Interaction of the PA2G4 (EBP1) protein with ErbB-3 and regulation of this binding by heregulin

The processes by which ErbB-3, an inactive tyrosine kinase, exerts its biological effects are poorly understood. Using the yeast two-hybrid system, we have isolated an ErbB-3 binding protein (Ebp1) that interacts with the juxtamembrane domain of ErbB-3. This protein is identical to that predicted to be encoded for by the human PA2G4 gene. Ebp1 is the human homologue of a previously identified cell cycle-regulated mouse protein p38-2G4. Two transcripts of ebp1 mRNA (1.7 and 2.2 kb) were detected in several normal human organs. The interaction of Ebp1 with ErbB-3 was examined in vitro and in vivo. The first 15 amino acids of the juxtamembrane domain of ErbB-3 were essential for Ebp1 binding in vitro. Treatment of AU565 cells with the ErbB-3 ligand heregulin resulted in dissociation of Ebp1 from ErbB-3. Ebp1 translocated from the cytoplasm into the nucleus following heregulin stimulation. These findings suggest that Ebp1 may be a downstream member of an ErbB-3-regulated signal transduction pathway. © 2000 Cancer Research Campaign

Epidermal growth factor receptor (EGFR) or ErbB receptor family members, including EGFR, ErbB-2, ErbB-3, and ErbB-4, are glycosylated transmembrane proteins with tyrosine kinase domains (Ullrich et al, 1984;Bargmann et al, 1986;Kraus et al, 1989;Plowman et al, 1990). Ligands, such as EGF or heregulin (HRG) neu differentiation factor (NDF)/neuregulin (NRG), activate ErbB family members by inducing homo-or heterodimerization amongst receptor members, resulting in stimulation of tyrosine kinase activity and initiation of signalling pathways (Goldman et al, 1990;Kim et al, 1994). ErbB family members are frequently overexpressed in human carcinomas including those of the breast, stomach, lung, ovary and pancreas (Slamon et al, 1987;Gullick, 1991;Lemoin et al, 1992). Overexpression of EGFR and ErbB-2 is closely associated with poor prognosis in breast cancer (Slamon et al, 1989). ErbB-3, a 180 kDa glycosylated transmembrane protein, was identified based on its amino acid identity to other EGFR family members (Kraus et al, 1989). The tyrosine kinase activity of ErbB-3 is at least two orders of magnitude less than that of the EGFR due to amino acid substitutions in the conserved tyrosine kinase domain .
Apart from receptor phosphotyrosine-SH2 or PTB domain interactions, increasing data suggest that there are non-phosphorylated receptor-substrate interactions controlled by ligand stimulation. For example, eps8, a 92 kDa SH3 domain containing protein, has been identified as a substrate for the EGFR based on its association with the EGFR juxtamembrane domain after EGF stimulation (Fazioli et al, 1993). Interestingly, a mutant EGFR, lacking a kinase domain, still associates with eps8, suggesting the possibility of pre-existing substrate-receptor complexes in the absence of tyrosine kinase activity (Castagnino et al, 1995). Furthermore, the association of a nuclear localizing zinc-finger protein, ZPR1, with the cytoplasmic tyrosine kinase domain of the EGFR, is decreased by EGF stimulation of EGFR tyrosine kinase activation (Galcheva-Gargova et al, 1996).
The molecular mechanisms by which HRG/NDF-induces signalling through the kinase inactive ErbB-3 receptor are not clear. The yeast-two hybrid system (Fields and Song, 1989) was used to characterize proteins that interact with ErbB-3 in the absence of tyrosine phosphorylation. We report that the protein (Ebp1) predicted to be encoded by the recently cloned human PA2G4 gene binds ErbB-3. This protein is the human homologue of the mouse p38-2G4 protein, a cell cycle-regulated DNA binding protein (Radomski and Jost, 1995;Lamartine et al, 1997). The gene encoding a curved DNA binding protein from Saccharomyces pombe also shows homology to members of the PA2G4 gene family (Yamada et al, 1994). In this paper we show Interaction of the PA2G4 (EBP1) protein with ErbB-3 and regulation of this binding by heregulin that Ebp1 associates with the juxtamembrane domain of ErbB-3 in a tyrosine kinase-independent manner. Ebp1 dissociates from ErbB-3 after HRG activation and translocates to the nucleus.

