A natural ErbB4 isoform that does not activate phosphoinositide 3-kinase mediates proliferation but not survival or chemotaxis.

ErbB4 is a member of the epidermal growth factor receptor (ErbB) family that mediates cellular responses activated by neuregulins (NRG) and other epidermal growth factor-like growth factors. Two naturally occurring ErbB4 isoforms, ErbB4 CYT-1 and ErbB4 CYT-2, have previously been identified. Unlike ErbB4 CYT-1, ErbB4 CYT-2 lacks a phosphoinositide 3-kinase (PI3-K)-binding site and is incapable of activating PI3-K. We have now examined the consequences of the inability of this isoform to activate PI3-K on cell proliferation, survival, and chemotaxis in response to NRG-1beta: (i) NRG-1beta stimulated proliferation of cells expressing either ErbB4 CYT-1 or ErbB4 CYT-2. Consistent with the mitogenic responsiveness, analysis of downstream signaling showed that Shc and MAPK were phosphorylated after stimulating either isoform with NRG-1beta. (ii) NRG-1beta protected cells expressing ErbB4 CYT-1 but not cells expressing ErbB4 CYT-2 from starvation-induced apoptosis as measured by effects on cell number and 4', 6-diamidino-2-phenylindole staining. Furthermore, in cells expressing ErbB4 CYT-2, Akt, a protein kinase that mediates cell survival, was not phosphorylated. (iii) NRG-1beta stimulated chemotaxis and membrane ruffling in cells expressing ErbB4 CYT-1 but not in cells expressing ErbB4 CYT-2. In summary, ErbB4 CYT-2 can mediate proliferation but not chemotaxis or survival. These results suggest a novel mechanism by which cellular responses such as chemotaxis and survival may be regulated by the expression of alternative receptor-tyrosine kinase isoforms that differ in their coupling to PI3-K signaling.

of tyrosine residues within their cytoplasmic tails. This creates specific binding sites for intracellular signal transduction molecules such as Shc, Grb2, and PI3-K that interact with phosphotyrosines via their Src homology 2 or phosphotyrosine-binding domains. These interactions result in activation of downstream signal transduction pathways by subsequent phosphorylation events, conformational changes, and/or translocation of the downstream signaling molecules (2,3). Thus, the nature of the cellular responses elicited by a growth factor is a result of the specific interactions of intracellular signaling molecules with the RTKs.
The ErbB subfamily of RTKs consists of ErbB1 (also known as epidermal growth factor receptor or HER1), ErbB2 (c-Neu or HER2), ErbB3 (HER3), and ErbB4 (HER4) (4 -8). ErbBs are receptors for the ligands of the epidermal growth factor family (9,10). ErbB2, however, has no known ligand and may solely function as a heterodimeric partner with the other ErbBs (11). The ErbB receptors are important regulators of cell behavior during development, as demonstrated by the severe developmental defects that occur in null mice with targeted erbB genes (reviewed in Ref. 12). Furthermore, overexpression, amplification, and rearrangement of erbB genes have been implicated in the initiation and progression of several human malignancies (13)(14)(15).
ErbB4 is a 180-kDa glycoprotein (8) that mediates the effects of the neuregulin (NRG) family of growth factors currently consisting of four cloned NRG genes designated NRG-1, NRG-2, NRG-3, and NRG-4 (12, 16 -18). Moreover, recent reports indicate that ErbB4 may also function as a receptor for betacellulin (19), heparin-binding epidermal growth factor-like growth factor (20), epiregulin (21), and epidermal growth factor (22). Activation of ErbB4 with its ligands leads to various cellular responses, such as proliferation, survival, chemotaxis, and differentiation (12). ErbB4 is expressed mainly in developing neural tissues and myocardium in vivo (23,24). Several neural cell types and cardiac myocytes also express ErbB4 in vitro (25)(26)(27). Consistent with a critical role in regulating cellular functions in brain and heart, null mice with a targeted erbB4 gene die between embryonic days 10 and 11 with defects in hindbrain and heart trabeculae (23). ErbB4 is also expressed by some epithelia and may participate in epithelial-mesenchymal interactions, for example during the differentiation of the mammary gland at pregnancy (28). Furthermore, there is evidence suggesting that ErbB4 has oncogenic activities in human tumors. For example, ErbB4 is expressed by several tumor cell lines (8), and overexpression of ErbB4 together with ErbB2 is associated with poor prognosis in childhood medulloblastoma (29). The molecular signaling mechanisms by which ErbB4 stimulates these activities are largely unknown.
Recently, it was demonstrated that two pairs of naturally occurring ErbB4 isoforms exist in human and mouse tissues (30 -32). These isoform pairs are differentially expressed in tissues such as heart and brain, suggesting that they may have specific functions (30,31). One pair of ErbB4 isoforms differs within the extracellular juxtamembrane domain (isoforms JM-a and JM-b) and in susceptibility to being cleaved in response to phorbol ester treatment (30). The other pair of ErbB4 isoforms differs in the cytoplasmic tail domain by the presence (CYT-1) or absence (CYT-2) of a 16-amino acid sequence including a Tyr-Xaa-Xaa-Met consensus-binding site for PI3-K (31,32). When overexpressed in an NIH 3T3 cell line, the isoforms are functionally different in that ErbB4 CYT-1 but not ErbB4 CYT-2 is capable of binding and activating PI3-K (31). The schematic structure of the ErbB4 isoforms, the sources of their identification, and the alternative nomenclature introduced by Sawyer and co-workers (32) are shown in Fig. 1.
