promega). For gel shift assays, in vitro expressed proteins were incubated with a 32 P-labeled DRE fragment derived from annealing OL73 and OL74 in the presence or absence of 10 M ␤-naphthoflavone (␤NF). The presence of the AHR in the DRE binding complex was confirmed using a high affinity rabbit polyclonal antibody raised against recombinant AHR (BEAR-3). Co-immunoprecipitation experiments were performed by co-incubating ϳ10 fmol of reticulocyte lysate-expressed proteins with 2 g of antibody in 500 l of cold MENG buffer (25 mM MOPS, pH 7.5, 0.025% sodium azide, 1 mM EGTA, 10% glycerol, pH 7.5), supplemented with 15 mM NaCl, 0.1 mM dithiothreitol, and 0.1% Nonidet P-40. Bound protein-antibody complexes were precipitated with protein A-Sepharose (Sigma) for 1.5 h at 4°C, washed four times with cold MENG buffer, eluted in 2ϫ SDS sample buffer, and analyzed by SDS-PAGE.
Western blots, EROD assays, and photoaffinity labeling of the AHR were performed using methods described previously (7,29,31). Microsomal and cytosolic fractions were isolated from ϳ0.5 g of mouse liver that was homogenized in ice-cold MENG buffer followed by two high speed centrifugations at 10,000 ϫ g and 100,000 ϫ g. The microsomal pellet was resuspended in 1 ml of resuspension buffer (15 mM Tris-Cl, pH 8.0, 250 mM sucrose). The supernatant, containing cytosolic protein, was saved for later AHR analysis. Western blot analysis was performed as described using the BEAR-3 antibody and a secondary antibody labeled with alkaline phosphatase (29). The EROD assays were performed in a 96-well format. In each well, one-one thousandth of the total microsomal preparation was diluted into 200 l in MENG buffer. To start the reaction, 3 l of 0.1 mM 7-ethoxyresorufin and 20 l of 5 mM NADPH were added. The production of 7-hydroxyresorufin was measured by fluorometry (fMax; Molecular Devices) at excitation of 510 nm, emission of 590 nm every 30 s over 10 min at 25°C. Total protein concentration was determined using the BCA assay (Pierce). The results are expressed as relative fluorescence units/min as calculated using SoftMaxPro software (Molecular Devices) and normalized to total protein levels. Photoaffinity labeling of the AHR protein with 2-azido-3-iodo[ 125 I]iodo-7,8-dibromodibenzo-p-dioxin ([ 125 I]Br2DpD) was performed essentially as described (31). Briefly, for labeling of the cytosolic AHR, liver cytosolic protein (300 g/ml) was incubated with [ 125 I]Br2DpD at the indicated concentrations for 2 h at 4°C. This was followed by the addition of one-tenth volume of 1%/0.1% activated charcoal/gelatin for 30 min at 4°C to remove unbound ligand. Following centrifugation, supernatants were exposed to UV light at 310 nm (80 watts, 4 cm) for 30 s. Protein was precipitated in acetone and separated by 7.5% SDS-PAGE. The location of labeled receptor was determined by autoradiography, and the band was excised and counted by ␥ emissions. The quantity of bound ligand is expressed as dpm/lane. The in vitro translated proteins were labeled by diluting 10 l of the translation reaction in 0.5 ml of MENG. We ensured that the 10 l of translation reaction contained equal amounts of translated receptor by 35 S calibration. The receptor was then labeled with 1 nM [ 125 I]Br2DpD for 30 min at 20°C followed by 4°C for 5 min. Unbound ligand was removed by charcoal/gelatin, and the quantity of bound ligand was determined in the same manner as above. The specific activity of the radioligand is 2176 Ci/mmol, and 1 pmol ϭ 4,830,700 dpm. For Scatchard analysis, it was assumed that free probe was equal to total counts added and that total bound was equal to specific bound (this assumes that nonspecific binding is negligible following SDS-PAGE and band excision) (30).