Library screening
A human fetal brain derived cDNA library was cloned into the transcription activator domain vector pJG4-5 (TRP1, amp r ) (also provided by Dr Brent). Expression of fusion proteins is under the control of a GAL1 promoter. The yeast strain EGY48 (ura3,his3,trp1,) was cotransformed with pEG202/erbB3, the cDNA library, and a pSH18-34 (URA3, amp r ) reporter plasmid which provides a LacZ gene with LexA binding sites. 1.4 × 10 6 primary transformants were tested for Leu gene expression on Gal/SD-His, Trp, Ura, Leu plates. Among 86 Leu-positive clones, 12 yeast clones expressed the lacZ gene in a galactose-dependent manner as tested by a combination of filter lift β-galactosidase assays and β-galactosidase assays on Xgal plates.

Determination of complete ebp1 cDNA sequences
The ebp1 cDNA isolated in the two-hybrid system was radiolabelled by random priming (Pharmacia) and used to screen a human small-cell lung carcinoma (1688) cDNA library cloned into the pBK phagemid (a gift of Dr A Doyle, University of Maryland). After secondary cDNA library screening, we obtained a phage with a 1.3 kb insert, but without an in-frame termination codon. To obtain the sequence of the remaining 3′-end of the ebp1 cDNA, total RNAs were isolated from AU565, EP62.1 and MDA MB468 human mammary carcinoma cell lines. RNA (1 µg) was reversetranscribed using oligo-dT primers and Moloney-murine leukaemia virus reverse transcriptase at 42°C for 15 min, 95°C for 5 min, and 5°C for 5 min. A first round of PCR was performed using a primer (sense; 5′-GGGAATTCTGCGCCAAACAT-GAACTGCT-3′) specific for the 3′-end of the 1.3 kb ebp1 insert, and Adaptor dT (antisense; 5′-GGGAATTCGTCGACATC-GATTTTTTTTTTTTTTTTT-3′) at 95°C for 1 min, 55°C for 1 min and 72°C for 1.5 min, for 30 cycles. A second round of PCR was performed using a primer (sense; 5′-GGGAATTCATGAGAAG-GAGGGTGAATTT-3′) specific for the 3′-end of the 1.3 kb ebp1 insert, and Adaptor dT (antisense) under the same reaction conditions. The resulting PCR product (0.6 kb) was cloned into the pCRII vector using a TA cloning kit (Invitrogen) and sequenced with an Applied Biosystems model 373A Automatic Sequencer.

Northern blotting
Total RNA was isolated from the human mammary carcinoma cell lines AU565, EP62.1, MDA MB231 and MDA MB468, a human lung epithelial cell line E6, and a sarcoma cell line SK-LMS-1. Total RNA (20 µg) was analysed on 1.2% formaldehyde agarose gels. After transfer of RNA onto nitrocellulose membranes, the ebp1 cDNA isolated in the two-hybrid system was radiolabelled by random priming and hybridized to blots at 65°C overnight. A Human Multiple Tissue Northern blot (Clontech, Palo Alto, CA, USA) was also hybridized with the radiolabelled ebp1 cDNA probe at 65°C overnight.
GST-wt or mutant erbB3 fusion proteins were expressed in the BL21 bacterial strain, purified on glutathione Sepharose beads, and examined by Coomassie blue staining of sodium dodecyl sulphate polyacrylamide gels (SDS-PAGE). 35 S-labelled Ebp1 was obtained by in vitro transcription/translation using a TNT/T7coupled reticulocyte lysate system (Promega). Equal amounts of GST or GST-erbB3 fusion proteins were incubated with Ebp1 at 4°C for 2 h and analysed by SDS-PAGE and autoradiography.

