The interaction of an epidermal growth factor/transforming growth factor alpha tail chimera with the human epidermal growth factor receptor reveals unexpected complexities.

It has been assumed that substitution of homologous regions of transforming growth factor α (TGF-α) into epidermal growth factor (EGF) can be used to probe ligand-receptor recognition without detrimental effects on ligand characteristics for the human EGF receptor (EGFR). We show that a chimera of murine (m) EGF in which the carboxyl-terminal tail is substituted for that of TGF-α (mEGF/TGF-α44-50) results in complex features that belie this initial simplistic assumption. Comparison of EGF and mEGF/TGF-α44-50 in equilibrium binding assays showed that although the relative binding affinity of the chimera was reduced 80-200-fold, it was more potent than EGF in mitogenesis assays using NR6/HER cells. This superagonist activity could not be attributed to differences in ligand processing or to binding to other members of the c-erbB family. It appeared to be due, in part, to choice of an EGFR-overexpressing target cell where high receptor number compensated for the low affinity of the ligand; it also appeared to be related to the ability of the chimera to activate the EGFR tyrosine kinase. Thus, when EGFR autophosphorylation was measured, mEGF/TGF-α44-50 was more potent than EGF, despite its low affinity. When tested using chicken embryo fibroblasts, substitution of the TGF-α carboxyl-terminal tail into mEGF failed to enhance its binding affinity for chicken EGFRs; however, the chimera was intermediate in potency between TGF-α and mEGF in mitogenesis assays. Our results suggest a contextual requirement for EGFR recognition which is ligand-specific. Further, the unpredictable responses to chimeric ligands underline the complex nature of the processes of ligand recognition, receptor activation, and the ensuing cellular response.

The epidermal growth factor receptor (EGFR) 1 ligand family, of which EGF and transforming growth factor ␣ (TGF-␣) are the best characterized members, bind to their cognate receptor in a mutually competitive fashion and have association con-stants in the range of 10 9 -10 10 M Ϫ1 (1). The structural homology of these proteins arises from the presence of six highly conserved cysteine residues and two glycine residues (2). Of the remaining residues, only four show significant homologies: the invariant Arg 41 and Tyr 37 and the semiconserved Leu 47 and Tyr 13 (EGF numbering). Mutational and chemical analyses of Arg 41 have demonstrated that this is probably the most important receptor contact residue, contributing at least 3 orders of magnitude to the overall binding free energy (3). Leu 47 , which lies in the flexible C-tail of the growth factor, is also important; both site-directed mutagenesis (4) and controlled proteolysis (5) have shown that loss of Leu 47 decreases receptor binding affinity by around 2 orders of magnitude. The requirement of aromaticity at positions 13 (6) and 37 (7,8) is less stringent; however, these residues also appear to contribute to receptor binding.
The fact that the EGF ligand family exhibits only a limited pattern of common surface residues suggests that their binding may not be identical. This possibility is readily apparent for the chicken EGFR, which displays differential affinity for EGF and TGF-␣ (9). In the case of the human EGFR, which displays comparable affinities for EGF and TGF-␣ (1,9), the ability of the monoclonal antibody 13A9 (10) to prevent binding of TGF-␣, but not EGF, to the human EGFR suggests that their binding is not identical. This is further supported by the observation that insertional mutagenesis into domain II of EGFR reduces binding of TGF-␣ but not EGF (11).
We have explored the basis for the differential recognition of EGF and TGF-␣ and have shown that the B-loop ␤-sheet is a major distinguishing feature that contributes to ligand recognition by the chicken EGFR (12). Further, the "hinge residue" that lies at the end of the ␤-sheet is a ligand-specific residue for the human EGFR (13). From a consideration of the nature of the hinge residue and its relation to the sequence of the B-loop ␤-sheet, we have postulated that the EGF family can be subdivided into EGF-like and TGF-␣-like and that this is of functional significance in receptor recognition (13). Thus, the receptor binding surface comprises a discontinuous epitope involving residues from the structural core of the EGF motif as well as the C-tail.
