Murine epidermal growth factor peptide (33-42) binds to a YIGSR-specific laminin receptor on both tumor and endothelial cells.

A laminin-antagonist peptide, comprising amino acids 33-42 of murine epidermal growth factor (mEGF-(33-42)), interacts with a breast cancer- and endothelial cell-associated receptor, which is specific for the laminin B1 chain sequence, CDPGYIGSR-NH2 (Lam.B1-(925-933)), and is immunologically similar to a previously described 67-kDa laminin receptor. In whole cell receptor assays, mEGF-(33-42), Lam. B1-(925-933), and laminin all have IC50 values for displacement of 125I-laminin in the range 1-5 nM. Cell attachment to solid-phase laminin is also blocked by all three ligands, but in contrast to the receptor assays, mEGF-(33-42) or Lam.B1-(925-933), while equipotent with each other, were less effective than laminin. The concentrations of the peptides required to produce half-maximal inhibition of attachment were in the range 230-390 nM, but those for laminin were 1000-fold lower, in the range 0.2-0.3 nM. Like laminin, solid-phase mEGF-(33-42) supports cell attachment, and this ability is blocked by anti-67-kDa receptor antibodies. Modeling studies suggest that both peptides present a tyrosyl and an arginyl residue on the same face of a right-handed helical fold with elliptical cross-section.

Laminin, a basement membrane glycoprotein of approximately 900 kDa, has multiple bioactive domains that bind to several integrin and nonintegrin receptors (1). Alterations in cellular interactions with laminin are implicated in the progression of several angiogenic diseases, notably diabetic retinopathy (2) and metastatic cancers (3).

Antibodies
The peptide (PTEDWSAQPATEDWSAAPTA, peptide Pro-20-Ala), corresponding to a linear sequence from the C-terminal end of the human laminin receptor (15), was used as the antigen template. The peptide was synthesized as a multiple antigen presentation derivative (23), using standard Fmoc protocols.
New Zealand White rabbits were immunized subcutaneously with 500 g of antigen in adjuvant (Alum Imject, Pierce, Chester, UK), with boosts of 800 g (at 21-day intervals). Test bleeds were taken 2 days after each boost, and serum was prepared.
The IgG fraction of antiserum was purified using immobilized protein G-Sepharose columns (Pharmacia Biotech, Uppsala, Sweden). The titer and specificity of antiserum as well as IgG fractions were confirmed using an enzyme-linked immunoadsorbant assay, as described (15) except that the multiple antigen presentation antigen was dried onto plates.
The R1 monoclonal antibody against the human EGF receptor was kindly provided by the Hybridoma Development Unit of the Imperial Cancer Research Fund (London, UK). This antibody has previously been shown to specifically block the biological effects of the EGF receptor (24).

Cell Culture
T-47D human breast cancer cells and SK HEP-1 human endothelial cells (25) were obtained from the European Animal Cell Culture Collection (Porton Down, UK), media, and fetal calf serum were from ICN Biomedicals (High Wycombe, UK). T-47D and SK HEP-1 cells were routinely passaged in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum.
Bovine retinal capillary endothelial cells (BRCE) were kindly provided by Dr. Usha Chakravarthy (Department of Opthalmology, Royal Victoria Hospital, Belfast, UK) and were cultured in DMEM containing 10% fetal calf serum.

Radiolabeled Laminin Receptor Assays
Affinity-purified murine laminin (Sigma) was radioiodinated using 125 I-labeled sodium iodide (Amersham, Buckinghamshire, UK) and immobilized chloramine-T (Iodobeads, Pierce), according to the manufacturer's instructions. After the reaction, excess Na 125 I and unincorporated 125 I 2 was separated from the iodinated protein by gel filtration on a GF-5 exclusion column (Pierce). Iodinated laminin fractions were recovered at a specific activity of approximately 1.2 Ci⅐g Ϫ1 protein (864 Ci⅐mmol Ϫ1 ).
