Epiregulin Is a Potent Pan-ErbB Ligand That Preferentially Activates Heterodimeric Receptor Complexes*

The ErbB signaling network consists of four transmembrane receptor tyrosine kinases and more than a dozen ligands sharing an epidermal growth factor (EGF) motif. The multiplicity of ErbB-specific ligands is incompletely understood in terms of signal specificity because all ErbB molecules signal through partially overlapping pathways. Here we addressed the action of epiregulin, a recently isolated ligand of ErbB-1. By employing a set of factor-dependent cell lines engineered to express individual ErbBs or their combinations, we found that epiregulin is the broadest specificity EGF-like ligand so far characterized: not only does it stimulate homodimers of both ErbB-1 and ErbB-4, it also activates all possible heterodimeric ErbB complexes. Consistent with its relaxed selectivity, epiregulin binds the various receptor combinations with an affinity that is approximately 100-fold lower than the affinity of ligands with more stringent selectivity, including EGF. Nevertheless, epiregulin’s action upon most receptor combinations transmits a more potent mitogenic signal than does EGF. This remarkable discrepancy between binding affinity and bioactivity is permitted by a mechanism that prevents receptor down-regulation, and results in a weak, but prolonged, state of receptor activation.

Various biological processes are controlled by intercellular interactions that are mediated by polypeptide growth factors. Examples include embryonic development, neuronal functions, hematopoiesis, and pathological situations, like wound healing and malignant transformation. The mechanism transmitting extracellular signals ultimately starts with binding of the growth factor to a cell surface receptor, that in many cases carries an intrinsic tyrosine kinase activity (1). These receptors fall into several subgroups sharing structural and functional characteristics. Each subgroup of receptors specifically recognizes a family of structurally homologous growth factors. Perhaps the most striking multiplicity of related growth factors is exemplified by the epidermal growth factor (EGF) 1 family of molecules (2). This six cysteine-containing motif of 45-60 amino acids is shared by all members of the family, and it functions as the receptor binding portion of the molecule. Currently there are four known receptors for EGF-like ligands, constituting the ErbB subgroup of receptor tyrosine kinases (also known as HER, or type I receptor tyrosine kinases (3)). Whereas ErbB-1 binds many ligands, including EGF, transforming growth factor ␣ (TGF␣), and amphiregulin, both ErbB-3 and ErbB-4 bind to a family of isoforms, collectively known as neuregulins (also called Neu differentiation factors, heregulins, glial growth factors, and acetylcholine receptor inducing activity) (4). A related group of molecules, termed NRG2, binds to the same two receptors (5)(6)(7), and a third molecule, NRG3, exclusively binds to ErbB-4 (8). Two other ligands, betacellulin (9), and the heparin-binding EGF-like growth factor (10,11) bind to both ErbB-1 and ErbB-4. Interestingly, the most oncogenic member of the ErbB family, ErbB-2, binds none of the EGF-like ligands with high affinity. However, recent studies indicate that ErbB-2 functions as a shared low affinity receptor that binds the apparently bivalent EGF-like ligands with low affinity, once they are presented by either one of the high affinity receptors (12).
Despite shared receptor specificity, it is clear that the multiple EGF-like ligands play distinct physiological roles: gene targeting experiments showed that loss of function of ErbB-1 (13)(14)(15) more severely impairs embryonic development than inactivation of one of its ligands, TGF␣ (16). On the other hand, targeting of either neuregulin (17), ErbB-2 (18), or ErbB-4 (19), resulted in the same embryonic cardiac defect, indicating that activation of an ErbB-2/ErbB-4 receptor combination is exclusively mediated by neuregulin in the developing heart. That ligand multiplicity related to tissue-specific expression is suggested by distinct spatial and temporal patterns of expression of the various ligands (reviewed in Ref. 2), and also by experiments with transgenic mice demonstrating tissue selectivity of specific ErbB-1 ligands (20). Part of the physiological selectivity of ligands with shared receptors may be attributed to their domains that flank the EGF-like motif, including the presence of heparin-binding sites, sugars, and specific protein motifs.
In this study we addressed the functional identity of epiregulin, a recently identified ligand of ErbB-1 (21,22). Like TGF␣, this ligand was originally isolated from the medium of transformed fibroblasts, and its transmembrane precursor carries only short sequences that flank the EGF-like motif. Epiregulin expression is relatively restricted; except for macrophages and placenta, other human tissues contain very low or no epiregulin transcripts, but most types of epithelial tumors are characterized by high expression of the growth factor (23). Although epiregulin competed with EGF on the binding to ErbB-1, it displayed relatively low affinity to ErbB-1-overexpressing cells (21). On the other hand, the factor displayed dual biological function in vitro: it stimulated proliferation of fibroblasts, smooth muscle cells, and hepatocytes, but inhibited growth of several tumor-derived epithelial cell lines (21). These observations, and the emerging broader than expected specificity of EGF-like ligands to ErbB proteins (reviewed in Ref. 24), prompted us to analyze the selectivity of epiregulin to ErbB proteins. Here we report that epiregulin is a pan-ErbB ligand that activates all ligand-stimulatable combinations of ErbB proteins with variable efficiency. Strikingly, in a model cellular system, epiregulin more potently activates mitogenesis than does EGF, although the affinity of EGF to ErbB-1 is approximately 100-fold higher. This superiority of epiregulin is independent on the presence of other ErbB proteins, and appears to result from a relatively inefficient mechanism of receptor inactivation.
