Interactions among vertebrate helix-loop-helix proteins in yeast using the two-hybrid system.

The helix-loop-helix (HLH) motif is contained in a rapidly growing family of transcription factors and has been shown to mediate dimerization among heterologous HLH-containing proteins. E12 is a widely expressed HLH protein that preferentially forms heterodimers with cell type-specific HLH proteins such as MyoD, myogenin, and the achaete-scute gene products. As a first step toward screening for novel cell type-specific partners of E12, we used a modification of the two-hybrid assay for detection of protein-protein interactions in vivo to determine whether dimerization of HLH proteins with E12 can occur in yeast. Using the GAL4 DNA-binding domain fused to the E12 HLH motif and the GAL4 transcription activation domain fused to MyoD, we show that E12 and MyoD can efficiently dimerize in yeast and reconstruct a hybrid transcription factor that activates reporter genes linked to the GAL4 DNA-binding site. The GAL4 DNA-binding domain fused to E12 was used to screen a mouse T-cell cDNA library in which the cDNA was fused to the GAL4 activation domain. Several cDNA clones encoding proteins that interact with E12 were isolated, one of which corresponded to the HLH protein Id-2. Given the ability of E12 to dimerize preferentially with cell type-specific HLH proteins, this strategy should be useful for cloning novel partners for E12 from a variety of cell types.

1989a). The HLH motif is composed of two predicted amphipathic a-helices separated by an unstructured loop and mediates dimerization of heterologous HLH proteins (Murre et al., 1989b). Dimerization brings together the basic regions of HLH proteins, which lie immediately amino-terminal to the HLH motif, to form a composite DNA-binding domain. Members of the HLH family of transcription factors bind to a palindromic DNA consensus sequence CANNTG, known as an E-box, which is present in the control regions of numerous tissue-specific genes (Blackwell and Weintraub, 1990;Murre et al., 1989a).
Cell type-specific HLH proteins such as MyoD (Davis et al, 1987), myogenin (Edmondson and Olson, 1989;Wright et al., 1989), and products of the Drosophila achaete-scute complex dimerize preferentially with widely expressed HLH proteins of the E-protein class, which includes E12, E47, and HEB (Hu et al., 1992;Murre et al., 1989a;Brennan and Olson, 1990). Transcriptional activation by these types of HLH proteins can be prevented by HLH proteins lacking basic regions, which form heterodimers unable to bind the E-box consensus sequence. Among these inhibitory HLH proteins are Id-1 (Benezra et al., 1990) and Id-2 , and the extramacrochuetae gene product from Drosophila (Ellis et al., 1990;Garrell and Modolell, 1990). This class of HLH proteins has been implicated in repression of cell type-specific transcription through inactivation of E-proteins or their partners, which are required for activation of cell type-specific genes.
The presence of E-proteins in a wide range of cell types suggests the existence of cell type-specific partners for dimerization and transcriptional activation. To date, however, the only examples of tissue-specific HLH proteins in vertebrates are the myogenic HLH proteins from skeletal muscle (reviewed by Olson (1990)) and the mammalian achaete-scute homologs from certain types of neural cells (Johnson et al., 1990). Numerous attempts to clone cell type-specific HLH proteins by low stringency screening of cDNA libraries or polymerase chain reaction amplification have been unsuccessful, suggesting that if such proteins exist, they may have diverged significantly from known HLH proteins. Given the ability of several cell type-specific HLH proteins to dimerize preferentially with E-proteins, an alternate strategy to isolate such proteins would be through a functional assay based on dimerization. Previously, we have shown that interactions between heterologous HLH proteins can be detected in transfected tissue culture cells using a modification of the twohybrid system for detection of protein-protein interactions in vivo (Chien et al., 1991;Fields and Song, 1989;Chakraborty et al., 1992). Here we demonstrate that vertebrate HLH proteins will dimerize in yeast and such interactions are sufficiently stable to allow screening of cDNA libraries for novel HLH proteins based on HLH-mediated dimerization. Using the GAL4 DNA-binding domain fused to the HLH motif of E12, we show that E12 and MyoD can efficiently dimerize in yeast and reconstitute a hybrid transcription factor that can activate expression of a reporter gene linked to the GAL4 DNA-binding site. We use the DNA-binding domain of GAL4 fused to E12 to screen a mouse T-cell cDNA library for potential dimerization partners for E12. One of the cDNAs, whose product reconstructed a functional transcrip-tion factor when dimerized with the GAL4-E12 chimera, was found to encode Id-2. These results demonstrate that yeast can support HLH-mediated dimerization, and they suggest that the two-hybrid system may be a powerful strategy for isolation of novel partners for E12 from diverse cell types.