Near confluent cultures of T-47D or SK HEP-1 cells were removed from flasks using 0.02% EGTA in calcium-free phosphate-buffered saline (CFS) and passed through a G-25 syringe needle to produce singlecell suspensions. Aliquots of each cell type (10 6 cells⅐ml Ϫ1 ) were dispensed into separate Eppendorf tubes (1 ml each) and pelleted. The cells were then resuspended in 1 ml of ice-cold serum-free DMEM containing 0.1% bovine serum albumin, either laminin or synthetic peptide (at the concentrations indicated), and iodinated laminin at a final 125 I-laminin concentration of 0.1 nM (approximately 50,000 cpm). These mixtures were incubated overnight at 4°C.
The tubes were then microcentrifuged at high speed, and the supernatant was removed. After washing the pellet with 500 l of CFS, the remaining radioactivity was determined using a ␥ radiation counter.
Nonspecific binding was determined by incubating cells with a 1000fold molar excess of unlabeled laminin. All estimations were carried out in triplicate.
IC 50 (concentration of peptide required for 50% inhibition of radioligand binding), EC 50 (effective concentration for 50% inhibition of cell attachment), and their S.D. values were calculated using computerized curve fitting (27).

Laminin Attachment Assay
Non-tissue culture grade 96-well plates (Sterilin Ltd., Middlesex, UK), coated with 2.5 g of murine laminin (Sigma) in 50 l of CFS/well, were air-dried overnight at room temperature. Preliminary experiments indicated that cell attachment was concentration-dependent; maximal binding occurred at a laminin coating of 2.5 g/well (data not shown). After rinsing with CFS (100 l), the plastic was saturated with casein (0.2% in CFS). Plates were incubated at room temperature for 45 min and then washed extensively with CFS (3 ϫ 100 l).
After removal of culture media, cells were detached from monolayers with EGTA (0.02% in CFS) at 37°C. The cells were then centrifuged at low speed, and the pellet was resuspended in serum-free DMEM. Peptides and soluble laminin (at the concentrations indicated) were incubated with aliquots of 10 6 cells (37°C for 1 h) in a volume of 1 ml. 100-l samples of each cell suspension were then added to the precoated multiwell plates and incubated for a further 60 min. Incubation media were removed, and the wells washed with CFS (3 ϫ 100 l) to remove nonadherent cells.
Attached cell numbers were evaluated spectrophotometrically at 620 nm after fixing with 10% formaldehyde and staining with crystal violet (28). All incubations were carried out at least in triplicate.
Briefly, non-tissue culture grade 96-well plates were coated with 100 g/well streptavidin (Sigma) in 0.2 M carbonate buffer (pH 9.8). Following an overnight incubation at 37°C, the wells were washed with CFS (3 ϫ 100 l), and the plastic was blocked with casein (0.2% in CFS). The plates were then incubated at room temperature for 45 min and washed with CFS as previously detailed.
Biotinylated mEGF-(33-42) in CFS was then aliquoted into the wells (1.6 g/well), and the plates were incubated for 3 h at 37°C. After a further block with 0.2% casein, the wells were washed with CFS (3 ϫ 100-l aliquots).
Control cells were prepared as above and preincubated for 1 h at 37°C in peptide-free medium, test cells were preincubated with serial dilutions of either the anti-laminin receptor polyclonal or anti-EGF (R1) receptor monoclonal antibodies, prior to addition to coated wells. Subsequent procedures were as detailed for the laminin attachment assay.
This structure was minimized to generate an initial structure for the dynamics run. Partial atomic charges were calculated using the Pullman method (31), and a distance-dependent dielectric of 4 ϫ r was employed, with an 8-Å cutoff for nonbonded interactions. Asp (P8) and Arg (P9) were assigned charges of Ϫ1 and ϩ1, respectively. During the simulation, the SHAKE algorithm (32) was used for nonpolar hydrogen atoms using a 100-fs interval with a dynamics calculation interval of 1 fs. The temperature was raised from 0 K to 350 K in seven equal steps of 200 fs, with coupling to the temperature bath every 10 fs. The simulation was continued for a further 40 ps, with coordinates, temperatures, and potential energies being recorded at 200-fs intervals. During this simulation, both the "local temperatures" of the backbone and side chain atoms and also the root mean square deviation of the C-␣ atoms from the starting mEGF-like conformation were monitored. A conformer from late in the simulation was chosen as being representative and was refined by minimization to give the final mEGF- (33)(34)(35)(36)(37)(38)(39)(40)(41)(42) fold.