Peptide Synthesis-Epiregulin was synthesized on an Applied Biosystems (ABI) model 431 peptide synthesizer fortified with UV feedback monitoring at 301 nm, and using Fmoc (9-fluorenylethoxycarbonyl-)-Rink amide AM resin. Only the EGF-like domain of the murine epiregulin (21) was synthesized. The conventional ABI monitor previous peak algorithm was employed up to five times with a cut-off of 3.5% of the first deprotection. A secondary deprotection was performed and followed by double coupling. Acetic anhydride/1-hydroxybenzotriazole capping was utilized at the end of each coupling, followed by washing with 1:1 trifluoroethanol/dichloromethane. The peptide was deprotected and removed from the resin as described (27), with the following modifications: methoxyindole (2%) was added to reagent K, and the reaction time was changed to 3.5 h. Small quantities of the reduced peptides were purified by reverse-phase high performance liquid chromatography and examined by matrix-assisted laser desorption ionization mass spectral analysis. The crude reduced protein was dissolved in a Tris-HCl buffer, pH 6.0, containing guanidium HCl (6 M) and diluted to a concentration of 0.06 mg/ml in methionine-containing buffer (10 mM) that included 1.5 mM cystine, 0.75 mM cysteine, and 100 mM Tris, pH 8.0. The mixture was stirred for 48 h at 4°C, and the oxidized protein isolated on a C-4 VYDAC 10 micron preparative column (22 ϫ 250 mm) using a 0.1% trifluoroacetic acid/water/acetonitrile gradient. The oxidized protein was lyophilized and characterized by mass spectrometry and amino acid analysis, and shown to be homogeneous. Electrospray mass spectrometry was used to verify the mass of the synthetic peptide.
Cell Lines-MDA-MB453 cells were purchased from the American Type Culture Collection (Rockville, MD). The Chinese hamster ovary (CHO) cell lines expressing various ErbB proteins or their combinations were described previously (28). The establishment of a series of interleukin 3 (IL-3)-dependent 32D myeloid cells expressing all combinations of ErbB-1, ErbB-2, and ErbB-3 has been described (29). To generate an ErbB-4-overexpressing derivative of 32D cells, we used an LTR-erbB-4 expression vector that was electroporated into 32D cells as described (30). Cell lines co-expressing ErbB-2 or ErbB-3, together with ErbB-4, were established by transfection of the pLXSHD reteroviral vector (31), directing ErbB-4 expression, into ErbB-2-or ErbB-3-expressing cells (D2 and D3, respectively) by using electroporation (Bio-Rad Genepulser, set at 400 volts and 250 millifarad). After a 24-h long recovery, cells were selected for 4 -5 weeks in medium containing histidinol (0.4 mg/ml). Clones expressing the two receptors were identified by using Western blotting, and isolated by limiting dilution. Due to differences in promoter potency, the selected cell line that singly expresses ErbB-4 (E4 cells) contained approximately 10 -12-fold more ErbB-4 molecules than cell lines expressing the combinations of ErbB-4 with ErbB-2 (D24 cells) or with ErbB-3 (D34 cells). A cell line expressing only approximately 5 ϫ 10 4 ErbB-4 molecules per cell was established by using previously described procedures (29) and denoted D4.
Radiolabeling of Ligands, Covalent Cross-linking, and Ligand Binding Analyses-Growth factors were labeled by using IODO-GEN as described (32). The specific activity was approximately 5 ϫ 10 5 cpm/ng. For covalent cross-linking analysis, cells (10 6 ) were incubated on ice for 1.5 h with 125 I-EGF, 125 I-NDF-␤1, or 125 I-epiregulin (each at 100 ng/ml). The chemical cross-linking reagent BS 3 was then added (1 mM), and after 90 min on ice, cells were pelleted and solubilized in solubilization buffer. For analyses of ligand displacement with 32D cells, 10 6 cells were washed once with binding buffer, and then incubated for 2 h at 4°C with a radiolabeled ligand (1 ng/ml) and various concentrations of an unlabeled ligand in a final volume of 0.2 ml. Nonspecific binding was determined in the presence of a 100-fold molar excess of the unlabeled ligand. To terminate ligand binding, each reaction tube was washed once with 0.5 ml of binding buffer and loaded on top of a 0.7-ml cushion of bovine serum. The tubes were spun (12,000 ϫ g, 2 min) to remove the unbound ligand. Ligand displacement from CHO cells was analyzed with cell monolayers grown in 24-well dishes. Monolayers were washed once with binding buffer and then incubated for 2 h at 4°C with 1 ng/ml of the radiolabeled ligand, along with increasing concentrations of an unlabeled growth factor. Then, cells were washed three times with ice-cold binding buffer. Labeled cells were lysed for 15 min at 37°C in 0.5 ml of 0.1 N NaOH solution containing 0.1% sodium dodecyl sulfate, and the radioactivity was determined. Nonspecific binding was calculated by subtracting the binding of radiolabeled ligands to untransfected CHO cells, or by performing the binding assays in the presence of a 100-fold excess of the unlabeled ligand.
