Aromatic hydrocarbon receptor polymorphism: development of new methods to correlate genotype with phenotype.

Differential CYP1A1 inducibility, reflecting variations in aromatic hydrocarbon receptor (AHR) affinity among inbred mouse strains, is an important determinant of environmental toxicity. We took advantage of the Ahr polymorphism in C57BL/6 and DBA/2 mice to develop an oligonucleotide-hybridization screening approach for the rapid identification of DNA sequence differences between Ahr alleles. Oligonucleotides containing single-base changes at polymorphic sites were immobilized on a solid support and hybridized with C57BL/6 or DBA/2 AHR cDNA radiolabeled probes. The observed hybridization patterns demonstrate that this approach can be used to detect nucleotide differences in the Ahr coding region with very high accuracy. In parallel experiments, we used a yeast two-hybrid system to assess phenotypic differences in AHR function. AHR activation, as measured by beta-galactosidase reporter activity in Saccharomyces cerevisiae strain SFY526, was determined following treatment with varying doses of the AHR ligand beta-naphthoflavone (BNF). We found that the C57BL/6 AHR has about a 15-fold higher affinity for BNF than the DBA/2 AHR, in much better agreement with results reported for whole-animal studies than the values observed by in vitro ligand-binding assays. Using C57BL/6 and DBA/2 AHR chimeric proteins, we also confirmed the previously reported observation that an A375V change is principally responsible for the high- to low-affinity AHR phenotype. There has been no straightforward method to reliably and reproducibly phenotype large numbers of humans for CYP1A1 inducibility or AHR affinity. Screening human AHR cDNAs by oligonucleotide-hybridization and yeast two-hybrid methodologies will be invaluable for the rapid and unequivocal determination of changes in DNA sequence and receptor-ligand affinities associated with human AHR polymorphisms.

One of the most challenging questions facing environmental health research today is the identification of genotype changes associated with phenotypes of increased resistance or susceptib ity to toxic environmental agents. Environmental pollutants generate varying degrees of response in an exposed population, often due to polymorphisms in the genes controlling the molecular mechanisms of the toxic response. For example, a clear genetic component exists in the susceptibility of mouse strains to many halogenated aromatic hydrocarbons (HAHs) and nonhalogenated polycyclic aromatic hydrocarbons (PAHs), such as 2,3,7,8-tetrachlorodibenzop-dioxin (dioxin) and benzo(a)pyrene (BaP), respectively. These two environmental compounds exhibit a high degree of toxicity. BaP is a potent carcinogen in animals and a suspected carcinogen in humans (1). Dioxin is associated with a variety of systemic effects, including immunosuppression, cleft palate, and tumor promotion in animals, and chloracne, immunosuppression, and possibly cancer and heart disease in humans (2,3). There is no instance known so far in which dioxin or BaP toxicity or carcinogenesis in laboratory animals is not mediated by the aromatic hydrocarbon receptor (AHR), a transcrip-tional regulator of xenobiotic-metabolizing enzymes encoded by the mammalian AHR gene (4,5).
A mouse Ahr polymorphism is known to be responsible for the variation in susceptibility to PAHs and HAHs (6)(7)(8). Four distinct mouse Ahr alleles have now been characterized, each encoding receptor proteins with differing ligand affinities (9). The AhA'-1, Ahrb-2, and Ahrb-3 alleles code for AHRs with higher affinity than the AHR encoded by the Ahrt allele (). The AhrbJ1 allele occurs in C57, C58, and MA/My inbred strains; the Ahtb-2 allele is carried by the C3H, BALB/cBy, and A inbred strains; the Ahr'-3 allele exists in Mus caroli, Mus spretus, and MOLF/Ei; and the AhAd allele occurs in AKR, DBA/2, and 129 strains (9).