LexA transcription factor-based yeast two-hybrid screening
The cytoplasmic domain of erbB3 was subcloned into the pEG202 LexA DNA binding domain vector and the expression of a LexA-erbB3 fusion protein (around 100 kDa) was determined in lysates of pEG202/erbB3 transformants by immunoprecipitation and Western blotting with an antibody against ErbB3. Tyrosine phosphorylation blots of ErbB3 immunoprecipitates from pEG202/ ErbB3 transformants revealed that ErbB3 was not tyrosine phosphorylated in yeast (Yoo and Hamburger, 1999). This screen was therefore biased towards the identification of proteins that bind to ErbB3 in the absence of tyrosine phosphorylation. A human fetal brain cDNA library was screened for proteins which interact with the cytoplasmic domain of ErbB3. We isolated 12 positive clones from 1.4 × 10 6 primary transformants.
The ebp1 clone isolated from the yeast two-hybrid screening contained a translation initiation codon, a 954 bp ORF, but not a termination codon. This cDNA also showed a 98% identity with the nucleotide sequence of the mouse gene Pa2g4 (EMBL X84789) The first methionine of this cDNA clone corresponded to the start codon of the mouse protein p38-2G4. To obtain a full A H e a r t B r a i n P l a c e n t a L u n g S k e l e t a l m u s c l e K i d n e y P a n c r e a s L u n g length ebp1 cDNA, we screened a cDNA library from a human small-cell lung carcinoma (1688) cell line and performed reverse transcription PCR (RT-PCR) of mRNA from the AU565 human mammary carcinoma cell line (see Methods and Materials). A 1648 nucleotide insert was obtained. This insert contained a termination signal and another ATG initiation site 5′ to the ATG of the original ORF. This larger cDNA shared a 98% nucleotide identity with the previously isolated human PA2G4 (Genbank U59435) clone. This cDNA encodes a predicted protein (Ebp1) of 394 amino acid residues which is 100% identical to the protein predicted to be encoded by the PA2G4 gene. This protein is also identical to p38-2G4, apart from the additional N terminal 54 amino acids. Like the mouse protein, Ebp1 is a basic protein with a putative nuclear localization signal in the C-terminus of the protein (aa 368-373). The protein also contains five putative phosphorylation sites (S/T-X-X-D/E) for casein kinase II, and six putative sites (S/T-X-R/K) for protein kinase C (PKC)-mediated phosphorylation. An amphipathic helical domain is also predicted (amino acids 258-313). An expression plasmid pBK/Ebp1 which encodes amino acids 1-373 of ebp1 was obtained from a human small-cell lung carci-noma (1688) cDNA library. To determine the size of Ebp1 expressed in mammalian cells, lysates of COS-7 cells or pBK/Ebp1-transfected COS-7 cells were immunoblotted with a rabbit antibody prepared against GST-EBP1. This antibody detected an endogenous 50 kDa Ebp1 protein and an exogenous protein that migrated at approximately 48 kDa (Figure 1). Preincubation of anti-GST-EBP1 antibody with excess amounts of purified GST-EBP1, but not with GST alone, diminished protein detection (Figure 1).

mRNA expression of ebp1
The distribution of ebp1 mRNA in various normal human organs was determined by Northern blot analysis (Figure 2A). Two ebp1 transcripts of approximately 1.7 and 2.2 kb were observed in all organs that were examined, including heart, brain, lung, pancreas, skeletal muscle, kidney, placenta and liver. The expression of ebp1 mRNA in various ErbB-expressing carcinoma cell lines was also examined ( Figure 2B). Two ebp1 transcripts of 1.7 and 2.2 kb were detected in the human breast carcinoma cell lines AU565 (ErbB-1, 2, 3+), MDA MB468 (ErbB-1, 3+) and MDA MB231  (Pierce et al, 1992), an ErbB-2 overexpressing, ErbB-3 expressing human lung epithelial cell line E6 (Noguchi et al, 1993) and a heregulin-producing sarcoma cell line SK-LMS-1.