Receptor Down-regulation Assay-Ligand-induced receptor downregulation was measured as follows: cells grown in 24-well plates were incubated at 37°C for up to 90 min without or with various ligands in binding buffer. The cells were then put on ice, rinsed twice with binding buffer, and surface-bound ligand molecules removed by using a 7-min long incubation in 0.5 ml of solution of 150 mM acetic acid, pH 2.7, containing 150 mM NaCl (33). The number of ligand-binding sites that remained exposed on the cell surface was then determined by incubating cells at 4°C with radiolabeled EGF (20 ng/ml) for 90 min.
Lysate Preparation and Western Blotting-For analysis of total cell lysates, gel sample buffer was added directly to cell monolayers or suspensions. For other experiments, solubilization buffer was added to cells on ice. The adherent CHO cells were scraped with a rubber policeman into 1 ml of buffer, transferred to microtubes, mixed harshly, and centrifuged (10,000 ϫ g, 10 min at 4°C). Samples were resolved by gel electrophoresis through 7.5% acrylamide gels, and electrophoretically transferred to nitrocellulose membranes. Membranes were blocked for 2 h in TBST buffer (0.02 Tris-HCl buffered at pH 7.5, 0.15 M NaCl, and 0.05% Tween 20) containing 1% milk, blotted for 2 h with 1 g/ml primary antibody, washed, and reblotted with 0.5 g/ml secondary antibody linked to horseradish peroxidase. Immunoreactive bands were detected with an enhanced chemiluminescence reagent (Amersham Corp.).
Cell Proliferation Assays-Cells were washed free of IL-3, resuspended in RPMI 1640 medium at 5 ϫ 10 5 cells/ml, and treated without or with growth factors or IL-3 (1:1000 dilution of medium conditioned by IL-3-producing cells). Cell survival was determined by using the MTT assay as described previously (29). MTT (0.1 mg/ml) was incubated for 2 h at 37°C with the analyzed cells. Living cells can transform the tetrazolium ring into dark blue formazan crystals, that can be quantified by reading the optical density at 540 -630 nm after lysis of the cells with acidic isopropyl alcohol (34).
Cellular Differentiation Assays-MDA-MB453 human mammary cancer cells were plated in chamber slides (Lab-Tek) and then incubated for 4 days in the absence or presence of ligands (50 ng/ml). Cells were stained with oil red O, to visualize neutral lipids, as described previously (35).

Induction of Cellular Differentiation and Tyrosine
Phosphorylation by Epiregulin in the Absence of ErbB-1-The duality of epiregulin's activity, namely, mitogenicity for some normal cells and growth inhibition of epithelial tumor cells (21), may depend on expression patterns of ErbB proteins, and thus may be explained by epiregulin's interaction with receptor species other than ErbB-1. As an initial test of this paradigm we examined the biological effect of epiregulin on MDA-MB453 mammary tumor cells, which are devoid of the EGF-receptor (ErbB-1), but can undergo phenotypic differentiation in response to EGF-like ligands (36). Evidently, these cells underwent growth arrest in response to long-term incubation with epiregulin, and displayed phenotypic differentiation that included cell flattening, and appearance of neutral lipid-containing vesicles (Fig. 1A). EGF, at 1-200 ng/ml, was inactive in inducing cell differentiation (data not shown), whereas similar phenotypic alterations were induced also by NDF/neuregulin, a ligand that interacts with both ErbB-3 and ErbB-4 (37). Consistent with their biological effects on MDA-MB453 cells, both epiregulin and NDF, but not EGF, were able to stimulate tyrosine phosphorylation of a 180-kDa protein at concentrations higher than 10 ng/ml (Fig. 1B). In conclusion, epiregulin action on the mammary epithelial cell line we examined is independent of ErbB-1, and is distinct from the effect of EGF.
Epiregulin Is a Relatively Potent Stimulator of ErbB-1, but It Can Transmit Biological Signals Also through Combinations of Other Receptors-To directly address the specificity of epiregulin to ErbB receptors, we employed a previously described series of cell lines derived from the IL-3-dependent 32D myeloid cell line (29). Parental 32D cells express no ErbB protein, but as a result of transfection, the derivative lines singly express ErbB-1, ErbB-2, ErbB-3, or ErbB-4 (cell lines denoted D1, D2, D3, and E4, respectively). Likewise, co-expression of two ErbB proteins established cell lines with various combinations. For example, D13 cells co-express ErbB-1 and ErbB-3. Analysis of cell proliferation in the absence of IL-3, but in the presence of increasing concentrations of epiregulin, EGF, or NDF-␤1, revealed several interesting characteristics of epiregulin. First, the factor was more potent than EGF on cells singly expressing ErbB-1 (D1 cells, Fig. 2A), as well as on cells expressing combinations of ErbB-1 with either ErbB-2 (D12 cells) or ErbB-3 (D13 panel in Fig. 2A). Not only were the dose-response curves of epiregulin shifted to the left, but this ligand exerted in D1 and D13 cells a higher maximal response than EGF. Consistent with the catalytic inactivity of ErbB-3 (38), and the inability of ErbB-2 to bind any of the ErbB ligands with high affinity (12), cells singly expressing ErbB-3 or ErbB-2 (D3 and D2 cell lines, respectively) did not respond to epiregulin ( Fig. 2A). For control, we verified that D2 cells are stimulatable by a mAb to ErbB-2 (25) (Fig. 2A), and D3 cells retained response to IL-3 (Fig. 3). Surprisingly, E4 cells that highly overexpress ErbB-4 exhibited mitogenic response to both epiregulin and EGF at concentrations above 5 ng/ml ( Fig. 2A). In fact, the response to EGF was reproducibly slightly higher than the mitogenic effect of epiregulin on these cells. Due to the use of different promoters, ErbB-4 expression in the E4 cell line was more than 10-fold higher than that of ErbB-1 in D1 cells (see "Experimental Procedures"). To address the possibility that epiregulin and EGF act through ErbB-4 only when this receptor is overexpressed, we analyzed a second cell line, D4, whose ErbB-4 expression is comparable with the level of ErbB-1 expression in D1 cells. When tested on D4 cells, both ligands displayed mitogenic activity (Fig. 2B), along with an ability to stimulate tyrosine phosphorylation (Fig. 2C). Nevertheless, in terms of the maximal response to IL-3, both epiregulin and EGF were more active on the ErbB-4-overexpressing cell line than on the low expressor D4 cells, implying that the level of expression of ErbB-4 affects the level of cell proliferation, but not ligand specificity.