Generation of Mouse Embryonic Fibroblasts-We have previously generated Ahr null mice that will be hereafter referred to as Ahr Ϫ/Ϫ (formal genetic designation is Ahr tmBra1 ). To generate matched mouse embryonic fibroblasts, Ahr ϩ/Ϫ mice that have been backcrossed to C57BL/6 for 16 generations were intercrossed to generate littermate ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ embryos. Embryos were isolated from their yolk sacs, and heads and livers were removed by dissection. DNA was isolated from the heads of individual embryos and was used for genotyping as described (10). At passage 2 (P2), Ahr ϩ/ϩ and Ahr Ϫ/Ϫ fibroblasts were maintained on a "3T3 protocol" until passage 25 and then passaged regularly at subconfluence as described (32). Briefly, in the 3T3 protocol, cells were grown in 6-cm dishes in Dulbecco's modified Eagle's medium-high glucose, supplemented with 10% fetal bovine serum, 0.1 mM nonessential amino acids, 2 mM L-glutamine, 10 mM HEPES, 100 units/ml penicillin, and 100 g/ml streptomycin. Every 3rd day, cells were trypsinized and replated at a density of 3 ϫ 10 5 . At passage 28, individual clones were isolated from one line of each genotype and passaged regularly at subconfluence. The Ahr Ϫ/Ϫ subcloned cell line used for these experiments was designated B4.4.
Transfection of Mouse Embryonic Fibroblasts-Transient transfections of mouse embryonic fibroblasts were performed using Fugene6 reagent (Roche Applied Sciences) with 1 g of total DNA and a 3:1 Fugene6/DNA ratio. When TCDD was used, cells were allowed to re-cover for 1 day after transfection. Following this period, 1 nM TCDD was added directly to the medium in Me 2 SO (final Me 2 SO concentration 0.1%) and cells were harvested after 2 h. Immunofluorescence was performed as described (26) using BEAR-3 and a fluorescein isothiocyanate-conjugated goat anti-rabbit secondary antibody (Jackson Immunoresearch, West Grove, PA) for fluorescence detection.
Generation of Ahr nls/nls Mice-A 15-kb region of homology surrounding exon 2 of AHR was isolated from a 129SvJ genomic library (Genome Systems) as described (7,33). To make the Ahr nls allele, a 9-nucleotide replacement was introduced by using the PCR product of OL1500 and OL942 as megaprimer with 22 rounds of amplification with OL941 (23). The SphI fragment from the mutant PCR product was used to replace exon 2 in an 8-kb BamHI genomic fragment. A 5.5-kb region containing the mutated exon 2 was amplified with OL1352 and OL1353 and cloned into the KpnI site of ploxPNT (34). A 7-kb SphI fragment from the 5Ј region of exon 2 was cloned into the NotI/XhoI site of this construct to make the final targeting construct that was designated pNTAHRnls (PL1075). For homologous recombination in ES cells, ϳ10 g of targeting construct was electroporated into GS1 cells (Genome Systems), and selection was performed using 200 g/ml G418 and 1 mM Ganciclovir. Clones were screened by Southern blot of BamHI-digested genomic DNA using a probe 3Ј to the end of the targeting construct (PL311). Correctly targeted clones were injected into 3.5-day postcoital C57BL/6 blastocysts, and resulting chimeras were backcrossed to C57BL/6 to determine contribution of the ES clones to the germ line. For experimental analysis, animals were backcrossed to C57BL/6 mice that are congenic for the Ahr d allele (35). Mice were genotyped using the PCR primers OL941 and OL942. The PCR was carried out for 35 cycles (95°C, 1 min; 63°C, 1 min; 72°C, 1 min) in buffer containing 4 mM MgCl 2 ; a BamHI digest cuts the 380-bp PCR product from the Ahr nls allele into two small fragments of 240 and 140 bp that were detected on a 2% agarose gel.
Animals-Animals were housed in a selective pathogen-free facility on corncob bedding with food and water ad libitum according to the rules and guidelines set by the University of Wisconsin. When TCDD was used, 5-week-old animals were injected once intraperitoneally with 100 g/kg TCDD in p-dioxane or with p-dioxane alone. After 6 days, animals were weighed and sacrificed, and organs were immediately removed and weighed. For angiography, ϳ1 ml of Omnipaque 300 (Nycomed, Inc., Princeton, NJ) was injected into the hepatic portal vein postmortem. Continuous x-ray images were obtained over ϳ10 s using an OEC 9800 Portable Vascular C-ARM (Medical Systems, Inc., Salt Lake City, UT).