PI3-Ks are lipid kinases that phosphorylate the 3Ј position of the inositol ring of phosphoinositides at the plasma membranes (33). As a consequence, membrane-binding sites for downstream signaling molecules such as the serine-threonine kinase Akt are generated (34). The PI3-Ks have been divided into separate classes based on their structure and substrate specificity (35). RTKs activate class I A PI3-Ks that are heterodimeric proteins consisting of a regulatory subunit, either p85, p55, or p50, and a catalytic subunit, p110. The Src homology 2 domains of the regulatory subunits mediate the association of the catalytic subunits with RTKs by binding to phosphotyrosines flanked by specific amino acids (36). Studies using several different approaches have implicated that RTK-activated PI3-Ks participate in the signal transduction pathways leading to numerous different cellular responses (reviewed in Ref. 37). For example, inhibiting PI3-K activation by chemical inhibitors or by mutating the receptor-binding site for PI3-K interferes with growth factor-stimulated survival (38), membrane ruffling and chemotaxis (39,40), and activation of signaling molecules such as Akt (41,42), BAD (43,44), and Rac (45). Interestingly, B-lymphocytes isolated from null mice with targeted p85␣ gene have either an attenuated survival or mitogenic response, depending on the survival/growth factor tested (46).
In this report, we have analyzed the biological consequences of the inability of a naturally occurring RTK isoform, ErbB4 CYT-2, to activate PI3-K. When ErbB4 CYT-2 was expressed in an NIH 3T3 cell line, cell proliferation in response to NRG-1␤ was comparable with cells expressing ErbB4 CYT-1, consistent with the unimpaired ability of the CYT-2 cells to activate the Shc/MAPK pathway. However, the ability of NRG-1␤ to induce survival and chemotaxis in CYT-2 cells was markedly reduced. The lack of survival was consistent with the inability of CYT-2 cells to activate the PI3-K/Akt pathway. The results demonstrate that the signaling pathways leading to proliferation can be dissociated from those leading to survival and chemotaxis and support a central role for PI3-K in both survival and chemotactic signaling.

EXPERIMENTAL PROCEDURES
Cell Lines-Clone 7 NIH 3T3 cells expressing ErbB4 CYT-1 or ErbB4 CYT-2 were prepared by cotransfection of a plasmid encoding a neomycin resistance gene together with a plasmid encoding an ErbB4 isoform as described previously (31). Several independently produced clones expressing either CYT isoform were prepared. The wild type (parental) clone 7 NIH 3T3 cells, NIH 3T3 cells expressing ErbB1, as well as one NIH 3T3 clone expressing ErbB4 CYT-1 designated ErbB4 CYT-1 number 1 (previously CYT-1 amg) were kindly provided by Dr. K. Zhang (Amgen Inc., Thousand Oaks, CA) (47). The wild type parental cells and a clone generated by transfecting the NIH 3T3 cells with the neomycin resistance gene encoding plasmid alone (designated Neo) served as negative controls. The cells were maintained in DMEM, 10% fetal bovine serum (FBS), 1% glutamine/penicillin/streptomycin supplement (Irvine Scientific), and 4.5 g/liter glucose. The culture media of transfected clones were supplemented with 500 g/ml G418 (Geneticin; Calbiochem).
Western Blot Analysis of ErbB4 Levels-Cells were grown to confluence on 100-mm dishes and lysed into a buffer containing 1% Triton X-100, 10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin. The lysates were centrifuged, and 75-g samples of total protein from the supernatants were separated on 6% SDS-PAGE gels. The separated proteins were transferred onto nitrocellulose filters (Schleicher & Schuell). Nonspecific binding to the filters was blocked by a 1-h incubation in a buffer containing TBS-T (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween-20), 5% nonfat dry milk, and 10 mg/ml bovine serum albumin (Sigma). The ErbB4 protein was subsequently detected by successive 1-h incubations with a polyclonal anti-ErbB4 antibody (sc-283; Santa Cruz; 1:100 dilution) and a peroxidase-conjugated goat anti-rabbit IgG antibody (Jackson Immunoresearch Laboratories; 1:10,000 dilution),  (8) is designated ErbB4 JM-a CYT-1, and the isoforms identified thereafter were designated JM-b and CYT-2. The alternative nomenclature for ErbB4 cytoplasmic isoforms suggested by Sawyer and co-workers (32) is also indicated, as are the sources for sequencing of the various ErbB4 isoforms. For isoforms JM-a and CYT-1, reference is only given to the original cloning of human ErbB4 (8). JM-a and CYT-1 sequences have also been reported for mouse (30,31,82) and rat (76,83). For isoforms JM-b and CYT-2, references are given to all the articles reporting sequences from various species. CYT-2 sequence was obtained from mouse (84) before the identification of alternative cytoplasmic isoforms (31,32). cDNA and partial cDNA are used to indicate the origin of the sequence from a single cDNA library clone encoding the whole open reading frame or from partial cDNA library clone(s), respectively. RT-PCR indicates that the sequence has been obtained from a RT-PCR product generated with primers flanking either the juxtamembrane or the cytoplasmic domain.
Radioiodination and Cross-linking-Recombinant human NRG-1␤ (R & D; residues 176 -246 of human heregulin-␤1) was iodinated using IODO-BEADS (Pierce) as described previously (20). Na 125 I (NEN Life Science Products) was used at 100 Ci/g of protein, and a specific activity of 120,000 CPM/ng was achieved. For cross-linking experiments, cells were plated in 60-mm dishes and grown to confluence. The cells were incubated in the presence of 25 ng/ml 125 I-NRG-1␤, and the bound 125 I-NRG-1␤ was immobilized to cell surface receptors by crosslinking with 200 M disuccinimidyl suberate (Pierce) as described (20). Cross-linked complexes were separated on 6% SDS-PAGE gels and visualized by autoradiography.