Modeling of Lam.B1-(925-933) Using Molecular Dynamics at 350 K-For the laminin B1 nonapeptide, the mEGF-(33-42) structure (S 0 ) was used as a starting point. The C-terminal cysteine-(S-Acm)-amide residue at P10 was removed, and the exposed arginyl residue (P9) was capped with an amide. Since this laminin nonapeptide, unlike mEGF-(33-42), was not N-terminally acetylated, the N-terminal acetyl capping group was replaced with an ammonium group. The rest of the residues were then "mutated" to reflect the required laminin sequence on a residue-by-residue basis with the proline geometry being adjusted by the software. This structure was minimized without charges, as for mEGF-(33-42), and was subjected to a dynamics simulation, extended to 51.4 ps, with identical conditions to the mEGF-(33-42) decapeptide.
Modeling of Random mEGF- (33)(34)(35)(36)(37)(38)(39)(40)(41)(42) Using Molecular Dynamics at 350 K-In order to generate a possible structure for the random peptide, the same starting molecular conformation (S 0 ) of mEGF-(33-42) was used as for the generation of the previous structures. This was mutated to reflect the desired sequence, and the S-ACM groups were added as before, with manual adjustment of the side chain torsion angles. The starting conformation was then minimized and subjected to the same dynamics simulation as the other two structures. The simulation extended to 41.4 ps.

RESULTS
Laminin Receptor Assay-Laminin receptor assays were carried out on live cells in suspension at 4°C to prevent possible interference from laminin-induced receptor up-regulation or temperature-dependent laminin polymerization (16,33).
Both the Lam.B1-(925-933) and mEGF-(33-42) peptides were found to be approximately equipotent with unlabeled native laminin in their respective abilities to compete with binding of 125 I-labeled laminin to receptors on the human breast cancer cell line T-47D, and the immortalized human endothelial cell line SK HEP-1. For example, Lam.B1-(925-933), mEGF-(33-42), and laminin displaced the radiolabeled ligand from T-47D cells with respective IC 50 values of 1 nM, 3 nM, and 1 nM (see Fig. 1A, Table I). In SK HEP-1 cells, the two synthetic peptides exhibited receptor binding activity close to that of native laminin (IC 50 value range Ϸ 1.5-2 nM, Table I). This similarity in binding affinities was confirmed in receptor assays of early passage BRCE cells, where laminin, Lam.B1-(925-933), and mEGF-(33-42) were equally active with IC 50 values in the range 3-4.3 nM (Table I).
In contrast, both neuromedin B and random mEGF-(33-42) failed to bind to the laminin receptor on any of the three cell types (Fig. 1A, Table I).
Laminin Attachment Assay-Preliminary investigations determined that cell binding to solid-phase laminin becomes increasingly irreversible with time. After addition to the laminin substratum, cells rapidly become resistant to detachment by subsequently added peptides, such that after 15 min of adhesion, additions of mEGF- (33)(34)(35)(36)(37)(38)(39)(40)(41)(42)  In the final assay configuration (described under "Materials and Methods"), treatment of the T-47D human breast cancer cells with solutions of laminin, Lam.B1-(925-933), or mEGF-(33-42) all resulted in inhibition of cell attachment to the laminin-coated wells. However, in contrast to the results obtained in liquid phase laminin receptor assays, binding to solid phase laminin was more resistant to displacement by the two synthetic peptides compared with displacement by native lami-nin (Fig. 1B, Table II). In all cell lines tested, concentrations of soluble laminin required for 50% inhibition of attachment (EC 50 values) were in the range 0. Neither neuromedin B (an unrelated decapeptide included as a negative control) nor the random mEGF-(33-42) peptide had any effect on cell attachment to the laminin substratum when each was tested up to concentrations of 1 mM (Fig. 1B, Table II). Attachment of cells to laminin was found to be blocked by the anti-67-LR antibody (anti-peptide Pro-20-Ala), but was unaf- fected by similar amounts of the R1 anti-EGF receptor antibody (results not shown).