In order to directly compare the null, wild type, and Ahr nls/nls alleles, genetic background had to be taken into consideration. Mice used in these experiments that are defined as harboring the Ahr Ϫ allele have been previously backcrossed to C57BL/6 for 16 generations. Because C57BL/6 carries the high affinity Ahr b allele and the Ahr nls allele was created in 129 ES cells carrying the lower affinity Ahr d allele, we used a C57BL/6 strain congenic for DBA2-derived Ahr d allele to make subsequent backcrosses of both the Ahr Ϫ and Ahr nls alleles (generous gift from Alan Poland). This breeding generated wild type Ahr d allele controls on the C57BL/6 background (35). For clearer presentation, the wild type Ahr d allele is referred to as Ahr ϩ ("wild type") throughout this paper. Mice harboring the Ahr nls allele were backcrossed for three generations to C57BL/6. To further minimize the influence of genetic background, all experiments were repeated from multiple independently derived sublines. Only a single representative experiment is presented.
Palate Cultures-Suspended palate organ cultures were performed essentially as described (36,37). Palatal shelves were dissected from embryonic day 12.5 embryos of the Ahr ϩ/ϩ , Ahr ϩ/Ϫ , and Ahr nls/nls genotypes. The palates were placed into cold phosphate-buffered saline and then transferred to prewarmed 1:1 Dulbecco's modified Eagle's medium/F-12 medium supplemented with 1% L-glutamine, 1% ascorbate, and 1% penicillin/streptomycin and immediately treated with either 0.03% Me 2 SO or 3.3 nM TCDD in 0.033% Me 2 SO (v/v). After exposure for 4 days, the palates were fixed in 10% formalin and stained with hematoxilin for visualization. Palates were scored as not fused or fused. Ambiguous partially fused palates accounted for only a small percentage and were not included in the data set.
Statistics-In the situation where multiple comparisons could be made, an analysis of variance was performed, and Tukey's test was used to determine differences with a p Յ 0.05 (38). When only two groups were being compared, a two-tailed t test was performed with the level of significance set at p Յ 0.05 (Fig. 5) (38).

Preliminary Characterization of the Ahr nls Mutation in
Vitro-In order to generate an AHR protein deficient in nuclear localization, we replaced amino acids Arg 37 , His 38 , and Arg 39 in the mouse AHR cDNA (pSV-AHR) with nonconservative amino acids Ala, Gly, and Ser, respectively, to generate the plasmid pSV-AHR nls (Fig. 1A). Given the overlap between the NLS and the DRE recognition sequences, this mutant was also predicted to have deficiencies in DRE binding. As a preliminary examination of the characteristics of the mutant AHR nls protein, we translated the cDNA in a reticulocyte lysate system. To demonstrate that the mutant receptor bound ligand normally, we performed labeling experiments with the photoaffinity ligand 2-azido-3-[ 125 I]iodo-7,8-dibromodibenzo-p-dioxin. We observed . Labeled proteins were separated by PAGE and visualized by autoradiography. A representative experiment with duplicate samples is shown. C, gel shift analysis of AHR, ARNT, and AHR nls recombinant proteins. Recombinant proteins were expressed alone or in the presence of ARNT or ARNT ϩ ␤NF and analyzed for binding to a 32 P-labeled oligonucleotide with a single DRE element by nondenaturing PAGE. The anti-AHR antibody (AHR-Ab) was added to indicated reaction to block formation of the complex. D-G, subcellular localization of recombinant AHR and AHR nls proteins. Ahr Ϫ/Ϫ cells were transiently transfected with either AHR (D and E) or AHR nls (F and G) cDNA (pSV-AHR and pSV-AHR nls , respectively). Cells were treated with either 1% Me 2 SO (C and E) or 1 nM TCDD plus 1% Me 2 SO (D and F), fixed in cold methanol, and analyzed for localization of AHR using anti-AHR antibody and detected with fluorescein isothiocyanate-conjugated secondary antibody. Digital images were collected by standard fluorescence microscopy. that the mutant receptor bound this ligand with a capacity that was similar to the wild-type protein (Fig. 1B). To determine the DNA binding capacity of the AHR nls protein, we performed gel shift analysis with a consensus DRE oligonucleotide (Fig. 1C). In the absence of agonists and in the absence or presence of ARNT, neither the AHR nor the AHR nls proteins interacted significantly with a 32 P-labeled DRE oligonucleotide. The addition of the agonist ␤NF induced formation of the wild-type recombinant AHR⅐ARNT⅐DRE complex that could be blocked by the addition of the anti-AHR antibody. However, the addition of ␤NF failed to induce formation of a recombinant AHR nls / ARNT/DRE complex, indicating that DNA binding is disrupted by the amino acid replacements (Fig. 1C).