Although the effect of epiregulin on cells coexpressing ErbB-3 with ErbB-1 (D13 cells) was higher than that of EGF, the response to NDF was much higher, presumably because NDF better recruits ErbB-3 into heterodimers (29,39). Nevertheless, it is clear that also epiregulin can recruit ErbB-3 into heterodimers, as reflected by its activity on cells coexpressing a combination of ErbB-3 with either ErbB-2 (D23 cells, Fig. 2A) or ErbB-4 (D34 cells, Fig. 2A). This ability of epiregulin distinguishes it from EGF, whose signaling through the ErbB-2/ ErbB-3 heterodimer occurs only at extremely high concentrations ( Fig. 2A and Ref. 40 and 41), and is completely inactive in stimulating an ErbB-3/ErbB-4 heterodimer ( Fig. 2A). Moreover, although EGF is slightly more potent than epiregulin on ErbB-4-expressing cells (E4 panels in Figs. 2A and 3), epiregulin is superior when ErbB-2 is coexpressed with ErbB-4 (D24 panels in Figs. 2A and 3), suggesting that this ligand is a better stimulator of the ErbB-2/ErbB-4 heterodimer. Taken together, the results presented in Fig. 2 imply that relative to EGF, epiregulin is a better agonist of ErbB-1-containing homo-and heterodimers. In addition, recruitment of ErbB-2, ErbB-3, and ErbB-4 into heterodimers is more efficient in the case of epiregulin. However, homodimers of ErbB-4 are better activated

FIG. 1. Induction of cellular differentiation and tyrosine phosphorylation by epiregulin in mammary cells lacking ErbB-1/EGF receptor. A,
MDA-MB453 human mammary cancer cells, that express no ErbB-1, were plated in chamber slides and then incubated for 4 days in the absence (CONTROL) or presence of epiregulin (50 ng/ml). Cells were stained with oil red O, to visualize neutral lipids. Note the appearance of lipid droplets (yellow) in epiregulintreated cells. The magnification used was ϫ 600. B, following an overnight starvation, 10 6 MDA-MB453 cells were incubated for 2 min at 37°C without or with EGF, NDF-␤1, or epiregulin, at the indicated concentrations. Whole cell lysates were then prepared, resolved by gel electrophoresis, and immunoblotted with an antibody to phosphotyrosine (PY20). Bound antibody was detected by using a chemiluminescence-based method.
by EGF, and neither homodimers of ErbB-2 nor ErbB-3⅐ErbB-3 complexes are stimulatable by the two ligands.
These conclusions were further supported by long-term sur-vival experiments that are presented in Fig. 3. In this type of analysis cells are maintained in the absence of IL-3, but in the presence of epiregulin (or other ligands) for several days, and  L26, open circles). The MTT assay was performed 24 h later. Results are presented as fold induction over the control untreated cells, and are the mean Ϯ S.D. of four determinations. Each experiment was repeated at least twice. Cells singly expressing ErbB-3 (D3 cells) responded to none of the ligands we tested, but these cells retained response to IL-3. B, D4 cells were tested for cell proliferation by using the MTT assay as described above, except that the indicated ligands were used at 100 ng/ml. For control, cells were incubated in the absence of IL-3 or ligands. C, ligand-induced tyrosine phosphorylation was analyzed in D4 cells by incubating 10 6 cells without or with the indicated ligands (each at 100 ng/ml). Following 2 min at 37°C, whole cell lysates were prepared and analyzed by immunoblotting with a mAb to phosphotyrosine. Antibody detection was performed with a chemiluminescence kit. Only the 180-kDa region of the blot is shown. their survival determined by using the MTT assay. Consistent with the dose curves of the short-term mitogenic assay, at a saturating concentration epiregulin acted as a slightly better survival factor than EGF for cells expressing ErbB-1, either alone or in combination with ErbB-3 (Fig. 3). Also consistent with the data of Fig. 2 was the observation that EGF exerted a better survival activity on ErbB-4-overexpressing cells (E4 panel in Fig. 3). Interestingly, the presence of ErbB-2, together with either ErbB-4 or ErbB-3, enabled epiregulin to become a potent stimulator of cell proliferation, whereas EGF acted primarily as a survival factor under these circumstances (D23 and D24 panels in Fig. 3). Although survival of cells coexpressing ErbB-3 and ErbB-4 was only slightly extended by epiregulin (D34 panel in Fig. 3), this effect was higher than that of EGF, reinforcing the relative preference of epiregulin for heterodimeric receptor combinations.
