Solution structure of the mEGF/TGFa 44250 chimeric growth factor Stephen G. Chamberlin 1, *, Lorraine Brennan 2, †, Sarah M. Puddicombe 1 , Donna E. Davies 1 and David L. Turner 2 1 Cancer Research Campaign Medical Oncology Unit, Southampton General Hospital, Southampton, UK; 2 Department of Chemistry, University of Southampton, Highfield, Southampton, UK The solution structure of the growth factor chimera mEGF/ TGFa 44250 has been determined using an extended version of the DYANA procedure for calculating structures from NMR data. The backbone fold and preferred orientation of the domains of the chimera are similar to those found in previous studies of EGF structures, and several H-bonds used as input constraints in those studies were found independently in the chimera. This shows that the modified activity of the chimera does not result from a major structural change. However, the improved precision of the structure presented here allows the origin of some unusual chemical shifts found in all of these compounds to be explained, as well as the results obtained from some site- specific mutants. Further studies of the properties of this chimeric growth factor should help to elucidate the mechanism(s) of hetero- and homodimerization of the c-erbB receptors. Keywords: NMR; EGF structure; growth factor; INDYANA; simulated annealing. Epidermal growth factor (EGF) [1,2] and transforming growth factor alpha (TGFa) [3] are members of a family that also includes heparin-binding EGF-like growth factor [4], amphiregulin [5], betacellulin [6], epiregulin [7] and the heregulins [8,9]. These growth factors play important roles in cell growth and differentiation [10] through their interaction with members of the c-erbB family of receptor tyrosine kinases [11]. They are characterized by a three- looped EGF motif imposed by three highly conserved intramolecular disulfide bonds, as well as by the presence of a number of other conserved residues that have been shown to be required for biological activity [12,13]. EGF and TGFa both show marked specificity for the EGF receptor (EGFR, c-erbB1) with binding resulting in receptor dimerization, activation of the intrinsic receptor tyrosine kinase, and initiation of intracellular signal transduction [14]. Although the EGFR is the primary site of ligand contact, recent studies have shown that the receptor dimers that form as a consequence of this interaction can be either EGFR/EGFR homodimers or EGFR/c-erbB (2,3 or 4) hetero- dimers [15,16]. As a result, most structure –activity studies with EGF and TGFa have failed to address the relative contribution of specific residues to the homodimerization or heterodimerization processes. This omission has been highlighted in recent studies using mEGF/TGFa 44250 ,a 49-amino-acid residue growth factor chimera in which residues 1–42 correspond to the sequence of murine EGF (mEGF 1–42) and residues 43–49 correspond to the C-terminal tail of human TGFa (hTGFa 44–50); this chimera was previously shown to be a superagonist when compared to EGF in mitogenesis assays using NR6/HER fibroblasts even though its relative receptor binding affinity was 1/100th that of EGF [17]. Detailed receptor binding studies confirmed that the chimera binds only weakly to the majority of cell surface EGFRs. However, a subset of sites can be detected for which the chimera retains an affinity similar to that of EGF. As these high affinity sites appear to be due to the formation of heterodimeric EGFR/c-erbB complexes [18,19], it seems likely that there are different ligand requirements for the formation of EGFR homodimers and heterodimers. In order to interpret the mechanism(s) underlying the altered receptor binding properties of mEGF/TGFa 44250 fully, it is essential to establish whether the conformation of the chimera differs from that of EGF. Several growth factors have been studied by NMR previously, because these compounds are not amenable to crystallization [20– 29]; they form looped structures stabilized by three disulfide bridges, with a pronounced antiparallel beta sheet formed in the longest loop. These characteristics present a challenge for solution structure determination, and the relative orientation of the N- and C-terminal regions is particularly difficult to define. The 1 H NMR spectrum of the chimera appears to be broadly similar to those published for EGF, including a broad line of single-proton intensity at about 0.5 p.p.m., hence the conformation is likely to be similar. However, chemical shift calculations based on published structures do not agree well with observed values. A preliminary solution structure of the chimera [30] confirmed the similarity to native forms but left open the question of the precise details of the structure that give rise to the characteristic patterns of chemical shifts. Therefore, the spectra were re-examined and a much larger number of constraints was used to determine a refined structure, which is presented here. Artificial hydrogen-bond constraints are often used to *Present address: Department of Chemistry, Leigh Hall, University of Florida, Gainesville, FL, USA. †Present address: Department of Biochemistry, University College Dublin, Belfield, Dublin 4, Ireland. Correspondence to D. L. Turner, Department of Chemistry, University of Southampton, Highfield, Southampton SO17 1BJ, UK. Fax: 1 44 023 80593781, Tel.: 1 44 0 23 80593330, E-mail: dLt@soton.ac.uk (Received 13 July 2001, revised September 2001, accepted 5 October 2001) Abbreviations: EGF, epidermal growth factor; TGFa, transforming growth factor alpha; upv, upper limit volumes; lov, lower limit volumes. Eur. J. Biochem. 