We used indirect immunofluorescence to determine the influence of the NLS mutation on the subcellular localization of the AHR in mammalian cells. In order to perform this experiment in the absence of endogenous wild-type AHR, we generated "immortalized" fibroblasts from Ahr Ϫ/Ϫ embryos using a 3T3 protocol (32). The Ahr Ϫ/Ϫ fibroblasts were transiently transfected with AHR or AHR nls cDNAs (pTgTAHRT7 or pSV-AHR nls , respectively). Following treatment with either vehicle or 1 nM TCDD, cells were fixed, and the localization of the AHR or AHR nls protein was visualized using anti-AHR antibody and a fluorescein isothiocyanate-conjugated secondary antibody. We observed that wild-type AHR localizes to the cytosol in untreated fibroblasts but moves to the nucleus within 2 h after treatment of cells with 1 nM TCDD, (Fig. 1, D and E). As predicted, the AHR nls protein was constitutively cytoplasmic, and its localization was not affected by TCDD (Fig. 1, F and G). This block in nuclear localization was also observed in COS7 and murine hepatoma cells (data not shown).
To demonstrate the capacity of the AHR nls protein to form normal complexes with its cognate cytosolic chaperones, we performed coimmunoprecipitation experiments with HSP90 and ARA9. For these experiments, proteins were expressed in vitro in reticulocyte lysate. For interaction analysis with HSP90, 35 S-labeled AHR or AHR nls were immunoprecipitated with antibodies to HSP90 ( Fig. 2A). For interaction analysis with ARA9, 35 S-labeled ARA9 was incubated in the presence of unlabeled AHR and AHR nls and precipitated with anti-AHR antibody (Fig. 2B). Results from three independent experiments indicated that both the AHR and AHR nls proteins formed complexes with the chaperones ARA9 and HSP90 with similar capacity. The results from representative experiments are shown in Fig. 2. Together these results indicate that AHR nls is capable of interacting with both HSP90 and ARA9 in a manner similar to the wild type AHR yet is no longer capable of translocating to the nucleus and interacting with DNA at DRE elements ( Figs. 1 and 2).
Generation of Ahr nls/nls Mice and Their Controls-To generate mice with mutant AHR nls protein, we used PCR to alter the above mentioned sequences of the basic region of AHR (exon 2) in a 15-kb region of homologous genomic DNA from the Ahr locus (Fig. 3A). The targeting construct was designed to insert the neomycin cassette into a similar region of Ahr as was done previously to generate the null mutation (Ahr Ϫ/Ϫ ) (7). After selection in both G418 and Ganciclovir, surviving clones were screened by Southern blot, and correctly targeted clones were used to generate chimeras and ultimately germ line transmission of the AHR nls mutation to mice. A representative map of the resulting Ahr nls allele is depicted in Fig. 3A beside the Ahr Ϫ/Ϫ allele and wild type alleles. The resulting Ahr nls/nls mice (formal genetic designation Ahr tmBra2 ) are born at a frequency consistent with a nonlethal allele, and the animals are fertile ( Fig. 3B and data not shown).