Ahr nucleotide differences, and corresponding AHR amino-acid changes between the C57BL/6 (B6) and DBA/2 (D2) mouse strains that carry the AhA'-1 and Ahkd alleles, respectively, have been studied extensively (9)(10)(11)(12)(13)(14)(15)). An A375V change has been reported to confer most of the observed phenotypic differences between B6 and D2 (15). With hepatic cytosol from B6 and D2 mice, ligand-affinity differences were found to range between 4-and 10-fold for the B6 and D2 AHRs (1618, whereas ligand-affinity differences range between 2and 6-fold for the B6 and D2 AHRs when cDNA-expressed AHRs are studied (9,15). These in vitro ligand-binding assays appear not to accurately reflect the in vivo variability in this polymorphism, however, because the B6-D2 differences in hepatic aryl hydrocarbon hydroxylase (CYPlA1) inducibility in the intact mouse are much larger-in the order of 15to 30-fold (6-8,15P. Many of the toxic effects ofAHR ligands observed in mice also occur in exposed human populations, in which a large amount of interindividual variability is seen; however, a DNA-based explanation of the human AHR polymorphism is still to be determined. CYPlAl inducibiity varies by more than 30-fold in humans (20)(21)(22)(23), and Scatchard plot analyses of 106 human placental samples have shown a >35-fold difference in AHR affinity between the highest and lowest phenotype (24,25), yet there has not been any direct correlation established between the degree of CYPlA1 inducibility and the variation in AHR affinity in humans  nucleotide or amino acid change in the human AHR has been unequivocally associated with ligand variability. Substitution of alanine for valine at position 381 (equivalent to position 375 in the mouse) was shown to increase ligand affinity several-fold in an in vitro cDNA-expression binding assay (15); however, the human AHR protein from the HepG2 and A431 cell lines, as well as from at least eight individuals, contains valine at position 381 in all cases (15,22, nucleotide change in exon 2 of the human AHR gene has been reported (29), but no phenotypic changes associated with this polymorphism were examined.
Progress in correlating any human AHR genotype with variations in the AHR phenotype has been hindered by the lack of efficient methods for screening large numbers of individuals. Recently, reverse transcriptase-polymerase chain reaction (RT-PCR) has been successful for rapidly sequencing portions of the AHR gene (22,29), but this approach has been limited to small groups of individuals and may be too cumbersome for identification of nucleotide changes in large populations. Approaches that rely on oligonudeotide hybridization (30), however, may provide a solution to this problem, particularly in light of the recent advances in DNA chip technology (31).
Detection of nudeotide differences must be coupled to the analysis of resulting functional changes in order to assess the relevance of any given polymorphism to environmentally related disease. In the case of the AHR, the use of mammalian tissue culture systems for this purpose has clearly proven to be unreliable. For example, expression of exogenous AHR in AHR-deficient CV-1 African green monkey kidney epithelial cells and mouse hepatoma c35 cells results in high basal CYPlA1 activity and low inducibility (32)(33)(34), possibly due to the presence of an endogenous AHR ligand (35); therefore, the expression of exogenous AHR in cell culture would not be useful for phenotypic characterization of the human AHR. A widely used method for the phenotypic analysis ofhuman AHR function has been the mitogen-activated lymphocyte culture assay for CYPlA1 induction (36). Implementation of this method is not trivial, with many laboratories experiencing difficulties in reproducibility (37,38). These findings underscore the importance of developing a reliable, efficient, and noninvasive test to determine AHR phenotype in large human populations.
The yeast Saccharomyces cerevisiae may provide the means for such a quantitative phenotyping test; in contrast to AHR-deficient mammalian cell lines, expression of exogenous AHR in yeast generates a highly sensitive detection system for the analysis of AHR activation (34,(39)(40)(41). In this report, we have used the well-characterized Ahr polymorphism between B6 and D2 mice to optimize an oligonucleotide-hybridization screening approach for the identification of Ahr nucleotide changes and to develop a yeast two-hybrid approach for the functional analysis of the corresponding AHR protein.