268, 6247–6255 (2001) q FEBS 2001 define secondary structural elements. However, indirect experimental evidence of the existence of H-bonds such as exchange rates and temperature dependence of amide proton chemical shifts presents difficulties because the H-bonds do not necessarily exist simultaneously in a dynamic structure and the acceptor may not be unique [31]. Predefined H-bonds were not used in this study, so that the quality of the structure can be tested in relation to exchange rate data. Interproton distances derived from NOEs may also be impossible to fit to a single set of coordinates if the molecule is conformationally heterogeneous, but a more serious problem arises in the process of selecting NOEs and converting their intensities to distances, as this often involves a degree of subjectivity. Therefore, we have calculated structures using interproton constraints that are derived directly from NOE volumes, with experimental errors used to estimate both upper and lower bounds. This procedure has been applied successfully to rotating frame NOE (ROESY) data [32,33] as well as to NOESY data [34–37], and is implemented here in an extended version of the program DYANA [38], referred to as INDYANA (intensity- DYANA) [36], in which the conversion from the measured NOE intensities to distances is fully automatic. The use of minimum-distance constraints yields a large increase in the amount of experimental information because upper limits for NOE intensities may be obtained even in the presence of degenerate chemical shifts or overlapping cross peaks. Furthermore, consistency with experimental data is improved because structures based on upper-limit distances alone allow protons to come into van der Waals contact with each other even if the experimental spectra show clearly that there is no NOE between them. MATERIALS AND METHODS Growth factor production The chimera, mEGF/TGFa 44250 , and wild-type mEGF were produced in Pischia pastoris using the pPIC9 vector from Invitrogen BV, Leek, the Netherlands. Following growth to mid-log phase in buffered minimal medium containing 1% (v/v) glycerol as a noninducible carbon source, cells were concentrated 10-fold before induction by daily addition of 0.5% (v/v) methanol for 3 days. After purification and characterization as previously described [17], this protocol yielded 38 mg of growth factor per litre of medium. NMR Experiments A3m M solution in 90%H 2 O/10% 2 H 2 O at pH 3 was used for the NMR experiments. Spectra were recorded on a Varian VXR500 spectrometer operating at 499.84 MHz. A NOESY spectrum [39] was recorded with a 100-ms mixing time at 20 8C, with 4096 points and a spectral width of 7 kHz for each transient, and 1024 increments with TPPI [40] to give a spectral width of 14 kHz in the second dimension. A TOCSY spectrum [41] with 60 ms of spin lock and a DQF-COSY spectrum were recorded under the same conditions. Determination of volume constraints Each NOESY cross peak, or cluster of overlapping peaks, was integrated using the program XEASY [42] together with areas of baseline either side of the peak in the F 1 dimension to correct any offset. Additional volumes were measured at positions predicted on the basis of preliminary calculated structures, even if there was no visible cross peak. The upper limit volumes (upv ) and lower limit volumes (lov ) were estimated as described previously [36], with a minimum uncertainty defined as three times the standard deviation of all of the baseline integrals, which is roughly equivalent to the intensity of the smallest recognizable peaks. Several of the weakest peaks yielded a negative lov that gave no meaningful upper distance limit. The lov was also discarded if the cross peak comprised contributions from degenerate protons; this is less restrictive than the ‘sum’ function in X-PLOR [43], but it involves no additional complexity in computation. Constraints involving resolved prochiral protons are handled by INDYANA in a manner similar to the original DYANA program in the absence of a stereospecific assign- ment, except that the fixed distances between the protons and pseudoatoms are included in the target function calculation together with the converted volume. Degenerate methylene or isopropyl proton signals are treated similarly, but the single available upv applies to both, and the lov, used as a basis for a fixed distance offset, is set to one half of the measured lov. Aromatic protons, such as Tyr Hd