Isolation and characterization of rat cDNA clones for two distinct thyroid hormone receptors.

Two distinct v-erbA-related cDNA clones representing the products of different genes were isolated from a rat liver cDNA library. The first, rc-erbA-alpha, was 82% identical to v-erbA and encoded a polypeptide with a calculated molecular mass of 45,000 daltons. This cDNA clone arises from the same gene product as a v-erbA-related cDNA isolated from rat brain by Thompson et al. (Thompson, C. C., Weinberger, C., Lebo, R., and Evans, R. (1987) Science 237, 1610-1614). The second cDNA clone, rc-erbA-beta, was 76% identical to v-erbA and encoded a polypeptide with a calculated molecular mass of 52,000 daltons. Both rc-erbA-alpha and rc-erbA-beta translational products bound 3,5,3'-triiodo-L-thyronine with affinities equal to each other (Kd approximately equal to 0.4 nM) and comparable to the nuclear thyroid hormone receptor extracted from rat liver. The relative affinities of a series of thyroid hormone analogs for both translational products were also identical. In various tissues and cell lines, the relative levels of rc-erbA-beta RNA, but not rc-erbA-alpha RNA, correlated with measurements of nuclear 3,5,3'-triiodo-L-thyronine binding sites. Based on this correlation, we suggest that rc-erbA-beta may encode the "classical" nuclear thyroid hormone receptor, whereas rc-erbA-alpha may encode an isoreceptor species with differing functional properties.


Isolation and Characterization of Rat cDNA Clones for Two Distinct
TWO distinct v-erbA-related cDNA clones representing the products of different genes were isolated from a rat liver cDNA library. The first, rc-erbA-a, was 82% identical to v-erbA and encoded a polypeptide with a calculated molecular mass of 45,000 daltons. This cDNA clone arises from the same gene product as a v-erbA-related cDNA isolated from rat brain by Thompson et al. (Thompson, C. C., Weinberger, C., Lebo, R., and Evans, R. (1987) Science 237, 1610-1614). The second cDNA clone, rc-erbA-b, was 76% identical to v-erbA and encoded a polypeptide with a calculated molecular mass of 52,000 daltons. Both rc-erbA-a and rc-erbA-b translational products bound 3,5,3'-triiodo-~-thyronine with affinities equal to each other (& = 0.4 nM) and comparable to the nuclear thyroid hormone receptor extracted from rat liver. The relative affinities of a series of thyroid hormone analogs for both translational products were also identical. In various tissues and cell lines, the relative levels of rc-erbA-b RNA, but not rc-erbA-a RNA, correlated with measurements of nuclear 3,5,3'-triiodo-~-thyronine binding sites. Based on this correlation, we suggest that rc-erbA-8 may encode the Kclassical" nuclear thyroid hormone receptor, whereas rc-erbA-a may encode an isoreceptor species with differing functional properties.
High affinity, limited capacity binding sites for 3,5,3'triiodo-L-thyronine (TJ' localized in the nuclear compartment of target tissues were first described by Oppenheimer and colleagues (1) in 1972 based on in vivo displacement techniques. A reasonable body of evidence now exists which links this nuclear T 3 binding activity with hormone action including: (a) correlation between nuclear binding affinity and bioactivity for a wide variety of thyroid hormone analogs, ( b ) absence of the T3 binding activity in tissues unresponsive * These studies were supported by Grant DK26919 from the National Institute of Arthritis, Diabetes, Digestive and Kidney Diseases. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. to hormone, and ( c ) correlation between the degree of receptor occupancy and biological activity (reviewed in Ref. 2). Despite the fact that the T3 receptor has not been purified to homogeneity, it has been characterized as a protein with a molecular mass of approximately 50,000 daltons (3-6), which can bind to DNA in uitro (7). In view of the rapid effects of T3 on gene expression, it is generally thought that the thyroid hormone receptor acts by binding to specific DNA sequences adjacent to or within target genes to regulate their expression. In the case of the rat growth hormone gene, a binding site for the thyroid hormone receptor in the 5"flanking region has recently been reported (8,9). This binding site correlates at least in part with sequences of the growth hormone gene which are capable of conferring hormonal responsiveness on a linked heterologous promoter (9-12). Thus, the current model of thyroid hormone action is well supported in this case.