Materials and Methods
Sequencing by hybridization. Twelve sets of four 21-residue oligonucleotides (Table 1), for a total of 48 target oligonucleotides, were  Oligonucleotides (21-mersl were synthesized for 10 polymorphic sites in the coding region of the mouse Ahr gene, 1 polymorphic site outside the coding region, and 1 control site of sequence identity between C57BL/6 and DBAN2. The sequence shown corresponds to the Ahr-1 allele. In each case, the middle nucleotide (in bold) was replaced by each of the other three nucleotides, one of which matches that of the Ahd allele (shown below each sequence); the other two oligonucleotides of each set are not shown. At positions 2679, 2687, 3330, and 3336-due to the proximity of two Ahrb-/Ahid polymorphic sites-the Ahrdconsensus oligonucleotides contain an additional nucleotide change.
commercially synthesized (Bio-Synthesis, Inc., Lewisville, TX) for hybridization experiments. Each set corresponded to 1 of the 11 sites at which the AhM' and Aht4 alleles differed, plus 1 control site at which both alleles were identical. The four oligonudeotides in a set differed only by the middle 11th residue, which contained A, C, G, or T. For each set, one oligonucleotide matched Ahrb-1, another matched Ahr*', and the other two matched neither.
The sequences and theoretical dissociation temperatures of 23 of the 48 oligonudeotides are shown in Table 1. Five of the 11 polymorphic sites encode amino acid sequence changes. Oligonucleotides were transferred to a Nytran membrane (Schleicher & Schuell, Keene, NH) using a slot-blotting apparatus (Bio-Rad, Richmond, CA). Each oligonudeotide (1 pg) was applied to a slot by filtration under vacuum, washed with 0.5 M sodium phosphate buffer (pH 7.0), and UV cross-linked to the membrane. Membranes were hybridized with B6, D2, or each of eight chimeric AHR cDNA probes (containing combinations of the polymorphic sites from the Ahrb-1 and Ah?4alleles). The chimeric cDNAs were constructed using standard recombinant DNA techniques (12,35), and encode the amino acid changes shown in Figure 1. cDNA probes were labeled with [32P]dCTP (Amersham, Arlington Heights, IL) by nick translation (Gibco-BRL, Inc., Rockville, MD). Membranes were prehybridized in a buffer containing 6X SSC (0.9 M NaCI, 0.09 M sodium citrate), 1% sodium dodecylsulfate, and 100 jg denatured calf thymus DNA per milliliter. Hybridizations were carried out in the same buffer, with the addition of 1 x 106 dpm/ml [32P]-labeled probe. Following overnight hybridization at temperatures ranging from 460C to 600C, membranes were washed at the hybridization temperature in 6X SSC, 0.1% sodium dodecylsulfate, and exposed to x-ray f'ilm. Hybridization intensities were quantified by phosphorimaging in a Storm 860 (Molecular Dynamics).
Yeast plarmid constructs and yeast transformation. The B6, D2, and eight chimeric AHR cDNAs, cloned in the eukaryotic expression vector pCDNAI/ Amp (35), were excised from this vector and cloned into the trp' yeast expression vector pGBT9 (ClonTech, Palo Alto, CA) in-fr-ame with the DNA binding domiain of GAL4, resulting in each of 10 different pGBT9AHR plasmids. The AHR nuclear translocator (ARNT) fulllength cDNA was excised from pCDNAneol ARNT and cloned into the keu+ yeast vector pGAD424 (ClonTech) in-frame with the GAL4 transactivation domain, resulting in plasmid pGAD424ARNT. S. cerevisiae strain SFY526 (leu, tPp-) was grown in YEPD medium (10 g/l yeast exract, 2 g/l peptone, 20 g/l dextrose, 253 mg!1 adenine, 133 mg/l uracil) and transformed by electroporation with pGAD424ARNT and plated on yeast minimal medium (1 M sorbitol, 1.7 g yeast nitrogen base/1, 1 g/l ammonium sulfate, 10 g/l succinic acid, 6 g/l NaGH, 20 g/l dextrose, 2% (w/v) agar, and a cocktail of essential amino acids minus leucine). After incubation at 300C, individual colonies were grown in liquid medium (same medium as above but without sorbitol or agar), transformed with each of the pGBT9AHR plasmids by electroporation, and plated on minimal medium minus leucine and tryptophan. Several individual colonies from these plates were then used for the AHR activation studies.