Quantitation of Cell Numbers after NRG-1␤ Treatment-To determine the effect of NRG-1␤ on the proliferation of cells in the presence of 1% serum, cells were plated at a density of 50,000 cells/well in 24-well plates in DMEM containing 10% FBS. The following day the media were replaced by DMEM, 1% FBS with or without 20 ng/ml NRG-1␤. Cell number was subsequently calculated at 2, 5, and 7 days after the initiation of the culture. This was carried out by trypsinizing the cells and measuring their numbers with an automatic cell counter (Coulter Counter; Coulter Electronics).
To assess the effect of NRG-1␤ on cell number after starvation in the absence of serum (survival assay), cells were plated at a density of 50,000 cells/well in 24-well plates in DMEM containing 10% FBS. The following day the media were replaced by DMEM containing no serum and 0, 1, 10, or 100 ng/ml NRG-1␤. After a further 3-day culture in the presence or absence of growth factor, the cell numbers were quantitated using a Coulter Counter as above.
Immunoprecipitation and Western Blot Analysis of Protein Phosphorylation-To analyze phosphorylation of signal transduction proteins, cells were grown to confluence and starved overnight in serum-free DMEM. The cells were stimulated with or without 50 -100 ng/ml NRG-1␤ for 10 min on ice followed by 5 min at 37°C, with the exception of the time course analysis of Erk activation, for which cells were stimulated at 37°C as indicated. The growth factor treatments were initiated on ice and then continued at 37°C to better synchronize conditions between cells on separate dishes. The results were similar when cell were stimulated for 5 or 10 min at 37°C without a preceding incubation on ice (data not shown). After stimulations, the cells were lysed in lysis buffer containing 1% Triton X-100, 10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM sodium orthovanadate, 10 mM sodium fluoride, and 10 mM sodium pyrophosphate.
To detect tyrosine phosphorylation of Shc, aliquots of the lysates corresponding to 500 g of total protein were immunoprecipitated with a polyclonal anti-Shc antibody (H-108; Santa Cruz; 2 l/500 l). The immunoprecipitated samples were separated in 10% SDS-PAGE gels and transferred to nitrocellulose filters. The tyrosine phosphorylation of the precipitated Shc proteins was subsequently analyzed by Western blotting with an anti-phosphotyrosine antibody (4G10; Upstate Biotechnology Inc.; 2 g/ml), a peroxidase-conjugated goat anti-mouse IgG secondary antibody (Cappel; 1:50,000 dilution), and ECL.
To detect association of p85 with tyrosine phosphorylated proteins, aliquots of the lysates corresponding to 500 g of total protein were immunoprecipitated with the 4G10 monoclonal anti-phosphotyrosine antibody (10 g/500 l). The immunoprecipitated samples were separated in 8% SDS-PAGE gels and transferred to nitrocellulose filters. The presence of p85 among the precipitated tyrosine phosphorylated proteins was subsequently analyzed by Western blotting with a monoclonal anti-p85 antibody (Upstate Biotechnology Inc.; 1:100 dilution), the peroxidase-conjugated anti-mouse IgG secondary antibody (Cappel; 1:20,000 dilution), and ECL.
The phosphorylation of Erk1 and Erk2 at Thr 202 and at Tyr 204 and the phosphorylation of Akt at Ser 473 were analyzed by direct Western blotting with phosphospecific polyclonal antibodies, anti-phospho-MAPK, and anti-phospho-Akt, respectively (both from New England Biolabs; 1:1000 dilution). For these experiments, aliquots corresponding to 50 -75 g of total protein were separated in 10% SDS-PAGE gels. The phospho-specific antibodies bound to antigens on the nitrocellulose filters were detected by an incubation with peroxidase-conjugated goat anti-rabbit IgG antibody (Jackson Immunoresearch Laboratories; 1:10,000 dilution) and ECL. Akt protein levels were determined by subsequent incubations of the filter with polyclonal anti-Akt antibody (Upstate Biotechnology; 1:500 dilution) and peroxidase-conjugated rabbit anti-goat IgG antibody (Chemicon; 1:10 000 dilution) followed by ECL.
DAPI Staining-To determine the effect of NRG-1␤ on the survival of serum-starved cells, nuclear morphology was analyzed by staining with DAPI (Sigma) as a method to identify apoptotic cells. Two different experimental designs were used to analyze the survival effect of NRG-1␤ in serum-starved low and high density cultures. For low cell density experiments, cells were plated at 100,000 cells/35-mm dish in DMEM containing 10% FBS. The following day the media were replaced by DMEM containing 0.5% FBS. After a 48-h incubation in the low serum medium, the media were replaced by serum-free DMEM with or without 40 ng/ml NRG-1␤ and cultured for another 48 h. For high cell density experiments, cells were plated at 400,000 cells/35-mm dish in DMEM containing 10% FBS. The following day, the cells were washed once with DMEM and then incubated for 72 h in serum-free DMEM with or without 40 ng/ml NRG-1␤. To deprive the cells from endogenously produced soluble survival factors, the cells were washed with DMEM, and the media were replaced by fresh DMEM with or without NRG-1␤ once every 24 h during the 72-h starvation/growth factor treatment period. After the incubations in the presence or absence of NRG-1␤, cells from both high and low density experiments were fixed with 3.7% formaldehyde in PBS for 30 min, washed once with PBS, and permeabilized with 1% Triton X-100 in PBS for 4 min at room temperature. After another wash with PBS, the cells were incubated with 0.5 g/ml DAPI in PBS for 60 min at room temperature. After rinsing with PBS, coverslips were mounted on top of the stained cells using Aquamount (BDH), and the cells were viewed in microscope under UV light and photographed. Chemotaxis Assay-The effects of NRG-1␤ and FBS on cell migration were measured by modified Boyden chamber assays as described previously (48). Briefly, cells starved overnight in serum-free DMEM were trypsinized, washed with PBS, and suspended in DMEM to a final concentration of 500,000 cells/ml. 50 l of the cell suspension was applied to wells in the upper chambers. Chemoattractants NRG-1␤ and FBS were diluted to various concentration in DMEM and applied to the wells in the lower chambers. A micropore filter (pore size, 8 m; Poretics Products Inc.) coated with 6.7 g/ml human fibronectin (Becton-Dickinson) in PBS was placed between the cells and the chemoattractants. After a 4-h incubation of the chambers at 37°C, the cells migrated through the filter were fixed with 3.7% formaldehyde in PBS, stained with Harris hematoxylin (Sigma), and mounted with Permount (Fisher). All experiments were performed in quadruplicate for each concentration of the chemoattractant.