Although structural analysis by two-dimensional 1 H NMR shows that the C-tail is flexible in the solution conformations of either EGF (14) or TGF-␣ (15), recent NMR analyses have shown that the conformation of the C-tail of TGF-␣ becomes more ordered upon receptor binding (16). Although substitution of the C-tail of TGF-␣ into human EGF has been reported to restore high affinity binding of hEGF to the chicken EGFR (17), its effects on binding to the human EGFR have not been evaluated. In view of the possibility of nonidentical binding of EGF and TGF-␣, we examined the consequences on binding and activation of the human EGFR when a similar substitution * This work was supported by the Cancer Research Campaign, United Kingdom. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The first two authors contributed equally to this work. ‡ To whom correspondence should be addressed. Tel.: 44-1703-798-795; Fax: 44-1703-783-839. 1 The abbreviations used are: EGFR, epidermal growth factor receptor; TGF-␣, transforming growth factor ␣; C-tail, carboxyl-terminal tail; the prefix h-or m-denotes human or murine, respectively; [ 125 I]UdR, 5-[ 125 I]iodo-2Ј-deoxyuridine; HFF, human foreskin fibroblast; CEF, chicken embryo fibroblast; mEGF/TGF-␣ 44 -50 , a chimera in which the carboxyl-terminal tail of mEGF is replaced by residues 44 -50 of human TGF-␣. was introduced into mEGF. Our results demonstrate that binding of the C-tail of the growth factor to EGFR exhibits ligandspecific features; the substitution has unexpected effects on receptor binding and tyrosine kinase activation which have not been documented previously for such a chimeric ligand. This work has been presented in abstract form (18).

EXPERIMENTAL PROCEDURES
Materials-All chemicals were purchased from Sigma (Poole, Dorset, United Kingdom (UK)), unless otherwise stated. Tissue culture materials and recombinant hEGF were from Life Technologies, Inc. (Paisley, Renfrewshire, UK). Recombinant betacellulin was purchased from R&D Sytstems Europe Ltd. (Abingdon, Oxon, UK); its purity was cited as Ͼ97% as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and silver staining and its biological activity (ED 50 ) against EGF responsive mouse Balb/c 3T3 fibroblasts was in the range of 0.1-0.3 ng/ml. Yeast extract, Bacto-peptone, and yeast nitrogen base without amino acids were purchased from Difco Laboratories Ltd. ( CB3 and CB4 cells (Chinese hamster ovary cells transfected with c-erbB3 and c-erbB4, respectively (20)) were a gift from Prof. Y. Yarden, Weizmann Institute, Rehovot, Israel. Primary cultures of human foreskin fibroblasts (HFFs) were obtained from a surgically removed foreskin and maintained in short term culture for up to 20 passages. Primary cultures of chicken embryo fibroblasts (CEF3) were grown from skin obtained from 14-day-old chicken embryos. The SKOV3 ovarian carcinoma cell line was obtained from ECACC (Porton Down, Salisbury, UK). The yeast ␣-factor secretion vector encoding murine EGF, pWYG9/EGF, was a gift from Dr. J. Clare, Wellcome Research Laboratories, Beckenham, Kent, UK.
Production and Purification of mEGF and mEGF/TGF-␣ 44 -50 -The growth factors were produced in Saccharomyces cerevisiae using the ␣-factor secretion vector pWYG9/EGF (21); in the case of the chimeric growth factor, the carboxyl-terminal 11 residues of mEGF were replaced by the corresponding 7 residues of TGF-␣ ( Fig. 1) by site-directed mutagenesis (22). Recombinant S. cerevisiae was grown in complete defined medium (23) containing 1.34% yeast nitrogen base and 2% raffinose as the carbon source. Growth factor expression was induced for 48 h after the addition of galactose. Growth factors were purified from medium by Sep-Pak extraction, Mono Q anion exchange chromatography, and C 18 reversed phase chromatography as described previously (13). Purified proteins were subjected to laser desorption mass analysis and protein concentrations determined by absorbance at 280 nm where the A 280 of a 1 mg/ml solution of the growth factor in water was calculated as (((nW ϫ 5,650) ϩ (nY ϫ 1,331))/mw) where nW and nY ϭ number of tryptophan and tyrosine residues, respectively, and mw ϭ molecular weight of the peptide as determined by laser desorption mass analysis.
Receptor Binding Assays-Competitive binding assays with 125 I-labeled mEGF were performed using HN5 cells, NR6/HER, or CEF3 as described previously (12,13). Direct binding assays were also performed with NR6, NR6/HER, SKOV3, CB3, and CB4 cells using 125 Ilabeled betacellulin, 125 I-labeled mEGF, and 125 I-labeled mEGF/TGF-␣ 44 -50 (specific activities in the range of 0.74 -2.62 MBq/g). In these assays cells were seeded into 24-well plates and used when 90% confluent; labeled growth factor (200 l/well) was added to the cells in phosphate-buffered saline containing 1% (w/v) bovine serum albumin and 0.02% (w/v) sodium azide (to inhibit receptor internalization) and binding allowed to proceed for 5 h at 22°C. Nonspecific binding was determined using a 100-fold excess of the equivalent unlabeled ligand.