Characterization of Ahr nls Mutation in Vivo-Expression levels of each mutant AHR protein were determined by Western blot and photoaffinity labeling of receptor from liver protein extracts. Visual inspection of the Western blots suggested that the expression of receptor from the Ahr nls/nls mouse liver was approximately equivalent to that observed in the Ahr ϩ/Ϫ liver. Both of these samples expressed the receptor at a level that FIG. 2. Capacity of AHR nls to form complexes with HSP90 and ARA9. A, co-immunoprecipitation of AHR nls with HSP90. Five microliters of in vitro translated [ 35 S]methionine-labeled wild type AHR or AHR nls was incubated with 2 g of anti-HSP90 antibody and precipitated with Protein A-Sepharose. Reticulocyte lysates contain large amounts of HSP90, so no additional protein is added. B, co-immunoprecipitation of ARA9 with AHR nls . Five microliters of in vitro translated [ 35 S]methionine-labeled full-length ARA9 was incubated with either wild type AHR or AHR nls . Protein complexes were incubated with 2 g of the AHR-specific antibody and precipitated using Protein A-Sepharose. All proteins in A and B were eluted from Protein A-Sepharose and separated by SDS-PAGE gels and visualized by autoradiography. Input, a loading control representing 100% of the radiolabeled ARA9 used in the assay. PreI, the preimmune control. was approximately one-half of what we observed from the Ahr ϩ/ϩ mouse liver (Fig. 4A). Since Western blots are a semiquantitative method, we provided an independent and more quantitative assessment of AHR expression levels by photoaffinity labeling the receptors with increasing amounts of the ligand 2-azido-3-[ 125 I]iodo-7,8-dibromodibenzo-p-dioxin (Fig.  4B). To quantitate bound photoaffinity ligand in the 104-kDa AHR bands, excised radioactive gel slices were quantified in a ␥ counter. In keeping with what we observed based upon Western blots, we found that the level of functional binding sites for this ligand in liver extracts of Ahr nls/nls mice approximated the levels found in Ahr ϩ/Ϫ animals and was half of the level found in Ahr ϩ/ϩ mice (Fig. 4B). In an effort to determine whether the ligand binding affinity was the same for receptor found in Ahr ϩ/Ϫ and Ahr nls/nls mice, we generated a saturation binding isotherm with increasing amounts of radioligand. A modified Scatchard analysis of the binding data revealed similar affinities between wild-type AHR and AHR nls proteins derived from liver cytosols (Fig. 4, C and D). As a result of these findings, we included heterozygous Ahr ϩ/Ϫ mice as controls in most experiments.
To determine whether the mutant AHR protein signals in vivo, Ahr nls/nls , Ahr ϩ/ϩ , and Ahr ϩ/Ϫ animals were injected intraperitoneally with 100 g/kg TCDD. After 6 days, liver microsomes were isolated and analyzed for EROD activity. Microsomes from wild type and Ahr ϩ/Ϫ mice show a similar basal EROD activity that was induced nearly 1000-fold by TCDD (Fig. 4E). In contrast, microsomes from Ahr nls/nls mice showed only basal EROD activity that was not altered by TCDD treatment. This result indicates that the AHR nls protein cannot activate gene transcription from DRE elements in vivo.
Ahr nls/nls Mice Are Resistant to TCDD Toxicity-To determine whether the NLS mutation influenced the role of the AHR in the toxicological response to TCDD, we exposed male Ahr nls/nls , Ahr Ϫ/Ϫ , Ahr ϩ/Ϫ , and Ahr ϩ/ϩ mice to TCDD and measured liver hypertrophy and thymic involution, two highly reproducible end points associated with TCDD toxicity in animals. Male, 5-6-week-old animals were treated with 100 g/kg TCDD in p-dioxane or an equivalent volume of p-dioxane alone. Mice were sacrificed 6 days later, and liver and thymus wet weights were recorded. Liver weights in Ahr ϩ/ϩ and Ahr ϩ/Ϫ animals increased significantly in response to TCDD, yet neither the Ahr Ϫ/Ϫ and Ahr nls/nls animals showed any TCDDinduced increase in liver weights in response to TCDD (Fig. 5A, p Յ 0.05). In addition, we observed that the relative liver weights of both the Ahr nls/nls and the Ahr Ϫ/Ϫ mice were reduced ϳ28% as compared with the Ahr ϩ/ϩ or Ahr ϩ/Ϫ controls (p Յ 0.05; Fig. 5A). Thymus weights in Ahr ϩ/ϩ and Ahr ϩ/Ϫ animals decreased significantly in response to TCDD, yet Ahr Ϫ/Ϫ and Ahr nls/nls , animals did not show any significant TCDD-induced decreases in thymus weights (p Յ 0.05; Fig. 5B). In this experiment, there was a trend suggesting that TCDD might have a subtle influence on relative thymus weights of Ahr Ϫ/Ϫ and Ahr nls/nls mice, but independent experiments did not support this relationship (data not shown).