Membrane Ruffling-Membrane ruffles were visualized by staining actin filaments with TRITC-labeled phalloidin (Sigma). Cells were cultured on glass coverslips in 24-well plates in DMEM, 10% FBS. Subconfluent cells were starved for 60 min in serum-free DMEM and subsequently stimulated for 15 min with DMEM containing 0 or 20 ng/ml NRG-1␤. After stimulation, the cells were fixed with 3.7% formaldehyde in PBS for 20 min at room temperature, washed once with PBS, and permeabilized with 100% acetone for 5 min at Ϫ20°C. The coverslips were washed three times with PBS, and the actin filaments were stained with 100 ng/ml of TRITC-phalloidin (Sigma) in PBS for 45 min at room temperature in a humidified atmosphere. After rinsing with PBS and distilled water, the coverslips were mounted with Aquamount, viewed in a microscope under UV light, and photographed.
To demonstrate that the ErbB4 protein expressed by the transfectants was functional, the cells were also analyzed by cross-linking with an iodinated ErbB4-ligand 125 I-NRG-1␤ (Fig. 2B). 125 I-NRG-1␤ did not form any detectable complexes with wild type or Neo control cell surface proteins (Fig. 2B,  lanes 1 and 2). On the other hand, 190-kDa complexes, consistent with the cross-linking of 7-kDa 125 I-NRG-1␤ to the 180-kDa ErbB4 receptor, were formed when any of the ErbB4 CYT-1 clones (15a, 42, and 1) or ErbB4 CYT-2 clones (2, 11, and 15b) were analyzed (Fig. 2B, lanes 3-8, respectively). The identity of the complexes smaller than 190 kDa is not currently known but may represent degradation products of the 190-kDa complex. The relative intensities of the cross-linked complexes (Fig. 2B) were in accordance with the relative expression levels of the ErbB4 protein analyzed by Western blotting (Fig. 2A). These data demonstrate that all the transfectants express functional ErbB4 proteins at their surfaces.
Both ErbB4 Cytoplasmic Isoforms Mediate Proliferation-The relative ability of NIH 3T3 cells expressing ErbB4 CYT-1 or ErbB4 CYT-2 to proliferate in response to NRG-1␤ was analyzed (Fig. 3). CYT-1 clone 42 and CYT-2 clone 11, which express comparable amounts of ErbB4, were chosen for the analysis as was CYT-2 clone 15b, which expresses relatively higher amounts of ErbB4. Cells were plated in 24-well plates in medium containing 10% FBS. The following day the media were replaced with media containing 1% FBS to ensure survival and either 0 or 20 ng/ml NRG-1␤. The cells were counted 2, 5, and 7 days after initial plating. ErbB4 CYT-1 clone 42 (Fig. 3A), ErbB4 CYT-2 clone 11 (Fig. 3B), and ErbB4 CYT-2 15b (Fig. 3C) responded to NRG-1␤ by an increase in cell number of about 1.6-fold for CYT-1 cells and about 2.5-fold for CYT-2 cells in 7 days when compared with cells not treated with NRG-1␤. The absence of dead cells was confirmed by trypan blue exclusion analysis at the 7-day time point. These results were confirmed when DNA synthesis was measured in cells made quiescent by a 7-day culture in DMEM ϩ 10% FBS without medium changes (not shown).
Both ErbB4 Cytoplasmic Isoforms Stimulate Phosphorylation of Shc and MAPK-Growth factor stimulation via ErbB4 activates downstream signaling pathways leading to proliferation. For example, activated ErbB4 binds and phosphorylates Src homology 2 domain containing signal transduction proteins such as Shc (49). The phosphorylation of Shc in response to NRG-1␤ treatment was examined by immunoprecipitation with an anti-Shc antibody followed by Western blotting with an anti-phosphotyrosine antibody (Fig. 4A). NRG-1␤ stimulation of cells for 10 min on ice followed by 5 min at 37°C resulted in the phosphorylation of the p52 and p46 Shc isoforms in CYT-1 clones 42 and 1 (Fig. 4A, lanes 1-4) and in CYT-2 clones 11 and 15b (Fig. 4A, lanes 5-8). Western blotting with anti-Shc antibody demonstrated that NIH 3T3 cell transfectants expressed all three Shc isoforms, p66, p52, and p46 (Fig. 4A, lane 9).