Molecular Modeling-All three of the modeled peptides showed significantly different folds from the native mEGF starting conformation, as measured by the C-␣ root mean square deviation values (Table III). Stable structures were obtained at 350 K for mEGF-(33-42), Lam.B1-(925-933), and the random control peptide after 20, 30, and 30 ps, respectively; "sweep" displays of the ␣-carbon atoms of the individual molecules showed stable folds for the central residues with less stability at the ends of the molecules (Fig. 3). The representative minimized structures (obtained at 22.6, 36.6, and 32.4 ps for mEGF-(33-42), Lam.B1-(925-933), and the random peptide, respectively) are shown in Fig. 4.
In the modeling studies, both active peptides adopted a righthanded helical conformation of similar geometry, making just over a complete turn (Table III, Fig. 4). This was stabilized by a hydrogen bond (not shown) between the P2 carbonyl oxygen and the P5 amide proton in both cases. The N and C termini of the two modeled structures diverged, perhaps as a result of the presence of the bulky S-acetamidomethyl and the blocking acetyl groups in mEGF-(33-42), not present in Lam.B1-(925-933). However, five residue pairs are virtually superimposable in the two structures (Fig. 4, A and B); these include the conserved Gly (P4), Tyr (P5), and Gly (P7), and the nonconserved residue pairs Ile/Ser (P6) and Ser/Asp (P8). In addition, the distance of the aromatic ring of Tyr (P5) from the guanidino group of Arg (P9) is similar in both (Table III).
The orientation of the peptide bonds of residues at P4 -P6 was very similar throughout the simulation for mEGF-(33-42) and Lam.B1-(925-933) (Table III), allowing for the possibility of similar hydrogen bond formation with the receptor in these two molecules. However, the orientation of the peptide linkages between residues at P3 and P4 and between P6 and P7 was not shared.
In contrast to these helical structures, native mEGF has a disulfide bridge between Cys 33 and Cys 42 (P1 and P10 in synthetic mEGF-(33-42)), which causes a fold reversal after Tyr 37 (P5) (30), forcing the C-␣ atoms of the last four residues (P7-P10) to adopt a completely different fold (Fig. 4D). Although the backbone fold of the second through sixth residues was retained and the side chain orientation for Tyr (P5) in this region was also conserved (Table III)
The random peptide adopted a significantly tighter helical conformation, stabilized by three backbone hydrogen bonds between residues at positions P3-P9 in a typical ␣-helical pattern. In addition, unlike the active peptides, this more compact helical conformation continued beyond P6 up to the penultimate residue and resulted in a much smaller Tyr (P5) to Arg (P9) distance (Table III). The termini were drawn together by a hydrogen bond (not shown) between the carbonyl oxygen of the N-terminal acetyl groups and a C-terminal amide proton.

DISCUSSION
The results presented show that the antiangiogenic activities of mEGF-(33-42) are mediated by the YIGSR-specific 67-kDa laminin receptor. Both mEGF-(33-42) and Lam.B1-(925-933) have affinities similar to that of native laminin in whole cell receptor binding assays (Fig. 1A, Table I), results being similar in the three cell lines tested. The IC 50 values for laminin, Lam.B1-(925-933), and mEGF-(33-42) (range Ϸ 1-5 nM) are in good agreement with laminin receptor affinities previously reported for the whole laminin molecule (K d ϭ 1-2 nM) (36,37). To our knowledge, this is the first report of laminin receptor radioligand displacement assays using these synthetic peptides.
Treatment of cells with solutions of laminin, Lam.B1-(925-933), or mEGF-(33-42) inhibits cell attachment to solid-phase laminin substrata (Fig. 1B, Table II). Unlike 4°C receptor assays, however, where all three ligands are found to be equipotent, attachment assays reveal large differences in the effective concentrations required to half-maximally inhibit adhesion to solid-phase laminin. The EC 50 values for the synthetic peptides were approximately 1000-fold higher than the EC 50 values for native laminin, and the peptides' EC 50 values were approximately 100-fold higher than their IC 50 values obtained from the receptor assays (Tables I and II).
In the case of the small synthetic ligands, the concentrations required to inhibit attachment suggest that the laminin receptors must be almost saturated in order to prevent even 50% of maximal attachment. By contrast, laminin (in solution) is relatively more efficient in inhibiting cell binding to solid-phase laminin when compared with its ability to displace 125 I-labeled laminin binding to suspended cells (EC 50 for attachment is approximately 10-fold lower than the IC 50 ). This suggests that soluble laminin need only occupy a minority of the cells' complement of receptors to inhibit binding to solid-phase laminin.