Processing of mEGF and mEGF/TGF-␣ 44 -50 by NR6/HER Cells-Confluent and quiescent NR6/HER cells were prepared in 96-well trays exactly as in mitogenesis assays. 125 I-Labeled EGF or 125 I-labeled mEGF/TGF-␣ 44 -50 (specific activity 2.6 -2.62 MBq/g) was added to the cells in complete mitogenesis assay buffer to give final concentrations in the range of 0.015-1.0 nM. After incubation for 8 h at 37°C, the medium was removed and passed over an Econopac 10DG desalting column to separate intact growth factor from degradation products. Control experiments demonstrated that Ͼ90% of labeled growth factor was voided from the column prior to exposure to cells; furthermore, C 18 reversed phase chromatography confirmed that the labeled EGF or mEGF/TGF-␣ 44 -50 contained in the void volume after exposure to NR6/HER cells for 8 h was chromatographically indistinguishable from that which had not been exposed to cells (in all cases one major (85-90%) and one minor peak (10 -15%) were obtained representing the mono-and di-iodo forms of the labeled growth factor (data not shown)). Thus, label contained in the void volume was taken as a measure of intact growth factor left in medium after exposure to cells.
EGFR Tyrosine Phosphorylation Assays-The ability of mEGF and mEGF/TGF-␣ 44 -50 to induce autophosphorylation of EGFR in HN5 cells was determined as described previously (13). The bands on the Western blots were quantified by densitometry.

RESULTS
Both mEGF and mEGF/TGF-␣ 44 -50 were efficiently expressed by S. cerevisiae and were readily purified to yield single peaks by C 18 reversed phase chromatography. In the case of mEGF, laser desorption mass analysis of the purified protein identified a single peak of mass 5774.8 closely corresponding to the predicted mass (Mϩ1) of mEGF 1-51 (5771.6); the A 280 of a 1 mg/ml solution was then calculated to be 3.1, and this was used to quantify the purified protein. For mEGF/TGF-␣ 44 -50 , the purified protein comprised a major peak of estimated mass 5249.4 in agreement with the predicted mass of the full-length chimera (5250.0). On the basis of these results, the A 280 of a 1 mg/ml solution was calculated to be 1.27. In the chimera preparations, a minor peak of mass 5195.4 was also detected; this was most similar to the predicted mass of a form of mEGF/ TGF-␣ 44 -50 in which the carboxyl-terminal alanine was lost and the methionine at position 21 was oxidized to a sulfoxide (Mϩ1 ϭ 5194.9). Oxidation of Met 21 has been observed previously in preparations of recombinant mEGF, but this modification was reported to have no effect on biological activity (24). Reanalysis of the chimera preparations by analytical C 18 reversed phase chromatography indicated that the contaminant represented about 8% of the total growth factor protein. Comparison of the relative binding affinities of mEGF and mEGF/TGF-␣ 44 -50 showed that the chimera (IC 50 ϭ 98.9 Ϯ 7.5 nM) was much poorer than mEGF (IC 50 ϭ 1.4 Ϯ 0.1 nM) in its ability to compete with 125 I-labeled EGF for binding to human EGFRs expressed by the genetically modified NR6/HER cell (Fig. 2). In similar experiments with the human squamous carcinoma cell line HN5, IC 50 values were 85.9 Ϯ 9.1 nM for the chimera compared with 0.42 Ϯ 0.05 nM for mEGF (data not shown).
When the mitogenic potencies of mEGF and mEGF/TGF-␣ 44 -50 were measured using NR6/HER cells (Fig. 3A), we were surprised to find that the chimera was a superagonist compared with mEGF (EC 50 16.4 Ϯ 1.3 and 120.2 Ϯ 6.2 pM, respectively) i.e. the chimera was approximately 7-fold more potent than mEGF. This shift in mitogenic potency was far in excess of that expected from its receptor binding activity; therefore, we also assayed the mitogenic activity of the chimera using HFFs, which naturally express human EGFRs. In this case, the mitogenic activity of the chimera (EC 50 ϭ 4,150 Ϯ 750 pM versus 56 Ϯ 10 pM for mEGF) closely paralleled its low receptor binding affinity (Fig. 3B).