Receptor Phosphorylation and MAP Kinase Activation by Epiregulin-Signaling by all EGF-like ligands is mediated by rapid tyrosine phosphorylation of the respective receptors, and is ultimately funneled to the mitogen-activated protein kinase (MAP-kinase/Erk) pathway (42). The biological differences we observed between epiregulin, EGF, and NDF in subsets of 32D cells suggested that these ligands may differ in signaling potency, and especially in their ability to recruit the MAPK pathway. To analyze receptor phosphorylation and MAPK activation we probed blots of whole extracts, prepared from ligandstimulated cells, with antibodies to phosphotyrosine, or with a murine mAb that specifically recognizes the active, doubly phosphorylated form of the ERK1 and ERK2 MAPKs (26). Surprisingly, the more mitogenic ligand of ErbB-1, epiregulin, exhibited weaker, but not less sustained, tyrosine phosphorylation of proteins at the 180-kDa range corresponding to ErbB-1 in D1 cells (Fig. 4A). Although both EGF and epiregulin stimulated MAPK phosphorylation in these cells, the patterns of activation differed: a comparable increase in the activity of both forms of the kinase was induced by epiregulin, whereas primarily the lower form was activated after stimulation with EGF. Importantly, although stimulation by EGF was more uniform at intermediate time intervals (10 -20 min), it completely disappeared after 30 -60 min, at which time the effect of epiregulin was still detectable. By contrast, ErbB-4 phosphorylation was stronger with epiregulin than with EGF (E4 panel in Fig. 4A), although the latter is a slightly more efficient mitogen for the ErbB-4-overexpressing E4 cells ( Figs. 2A and  3). These differences between ErbB-1 and ErbB-4 phosphorylation are cell-type independent, because they were reproduced in a series of CHO cells expressing ErbB-1 (CB1 cells) or ErbB-4 (CB4 cells), on a low background of the endogenous hamster ErbB-2 (Fig. 4B). Analysis of 32D cells expressing a combination of ErbB-2 with ErbB-3 (D23 cells) revealed that both forms of MAPK were rapidly stimulated by epiregulin, but phosphorylation of both ErbBs and MAPKs by EGF occurred only at very high ligand concentrations, in agreement with recent reports (40,41). The maximal activation of MAPK in these cells was observed upon stimulation with NDF, a ligand whose mitogenic effect was almost equivalent to that of IL-3 (Fig. 3). A relatively sustained stimulation, and appearance of an activated Erk-2, were observed upon activation of both D23 and D24 cells by their most potent ligand, namely, NDF, implying that these features may characterize the stronger mito- genic signals. In conclusion, the relative strength of mitogenic signals of EGF-like ligands better correlates with the duration of MAPK activation (especially the modification of Erk-2) than with the intensity of ErbB phosphorylation.
Low Affinity Interaction of Epiregulin with ErbB-1 and Other ErbB Proteins-The relatively weak stimulation of ErbB-1 phosphorylation by epiregulin (D1 panel in Fig. 4) suggested low affinity interaction of epiregulin with ErbB-1 on D1 cells. This possibility was addressed by employing two assays: covalent cross-linking of a radiolabeled epiregulin to the surface of ErbB-expressing 32D cell derivatives (Fig. 5), and ligand displacement analyses that were performed with both 32D-and CHO-derived cell lines (Fig. 6). Epiregulin was radiolabeled with 125 I and covalently cross-linked to the surface of 32D cells by using the BS 3 covalent cross-linking reagent. The specificity of labeling by epiregulin was evident by the absence of covalent cross-linking to ErbB-2 and ErbB-3, when these receptors were singly expressed (D2 and D3 cells, respectively, Fig. 5), and by displacement of radioactive epiregulin by a large excess of the unlabeled ligand (data not shown). Interestingly, only a very weak signal was observed when radiolabeled epiregulin was covalently cross-linked to cells singly expressing ErbB-1, although these cells displayed a strong cross-linking signal with 125 I-EGF, whose specific radioactivity was comparable to that of 125 I-epiregulin (Fig. 5). A slightly stronger signal was observed when cells coexpressing ErbB-1 and ErbB-2 were analyzed, implying that the corresponding heterodimer cooperatively interacts with epiregulin. The combination of ErbB-1 with ErbB-3 was less efficient than that of ErbB-1 with ErbB-2, although the numbers of ErbB-1 molecules on D1, D12, and D13 cells were similar. By contrast with ErbB-1, affinity labeling of ErbB-4 in the overexpressing E4 cell line was very efficient in the case of both epiregulin and NDF, but relatively weak labeling was observed with EGF (Fig. 5), in accordance with receptor phosphorylation signals (Fig. 4A). Similar observations were made with the D4 and CB4 cell lines (data not shown). Interestingly, we were unable to detect covalent crosslinking of epiregulin to cells coexpressing ErbB-3 with either ErbB-2 or ErbB-4 (D23 and D34 lanes in Fig. 5, note that ErbB-4 expression in D24 and D34 cells is approximately 10fold lower than in E4 cells), although these combinations reacted with NDF. By contrast, the ErbB-2/ErbB-4 combination displayed a clearly detectable signal with 125 I-epiregulin, reflecting the relatively high mitogenic response of D24 cells to epiregulin ( Figs. 2A and 3).