The inhibitory effect of the peptides on cell binding to solidphase laminin is both time-and concentration-dependent. Thus, the peptides are less effective when added simulta-neously with the cells at the time of plating onto the laminin substratum, and cells that are allowed to bind to solid-phase laminin become increasingly resistant to detachment by subsequent treatment with mEGF- (33)(34)(35)(36)(37)(38)(39)(40)(41)(42) or Lam.B1-(925-933). This suggests that binding to solid-phase laminin is not a simple equilibrium binding process but instead becomes increasingly irreversible with time. This may be a consequence of ancillary binding by the attached cell to multiple sites on native laminin that are distinct from the 67-LR binding domain; laminin contains a variety of both integrin and nonintegrin binding sites (1). Initially the process of formation of these supplementary bonds could be interfered with by co-incubation with native laminin in solution, but the small peptides, which only contain the 67-LR binding domain, would be ineffective.
Despite the presence of ancillary binding sites on laminin, saturation of the YIGSR-specific high affinity sites alone (by pretreatment with either Lam.B1-(925-933) or mEGF-(33-42)) nevertheless renders the cells incapable of attaching to solidphase laminin. This suggests that binding to these ancillary sites on the solid-phase laminin molecules is dependent upon, and consequent to, binding to 67-LR. The results also suggest that these supplementary bonds are not made during 4°C incubations with 125 I-laminin in solution.
The fact that the small linear peptides have receptor affinities equal to that of laminin would suggest that all three ligands for the 67-LR adopt a common conformation. Using unconstrained molecular dynamics, we have obtained struc-   (925-933); C, the random peptide (see "Materials and Methods" for modeling procedure) are shown, together with the NMR-defined structure of mEGF, entry 1EPI from the Brookhaven Protein Data Bank (D). In these diagrams of the peptides, the complete side chains are shown, superimposed on the C-␣ trace. Residues of high backbone structural homology between the different peptide molecules are highlighted by coloring the trace in green (P3-P6). For the native mEGF molecule, the region outside the sequence 33-42 is cyan, and the rest is in red except for residues 35-37, which, having a similar fold to the modeled mEGF-(33-42) peptide, are colored green. The disulfide bridge between residues 33 and 42 in the native mEGF is also highlighted by being displayed in yellow. Each molecule is presented with the N terminus to the left.
(925-933) at the N and C termini, but the overall fold is similar. The inactive random peptide, on the other hand, adopts a much tighter and more regular helical fold (Table III).
Unlike mEGF-(33-42), native mEGF showed very weak or no interaction with the 67-LR (data not shown). In contrast to both active synthetic peptides, native mEGF C-loop has a fold reversal after Tyr 37 (P5) (30), which results in the side chains of Ser (P6) through Cys (P10) occupying completely different positions. Thus, Arg (P9) of native mEGF (Fig. 4D) cannot occupy an equivalent position to that found in the synthetic peptides, and Tyr (P5) is significantly less solvent-exposed in native mEGF (35) than it is in the linear peptide structures (Table III, Fig. 4, A and B). Thus, the primary sequence homology is insufficient for receptor recognition; rather, we propose that the consensus sequence GYXGXR must adopt an open right-handed helical fold, presenting the aromatic side chain of a tyrosyl residue (P5) and the guanidino group of the arginyl residue (P9) on the same side of the fold.
Various studies have demonstrated that Lam.B1-(925-933) analogues exert antimetastatic effects via inhibition of attachment to basement membranes. The results presented here show that mEGF-(33-42) has a similar antimetastatic potential. As an antiangiogenic agent it has the advantage of being a pure antagonist of laminin in human and chick cells, whereas Lam.B1-(925-933) acts as an agonist or partial agonist (5). mEGF-(33-42) does not bind to the EGF-receptor and may therefore be a useful probe for studies of the incompletely characterized YIGSR-specific receptor. Further studies to investigate the role of individual residues in the mEGF- (33)(34)(35)(36)(37)(38)(39)(40)(41)(42) sequence are under way.