In view of the anomalous mitogenic response of the NR6/ HER cells, we examined whether this may have arisen from differences in ligand processing. Induction of mitogenesis depends on prolonged exposure of cells to growth factor (25), the magnitude of the response being dependent on the growth factor concentration during an 8-h time window. Although we found that NR6/HER cells consumed more EGF than mEGF/ TGF-␣ 44 -50 in 8 h (Fig. 5A), the difference between the concentrations of EGF and chimera was insufficient to account for the 2-3 log shift in mitogenic response over that predicted from the receptor binding assays. Fig. 5B shows that when the mitogenic response was plotted as a function of the measured intact ligand concentration at 8 h, the chimera was still more potent than mEGF.
To explore further the basis for the superagonistic activity of mEGF/TGF-␣ 44 -50 , we measured its ability to stimulate autophosphorylation of EGFR. When used at equimolar concentrations, the chimera was more potent than mEGF at inducing tyrosine phosphorylation of EGFR in HN5 squamous carcinoma cells, which overexpress the EGFR (Fig. 6, A and B). This indicated that the chimera had the ability to activate EGFR in excess of that anticipated from its receptor binding affinity.
Since betacellulin has been reported to bind to c-erbB4 as well as EGFR (26), we wanted to investigate whether the superagonist activity of the chimera could be attributed to alterations in its binding specificity. This was determined by comparing the activity of 125 I-labeled mEGF/TGF-␣ 44 -50 with that of mEGF and betacellulin in direct binding assays using a range of cell lines expressing different members of the c-erbB family (Fig. 7A). As expected, 125 I-labeled mEGF bound to NR6/HER cells as well as to SKOV3 cells (which express EGFR and overexpress c-erbB2); it did not to bind to the parental NR6 cell line, which does not express EGFRs, nor did it bind to CB3 or CB4 cells that have been transfected to express either c-erbB3 or c-erbB4 and lack EGFR. In contrast, 125 I-labeled betacellulin bound to all five cell lines (albeit weakly to NR6  and CB3 cells), confirming the broader specificity of betacellulin, which recognizes EGFR (present in NR6/HER and SKOV3 cells) as well as c-erbB4 (present in CB4 cells). The superior binding of 125 I-labeled betacellulin to NR6/HER cells compared with 125 I-labeled EGF, presumably reflects expression of c-erbB4 (or c-erbB3) by NR6 cells, as evidenced by binding of 125 I-labeled betacellulin to the parental cell line. The pattern of binding of 125 I-labeled mEGF/TGF-␣ 44 -50 closely paralleled that of mEGF but was reduced in magnitude consistent with its lower relative binding affinity determined in competitive binding assays. Failure of the chimera to bind to c-erbB4 was confirmed in a competitive binding assay with 125 I-labeled betacellulin; even when used at a 1,000-fold higher concentration over unlabeled betacellulin, the mEGF/TGF-␣ 44 -50 failed to compete for binding to c-erbB4 (Fig. 7B). These data suggest that the binding specificity of the chimera is exclusively toward EGFR.

DISCUSSION
Substitution of homologous regions of EGF and TGF-␣ has been used previously to explore the receptor binding regions of EGF and TGF-␣ by exploiting the differential affinities for the chicken EGFR for these growth factors (12,17). In our previous studies where we substituted the B-loop ␤-sheet of TGF-␣ into mEGF, the relative receptor binding affinity of the chimera for the human EGFR was unchanged by comparison with mEGF; however, its affinity for the chicken EGFR was enhanced by 1 order of magnitude. Hence, our finding that substitution of the C-tail of TGF-␣ into mEGF caused a reduction in binding affinity for the human EGFR was unexpected, particularly as this region of the growth factor is not a part of the structural core of the molecule. However, our data are consistent with several previous lines of evidence which suggest that binding of EGF and TGF-␣ is nonidentical.