We then compared the capacity of epiregulin, as opposed to EGF, to displace a cell-bound radioactive EGF from the surface of 32D or CHO cells singly expressing ErbB-1 (D1 and CB1 were incubated at 37°C for various time intervals (indicated in minutes) with epiregulin (100 ng/ml), EGF (100 ng/ml, except for D23 cells that were treated with 500 ng/ml), or NDF-␤1 (100 ng/ml): D1, CB1, E4, and CB4 cells that singly express ErbB-1 or ErbB-4, respectively, whereas D23, D24, and CB14 cells co-express a combination of the corresponding two receptors. In the end of the incubation period, whole cell lysates were prepared, cleared from debris and nuclei, resolved by gel electrophoresis, and subjected to immunoblotting with either an antibody to phosphotyrosine (P-TYR), or with an antibody specific to the active doubly phosphorylated form of MAPK, as indicated. Derivatives of CHO cells were analyzed only with antibodies to phosphotyrosine. Signal detection was performed by using a chemiluminescence kit. cells, respectively). In contrast with the mitogenic superiority of epiregulin for ErbB-1-expressing 32D cells, the apparent binding affinity of epiregulin, as reflected by the competition curves, was 2 orders of magnitude lower than that of EGF (D1 and CB1 panels in Fig. 6, A and B). The presence of ErbB-4 together with ErbB-1 did not significantly alter the ability of epiregulin to displace EGF from the surface of CB14 cells (Fig.  6B), although epiregulin was able to displace, albeit with low efficiency, a surface-bound 125 I-NDF from ErbB-4-expressing cells (D4 or CB4 cells, Fig. 6, A and B). The results of ligand displacement experiments that were performed with E4 cells were qualitatively similar (data not shown). NDF displacement by epiregulin, or EGF, was relatively efficient in D24 cells, but only weak competition was detectable in D34 cells, consistent with the relative mitogenic potency of epiregulin for D24 and D34 cells ( Figs. 2A and 3). Thus, affinity labeling (Fig. 5) and ligand competition analyses (Fig. 6) imply that epiregulin binds cooperatively to the combination of ErbB-2 with ErbB-4. By contrast, only very weak competition between epiregulin and NDF was observed in cells expressing ErbB-3, either alone or in combination with ErbB-2 or ErbB-4 (Fig. 6A), implying that ErbB-3, unlike ErbB-4, does not cooperate with ErbB-2 in epiregulin binding. This conclusion is consistent with the absence of a detectable cross-linking signal in D3, D13, and D23 cells (Fig. 5). In light of this inference the results obtained with D23 cells are interesting because epiregulin displayed only a slightly better ability than EGF to displace NDF from these cells, but its mitogenic activity was much stronger than that of EGF ( Figs. 2A and 3). In conclusion, receptor binding analyses indicated direct interaction between epiregulin and two receptors, ErbB-1 and ErbB-4. Although neither ErbB-3 nor ErbB-2 directly interact with epiregulin, the latter protein significantly cooperates with both direct receptors of epiregulin.
Epiregulin-induced Down-regulation of ErbB-1 Is Defective-The superior mitogenic activity of epiregulin is analogous to that of TGF␣. This latter ligand of ErbB-1 is a better agonist than EGF when tested in vitro in mitogenic, angiogenic, and motogenic assays (43,44). Apparently, the relatively potent activity of TGF␣, whose binding affinity is almost identical to that of EGF, is due to the absence of receptor down-regulation, which allows sustained cellular activation (45). To examine the possibility that epiregulin's superiority is due to a defective receptor inactivation process, we exposed CB1 cells to epiregulin, EGF, or TGF␣, and determined the extent of disappearance of ErbB-1 from the cell surface. Evidently, whereas EGF induced gradual disappearance of the surface-exposed ErbB-1, neither epiregulin nor TGF␣ led to a significant change in the level of surface ErbB-1 (Fig. 7), although at the concentrations we used both ligands were more mitogenic than EGF ( Fig. 2A,  and data not shown). In experiments that are not presented we found that the difference in receptor down-regulation was not due to defective endocytosis of epiregulin, whose rate of internalization was comparable to that of EGF and TGF␣. This observation raised the possibility that unlike EGF, which directs ErbB-1 to degradation in lysosomes, epiregulin binding to ErbB-1 is followed by receptor recycling, a route taken by the TGF␣-driven ErbB-1 (45,46). This notion was supported by an experiment that tested the effect of monensin, a well characterized inhibitor of receptor recycling (47), on down-regulation of ErbB-1. In the presence of the carboxylic ionophore both epiregulin and TGF␣ induced significant down-regulation of ErbB-1, but this compound was ineffective on the extensive down-regulation that was induced by EGF (Fig. 7, and data not shown). In conclusion, the relatively strong biological action of epiregulin through ErbB-1 may be due to continuous recycling of ErbB-1 back to the cell surface, thus allowing prolongation of epiregulin signaling. DISCUSSION The evolutionary pathway of the ErbB signaling module, from worms (48) and flies (49) to mammals, indicates that duplication of genes encoding EGF-like ligands preceded multiplication of receptor-encoding genes. Despite multiplicity of ligands and receptors, it is clear that the downstream signaling mechanisms, namely a linear cascade leading to MAPK activation, has been conserved. Thus, to gain functional diversity, variations on the common theme of ligand-ErbB-MAPK evolved throughout evolution. Examination of the interactions between one of the mammalian ErbB ligands, epiregulin, and various combinations of the four ErbB proteins uncovered two novel features of the evolved module, that are schematically presented in Fig. 8. First, epiregulin is a broad-specificity ligand that activates all eight ligand-stimulatable combinations of ErbBs. Second, despite its extremely low affinity, signaling by epiregulin is more potent than the bioactivity of a high affinity ligand, namely, EGF. The mechanisms underlying these two features, and their functional implications, are discussed below.