ErbB4 Isoform Deficient in PI3-K Activity
CYT-1 clones 42 and 1 (Fig. 4B, lanes 3-6) and in CYT-2 clones 11 and 15b (Fig. 4B, lanes 7-10) but not in the Neo control cells (Fig. 4B, lanes 1 and 2). Interestingly, the relative amount of phosphorylation of the Erk proteins, normalized to levels of ErbB4 expression (Fig. 2), was consistently slightly higher in the CYT-2 clones than in the CYT-1 clones (Fig. 4B, lanes 7-10  compared with lanes 3-6). The kinetics of MAPK activation were analyzed after 0 -40 min of stimulation with NRG-1␤ (Fig. 4C). Both Erk1 and Erk2 were activated within 1 min in cells expressing CYT-1, clone 1 (Fig. 4C, upper panel), or CYT-2, clone 15b (Fig. 4C, lower panel). Maximal MAPK activation occurred at 5 min. However, the signal seemed to be of longer duration in cells expressing CYT-2, lasting for 40 min compared with 10 min for cells expressing CYT-1. Taken together, these results demonstrate that ErbB4 CYT-2 is coupled to the Shc/MAPK pathway in the absence of PI3-K activity and that expression of this ErbB4 isoform is compatible with mediating a proliferative response.
ErbB4 CYT-1 but Not ErbB4 CYT-2 Mediates Cell Survival-PI3-K plays a critical role in mediating growth factor-stimulated cell survival (38). Cells expressing the two ErbB4 cyto-plasmic isoforms were analyzed for survival in response to NRG-1␤ treatment under serum-free conditions that induce apoptosis (Fig. 5). Cells were plated on 24-well plates at a density of 50,000 cells/well in medium containing serum, and the next day the media were replaced by serum-free medium containing 0, 1, 10, or 100 ng/ml NRG-1␤. After a further 72 h of incubation, the cells were counted. In the absence of the NRG-1␤, the cell number in all the clones tested was less than the 50,000 cells initially plated, thereby demonstrating that some cell death was occurring. Interestingly, the CYT-1 clones (15a, 42, and 1) seemed less sensitive to serum starvation in the absence of NRG-1␤ than the CYT-2 clones (11 and 15b) compared with the Neo control cells, possibly because of the constitutive activity of the overexpressed ErbB4 CYT-1 receptors (31). Addition of NRG-1␤ rescued cells expressing ErbB4 CYT-1 in a dose-dependent manner but not cells expressing ErbB4 CYT-2. CYT-1 clone 1, which expressed the highest ErbB4 protein levels (Fig. 2), showed the strongest survival response to NRG-1␤ (Fig. 5A), and CYT-1 clone 15a, which expressed the lowest ErbB4 levels (Fig. 2), showed the weakest response (Fig. 5A). In contrast to the CYT-1 clones, the number of cells in the Neo and CYT-2 clones remained virtually unchanged in response to NRG-1␤ (Fig. 5B). These results suggest that an activated ErbB4 CYT-1 receptor, but not an activated ErbB4 CYT-2 receptor, can mediate a survival signal that protects cells from starvation-induced apoptosis.
To analyze apoptosis in cells expressing the two ErbB4 CYT isoforms, the nuclear morphology of cells serum-starved in the presence or absence of NRG-1␤ was visualized by DAPI staining (Fig. 6, A and C) and an apoptotic index, defined as the percentage of nuclei with a condensed morphology, was measured (Fig. 6, B and D). Because cell density and starvation conditions may affect survival, the experiments were performed in both low and high density cultures and after different starvation periods. When cells were plated at relatively low cell density (100,000 cells/35-mm dish), growth arrested for 48 h in 0.5% FBS, and then cultured for 48 h in serum-free medium, the apoptotic indexes for two clones expressing ErbB4 CYT-1 (15a and 42) were 47 and 31%, respectively (Fig. 6, A and B). In the presence of 40 ng/ml NRG-1␤, the apoptotic indexes of the two CYT-1 clones were reduced to 12 and 11%, respectively. On the  1, 3, 5, and 7) or with (lanes 2, 4, 6, and 8) 50 ng/ml NRG-1␤ for 10 min on ice followed by 5 min at 37°C. Cell lysates were immunoprecipitated with an anti-Shc antibody and analyzed for tyrosine phosphorylation by Western blotting with an anti-phosphotyrosine antibody and ECL. The expression of p66, p52, and p46 Shc proteins was demonstrated by a direct Western blot analysis of lysate from CYT-1 1 clone with the anti-Shc antibody (lane 9). B, Neo control cells (lanes 1 and 2), cells expressing ErbB4 CYT-1 (lanes 3-6; clones 42 and 1), and cells expressing ErbB4 CYT-2 (lanes 7-10; clones 11 and 15b) were starved for 24 h in the absence of serum and stimulated without (lanes 1, 3, 5, 7, and 9) or with (lanes 2, 4, 6, 8, and 10) 50 ng/ml NRG-1␤ for 10 min on ice followed by 5 min at 37°C. Cell lysates were analyzed for tyrosine phosphorylation of p44 Erk-1 and p42 Erk-2 by Western blotting with a phospho-specific anti-MAPK antibody and ECL. C, cells expressing ErbB4 CYT-1 (upper panel; clone 1) and ErbB4 CYT-2 (lower panel; clone 15b) were starved for 24 h in the absence of serum and stimulated with 50 ng/ml NRG-1␤ for 0, 1, 5, 10, 20, or 40 min (lanes 1-6, respectively) at 37°C. Cell lysates were analyzed for Erk phosphorylation as in B. other hand, cells expressing ErbB4 CYT-2 were not protected as well by NRG-1␤. The apoptotic indexes of the CYT-2 2 and 15b clones were reduced from 49% to 29% and from 33% to 26%, respectively. Thus on the average, NRG-1␤ reduced the apoptotic index of ErbB4 CYT-1 clones by 3.2-fold but of ErbB4 CYT-2 clones by only 1.5-fold. The apoptotic index of the Neo control cells was not affected at all by the presence of NRG-1␤.