MATERIALS AND METHODS
Yeast Strains and Expression Vectors-Construction of the GAL4(AD)-MyoD fusion vector was accomplished by first cloning the 1.35-kilobase pair SstI fragment from pBS/xlmf25 (Scales et al., 1990) into the bacterial expression vector pRXl (Rimm and Pollard, 1989). The resulting plasmid (pRX/xlmf25) contained sequences encoding amino acids 2-286 of XMyoDa fused to the first 21 amino acids of TrpE and placed an EcoRI site immediately upstream of the SstI site at the 5' end of the XMyoDa-coding fragment. The xlmf25coding portion was excised from the expression vector by digestion with EcoRI, and the ends were made blunt by filling with Klenow fragment. The EcoRI fragment was cloned downstream of the GAL4 activation domain (AD) by ligation with pSE1107' that had been digested with BamHI and made blunt-ended by filling with Klenow fragment. The orientation of the MyoD-containing EcoRI fragment within the resulting plasmid, GAL(AD)-MyoD, was confirmed by sequencing across the GAL4-MyoD cloning junction. Immunoblotting showed that the GAL4-MyoD fusion protein was stably expressed in yeast (data not shown). The GAL4(DB)-PKC fusion plasmid was constructed by digesting a bovine protein kinase C (PKC) a-cDNA in pCDM8* (James and Olson, 1992) with BanI, filling in the ends with Klenow polymerase, and digesting with BamHI. The 1.1-kilobase pair fragment encodes amino acids 302-672 from bovine PKC-a. The fragment was gel-purified and then inserted into the SmaI-BamHI sites in pAS1. Details of this chimeric plasmid and the proteins with which it interacts will be described el~ewhere.~ GAL4(DB)-E12 fusion plasmid was constructed by digesting EMSV-E12 (Murre et al., 1989a) with AflIII, filling in the ends, gel-purifying the 440-bp fragment, and inserting it into the S m I site of pAS1.' This portion of E12 contains amino acids 508-654 and includes the bHLH motif. The GAL4(AD) cDNA fusion library was constructed from mouse T-cell poly(A)+ mRNA using a newly developed X vector, XACT' (Elledge et al., 1991).
Western Blot Procedure-Western blots were performed using a 1:lOOO dilution of anti-GAL(1-147) antibodies (generously provided by M. Ptashne) to detect expression of GAL4(DB) fusion proteins. Alkaline phosphatase-conjugated goat anti-rabbit antibody (Boehringer Mannheim 605230) was diluted 1:5000 and used to detect the primary antibody.
Color Development Assays-Yeast harboring both GAL4(DB) and GAL4(AD) fusion proteins were monitored for 0-galactosidase activity using plate and liquid assay methods. Yeast transformants were transferred to nitrocellulose filters, permeabilized in liquid nitrogen, and placed on Whatman No. 3 filter paper that had been soaked in Z-buffer (60 mM Na2HP04, 40 mM NaH'PO,, 10 mM MgCl', 50 mM 0-mercaptoethanol) containing 1.0 mg/ml 5-bromo-4-chloro-3-indolyl P-D-galactoside at 30 "C. Positive colonies appeared in 5 min to 10 h. The liquid assays were performed with whole cell extract to et al., 1989). quantitate positive colonies isolated in the primary screen (Sambrook Rescue of GAL(AD) Fusion Plasmids-Total yeast DNA was isolated as described (Hoffman and Winston, 1987)  preparation.
was precipitated with 0.5 volume of 7.5 M ammonium acetate and 1 volume of isopropanol. The pellet was washed twice in 70% ethanol and resuspended in 20 p1 of sterile distilled water and used to electroporate Escherichia coli C-600 cells containing the leuB mutation, which can be complemented by the yeast LEU2 gene present on the GAL4(AD) plasmid. DNA Sequencing-DNA sequencing was performed using doublestranded plasmid DNA with the Sequenase kit (United States Biochemical Corp.). Sequences were analyzed using the Beckman Microgenie sequence analysis package. contains amino acids 508-654 of human E12 (Kamps et al., 1990) fused to the carboxyl-terminal end of the GAL4 DNAbinding domain, encoded by residues 1-147 of GAL4. This portion of E12 contains the bHLH motif, which lies between residues 550-603, as well as the region immediately aminoterminal to the basic region, which has been shown to inhibit E12 homodimerization . Since GAL(DB)-E12 was anticipated to lack transcriptional activity on its own and would therefore be undetectable by selection, we confirmed that it was expressed in yeast by Western blot analysis of extracts of yeast transformed with the GAL(DB)-E12 expression vector. Antibody directed against GAL1-147 detected a protein of 35 kDa in GAL(DB)-E12 expressing cells, but not in untransformed yeast (Fig. 1B). This agrees with the predicted size of the GAL4-El2 fusion protein.