We also examined the closure of palatal shelves in an organ culture model system, where we can directly expose the tissue to TCDD (36,37). Palates were excised from embryonic day 12.5 Ahr ϩ/ϩ , Ahr ϩ/Ϫ , and Ahr nls/nls embryos, and the tissue was directly exposed to TCDD in culture for 4 days. In response to 0.03% (v/v) Me 2 SO alone, genotype did not affect palatal closure. However, when exposed to 3 nM TCDD in Me 2 SO, no palates from Ahr ϩ/ϩ or Ahr ϩ/Ϫ mice fused, whereas the TCDDexposed palates from Ahr nls/nls animals all fused (Table I). This result was not different from Me 2 SO-treated controls. Therefore, AHR nls protein fails to activate those events, leading to FIG. 4. The AHR nls protein binds ligand but does not activate DRE-mediated transcription in vivo. A, Western blot analyses showing protein levels in liver cytosolic extracts from three individual animals from each genotype Ahr ϩ/ϩ , Ahr ϩ/Ϫ , and Ahr nls/nls . Cytosolic protein (150 g) is loaded in each well of a 7.5% polyacrylamide gel. Visualization is performed with anti-AHR antibody. B, ligand binding capacity of cytosolic AHR proteins using 1 nM photoaffinity ligand. 150 g of the same protein extracts in A were incubated with [ 125 I]Br 2 DpD and activated at 310-nm light. Samples were washed to remove unbound ligand and separated by SDS-PAGE. Ligand binding was analyzed by autoradiography. C and D, ligand binding of the AHR from Ahr ϩ/Ϫ and Ahr nls/nls animals. Cytosolic extracts from Ahr ϩ/Ϫ (C) and Ahr nls/nls (D) animals were incubated with increasing amounts of [ 125 I]Br 2 DpD ligand. Saturation binding isotherms are plotted as a function of total ligand versus bound (dpm counted from excised gel pieces). Each point on the plot represents the average of two experiments. Insets, Scatchard analysis of binding data calculated according to "Experimental Procedures." B max values are expressed as relative dpm: Ahr ϩ/Ϫ B max ϭ 1290; Ahr nls/nls B max ϭ 1110. Because of significant overlap between curves and for clarity, results from Ahr ϩ/Ϫ (C) and Ahr nls/nls (D) are plotted separately. E, EROD activity analyses from microsomes in Ahr ϩ/ϩ , Ahr ϩ/Ϫ , Ahr Ϫ/Ϫ , and Ahr nls/nls animals. Microsomal isolations from liver extracts were incubated with ethoxyresorufin in the presence of NADPH. EROD activity was measured from pdioxane-treated (Ϫ) and p-dioxane plus TCDD-treated (ϩ) animals at excitation of 510 nm and emission of 590 nm. Fluorescent values were normalized to total protein levels. Error bars, S.D. values from three individual animals. cleft palate formation in response to TCDD.
Ahr nls/nls Mice Show Developmental Defects Similar to Ahr Ϫ/Ϫ Mice-Wild type and Ahr nls/nls animals were examined for developmental phenotypes that have been previously reported in Ahr Ϫ/Ϫ mice (6 -8, 10, 39 -43). Tissue wet weights were measured on major organs of 8-week-old male mice. We observed that relative liver weights in Ahr nls/nls animals were 27% smaller when compared with Ahr ϩ/ϩ controls ( Fig. 6A; p Յ 0.001). The reduced Ahr nls/nls liver weights were identical to those reported previously in Ahr Ϫ/Ϫ animals (7). Also similar to what has been reported for Ahr Ϫ/Ϫ animals, normalized spleen and heart weights were moderately increased in Ahr nls/nls compared with Ahr ϩ/ϩ (Fig. 6, B and C; p Յ 0.001). Importantly, relative weights of thymus, kidney, lung, testes, and brain remained unaffected by genotype (thymus weight is shown as an example) (Fig. 6D).