The nucleotide sequence(s) reported in thispaper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s)
Recently, it has been suggested that the cellular erbA gene product is the thyroid hormone receptor (13, 14). The basis of this hypothesis was the finding that the c-erbA polypeptide, expressed by in uitro transcription and translation from a cloned cDNA, bound to T3 with an affinity similar to that for the native receptor. In addition, the binding of a number of thyroid hormone analogs to the c-erbA polypeptide showed the same hierarchical order as that observed for the nuclear receptor. The relative molecular mass of the c-erbA polypeptide was 46,000 daltons from chicken and 52,000 daltons from human. Finally, the c-erbA mRNA sequence is homologous to the sequence of the steroid hormone receptors (15-19). Most notably, a region corresponding to a domain of 68 amino acid residues of the steroid receptors which is involved in binding to specific DNA sequences is highly conserved (20). Considering the similarity in the early cellular action of thyroid and steroid hormones, it was thus reasonable to suggest that these receptors evolved from a common ancestral gene.
In humans, multiple erbA-related genes have been found two genes are located on chromosome 17 (21) and two genes are found on chromosome 3 (14). The former have greater similarity (-82% identity at the nucleotide level) to the chicken v-erbA oncogene than the latter (~7 4 % ) (14, 22). At least one member of each of the v-erbA-related gene pairs is expressed, as evidenced by isolation of the corresponding cDNA clones. In this paper, we report that at least two C-erbA genes are expressed in rat, often in the same tissues. The binding of T3 and other thyroid hormone analogs to both of these gene products occurs with similar affinity. The relative abundance of one form (p) correlates well with nuclear T3 binding activity of various cells and tissues and may thus represent the "classical" T3 receptor. We propose that the other form (01) may represent an isoform of the classical T3 receptor.

EXPERIMENTAL PROCEDURES
cDNA Cloning and Sequencing-A rat liver cDNA library was constructed in the Xgtll vector from total hepatic poly(A)-containing RNA of a normal male adult Sprague-Dawley rat (23). First strand cDNA synthesis was carried out in two separate reactions using oligo(dT) and a random mixture of octanucleotides as primers for reverse transcriptase. Equal portions from the two reactions were mixed and used for second strand synthesis by the method of Gubler and Hoffman (24). Following methylation of internal EcoRI sites and addition of EcoRI linkers, double-stranded cDNA was size-selected on a Sepharose CL-4B column to enrich for molecules greater than 1000 bp in length. A library of 1.3 X lo6 independent recombinant phage was obtained.
The cDNA library was initially screened with a DNA fragment corresponding to nucleotides 277-1140 of the chicken v-erbA oncogene (numbering according to Debuire et al. (25)). This fragment was radiolabeled by the method of Feinberg and Vogelstein (26) and used to screen 1 X lo5 recombinant phage. Hybridization was performed in 50% formamide, 0.75 M NaCl at 42 "C and washing at 50 "C in 0.015 M NaCl, 0.1% SDS. A single positive plaque was isolated with an insert size of 1200 bp. This fragment was subcloned into the EcoRI site of pTZl8R (27) and used for rescreening 5 X lo5 additional plaques. A second positive plaque was purified which carried an insert of approximately 2300 bp. This fragment was likewise subcloned into pTZ18R and designated rc-erbA-a. The library was rescreened with the human placental erbA-8 clone of Weinberger et al. (14). A single additional positive plaque was identified with an insert of approximately 2450 bp which was subcloned into pTZ18R and designated rc-erbA-8.
DNA sequencing was performed by the dideoxy chain termination method of Sanger (28) using [35S]dATP. For rc-erbA-P, a series of deletion clones from either end was generated by unidirectional digestion with exonuclease 111 as described by Henikoff (29). All portions of the sequence were determined independently in both directions from overlapping clones except the first 128 bases of the 5"untranslated region, which was sequenced only in one direction.
Thyroid Hormone Binding Studies-Polypeptides encoded by rc-erbA-a and -8 were expressed by in vitro transcription-translation.