Pan-ErbB Specificity of Epiregulin-The four mammalian ErbB proteins can form 10 homo-and heterodimeric complexes, including an ErbB-3 homodimer, which is biologically inactive (29), and an ErbB-2 homodimer whose formation may be driven by receptor overexpression (50), or by a transmembrane oncogenic mutation (51). Epiregulin can signal through all but these two homodimeric combinations of ErbBs (Fig. 8). This broad specificity is unique; no other EGF-like ligand has such a wide selection of receptors. However, due to its broad selectivity, none of the receptors of epiregulin binds it with high affinity (Figs. 5 and 6).
One of the most surprising observations made in the course of the present study is the ability of both epiregulin and EGF to activate ErbB-4 when this receptor is singly expressed. This observation is reminiscent of several recent reports that identified betacellulin (9) and heparin-binding EGF (10, 11) as ligands of ErbB-4. Conceivably, ErbB-1 and ErbB-4 share some structural features at their ligand-binding sites, thus defining a subgroup of direct ErbB-1 ligands, including EGF, betacellulin, and heparin-binding EGF, but excluding TGF␣ and amphiregulin, as ligands with dual receptor specificity. Nevertheless, like all other interactions of epiregulin, binding to ErbB-4 is characterized by very low affinity; the corresponding dissociation constant is estimated to be in the micromolar range (D4 and CB4 panels in Fig. 6). The affinity of the other direct receptor of epiregulin, ErbB-1, is only 10-fold better, much higher than the nanomolar or lower apparent K d of EGF or NDF binding to their direct receptors (Fig. 6A). However, re-ceptor combinations containing ErbB-1 and ErbB-4 are not the only receptors for epiregulin; although this ligand does not interact with isolated components of the ErbB-2/ErbB-3 heterodimer, it can efficiently stimulate the respective receptor combination (D23 panels in Figs. 2A and 3). This is probably mediated by an extremely low affinity of epiregulin to ErbB-3 (D3 panel in Fig. 6), and by a cooperative effect of the coexpressed ErbB-2. This effect of the ligand-less ErbB-2 is extended to heterodimers containing the direct epiregulin receptors, namely ErbB-1 and ErbB-4; cooperativity is exemplified by the relatively strong binding of epiregulin to cells coexpressing ErbB-1 and ErbB-2 (but not to cells co-expressing ErbB-1 and ErbB-3, Fig. 5), and by the ability of ErbB-2 to augment epiregulin binding to ErbB-4 (compare D4 and D24 panels in   6). This binding effect is translated to enhanced signaling by the ErbB-2/ErbB-4 heterodimer relative to the ErbB-4 homodimer, and is apparently more relevant to epiregulin than to EGF (compare E4 and D24 panels in Fig. 2A). The mechanism underlying signal amplification by ErbB-2, a process that is significant to tumors overexpressing this receptor, has been previously attributed to its ability to decelerate dissociation of NDF and EGF from ErbB-2-containing heterodimers (25,52). The present study apparently extends this mechanism to epiregulin.
How does epiregulin recognize all six heterodimeric complexes of ErbBs? According to a ligand bivalence model (12), a notion supported by recent affinity labeling studies (53), and by measurements of the stoichiometry of ligand-receptor interactions in solution (54), epiregulin carries a high affinity binding site whose specificity is limited to ErbB-1 and ErbB-4. Another site that is structurally distinct and may localize to the Cterminal half of the ligand, binds with broad specificity but low affinity to other ErbB proteins, including ErbB-1 and ErbB-4 (thus allowing homodimer formation), as well as to ErbB-2 and ErbB-3, to confer heterodimer formation. Nevertheless, as is the case with EGF and NDF, the putative "low-affinity/broadspecificity" site of epiregulin apparently prefers ErbB-2 over other receptors. This model explains how ErbB-2 augments epiregulin signaling through the ErbB-2/ErbB-3 and ErbB-2/ ErbB-4 heterodimers.