In vitro and in vivo interactions of Ebp1 and ErbB-3
We investigated the interaction of the Ebp1 protein with ErbB-3 in vitro. Purified GST or a GST-ErbB-3 (aa 665-1339) fusion protein was incubated with the 35 S-labelled in vitro translation product of pBK/Ebp1 ( Figure 3A). GST-ErbB-3, but not GST alone, bound an approximately 48 kDa ebp1 translation product. The interaction between Ebp1 and ErbB-3 was further investigated in vivo in the AU565 mammary carcinoma cell line. AU565 cells express ErbB-1, 2 and 3 protein and ErbB-4 mRNA (Yoo and Hamburger, 1998).
Protein immunoblot analysis with an HRP-conjugated GST-EBP1 antibody demonstrated the presence of an approximately 50 kDa Ebp1 protein in ErbB-3, but not ErbB-2, immunoprecipitates ( Figure 3B). Conversely, ErbB-3 was present in Ebp1 immunoprecipitates ( Figure 3C). As the juxtamembrane domain of the EGFR is more closely related to that of ErbB-3 (54% identity), we examined the ability of EGFR to bind Ebp1. NIH3T3 cells, overexpressing human EGFR, were transiently transfected with a Flag tagged Ebp1 cDNA. Cell lysates were immunoprecipitated with antibody to the EGFR. Flag-tagged Ebp1 did not co-immunoprecipitate with the EGFR ( Figure 3D).
To identify the region of ErbB-3 that is required for binding to Ebp1, deletion mutants of GST-ErbB-3 were constructed ( Figure 4A) Figure  4B). GST-ErbB-3 665-690 , which contains the first 26 amino acids of the cytoplasmic domain of erbB3, was the smallest mutant to retain the ability to bind to EBP1. GST-ErbB3 756-1339 , which contained the full cytoplasmic domain of erbB3 except the first 92 amino acids (aa 665-756), failed to bind EBP1, indicating that the 26 amino acids of the juxtamembrane domain of erbB3 (aa 665-690) were required for binding to EBP1. ErbB-3 binding to Ebp1 was competed by a purified peptide specific for ErbB-3 (aa 665-690) ( Figure 4C).
To further characterize the binding domain for Ebp1, a series of alanine scanning mutants of GST-ErbB-3 665-690 (A1-A5) was created, and tested for binding to Ebp1 ( Figure 4D). Mutants A1, A2 and A3 lost, but mutant A4 retained, the ability to bind Ebp1, indicating that some of the first 15 amino acids of the juxtamembrane domain of ErbB-3 (aa 665-680) were essential for binding to Ebp1. In addition, the binding of mutant A5 was somewhat reduced compared to that of the controls.

Ebp1 translocates into the nucleus after HRG incubation
A putative nuclear localization signal is present at the C-terminal end of the Ebp1 protein. We therefore examined the subcellular localization of Ebp1 in serum-starved and HRG-(10 ng ml -1 ) stimulated AU565 cells by immunoperoxidase staining using the anti-p38-2G4 monoclonal antibody. Ebp1 protein was primarily localized in the cytoplasm (> 90% of cells) of serum starved cells ( Figure 5A). Treatment of AU565 cells with HRG altered the subcellular distribution of Ebp1. Translocation of Ebp1 protein into the nucleus was observed after 5 min (60.5% in the nucleus), and continued up to 90 min (93.1% in the nucleus) after HRG treatment ( Figure 5B). Similar results were obtained using the rabbit antibody to GST-EBP1 (data not shown).
To substantiate our morphological findings, we next examined the effect of HRG on the binding of ErbB-3 to Ebp1. AU565 cells were treated with HRG (10 ng ml -1 ) or EGF (50 ng ml -1 ) for 15 min and the interaction of ErbB-3 and Ebp1 was analysed by immunoprecipitation ( Figure 5C). Treatment of AU565 cells with HRG caused a reduction (as determined by densitometry) of binding between Ebp1 and ErbB-3, providing biochemical evidence of HRG induced dissociation of Ebp1 from ErbB-3. Treatment of cells with EGF did not affect Ebp1 binding to ErbB-3. The ratio of Ebp1 to ErbB-3 for control, heregulin and EGFtreated cells was 1.6, 0.3 and 1.3 respectively.