In Ahr Ϫ/Ϫ mice, the DV remains open throughout adulthood.
To determine whether a similar phenotype exists in Ahr nls/nls animals, we performed time lapse angiography to observe the flow of contrast medium through the perfused liver. In the control Ahr ϩ/ϩ and Ahr ϩ/Ϫ animals, contrast medium flowed into the portal vein and immediately into the portal branches of the liver vasculature (Fig. 6, E and F, respectively). After filling the major branching veins of the liver, contrast entered the suprahepatic IVC and then filled the infrahepatic IVC. However, contrast medium in Ahr nls/nls mice flowed directly from the portal vein to the IVC. The shunt between the portal vein and the IVC was clearly visible as a short segment that runs perpendicular to both the portal vein and IVC within the liver (Fig. 6H). This vascular pattern is consistent with patent DV in Ahr Ϫ/Ϫ mice ( Fig. 6G

TABLE I
Closure in wild type and mutant palates following direct exposure to TCDD in organ culture Primary and secondary palatal shelves were dissected from embryonic day 12.5 animals and cultured in media containing either 0.03% Me 2 SO alone or 3 nM TCDD in 0.03% Me 2 SO for 4 days. Palatal shelves were scored as "fused" if fused along entire length of shelf at the end of the 4 days. Ambiguous partial fusion events occurred occasionally and were not included. XMEs through receptor-DRE interactions within the nucleus (44). However, this understanding of the DRE-linked transcriptional targets of AHR has not yet explained the variety of toxic events that occur in response to potent agonists like TCDD; nor have they been used to explain the aberrant liver/vascular development observed in Ahr Ϫ/Ϫ mice. Although it may be simple to assume that these receptor dependent processes are related to DRE-driven transcription, several reports have suggested that the AHR may signal through DRE-independent pathways (11, 13-16, 17, 45, 46, 48). Thus, we are still in need of a formal proof to demonstrate that the mechanism by which XMEs are up-regulated is similar to the mechanism by which the AHR plays a role in TCDD toxicity or regulates aspects of mammalian development.
In an effort to better understand the importance of DREindependent signaling, we are constructing mutant mouse models that directly test the roles of receptor functional domains on various aspects of AHR biology in vivo. The AHR nls mouse was designed specifically to examine the role that cytosolic events play in TCDD-induced toxicity and AHR-dependent liver development. In this endeavor, we chose to first generate animals with an AHR protein that can be activated by ligand but that is not capable of translocating to the nucleus or binding to DREs. The rationale for such a model is that it can be used to identify AHR-mediated end points that do not require nuclear localization of the AHR. Put another way, with this model we may be able to identify cytosolic signaling pathways related to AHR biology, or we can provide evidence to refute their existence.
Generation of this mouse model required consideration of the overlapping domain structure of the basic ␣-helix of AHR within the bHLH domain. This basic region encodes the nuclear localization motif that is revealed upon ligand binding, and it harbors the DRE contact residues by which the AHR interacts with its DNA half-site within these enhancers (20,28). In early characterization studies, we demonstrated that the NLS mutation resulted in an AHR protein that interacts normally with chaperones but that could not translocate to the nucleus or bind to DREs upon exposure to agonists. To this end, we used immunocytochemistry to demonstrate that the mutant protein remained cytosolic in the presence of an agonist. We then also used a battery of in vitro assays to demonstrate that the mutant protein was capable of interactions with its chaperones HSP90 and ARA9 ( Figs. 1 and 2). Once the predicted consequences of our mutations were confirmed, we used homologous recombination in ES cells to replace the endogenous bHLH region of the Ahr locus (exon 2) with a mutated exon carrying the NLS amino acid substitutions.