Plasmid DNA was linearized by digestion with S m I for clones in pTZ18R or with EcoRI for the 5'-deleted rc-erbA-a clone (see below). After phenol extraction and ethanol precipitation, plasmid DNA was transcribed using the conditions of Krieg and Melton (30) except that no bovine serum albumin was included and NaCl was added to 25 mM. In addition, the dinucleotide 7MeGpppG (Pharmacia LKB Biotechnology Inc.) was added to 0.5 mM to yield capped RNA molecules (31). RNA was purified by treatment with RNase-free DNase, phenolextraction, and chromatography on a Bio-Gel P-30 spun column. Translation of RNA was performed in the micrococcal nucleasetreated rabbit reticulocyte lysate system (Promega Biotec) in 30-pl reactions at 30 "C for 60 min. Ten pCi of [36S]methionine (Du Pont-New England Nuclear) was added for protein labeling experiments. To achieve efficient expression of rc-erbA-a, it was necessary to remove a large portion of the 5'-untranslated region. For this purpose, rc-erbA-m was linearized by cleavage with BamHI and then subjected to digestion with exonuclease I11 (29). Samples were then blunt-ended with SI nuclease, and the deleted inserts were purified from the vector by cleavage with EcoRI and electrophoresis on low melting agarose. DNA fragments of the appropriate length were subcloned between the SmaI and EcoRI sites of pTZ19R (27) and sequenced to determine the extent of deletion. The clone used for in vitro transcriptiontranslation studies, rc-erbA-a (-29), retained 29 bp of 5"untranslated region.
Thyroid hormone binding was performed using the procedure of Sap et al. (13). Aliquots (0.3-1 pl) of translational products were incubated for 16 h at 4 "C with varying concentrations of [1251]T3 (Du Pont-New England Nuclear, 2200 Ci/mmol) in a final volume of 10 p1 in a buffer containing 20 mM Tris-HC1, pH 7.5, 50 mM NaC1, 2 mM EDTA, 5 mM 2-mercaptoethanol, 10% (v/v) glycerol. Nonspecific binding was assessed by addition of 1 p~ unlabeled T, (Sigma) to the reactions. Bound T3 was determined by the nitrocellulose filter binding assay of Inoue et al. (32). Control reactions programmed with rc-erbA clones in the opposite orientation gave no binding above nonspecific background. For competition experiments, 1 nM [1251]T3 was mixed with varying concentrations of the appropriate analog (Sigma), and binding was analyzed as above.
RNA Extraction and Northern Analysis-Tissue RNA samples were purified by a modification of the guanidine-HCl method of Deeley et al. (33) from normal adult male Sprague-Dawley rats (BioLabs, St. Paul, MN). Poly(A)-containing RNA was extracted by two successive rounds of chromatography on oligo(dT)-cellulose (Collaborative Research). Total RNA was extracted from cultured cells by homogenization in 4 M guanidine thiocyanate followed by centrifugation through a 5.7 M CsCl cushion as described by Chirgwin et al. (34). RNA samples were subjected to electrophoresis on 1.5% agarose gels containing 2.2 M formaldehyde (35). Following transfer to Zeta-Probe nylon membranes (Bio-Rad) by electroblotting, RNA was hybridized with indicated probes as described by Narayan et al. (35). Sizes of RNA species were estimated using RNA markers from Bethesda Research Laboratories (RNA ladder).
Rat hepatocyte-hepatoma fusion cells were isolated by somatic cell fusion of primary rat hepatocytes and the rat hepatoma cell line FA0 (36). The FA0 cell line lacks hypoxanthine-guanine phosphoribosyl transferase and is resistant to 1.5 mM ouabain. Fusion cell lines which were selected in the presence of 8-azaguanine and ouabain were maintained on Ham's F-12 media with 10% fetal calf serum. T3 nuclear receptor levels in various cell lines were measured as described by Mariash et al. (37).