P3-Galactosidase assays. Minimal medium minus leucine and tryptophan was inoculated with individual yeast colonies and grown overnight at 30C to saturation.
Dioxin, 03-naphthoflavone (BNF), or BaP (dissolved in dimethylsulfoxide) was then added to fresh medium to final concentrations ranging from 10-10 to i0-5 M, and the medium was inoculated with an aliquot of the saturated yeast cultures to an GD600 (optical density) ranging between 0.05 and  Figure 2. Hybridization of C57BL/6 (B6) and DBA/2 (D2) cDNA probes to oligonucleotide arrays spanning all polymorphic sites in the Ahrb-1 and Ahrd1 alleles. The hybridization temperature was 56C. The allele, the nucleotide (NT), and its position in the AHR cDNA are indicated adjacent to a representative example of a hybridization experiment. Hybridization intensities were quantitated for all probes, and the mean values relative to the maximum for each oligonucleotide in a set are shown.
was stopped by addition of ZnCl2 to a final concentration of 1 mM; P3-galactosidase activity was assessed by measuring the absorbance at 578 nm and calculated using the equation: IP-galactosidase activity = (1,000 X A7lA6 x cell volume (ml) X time (mm)].

Results
Detection ofsingle Ahr nucleotide changes by oligonucleotide hybridization. To test whether the oligonucleotide-hybridization screening approach could be used to identify nucleotide changes reflecting amino acid differences, we hybridized B6, D2, and chimeric AHR cDNA probes to an oligonucleotide array containing single-base substitutions at all mouse Ahr polymorphic sites, plus the one control invariant site. Representative hybridization results are shown in Figure 2 for each of the 12 sets of four oligonucleotides. The relative hybridization intensities of all the probes were calculated for each oligonucleotide array position, and the means for a given nucleotide position are shown adjacent to the corresponding slot. In all cases, hybridization was strongest with the correct oligonucleotide, although we detected varying degrees of hybridization to the other three oligonucleotides. Hybridization to Ahrbsequences by 1 1 of the 12 B6 probes and hybridization to Ahid sequences by 10 of the 12 D2 probes occurred with the highest degree of specificity. Nonspecific binding was less than 19% of maximum for Environmental Health Perspectives . Volume 106, Number  Of the other two, two pro one at the correct T and t were observed for nudeotide At position 3336, very litti took place, possibly becausi cleotide also carried a secc change at position 3330. N quantitation showed that h the correct oligonucleotide, this position, was the stronge All hybridization experin Figure 2 were conducted at found that the dissociation ti each of the 48 oligonucleoti4 ly (Table 1). We tested whi hybridizations that generat results could be resolved by hybridization temperature. ;onudeotides at oligonucleotide. Ambiguity for the Ahrt-1 addition to the consensus oligonudeotide at position 2038 the correct C was resolved as the hybridization temperas containing G ture was increased to approach the dissociaugh with lower tion temperature. For the D2 cDNA 0, respectively. hybridization at this position, two of the tminent bands, nonspecific bands were resolved, but no dear he other at G, improvement in discrimination for the other position 2038.
two was seen. Hybridization of D2 cDNA to e hybridization the oligonucleotides spanning position 3336 ,e this oligonuwas unambiguous at 46°C, but the overall rnd nucleotide intensity was greatly decreased as the temperotwithstanding, ature was increased, reflecting the lower disybridization to sociation temperature due to the presence of containing A at a second nudeotide change in the consensus st.
oligonucleotide at this position. We condudnents shown in ed that the hybridization patterns clearly 56C, but we show that each probe has a strong preference emperatures for for the oligonucleotide with a perfect des differ widesequence match and that this oligonu-Lether the three cleotide-hybridization technology is capable ted ambiguous of detecting single-nudeotide changes with a modifying the high degree ofaccuracy.