Detection
GAL(AD)-MyoD contains all of MyoD fused to the carboxyl-terminal end of the GAL4 activation domain, encoded by residues 768-881 of GAL4. When yeast were transformed with either GAL(DB)-E12 or GAL(AD)-MyoD alone, there was no activation of lacZ above background (Table I). However, when the two hybrids were coexpressed, we observed high levels of lacZ expression. Typically, yeast colonies coexpressing the two hybrids stained intensely blue within 2-5 min following exposure to the chromogenic substrate for pgalactosidase. Activation of transcription by the combined hybrids represented a specific protein-protein interaction because no activation was observed when GAL(DB) fused to PKC-a was coexpressed with GAL(AD)-MyoD (Table   I).
These results demonstrate that specific protein-protein interactions between E12 and MyoD are readily detectable in yeast.
Since MyoD is a substrate for PKC (Li et al., 1992), the failure to detect activation with GAL(DB)-PKC plus GAL(AD)-MyoD suggests that this enzyme-substrate complex is relatively unstable i n vivo.

Screening of a T-cell cDNA Library with GAU-El2 "Bait"-
We next sought to determine whether GAL(DB)-E12 could be used as "bait" to screen for mammalian cDNAs in a yeast cDNA expression library. The yeast strain used for these B.

Activation of lac2 Reporter Gene by GAL4 Chimeras
Yeast transformants harboring the indicated expression plasmids were grown to mid-log phase in SD lacking Trp or Leu or both. Extracts were prepared and assayed for /3-galactosidase activity as described in the text. p-Galactosidase activity is expressed as percent of activity in extracts with GAL(DB)-E12 + GAL(AD)-MyoD. UAS. For this screen, we used a mouse T-cell cDNA library in which cDNA was fused to GAL(AD). Screening of 500,000 transformants by cotransformation with GAL(DB)-E12 for cDNAs that could rescue activation of the lacZ and HIS markers linked to the GAL4 UAS yielded seven colonies. When stained for lacZ expression, two colonies showed strong expression within 5 min and the additional five colonies showed expression after incubation overnight.
We chose to focus our attention on the two clones (designated E12-bpl and E12-bp2) that strongly activated lacZ expression. T o determine whether activation of lacZ was dependent on an interaction between the polypeptides encoded by these clones and the E12 portion of GAL(DB)-E12, we coexpressed these plasmids with the GAL4(DB)-PKC construct. Clone E12-bp2 showed activity in this transformation, whereas clone E12-bpl did not. These results suggest that clone E12-bp2 encodes a protein that interacts with GAL4(DB) or that it activates transcription of the reporter gene more directly, perhaps by binding to DNA sequences in the upstream region. Because this clone did not fulfill the criteria for specificity, we did not characterize it in greater detail.
El2-bpl Encodes Id-2-To further investigate the nature of the interaction between the protein encoded by clone E12bpl and GAL(DB)-E12 we determined the nucleotide sequence of the cDNA in the region of the junction with GAL(AD) (Fig. 2). Comparison of the partial nucleotide sequence of clone E12-bpl with the GenBank data base showed that it corresponded to mouse Id-2  and contained the sequence beginning in exon 1 and extending through the stop codon to nucleotide 887 in the 3"untranslated region fused in-frame with the GAL4(AD). This region of Id-2 contains all of the HLH motif. We conclude that E12 can dimerize with MyoD as well as Id in yeast and that the stability of these interactions is sufficient to support screening of cDNA libraries.
In the previously published sequence of mouse Id-2, there were two in-frame AUG codons within exon 1, the more 5' of which was predicted to represent the site for translation initiation . Clone E12-bpl began 43 nucleo-Asp Pro Arg vol Phe Thr Leu Trp Apq Leu Pro Pro Gly Leu Pro Pro Thr Ser Ser Wet LYS FIG. 2. Derived amino acid sequence of the junction between GAL(AD) and Id-2. The amino acid sequence of the cDNA insert in clone El2-bpl was determined by DNA sequencing using a primer upstream of the junction of GAL(AD) and the cDNA insert. Nucleotide sequence and deduced open reading frame for a portion of the sequence are shown, and the regions corresponding to GAL(AD) and Id-2 are indicated. The predicted initiating methionine for translation of Id-2 is indicated with an asterisk, and the position of a C that was reported in the Id-2 sequence described previously (32) is shown with an arrowhead below the sequence. tides upstream of the more 3' AUG (110 nucleotides downstream of the more 5' AUG). In sequencing the 5' region of the Id-2 cDNA insert near the junction with GAL4, we detected a discrepancy with the published cDNA sequence in which a single dC present in the previously published sequence was missing from our cDNA (see Fig. 2). We are certain that this dC cannot be present in our cDNA because it would shift the reading frame, resulting in translation termination of a protein lacking an HLH motif. We suggest therefore that the more 3' AUG of Id-2 may represent the true translational initiation site. Clone E12-bpl would therefore begin within the 5"untranslated region of the Id-2 mRNA.