RESULTS
Isolation of Two erbA-related Rat Liver cDNA Clones-Liver is a major target organ for thyroid hormone, containing approximately 4500 receptors/cell in the rat (38). In addition, the expression of several genes in liver, such as SI4 and malic enzyme, are known to be responsive to circulating thyroid hormone levels (for review, see Ref. 39). We therefore decided to screen a rat liver cDNA library for cDNA clones which cross-hybridized with erbA sequences. Using chicken v-erbA DNA and a human placental c-erbA cDNA as probes, we have ,/isolated two distinct cDNA clones representing the products of different genes (Fig. lA). The first was a cDNA with an insert size of 2300 bp which was about 82% identical to the chicken v-erbA oncogene at the nucleotide level. The restriction enzyme map of this cDNA was identical to that recently reported by Thompson et al. (40) for a rat erbA-homologous cDNA isolated from brain. Sequencing of three segments of this clone, representing 5'-flanking, coding, and 3"flanking regions of the mRNA, revealed that it was identical to the sequence of the brain cDNA (data not shown). Thus, we conclude that this cDNA arises from the same gene product as the rat brain cDNA and have designated this clone, rc-erbA-a, consistent with the nomenclature suggested by Weinberger et al. (14).
The second rat liver cDNA clone had an insert size of 2450 bp and a restriction map which was distinct from rc-erbA-a ( Fig. lA). To demonstrate that this cDNA arose from a different gene, the two cDNA clones were used as probes for Southern hybridization of rat genomic DNA cleaved with a variety of restriction enzymes (Fig. 1B). The probes used in this experiment were generated from the 3'-portion of the respective cDNA clones in order to avoid the highly conserved DNA-binding motif. The pattern of hybridizing bands observed under these moderate stringency conditions was unique for each probe. For several restriction enzymes, only a single band was detected with either probe. Thus, the two erbArelated cDNA clones arise from two different genes in rat.
The nucleotide sequence of the second cDNA clone was determined (Fig. 2). This cDNA clone was only about 76% identical at the nucleotide level to the chicken v-erbA, a value similar to that reported for a human placental c-erbA isolated by Weinberger et al. (14) and designated as the ,8 class. Thus, we designated this clone as rc-erbA-P. The nucleotide sequence contained a single long open reading frame. Two possible initiation codons are present at nucleotides 124 and 136. Based on analogy to the human placental cDNA in which only the initiation codon at position 136 is found and the somewhat weaker "context" of the AUG codon at position

Multiple Thyroid Hormone Receptors in the Rat
1 2 3 4 5 6 1 2 3 4 5 6 rc-erbA-a rc-erbA-P Restriction enzymes which did not digest either cDNA were EcoRI, HindIII, Sa& and SphI. B, restriction enzyme analysis of genomic DNA. Rat genomic DNA (IO pg) was digested with BarnHI ( l a n e 1 ), EcoRI ( l a n e 2). HincII ( l a n e 3), HindIII (lane 4 ) , SstI ( l a n e 5), or a combination of BamHI and EcoRI (lane 6), and products were separated on a l% agarose gel. Fragments were transferred to Zeta-Probe membranes and hybridized to 1 X lo6 cpm/ml of '*P-labeled rc-erbAa probe (left) or rc-erbA-8 probe (right). The rc-erbA-a probe used was a fragment from the HincII site to the 3'-end of the cDNA, and the rc-erbA-8 probe was a fragment from the PuuII site to the 3'-end of the cDNA. Filters were washed at 60 "C in 0.015 M NaCI, 0.1% SDS and subjected to autoradiography for 3 days.
124 (41,42), we have designated the initiation codon at 136. This suggestion will need to be confirmed in future studies. The open reading frame using this AUG encodes a polypeptide of 456 amino acid residues and a calculated molecular mass of 52,100 daltons. It is preceded by a 5"untranslated region of at least 265 nucleotides which contains five additional short open reading frames. The sequence is 87% identical to the human placental erbA-/3 cDNA at the nucleotide level.
Comparison of the sequences of the two rat cDNA clones reveals that they are unrelated throughout the 5"untranslated region and the N-terminal segments of the encoded polypeptides (Fig. 3A). Beginning a t amino acid 41 of the a sequence and amino acid 90 of the j3 sequence, the remaining amino acid sequences are 83% identical. The similarity is most noteworthy in two stretches of amino acids in the Cterminal half of the polypeptides: from amino acid residues 267 to 318 of the / 3 form, 50 out of 51 residues are identical and from residues 411 to the C-terminal end of 8, the two polypeptides are identical for 46 consecutive residues. In the 68-amino acid, cysteine-rich domain of c-erbA which is most related to the DNA-binding domains of the steroid hormone receptors, the two sequences are common a t 59 out of 68 residues (Fig. 3B). The differences are clustered in the second half of the domain at positions which show a high degree of variability between different classes of steroid hormone receptors.