When cells were plated at higher density (400,000 cells/ 35-mm dish) and cultured for 72 h in serum-free medium, the differential in NRG-1␤ effect on the survival of ErbB4 cells expressing the CYT-1 or CYT-2 isoform was even more prominent (Fig. 6, C and D). The apoptotic indexes of cells expressing ErbB4 CYT-1 were reduced in the presence of NRG-1␤ from 36% to 7% and from 38% to 8%, for the CYT-1 clones 15a and 42, respectively, an average of 5.1-fold. On the other hand, for ErbB4 CYT-2 clones 2 and 15b, the apoptotic indexes were reduced by NRG-1␤ from 64% to 49% and from 56% to 53%, respectively, for an average of only 1.2-fold. The more effective protection of cells expressing ErbB4 CYT-1 under the high density culture conditions could be due to the survival advantage of greater number of cells providing soluble or insoluble survival factors. It is also possible that a longer starvation period in the complete absence of serum (72 h versus 48 h) generates a greater experimental window that heightens the survival differences between the two cell types. Taken together, it was concluded that activated ErbB4 CYT-1, but not activated ErbB4 CYT-2, can mediate survival in cells undergoing starvation-induced apoptosis under different experimental conditions.
ErbB4 CYT-1 but Not ErbB4 CYT-2 Stimulates Phosphorylation of Akt-It has been suggested that cell survival is medi-ated by PI3-K utilizing a mechanism that includes phosphorylation of Akt (34). Lysates of cells expressing ErbB4 CYT-1 or ErbB4 CYT-2 were immunoprecipitated with an anti-phosphotyrosine antibody followed by Western blotting with an antibody against the p85 subunit of PI3-K. NRG-1␤ treatment resulted in association of p85 with tyrosine phosphorylation when ErbB4 CYT-1, but not ErbB4 CYT-2, was activated (data not shown), confirming our earlier results that ErbB4 CYT-1, but not ErbB4 CYT-2, can stimulate PI3-K activity (31). To test whether ErbB4 CYT-1 and ErbB4 CYT-2 differed in their ability to activate Akt, cells were stimulated with or without NRG-1␤, and the phosphorylation of Akt at Ser 473 was analyzed by Western blotting with a phospho-specific anti-Akt antibody (Fig.  7, upper panel). The phosphorylation of Akt was not enhanced in Neo control cells (Fig. 7, upper panel, lanes 1 and 2) nor in CYT-2 clones 11 and 15b (Fig. 7, upper panel, lanes 7-10) in response to NRG-1␤ treatment. On the other hand, NRG-1␤ stimulated the phosphorylation of Akt in CYT-1 clones 15a and 42 (Fig. 7, upper  panel, lanes 3-6). To control the loading of Akt protein, the same filter was reprobed with an anti-Akt antibody (Fig. 7, lower  panel). These results indicate that ErbB4 CYT-1, but not ErbB4 CYT-2, can be coupled to the PI3-K/Akt pathway.
ErbB4 CYT-1 but Not ErbB4 CYT-2 Mediates Chemotaxis and Membrane Ruffling-Growth factor-stimulated chemotaxis and actin rearrangements associated with cell motility require activation of PI3-K signaling (40). We previously demonstrated that nanomolar concentrations of the PI3-K inhibitor, wortmannin, inhibited the chemotaxis of cells expressing ErbB4 CYT-1 clone 1 induced by NRG-1␤, thereby demonstrating that PI3-K activity is required for migration mediated by FIG. 6. Analysis of apoptosis in serum-starved cells expressing ErbB4 cytoplasmic isoforms by DAPI staining. The effects of NRG-1␤ on the survival of serum-starved ErbB4 transfectants were analyzed under two different conditions. A, Neo control cells and cells expressing ErbB4 CYT-1 or ErbB4 CYT-2 were plated at 100,000 cells/35-mm dish. Next day the media were replaced by DMEM containing 0.5% FBS, and after 48 h of starvation, the media were replaced by serum-free DMEM containing 0 or 40 ng/ml NRG-1␤. After a further 48-h incubation in the presence or absence of NRG-1␤, nuclear morphology was visualized by DAPI staining. Representative examples of NRG-1␤-treated and untreated control, ErbB4 CYT-1 and ErbB4 CYT-2 cells are shown. The apoptotic nuclei have a characteristic condensed morphology and are small and white (arrow) as opposed to normal nuclei, which are larger and light blue with a more regular shape (arrowhead). B, quantitative analysis of apoptotic indexes of all the clones included in the experiment described in A. To obtain an apoptotic index, the number of apoptotic nuclei in a microscope field was counted and divided by the number of all the nuclei within the same area. Means and standard deviations of apoptotic indexes calculated from four independent fields were determined. C, wild type control cells and cells expressing ErbB4 CYT-1 or ErbB4 CYT-2 were plated at 400,000 cells/35-mm dish. Next day the media were replaced by serum-free DMEM containing 0 or 40 ng/ml NRG-1␤. After a further 72-h incubation, the nuclear morphology was assessed by DAPI staining as in A. D, quantitative analysis of apoptotic indexes of the experiment described in C.