TheAHR is ligand responsive in a yeast ts of hybridizatwo-hybrid system. To determine the Ler temperatures capacity of various ligands to activate the 7 position 2038 AHR in a yeast two-hybrid system, we D2 probes and transformed yeast strain SFY526 carrying position 3336 pGAD424ARNT with pGBT9AHR vecte. Regardless of tors encoding B6 or D2 AHR (Fig. 4). erature used, Cultures were treated with varying concenfor the correct trations of dioxin, BNF, or BaP ranging from 10-10 to 10-6 M and assayed for 0-M 0 f galactosidase activity. The AHR was activated by all three ligands in a dose-dependent manner. At 10-6 M dioxin treatment resulted in a sixfold increase of I-galactosidase activity, whereas BNF, and BaP~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.  we found that activation of the B6 AHR was greater than that of the D2 AHR, regardless of which ligand was used (data not shown). The low levels of induction by dioxin at doses 1,000-fold higher than those needed in mammalian cells are probably due to low permeability of yeast cells for very hydrophobic compounds. These results confirm work by others (34,(39)(40)(41), demonstrating that a yeast two-hybrid system can be used to assess ligand-dependent activation of the AHR.
Quantitation ofAHR ligand affinity in the yeast two-hybrid system. The observation that the AHR is ligand responsive in yeast (Fig. 4) led us to test whether the yeast two-hybrid assay could be used to assess functional differences in AHR phenotype, i.e. variability in ligand affinity. Using doses of BNF ranging from 10-10 to 10-5 M, we found that BNF activates the B6 AHR at a dose approximately 15-fold lower than the D2 AHR (Fig. 5A). To identify amino acid changes responsible for this difference in induction levels, dose-response experiments were conducted with the chimeric AHR constructs (illustrated in Fig. 1). We found a clear bimodality in dose-response curves of the different chimeric AHR proteins (Fig. 5B). Regardless of other changes, all chimeras carrying alanine (Ala)-375 responded in a fashion similar to that of the B6 receptor, whereas chimeras containing valine (Val)-375 responded like the D2 receptor. In addition, plateau values at the higher doses were 1.5to 2-fold higher for the B6-like chimeras (data not shown).
Dose-response data were fitted to a four-parameter function to estimate the dose capable of generating the half-maximal activation of the f-galactosidase reporter gene (EC50). Table 2 presents the EC50 values for B6, D2, and all eight AHR chimeras. AHR proteins carrying Val-375 were found to exhibit 14to 24-fold higher EC50 values than AHR proteins containing Ala-375.

Discussion
.'. We have shown in this report that oligonudeotide hybridization screening can be effectively used to identify mouse Ahr nudeotide TCDD BNF BaP differences with a high degree of success and reliability. From an array of 48 oligonu-*galactosidase activity following treat-deotides containing individual base substituast with three aromatic hydrocarbon tions at 11 polymorphic sites, specific ligands at three concentrations. hybridization was evident for 1 1 of the 12 ns: TCDD, 2,3,7,8-tetrachlorodibenzo-AO'l allele sites and 10 of 12 Ahrallele sites. NF, 3-naphthoflavone; BaP, benzo(a)-We have also shown that hybridization to the ssay results are reported as fold nave olsonotdf th e otOthe over dimethylsulfoxide-treated con-incorrect oligonudeotide for the other three bar represents the mean ± standard sites occurred-to varying degrees, causing three samples.
less-than-definitive results-for two reasons.
Volume 106, Number 7, July 1998 * Environmental Health Perspectives . Articles * Characterization of the AH receptor polymorphism use of suboptimal hybridization temperature and close proximity of two polymorphic sites. Our results demonstrate the power of this method to identify sequence polymorphisms without the need for DNA sequencing. Although this is the first report of the application of oligonudeotide hybridization screening of the Ahr gene, this approach has been used successfully to detect polymorphic alleles of several genes, induding S-globin, HLA-A, BRCAI, and ,B-thalassemia (42)(43)(44)(45)(46).