Thyroid Hormone Binding Activity of c-erbA Polypeptides-In order to confirm that both erbA-related cDNA species encode polypeptides capable of binding to T3, hormone binding experiments were carried out. The two cDNA inserts were subcloned into the pTZ18R vector in proper orientation behind the T7 RNA polymerase promoter sequence. The encoded polypeptides were then expressed by in vitro transcription followed by translation in a rabbit reticulocyte lysate. To achieve significant levels of expression of the rc-erbA-a cDNA, it was necessary to remove a large segment of the 5'untranslated region. Following this deletion, a major polypeptide product with an estimated molecular mass of 50 kDa was obtained (Fig. 4). The rc-erbA-8 cDNA was expressed without removal of its 5'-untranslated region and resulted in a major polypeptide product of 57 kDa. In both cases, a number of lower molecular weight bands of lesser intensity could be seen. These products could result from proteolytic cleavage of the initial product, initiation a t internal AUG codons or incomplete translation. The difference in size observed for the major translational products of the a and / 3 cDNA clones is consistent with that predicted from the nucleotide sequences (i.e. 45,000 daltons for predicted a polypeptide versus 52,000 daltons for B).
Both a and , 9 c-erbA clones encode polypeptides which bind T3 with high affinity. The dissociation constants estimated for TI binding were 0.4 and 0.49 nM, respectively (Fig. 5).
These values are in reasonable agreement with that reported for the nuclear T3 receptor extracted from rat liver (43). Competition for binding of [lZ5I]T3 to the c-erbA translational products by several thyroid hormone analogs was also very similar (Fig. 6). For both forms, competition with L-thyroxine required about eight times as much hormone as that observed for T3, whereas 3,3',5'-triiodo-~-thyronine was less than 100 times as effective. 3,5,3'-triiodothyroacetic acid, an analog which binds with a 1.6-fold higher affinity to the extracted liver receptor than T3 (2), was slightly more effective as a competitor than T3 itself. Again, both a-and B-derived polypeptides were indistinguishable with respect to their 3,5,3'triiodothyroacetic acid binding properties. Thus, the hormone binding properties of the two erbA-related polypeptides appear to be highly conserved.
Distribution of a and c-erbA Transcripts-Major target organs for thyroid hormone in the adult rat, such as anterior pituitary, liver, kidney, heart, and brain, contain between 2000 and 6000 nuclear T3 binding sites/cell (38). On the other hand, spleen and testis, two tissues generally considered as nonresponsive to T3, have few, if any, nuclear T3 binding sites. We compared the relative amounts of transcripts corresponding to the two forms of rc-erbA in these tissues. Poly(A)-containing RNA was extracted from each of the tissues and analyzed by electrophoresis in denaturing agarose gels. After transfer, filters were hybridized under conditions in which the two rc-erbA cDNA clones do not cross-hybridize (Fig. 7). In all tissues, three distinct RNA species with sizes of approximately 2600,5500, and 6600 nucleotides hybridized to the rc-erbA-a probe. Under the stringency conditions used, the rc-erbA-a probe only hybridizes to a single gene in the rat

GTC ACC CGT AAC CAG TGC CAG GAA TGT CGC TTT AAG AAA TGC ATC TAT GTT GGC ATG GCA ACA GAC CTG GTG CTG GAT GAC AGC AAG AGG ( 540) Val Thr Arg Asn Gln Cys Gln Glu Cys Arg Phc Lys Lys Cys Ile Tyr Val Gly Met Ala Thr Asp Leu Val Leu Asp Asp Ser Lys Arg 160
CI 170 180   genome and thus these three transcripts likely represent the products of a single gene transcript which is alternatively processed. As reported by Thompson et al. (40), these forms of erbA-related mRNA are most abundant in brain compared to other tissues. In addition, pituitary, heart, and kidney all express relatively high levels of the three a gene transcripts. Conversely, liver RNA has a very low level of these transcripts. In fact, the hepatic level is no higher than that observed in the nonresponsive tissue spleen. Based on densitometric comparison of different exposures, we estimate that the brain contains roughly 8-12 times more a transcripts than heart, pituitary, or kidney and roughly 70 times more a transcript than the liver or spleen. Thus, the tissue distribution of the rc-erbA-a transcripts is different from that previously determined for the nuclear thyroid hormone binding sites in rat. Hybridization of RNA from various tissues to the rc-erbA-/ 3 probe revealed a distinct pattern. A single RNA species with a size of approximately 6500 nucleotides was detected. The relative concentration of this species was high in pituitary, liver, kidney, heart, and brain, all tissues which contain relatively high levels of nuclear T3 binding sites. On the other hand, transcripts capable of hybridizing to rc-erbA-6 were undetectable in spleen and testis, even on longer exposures of this filter. Thus, this form of erbA-related transcript showed a reasonable correspondence to the previously reported nuclear T3 binding sites in these various tissues.