Cell migration is characterized by extension of lamellipodia (52). To analyze the effect of NRG-1␤ on actin reorganization and lamellipodia formation, cells expressing ErbB4 CYT-1 or ErbB4 CYT-2 were plated on coverslips, stimulated for 15 min with NRG-1␤ and stained with TRITC-labeled phalloidin to visualize the actin filaments (Fig. 9). In the absence of NRG-1␤, the actin filaments were mostly organized as stress fibers in cells expressing either ErbB4 isoform (Fig. 9, A and C). When stimulated with NRG-1␤, cells expressing ErbB4 CYT-1 (clone 42) responded by reorganizing the actin filaments into membrane ruffles at the cell edges and by formation of lamellipodia (Fig.  9B). In contrast, NRG-1␤ did not cause any significant changes in the actin organization of cells expressing ErbB4 CYT-2 (clone 11) (Fig. 9D), consistent with the inability of these cells to migrate in response to NRG-1␤. Similar results were obtained by independent clones ErbB4 CYT-1 1 and ErbB4 CYT-2 15b (not shown). These results suggest that ErbB4 CYT-1, but not ErbB4 CYT-2, can mediate chemotaxis as well as reorganization of actin filaments into membrane ruffles and lamellipodia. DISCUSSION We have analyzed the functional specificities of two naturally occurring ErbB4 cytoplasmic isoforms, ErbB4 CYT-1 and ErbB4 CYT-2. Structurally the two isoforms differ by 16 amino acids present in the cytoplasmic tail of ErbB4 CYT-1 that are not present in ErbB4 CYT-2 (Fig. 1A). These 16 amino acids include a Tyr-Thr-Pro-Met sequence that matches the consensus-binding sequence for the p85 subunit of PI3-K (36) and that is necessary for the activation of PI3-K by the ErbB4 ligand NRG-1␤ (31). In this report we have demonstrated that activation of PI3-K via ErbB4 is necessary for NRG-1␤-dependent survival and chemotaxis but not proliferation. Several previous reports have addressed the role of PI3-K signaling in mediating growth factor responses. These studies have used genetically engineered receptor mutants or drugs, often toxic, and other strategies that target PI3-K. To our knowledge, our study is the first report of the functional consequences of the inability of activating PI3-K using a naturally occurring receptor variant.
Survival was monitored by measuring cell numbers, visualization of nuclear morphology by DAPI staining, and analyzing Akt phosphorylation, after serum starvation of several independent cell clones in the presence or absence of NRG-1␤. NRG-1␤ protected cells expressing ErbB4 CYT-1 from apoptosis to a much greater extent than cells expressing ErbB4 CYT-2. Although NRG-1␤ reduced the apoptotic index of cells expressing ErbB4 CYT-1 by 3.2-5.1-fold, depending on culture conditions, this index was reduced by only 1.2-1.5-fold in cells expressing ErbB4 CYT-2. Consistent with these results, a platelet-derived growth factor (PDGF) receptor-␤ mutant engineered to lack a PI3-K-binding site is incapable of mediating PDGF-stimulated survival (38). Efficiency of the protection of cells expressing ErbB4 CYT-1 by NRG-1␤ was correlated with the expression levels of ErbB4 protein, indicating a specific response. NRG-1␤ stimulation of ErbB4 CYT-1, but not of ErbB4 CYT-2, also resulted in the phosphorylation of the serine-threonine kinase Akt. These results are in agreement with the suggested role of Akt in survival signaling as a downstream effector of PI3-K (38,41,42,53), and numerous reports have concluded that Akt is necessary for the PI3-K-dependent survival pathway after exposure to a range of apoptotic stimuli (34, 54 -58). The mechanism by which Akt is coupled to survival may include Akt-mediated phosphorylation of the BCL-2-like protein BAD (43,44), caspase-9 (59), or transcription factors (60,61). Whether any of these molecules are involved in the survival activity mediated by ErbB4 CYT-1 is currently unknown.
Chemotaxis is another biological property affected by the lack of an ErbB4 PI3-K-binding site. NRG-1␤ readily stimulated the migration of cells expressing ErbB4 CYT-1, but not at  1 and 2), cells expressing ErbB4 CYT-1 (lanes 3-6; clones 15a and 42), or ErbB4 CYT-2 (lanes 7-10; clones 11 and 15b) were starved for 24 h in the absence of serum and stimulated without (lanes 1, 3, 5, 7, and 9) or with (lanes 2, 4, 6, 8, and 10) 50 ng/ml NRG-1␤ for 10 min on ice followed by 5 min at 37°C. Cell lysates were analyzed for phosphorylation of Akt at Ser 473 by Western blotting with a phospho-specific anti-Akt antibody and ECL (upper panel). The same filter was reprobed with anti-Akt antibody to demonstrate the loading of the protein (lower panel). Arrows indicate the ϳ60-kDa bands detected with the anti-phospho-Akt (p-Akt) and anti-Akt antibodies. all of cells expressing ErbB4 CYT-2. The dose at which NRG-1␤ stimulated half-maximal migration seemed to correlate with the level of ErbB4 CYT-1 expression. Based on previous results using checkerboard analysis, the migration response of ErbB4transfected NIH 3T3 cells to NRG-1␤ appears to be directional chemotaxis as opposed to chemokinesis (20). NRG-1␤-induced chemotaxis of cells expressing ErbB4 CYT-1 was characterized by reorganization of actin filaments into membrane ruffles and lamellipodia. No such activity occurred with cells expressing ErbB4 CYT-2. These results are consistent with reports that PDGF receptor-␤ mutants engineered to lack the PI3-K-binding site are defective in stimulating membrane ruffling and chemotaxis (40,62) and that inhibition of PI3-K function blocks chemotaxis and ruffling stimulated by RTK ligands, such as PDGF, insulin-like growth factor-1, and insulin (39,63). The effects of PI3-K on chemotaxis may be mediated via Rac (45), a Rho-like GTPase that regulates the reorganization of actin filaments in membrane ruffles (64). The function of Rac in ErbB4-mediated migration remains to be determined. It has been shown, however, that activation of Akt and Rac-mediated membrane ruffling are separate pathways activated in parallel downstream of PI3-K (65,66).