There are two major points regarding controls that arose during the course of these experiments. The first relates to genetic background. Because embryonic stem cells are derived from 129Sv mice that carry the d allele of Ahr, we utilized the C57BL/6-Ahr d animals for backcrossing and to generate control animals for these experiments. Mice harboring the wild-type d allele are not used as frequently as those strains harboring the more common b alleles due to the fact that the d allele encodes a receptor with a lower binding affinity for TCDD (49). Importantly, this lower binding affinity exhibited by the d allele has been shown to be similar to the human AHR and thus may serve as a more relevant model for human toxicity (47). The major influence this parental allele had on our experimental design was that it necessitated the development of dosing protocols that would induce toxic end points in mice harboring this lower affinity receptor. Thus, it will be noted that the doses of TCDD used in these studies are about 5-10-fold higher than are commonly used in most mouse studies.
The second point relevant to control animals is that although the AHR from the d allele and the nls allele display similar affinities for ligand, the nls allele is expressed at a lower level (Fig. 4). Our photoaffinity labeling of the AHR found in the livers of Ahr nls/nls mice indicated that the mutant protein was expressed at a level that was approximately one-half of that observed in Ahr ϩ/ϩ mice. Fortunately, AHR expression in wildtype animals appears to be a direct function of gene copy number. 2 Therefore, mice that were heterozygous for the d allele and the null allele served as controls (i.e. Ahr ϩ/Ϫ ).
To determine whether nuclear events were required for the role of AHR in TCDD toxicity or normal liver development, we performed a number of toxicology experiments with the Ahr nls/nls mice. First, we examined the toxic response of the Ahr nls/nls mice by injecting the animals with TCDD and assaying for liver hypertrophy and thymic involution. We found that Ahr nls/nls mice failed to exhibit these obvious liver and thymic end points, whereas the corresponding Ahr ϩ/ϩ and Ahr ϩ/Ϫ controls were quite sensitive (Fig. 5, A and B). We also examined a TCDD-induced teratogenic response, namely the formation of cleft palate. To this end, we exposed embryonic day 12.5 palatal shelves directly to TCDD in culture. This protocol directly exposes palate tissue in culture to TCDD and eliminates issues related to maternal influence. Just like the hepatic and thymic responses, we observed that the palates from Ahr nls/nls animals were completely resistant to TCDD exposure, whereas the corresponding controls were sensitive (Table I). From these results, we conclude that nuclear localization is required for TCDD-induced liver hypertrophy, thymic atrophy, and cleft palate formation.
We also assessed the role of cytosolic events in the developmental aspects of AHR biology. In this regard, all of the known developmental defects we have previously observed in Ahr Ϫ/Ϫ mice are also observed in the Ahr nls/nls mice. To this point, the Ahr nls/nls mice display a similar reduction in relative liver weight and moderate increases in heart and spleen weights. In more recent experiments, we have also observed that the livers of Ahr nls/nls display microvesicular fatty changes around day 6 after birth (7). Just as in the null allele these fatty changes resolve by adulthood (data not shown). Finally, the presence of a patent DV was observed in both the Ahr nls/nls mice and in the Ahr Ϫ/Ϫ mice (Fig. 6). Just as in the toxicology experiments described above, we are led to conclude that the nuclear localization of the AHR is an essential step in the receptor pathways that regulate liver development/DV closure.
The Ahr nls/nls mice provided us with a model system to test the importance of cytosolic interactions in certain aspects of AHR biology. Given that it would be impossible to test each proposed model in our system (i.e. cSrc, ceramide, etc.), our strategy was to carry out experiments that would examine the relative importance of cytosolic versus nuclear signaling with regard to a few classic receptor-mediated end points (i.e. liver hypertrophy, thymic involution, cleft palate formation, and persistence of the DV). Although our results are not supportive of a role for cytosolic events for these end points, it is still possible that cytosolic signaling by the AHR is important in end points not examined here (e.g. chloracne, altered T-cell responses, etc.). It is also important to point out that we cannot rule out the possibility that cytosolic signaling does occur and that it is dependent upon the exact identity of residues Arg 37 , His 38 , and Arg 39 . In an effort to address such possibilities, we plan to make this mutant available to any laboratory interested in formally examining the influence of the nls allele on any proposed cytosolic mechanism. Finally, given that the NLS mutation eliminates nuclear localization, it remains possible that nuclear, yet non-DRE binding events are important in AHR biology. Nuclear events such as cross-talk with the hypoxia-inducible factor and interactions with steroid receptors are still potentially important mechanisms of TCDD action or AHR-mediated developmental signaling.