CAA GAC AGT TTC CTG TTG GCC TTT GAA CAC TAT ATC AAT TAC CGG AAG CAC CAT GTG ACA CAC TTT TGG CCC AAA CTC CTG ATG AAG GT Gln Asp Ser Phe Leu Leu Ala Phe Glu His Tyr Ile Asn Tyr Arg Lys His His Val Thr His Phe Trp Pro Lys Leu Leu Met Lys Val 400 410 ACG GAC CTG CGG ATG ATT GGA GCG TGC CAC GCC AGC CGC TTC CTG CAC ATG AAG GTG GAG TGC CCC ACC GAG CTC TTC CCG CCT CTC TT Thr Asp Leu Arg Met Ile Gly Ala Cys His Ala Ser Arg Phe Leu His Met Lys Val Glu Cys Pro Thr Glu Leu Phe Pro Pro
We have also analyzed the relative levels of erbA transcripts in several cell lines which contain different levels of nuclear T3 binding sites. For this purpose, we used the mouse 3T3-Ll preadipocyte cell line, which possesses approximately 1200 nuclear T3 binding sites/cell, and a series of rat hepatoma cell lines containing varying numbers of sites. These hepatoma cell lines included a clonal descendant of the cell line FAO, which had about 350 nuclear T 3 sites/cell. In addition, three cell lines formed by somatic cell fusion of the FA0 hepatoma with primary rat hepatocytes were tested. These three cell lines contained 750, 2400, and 7000 sites/cell. RNA was extracted from these cell lines and hybridized to the rc-erbAa probe (Fig. 8). Again, three transcripts were detected, but the relative abundance of these transcripts was inconsistent with results from T3 binding experiments. The 3T3-Ll cells contained a relatively low number of T S receptors and the highest level of RNA, whereas the fusion cell with the highest content of T3 binding sites contained one of the lowest levels of a transcripts. No consistent pattern of hybridization and T3 binding sites could be discerned.
By contrast, a reasonably good correlation between levels of T 3 binding sites and RNA capable of hybridizing to rc-  , and 7000 nuclear Ts receptors/cell, respectively. Autoradiograms were exposed for 9 h for rc-erbA-a probe and 46 h for rc-erbA-8.
erbA-B was found in the rat hepatoma cell lines. In this case, transcripts homologous to the / 3 probe were found at increasing levels in each of the hepatoma cell lines with successively higher contents of T3 binding sites. However, 3T3-Ll cells contained no detectable level of B transcript and yet displayed 1200 sites/cell. We surmise that these sites must be formed from the a transcripts. Note that in this case the autoradiogram for the a probe was exposed for only one-fifth the time of that for the B probe, despite the fact that the probes were labeled to identical specific activities. It thus appears that the a transcripts are far less effectively used to form sites detectable in the nuclear T3 binding assay than the /3 transcript.

DISCUSSION
We have isolated two different v-erbA-related cDNA clones from rat liver which encode thyroid hormone binding proteins. Comparison of the predicted amino acid sequences of the a and /3 forms of c-erbA between different species reveals that these two classes of potential thyroid hormone receptors have been conserved evolutionarily. The rat c-erbA-a is more closely related to the chicken embryonic cDNA reported by Sap et al. (13) and a human testis cDNA reported by Benbrook and Pfahl(44) than it is to the rat c-erbA-8. Likewise, the rat c-erbA-/3 is more closely related to the human placental cDNA reported by Weinberger et al. (14) than to the rat c-erbA-a. This trend is particularly noticeable in the DNA binding domains of the various forms of c-erbA (Fig. 3B). These data imply that the duplication event that gave rise to the a and B gene families occurred prior to the evolutionary separation leading to present day avian and mammalian species. We have not been able to detect further v-erbA-related genes in the rat genome by Southern hybridization at low stringency with any of the clones. However, in humans, there are at least two additional erbA-related genes (14,21,22). Whether these additional genes form functional polypeptides has not been determined.