Although lack of PI3-K coupling to ErbB4 affects survival and migration adversely, it has no adverse effect on proliferation. NRG-1␤ was a potent mitogen for NIH 3T3 clones expressing either ErbB4 CYT-1 or ErbB4 CYT-2, when analyzed in a medium containing 1% serum. In addition, both isoforms were capable of activating signaling pathways that lead to proliferation. This was demonstrated in that activation of both ErbB4 isoforms lead to phosphorylation of Shc, an adapter molecule that links RTKs to the Grb2-SOS-Ras-Raf-MEK-MAPK pathway (50,51), and to phosphorylation of the MAPKs Erk1 and Erk2, which play a central role in regulating mitogenesis (67,68). These findings are consistent with earlier studies showing that a mutant PDGF receptor-␤ that does not activate PI3-K nevertheless stimulates DNA synthesis (40). There are contrary reports, however, indicating that PI3-K is necessary for mediating DNA synthesis in response to growth factors (69 -71), as well as sufficient for stimulating cell cycle entry in quiescent fibroblasts (72,73). Interestingly, microinjection of anti-p110␣ antibodies (70) and analysis of B cells isolated from mice with a targeted p85␣ gene (46) indicate that PI3-K is necessary for mitogenesis downstream of some but not all agonists.
When interpreting our results, it should be kept in mind that the 16 amino acids lacking from the CYT-2 isoform may mediate interactions with other signal transduction molecules in addition to PI3-K. To analyze this possibility, experiments including mutating the p85-binding site in ErbB4 CYT-1 or rescuing the defects in ErbB4 CYT-2 signaling by reconstructing the minimal p85-binding site or using a constitutively active form of PI3-K will be necessary. It is also plausible that the inability to couple ErbB4 to PI3-K affects cellular activities not addressed in our experiments, such as invasion, adhesion, or vesicular trafficking (37,74).
Our studies have been carried out using a genetically engineered ErbB4 CYT-2 construct based on an RT-PCR product generated with primers specific for the ErbB4 cytoplasmic domain. However, a full-length protein corresponding to the predicted ErbB4 CYT-2 sequence has not been isolated from natural sources. Thus, we cannot exclude the possibility that the natural ErbB4 protein with the CYT-2 type of cytoplasmic domain has further structural or functional changes outside the cytoplasmic region. However, a recently described single cDNA clone from a chick whole embryo cDNA library (75) encodes almost all of the open reading frame of chick ErbB4 with the exception of the first N-terminal 152 amino acids (compared with human ErbB4 (8)). This open reading frame encodes the CYT-2 variant within the cytoplasmic tail. These data suggest that ErbB4 CYT-2 domain can be found in a functional ErbB4 protein at least in chick. There are three reports of single cDNA library clones that span both the juxtamembrane and cytoplasmic domains. These encode human ErbB4 JM-b CYT-1 (30,47), rat ErbB4 JM-a CYT-1 (76), and chick ErbB4 JM-a CYT-2 (75). These cDNA sequences together with the comparative RT-PCR analysis of ErbB4 isoform expression in various mouse tissues (30,31) suggests that at least three, and putatively all four, of the possible combinations of ErbB4 juxtamembrane and cytoplasmic domains exist in vivo.
The ErbB4 cytoplasmic isoforms are distributed in normal tissues in a tissue-specific manner, suggesting that they have isoform-specific functions (30,31). The lethal phenotypes of erbB4 null mice indicate that expression of the erbB4 gene is of critical importance for cellular functions during normal development of the heart and brain (23). It will be of interest to analyze the specific expression patterns of the ErbB4 isoforms in various developmental processes where correlations between the cellular behavior and the type of isoform present can be made. ErbB4 has also been implicated in malignant transformation. Overexpression of ErbB4 together with other ErbBs transforms cells in vitro (77), and overexpression of ErbB4 protein or erbB4 gene amplification have been observed in medulloblastoma, breast cancer, and ovarian granulosa cell tumor patients (29,78,79). The first RT-PCR analyses of ErbB4 cytoplasmic isoform expression in human malignancies indicate that both ErbB4 CYT-1 and ErbB4 CYT-2 are present in 9 of 12 samples of human infiltrating ductal carcinomas (32). Future experiments will reveal whether different ErbB4 isoforms also have differential transforming potencies and ErbB4 Isoform Deficient in PI3-K Activity whether their expression levels are of differential prognostic or predictive value in human neoplasias. The fact that both PI3-K and Akt have been identified as viral oncogenes (80,81) supports a hypothesis that ErbB4 CYT-1, capable of stimulating PI3-K, Akt, survival, and chemotaxis, is more likely to be associated with malignancies than ErbB4 CYT-2.
In summary we have demonstrated that the RTK ErbB4 exists in two isoforms, ErbB4 CYT-1 and ErbB4 CYT-2. Both isoforms are coupled to the Shc/MAPK signaling pathway and support proliferation, but only ErbB4 CYT-1 can couple to the PI3-K and Akt pathways to support chemotaxis and survival. We suggest that these naturally occurring receptor isoforms provide a valuable tool to analyze the biological activities of the ErbB4 receptor and to sort out the roles of PI3-K in cellular signaling.