DISCUSSION
There is considerable evidence for the involvement of cell type-specific bHLH proteins in the control of tissue-specific gene expression in vertebrates. (i) E12 and several related Eproteins that are widely expressed dimerize preferentially with cell type-specific HLH proteins such as MyoD and myogenin Murre et al., 1989aMurre et al., , 1989b. (ii) Id, a negative regulator of HLH protein DNA binding, is expressed in many undifferentiated cell types and is downregulated upon activation of differentiation, which has been proposed to release E-proteins to dimerize with cell typespecific partners (Benezra et al., 1990). (iii) Many tissuespecific genes contain E-boxes in their control regions that are essential for cell type-specific transcription (Murre et al., 1989a;Nelson et al., 1990;Olson, 1990;Sartorelli et al., 1992). (iv) The involvement of numerous cell type-specific bHLH proteins in cell fate specification in Drosophila suggests that these proteins constitute a conserved mechanism for regulation of cell type-specific transcription (Alonso and Cabrera, 1988;Caudy et al., 1988;Thisse et al., 1988).
Although HLH proteins contain several conserved amino acids within the two amphipathic a-helices of the HLH motif, the degree of nucleotide sequence homology between the genes encoding different subclasses of HLH proteins is in most cases not sufficient to allow cross-hybridization. Therefore, to facilitate identification of novel HLH proteins independent of sequence homology, we have used a modification of the two-hybrid assay, which is based only on the ability of proteins to dimerize with E12. There are several advantages to the two-hybrid system as a method for isolating novel cell typespecific partners for E12. It is a functional assay, based only on the ability of proteins to dimerize with E12, not on nucleotide sequence homology. The assay also does not depend on DNA binding activity of heterodimers formed between E12 and its putative partners. The assay is also highly sensitive and specific, and in our hands has not yielded significant background. The relative strength of protein-protein interactions can also be assessed by quantitative determination of lacZ activity in yeast extracts.
Using the E12 HLH motif linked to the GAL4 DNAbinding domain as "bait," we showed'that E12 and MyoD can dimerize in yeast and we identified Id-2 from a mouse T-cell cDNA library. The ability of E12 to dimerize with MyoD and Id in yeast indicates that HLH-mediated dimerization does not require post-translational modifications or enzymatic activities unique to higher eukaryotic cells. The use of GAL(DB)-E12 to screen yeast expression libraries for dimerization partners should therefore be a useful strategy for identification of novel HLH proteins from a variety of cell types.
In addition to serving as an interface for dimerization and DNA binding, there is evidence that the bHLH motif interacts with coregulators to induce muscle-specific transcription (Brennan et al., 1991;Davis et al., 1990). The specificity of interaction with these putative coregulators has been mapped to the amino acids in the center of the basic domains of myogenin and MyoD. The two-hybrid system may also be useful for cloning such proteins.