Comparison of the amino acid sequences between a and 0 polypeptides reveals two regions in the C-terminal half which are particularly highly conserved. These regions are also nearly identical in c-erbA polypeptides from chicken and human. Thus, the sequences in these regions have remained identical for a reasonably long period of evolutionary time.

Multiple Thyroid Hormone Receptors in the Rat
Based on the regions of the steroid hormone receptors which have been shown to be involved in hormone binding, it is likely that these two regions may form essential elements in the thyroid hormone binding domain. Alterations in the gagv-erbA hybrid polypeptide (relative to c-erbA) which result in the loss of hormone binding properties have recently been mapped to the C-terminal half of the molecule, consistent with this view (45). The strong degree of conservation of amino acid sequences in these regions between a and @ forms may explain the remarkable similarity in the hormone binding properties of these two polypeptides. Both proteins have indistinguishable T3 and thyroid hormone analog binding patterns. Thus, it is likely that both polypeptides function in vivo as thyroid hormone binding proteins. Based on these binding properties, it is reasonable to suggest that both proteins may be receptors for thyroid hormone. Direct verification of this hypothesis awaits further experiments for both forms of c-erbA.
The levels of rc-erbA-@ mRNA in various tissues and cell lines correlate reasonably well with measurements of nuclear T3 binding sites. A large amount of correlative data suggests that the nuclear T3 binding sites in liver and pituitary tumor cells are true receptors for thyroid hormone (for review see Ref. 2). This leads us to hypothesize that rc-erbA-@ represents the classical T3 receptor, that is, the receptor whose activity is measured in nuclear T3 binding assays and has been correlated with bioactivity. If rc-erbA-@ encodes the classical T3 receptor, then what does rc-erbA-a encode? One possibility is that rc-erbA-a encodes a second form of nuclear T3 receptor. However, there is little or no correlation between a mRNA levels and nuclear T3 binding activity. There are several possible explanations which could account for this observation. The c-erbA-a polypeptide may have a short half-life in vivo and thus not accumulate to levels consistent with its mRNA content. The a mRNA may be inefficiently translated in vivo under normal circumstances and require certain physiological conditions for effective utilization. Finally, it is possible that only a fraction of the a transcripts may encode fulllength receptor. c-erbA-a cDNA clones have been isolated from rat brain which lack T3 binding activity, but whose sizes are indistinguishable from full length mRNA by Northern analysis (46). Sequence analysis of these clones revealed that they differed by small amino acid substitutions in the hormone binding domain. Thus, it is possible that a sizable fraction of the rc-erbA-a encoded polypeptides do not correspond to thyroid hormone hormone-binding forms. The possible physiological roles of these alternate forms is not clear.
If a portion of the rc-erbA-a gene product encodes a second nuclear T3 receptor, then what are the functions of the a and @ forms? The amino acid differences in the DNA binding domains of the two receptors may indicate that each recognizes a different set of target genes. This region has been postulated to form two zinc binding "fingers" involved in interaction of receptor with DNA. Comparison of the differences between a and fl polypeptides in this region indicates that the first putative zinc finger is completely conserved. All of the nine amino acid changes between the two forms occur in the second finger or the region between the two fingers. If these differences are sufficient to alter the DNA binding specificity of the receptors, the two forms may be responsible for regulating the expression of different sets of genes in a tissue-specific or developmentally controlled pattern. On the other hand, the two c-erbA polypeptides may interact with the same DNA sequences, but have somewhat different properties relative to the activation process. For example, the two forms may interact differently with transcriptional factors in activating or inhibiting gene expression. Distinguishing between these possibilities will require further work on the DNA binding properties of the various forms of c-erbA and the functional significance of this interaction.