We envision the application of modifications of this method (e.g., DNA chip technology) for the rapid and unequivocal screening of the human AHR gene polymorphism in a large number of individuals. The complete 2,547-nudeotide coding sequence of the human AHR cDNA could be represented in an array of 849 21-mer sequential oligonucleotides, each beginning three nucleotides after the previous one. Any polymorphism, detected by absence of hybridization to the test probe, would be present in seven (21/3) sequential oligonucleotides. Inclusion of control and test probes labeled with different fluorescent dyes in the same hybridization reaction would provide a rapid and efficient means of detecting possible polymorphisms. Dozens of samples could be processed in a single day. Resolution of ambiguities and determination of the actual nucleotide present at a polymorphic site could then be determined simply by direct sequencing of the corresponding short stretch of DNA, thus greatly decreasing the cost and time needed for screening a large number of samples.
We also have also demonstrated in this report the usefulness of a yeast two-hybrid assay to quantitate the mouse AHR phenotype, i.e., its ligand affinity. The phenotypic variation, ranging from 15to 30-fold differences, between the B6 and D2 AHR proteins in the intact animal was shown to be accurately reflected in the yeast twohybrid system. As measured by P-galactosidase reporter activity in this assay, the EC50 for BNF is about 15-fold lower for B6 than D2, in good agreement with B6-D2 differences in CYPlAI inducibility in the intact animal, and two to four times greater than B6-D2 differences reported in in vitro ligand-binding studies (9,15). In addition, using chimeric AHRs, we were able to verify in this biological system the importance of the A375V change in generating the low-affinity phenotype.
The yeast two-hybrid system thus appears to represent more faithfully the biological difference in CYPlAl inducibility phenotype than in vitro ligand-binding assays. In vitro, the low-affinity forms of the AHR require stabilization by molybdate senting BNF activation of the B6 and the D2 AHR. All 21 data points for both B6 and D2 AHRs were fitted to the 4-parameter function fix) = (maxmin)/[l+ (xj)sl + min. (B) BNF dose-response curves for B6, D2, and eight B6-D2 chimeric AHRs. These experiments were repeated three times with comparable results; data from one representative experiment are shown. Curve fits were completed, as described above, using the data points shown. (17), whereas in the yeast system this artificial stabilization of the AHR is not required, perhaps providing a more accurate physiological model of AHR function. In addition, different AHR phenotypes might reflect aspects ofAHR activation other than ligand binding, such as translocation to the nucleus or dimerization with ARNT (5) that are not evaluated in the in vitro ligandbinding assays.
The existence of a human AHR polymorphism to explain the observed variability in CYPlAI inducibility (20)(21)(22) and AHR affinity (2426) has been hypothesized for many years; however, conclusive evidence to demonstrate this putative polymorphism is lacking (23,37,38). Among the B6, D2, C3H, and Mus spretus mouse lines, there are 12 AHR amino acid differences (9)(10)(11)(12)(13)(14)(15). While most amino acid changes were shown to contribute greater than twofold to receptor affinity, the opal mutation contributes two-to threefold, and the A375V change contributes fourto sixfold to the differences in AHR affinity (9,15). It is therefore very likely that the human AHR polymorphism will be made up of at least a dozen, and probably many more, amino acid differences reflecting the variability in ligand affinity. In this report, we have shown that a yeast-two hybrid assay can readily determine phenotypic differences observed between the B6 and D2 mouse AHR proteins, which leads us to propose that this system might be applicable to the analysis of large numbers of anticipated differences in human AHR proteins. The experiments reported herein only assessed ligand-binding and dimerization properties, and it is possible that functional polymorphisms in a human population may include other steps of the AHR signaling pathway (5). Expanding the yeast two-hybrid assay to include a reporter gene that responds to the transactivating activity of AHR/ARNT complexes (39) may provide a unique opportunity to identify additional steps in the AHR activation pathway that could be altered and might thus be reflected in the well-documented studies about human variation in CYPlAl inducibiity and AHR affinity.
Our results demonstrate the utility of the oligonudeotide hybridization screening approach for identification of the mouse Ahr genotype and of a yeast two-hybrid assay for quantification of the AHR phenotype. Used together, these two methods constitute a powerful novel approach to analyze the role that human AHR polymorphisms might play in environmentally related diseases caused by AHR ligands such as dioxin and BaP.