DISCUSSION
ErbB-3, despite its impaired tyrosine kinase activity, is a functional receptor for HRG. ErbB-3-HRG interaction leads to cellular proliferation or differentiation by stimulation of a cascade of signalling events (Bacus et al, 1993;Fedi et al, 1994;Sliwkowski et al, 1994). To better understand HRG-mediated ErbB-3 signalling pathways, we performed a yeast two-hybrid screening with the cytoplasmic domain of ErbB-3 to isolate proteins that interact with ErbB-3. We report here that Ebp1, the protein predicted to be encoded by the PA2G4 gene, is an ErbB-3-interacting protein, which translocates into the nucleus upon HRG stimulation.
The PA2G4 gene is a member of the recently described PA2G4 family (Lamartine et al, 1997). The prototype of this family, the mouse p38-2G4 gene, was isolated by generating monoclonal antibodies against DNA binding proteins. The sequence of the human PA2G4 cDNA predicts a protein of 394 amino acids and approximately 45 kDa molecular weight, which has a longer N-terminal region than mouse p38-2G4 protein. Both the in vitro translation product and the in vivo expressed protein of our slightly truncated ErbB-3-associated Ebp1 was analysed by ErbB-3 immunoprecipitation (IP) and sequential immunoblotting (WB) with antibodies against p38-2G4 (top) and ErbB-3 (bottom). Representative of three trials pBK cDNA clone migrate at a somewhat higher molecular weight than predicted. Using antibodies generated against both a GST-EBP1 fusion protein and the murine p38-2G4 protein (data not shown), we detected an endogenous full-length protein of approximately 50 kDa from AU565 and COS-7 cell lysates. Ebp1 contains five putative phosphorylation sites (S/T-X-X-D/E) for casein kinase II, and six putative sites (S/T-X-R/K) for PKCmediated phosphorylation. Thus, the changes in electrophoretic mobility may be due to post-translational phosphorylation. ErbB-3 bound Ebp1 via its juxtamembrane domain. In the experiments presented here, ErbB-2, which has a 38% amino acid identity in the 26 amino acid binding region, did not bind to Ebp1. The EGFR, singly transfected into NIH 3T3 cells, also did not bind to Ebp1. ErbB3 and EGFR have a 54% amino acid identity in the 26 amino acid binding region with 6/7 positively charged residues remaining the same. Thus, it is unlikely ErbB-3-Ebp1 binding is non-specific.
Ebp1 dissociated from ErbB-3 after heregulin treatment. Ebp1 did not dissociate from ErbB-3 after EGF treatment. This is somewhat surprising as EGF treatment of MDA 468 cells results in tyrosine phosphorylation of ErbB-3 (Kim et al, 1994). However, EGF or HRG treatment result in phosphorylation of different amino acid residues on ErbB receptors (Olayioye et al, 1998). In addition, HRG, but not EGF, treatment may have modified Ebp1 itself leading to dissociation from ErbB-3. Modification of Ebp1, rather than ErbB-3, by heregulin may lead to dissociation.
Although we have demonstrated nuclear localization of Ebp1 after HRG stimulation, we have not yet directly demonstrated the function of the Ebp1 protein in the nucleus. Amino acid homology comparison of the Ebp1 protein shows a 42% homology with a yeast curved DNA binding protein (Yamada et al, 1994). The mouse homologue p38-2G4 was originally cloned from Ehrlich ascites tumour cells using a set of monoclonal antibodies which specifically recognizes DNA binding proteins. Biochemical analysis suggested p38-2G4 bound directly or, via a protein complex, to DNA (Radomski and Jost, 1995). These results suggest that Ebp1 may be involved in the regulation of cell growth or differentiation in mammary epithelial cells through putative DNA binding properties after translocation into the nucleus.
Ligand-control of gene activation by directly inducing translocation of membrane receptor bound regulatory proteins to the nucleus is an emerging area in signal transduction. For example, a zinc finger protein, ZPR1, which associates with unstimulated EGFR, and MADR2, a substrate for the transforming growth factor-β (TGF-β) receptor, both dissociate from their receptors and translocate into the nucleus after ligand stimulation (Macias-Silva et al, 1995;Galcheva-Gargova et al, 1996). In Drosophila, Suppressor of Hairless (Su(H)) also directly associates with intracellular ankyrin repeats of the receptor Notch and mediates signalling by translocation into the nucleus after receptor stimulation (Jarriault et al, 1995). Although Ebp1 does not resemble these proteins structurally, these proteins may together constitute a new functional class of signalling molecules.
In conclusion, we report here that the Ebp1 protein, which binds to the juxtamembrane domain of ErbB-3, translocates into the nucleus after HRG stimulation. The biological relevance of ErbB-3 interaction with Ebp1 and the mechanisms of Ebp1 dissociation from ErbB-3 remain to be elucidated.