In a previous study, Kramer et al. (17) produced a panel of hEGF/TGF-␣ chimeras and reported that substitution of the C-tail of TGF-␣ into hEGF was responsible for conferring high affinity binding toward the chicken EGFR. We could not confirm this; in our own experiments using CEFs, we found the activity of mEGF/TGF-␣ 44 -50 in receptor binding assays to be low, comparable to that of wild type mEGF. However, the earlier studies assumed that interchange of segments of EGF and TGF-␣ would have no effect on the receptor binding affinity of the chimeras for the human EGFR, and their quantification depended on standardization in a receptor binding assay using human EGFRs. As our present results demonstrate, such an assumption is invalid for an mEGF/TGF-␣ tail chimera and is likely to be so for an hEGF/TGF-␣ tail chimera. Thus, the previous study overestimated the concentration of the chimera by 80 -200-fold; this would have been sufficient to give the apparent increase in binding affinity when tested on the chicken EGFR.
Although the low affinity of the mEGF/TGF-␣ 44 -50 chimera for the human EGFR might be explained by perturbations in the structure of the growth factor in the free or receptor-bound state, its low affinity is apparently at odds with its mitogenic potency for NR6/HER cells. Previous studies with a Leu 47 3 Val mutant of mEGF have attempted to explain the higher than expected mitogenic potency of the mutant EGF by differences in ligand processing (27). A similar explanation has been proposed to account for differences in cell response between mEGF and TGF-␣ (28). Our data for the mEGF/TGF-␣ 44 -50 chimera could not be explained in this way; even after correction for differences in processing of mEGF or mEGF/TGF-␣ 44 -50 by NR6/HER cells, the chimera was still much more potent than mEGF. This finding is not specific to the mEGF/ TGF-␣ 44 -50 chimera. In other related studies using an hEGF Leu 47 3 Ala mutant (29), we have also found discrepant results for mitogenic potency which could not be attributed to ligand processing. In these studies, mitogenic potency was found to correlate with ligand affinity when tested on cells expressing around 10 4 receptors (e.g. HFFs); however, when tested on cells that express more than 10 5 receptors (e.g. NR6/HER fibroblasts or NRK52E cells) the mutant growth factor was equipotent with EGF. We are currently using mathematical models describing formation of dimeric ligand-receptor complexes to understand the influence of ligands of differing efficacy on cells with a range of EGFR densities and its relationship to the production of activated receptors.
Although we consider the enhanced mitogenic potency of the mEGF/TGF-␣ 44 -50 chimera to be related to the high receptor number of NR6/HER cells, the chimera differed from any other mutant that we have examined previously in that it was more potent than wild type EGF or TGF-␣. The reason for the superagonist activity appears to reside in its greater ability to stimulate the EGFR tyrosine kinase, not just in excess of its receptor binding capacity, but over and above that attained by EGF. This contrasts with other mutant growth factors that we have produced where receptor-ligand affinity paralleled receptor phosphorylation (13).
One possible reason for enhanced activity of the chimera was that it possessed a broader receptor specificity, as has been shown previously for betacellulin. However, the mEGF/TGF-␣ 44 -50 chimera showed no demonstrable binding activity to receptors other than EGFR. We therefore suggest that the  In panel B, the ability of unlabeled mEGF/TGF-␣ 44 -50 (å), mEGF (E), or betacellulin (BTC, ࡗ) to compete with 125 I-labeled betacellulin for binding to c-erbB4 expressed by CB4 cells was measured. Data are the mean of two observations. potent activity of the chimera in phosphorylation assays may be due to unusual receptor binding kinetics that influence the formation and stability of activated receptors (either homodimers or heterodimers); alternatively, the chimeric nature of the ligand may result in production of higher receptor oligomers that are able to maintain the complex in an activated state with a longer half-life. We are currently performing detailed kinetic analyses to examine these possibilities.
In conclusion, we have demonstrated that the properties of chimeric EGFR ligands are not always predictable, as had been assumed previously (17). The mEGF/TGF-␣ 44 -50 chimera exhibited unusually low affinity for EGFR yet was highly mitogenic for some cell lines. The basis for this discrepancy could not be attributed to any single factor. Thus, although differences in ligand depletion made a small contribution to the overall response, it could not provide a complete explanation for the superagonist activity of the chimera. From the present studies and other related work, we propose that the mitogenic effect of low affinity ligands can be enhanced in cells such as NR6/HER which express high EGFR numbers. This we believe to be related to mass action effects on the formation of dimeric receptor species. In the case of the chimera this enhancement translates into a superagonist effect because of its ability to activate the EGFR tyrosine kinase more effectively than EGF. Our comprehensive characterization of this chimeric growth factor emphasizes the complex nature of the processes of ligand recognition, receptor activation, and the ensuing cellular response.