Mechanism of Signaling Superiority of Low Affinity Ligand-ErbB Interactions-In their original analysis of epiregulin interactions with various cell types, Toyoda and collaborators (21) found that this ligand was more mitogenic than EGF for several types of normal cells, although epiregulin binding to cells of another type (the A-431 epidermoid carcinoma line) displayed a 10-fold lower affinity. Potentially, this discrepancy could be due to the different repertoires of ErbB proteins expressed on the surface of the different lines of cultured cells that these authors examined. However, our studies with engineered myeloid cells excluded this possibility, because epiregulin's superiority was retained also by cells singly expressing ErbB-1. In fact, our results extend the discrepancy between binding affinity and bioactivity to signaling through ErbB-4. Thus, epiregulin is a relatively potent stimulator of mitogenesis through both ErbB-1 and ErbB-4, despite being a very low affinity ligand of these two receptors (D1 and E4 panels in Figs.  2A and 6A). The observation that ErbB-1 phosphorylation by epiregulin is weaker than the effect of EGF (Fig. 4A), implies that receptor activation is not the sole determinant of signaling potency. Instead, differences in the inactivation process may be critical: apart from differential recruitment of both tyrosinespecific phosphatases (55) and the negative regulator c-Cbl (56), endocytosis of ligand-receptor complexes is a major process that leads to inactivation of growth factor signaling (reviewed in Ref. 57). Our initial studies of this aspect of epiregulin's action indicated that this ligand, unlike EGF, mediates limited, if any, down-regulation of ErbB-1 (Fig. 7). Additional analyses raised the possibility that epiregulin undergoes internalization, but its receptor rapidly recycles to the cell surface (Fig. 7). Presumably, the very low affinity of epiregulin to ErbB-1 is insufficient to direct this receptor to lysosomal degradation, either because phosphorylation on tyrosine residues, which is essential for rapid internalization (58), is relatively inefficient, or because the ligand dissociates very rapidly. It is relevant that mutations of another receptor, that stabilize ligand-receptor interactions at the moderately acidic conditions of early endosomes, accelerate receptor degradation and prevent recycling (59,60), indicating that the strength of ligand binding is critical for receptor routing. This mechanism of epiregulin/ErbB-1 interactions is expected to promote a relatively weak level of receptor activation, but due to receptor recycling, repeated association-dissociation cycles may result in prolongation of signaling. In support of this model, we observed an overall lower activation of MAPK by epiregulin, but this was more prolonged than in the case of EGF (D1 panel in Fig. 4A). Variations of the proposed mechanism have previ- FIG. 7. Epiregulin-induced down-regulation of ErbB-1. CB1 cells were grown to 80% confluence in 24-well plates, rinsed with binding buffer, and incubated at 37°C for the indicated time intervals with one of the following ligands (each at 1 ng/ml): epiregulin (closed squares), EGF (circles), or TGF-␣ (triangles). Sister epiregulin-treated cells were similarly incubated, except that monensin (0.1 mM) was added to the medium (open squares). Thereafter, monolayers were rinsed twice with binding buffer, followed by a 7-min long incubation with a low pH stripping buffer that removes surface-bound ligands. The level of surface receptors, relative to the number of ligand-binding sites present before down-regulation, was determined by incubating cells for 1.5 h at 4°C with radiolabeled EGF. The results are expressed as the average fraction and range (bars) of the original binding sites that remained on the cell surface after exposure to the non-labeled ligand at 37°C.

FIG. 8. Summary of epiregulin-receptor interactions.
The horizontal gray bar represents the plasma membrane, and the 10 possible receptor dimers are shown schematically as double circular structures. Specific ErbB proteins are identified by their numbers. Two ErbB ligands, epiregulin and EGF, are compared and their relative strength of signaling, as revealed by using an IL-3-dependent series of cell lines, are represented by the thickness of the corresponding arrows. Broken arrows indicate very low bioactivity. For simplicity, the ability of EGF to stimulate an ErbB-2/ErbB-3 heterodimer at very high ligand concentrations is not represented. Note that no ligand binds with high affinity to ErbB-2 homodimers. Because no 32D cell derivative co-expressing ErbB-1 and ErbB-4 has been established, the data related to this heterodimeric combination was inferred from experiments with transfected CHO cells. All other receptor combinations were examined in 32D cell derivatives. ously been reported: in the case of TGF␣, whose binding affinity is comparable to that of EGF, the more rapid dissociation of the ligand-receptor complex in an acidic endosomal compartment drives ErbB-1 to recycling (45). This is contrasted with the lysosomal destination taken by an EGF-bound ErbB-1. As a result, signaling by TGF␣ is often more potent than that of EGF. An even closer example is provided by a mutant of EGF that was engineered to enhance the mitogenic potency of the growth factor for biotechnological applications (61). This mutant achieved mitogenic superiority through a combination of a 50-fold lower affinity, longer retention in culture supernatants, and a very limited receptor down-regulation.
In addition to the question how wide is the relevance of our findings to other growth factors whose binding affinities are very low, several other interesting questions are left open. The exceptionally broad specificity of epiregulin joins other observations that collectively imply non-redundancy of the multiple EGF-like ligands (reviewed in Ref. 62). Evidently, each ligand differs from other members of its family by a unique preference for certain ErbB proteins. This, however, does not explain how different ligands mediate mitogenesis on some cells, but differentiation (37), survival (63), or cell motility (10) on other types of cells, although in all cases the MAPK pathway is recruited. Even more difficult to reason is the inhibitory activity of epiregulin on certain epithelial cell lines (21), because all of its receptors turned out to be stimulatory for myeloid cells ( Figs.  2A and 3). Perhaps a cell type-specific component lying downstream of ErbBs determines the nature of cellular response. Another puzzling issue is the contrast between the broad selectivity of epiregulin for ErbBs, and its very limited pattern of expression (23). This and other questions will require in vivo studies of epiregulin's physiological role.