and H1, are a special case because, although rapid ring flips usually render the protons equivalent, the fixed distance between them is large enough for many NOEs to be assigned to one or other side of the ring. Each NOE may be ‘pseudo- stereospecifically assigned’ individually [44,45], which is achieved by having two sets of proton labels for each ring; one set that is treated by the program as stereo pairs and another that is recognized as unique. This procedure provides direct information about ring orientations, which is not possible if constraints are applied to pseudoatoms on the C 2 axis of a ring. This is in accordance with the observation that aromatic groups are usually well defined in crystal structures even when NMR shows that they undergo rapid 1808 flips. Automatic distance calibration A relation of the form r ¼ k/ n p V was used to convert NOE volumes into interproton distances. Different scaling factors, k, were used for different classes of proton: NOEs between methyl groups, a methyl and a single proton, or two single protons were treated separately. A fourth scaling factor was used for NOEs between amide protons and other single protons. This involves a fundamental change to the DYANA program as the conversion from volumes to distances occurs at the level of target function evaluation. The scaling factors are ‘non-Cartesian’ parameters of the fit that are optimized simultaneously with torsion angles, either by conjugate gradient minimization, or by simulated annealing. These additional variables can be thought of as nongeo- metric dimensions and they are assigned a weight, analo- gous to mass, to scale their rate of change for the Newtonian dynamics. In practice, the reduction in computational speed caused by the additional parameters is offset by an improvement in convergence that appears to result from the implicit flexibility of the distance constraints. As each calculated structure is defined by its own set of scaling factors in addition to the set of torsion angles that defines its 6248 S. G. Chamberlin et al.(Eur. J. Biochem. 268) q FEBS 2001 geometry, the set of solutions reflects any uncertainty in the calibration as well as alternative fits to the set of distances. The main advantages of the procedure are simple: the details of the calibration are precisely and unambiguously defined, it is fully automatic, and the family of calculated structures cannot be biased by predetermined calibration constants. Allowance for spin diffusion The monotonic relationship between NOE intensity and interproton distance may be spoiled by local variations in correlation time, fluctuations in conformation (including aromatic ring flips), or by spin diffusion through networks of closely spaced protons [46]. These complications are interrelated, but the NOEs expected for a given structure can be calculated approximately from the exponential of the matrix of theoretical cross relaxation rates. These values may be replaced by scaled experimental values and the logarithm of the matrix may be taken to obtain distances that take account of spin diffusion [47,48]. Typically, however, such calculations generate some nonphysical negative cross relaxation rates as a consequence of the failure of the approximation of a rigid molecule and because of inaccuracies in the starting structure. Although mathemat- ical convergence may be achieved by iterative calculations, the accuracy of the distances obtained remains uncertain. Therefore, we use relaxation matrix calculations simply to estimate the errors that might be induced by spin diffusion and then soften all distance constraints accordingly. In effect, this allows for the uncertainty in converting NOE volumes into distances that remains despite having optimized the calibration curve. The relaxation matrix calculations use the average inverse sixth power of interproton distances from an ensemble of structures and a single correlation time, optimized to fit the set of measured NOEs, with fast methyl group rotation and rapid rings flips taken into account [49–51]. The calculated values were replaced by scaled NOE intensities and the rmsd of the ratio between the distances found after back trans- formation and the maximum or minimum distances obtained by automatic calibration in INDYANA was then used to set a parameter for loosening distance constraints to ensure that the scaling factors, that include the variable effects of spin diffusion, do not result in excessively tight constraints. Torsion angle constraints Scalar couplings between NH and CaH protons ( 3 J HNHa ) were measured from one-dimensional spectra and the DQF-COSY spectrum and converted into constraints for the backbone torsion angle, f,usingtheprogram HABAS together with preliminary structures [52]. Structure calculation Disulfide bridges were built by modification of the standard cysteine residue in the DYANA library [53]. A pseudoatom was included instead of the HG atom and a bridge was formed by superimposition of the pseudoatoms on the SG atoms of the other Cys residue, with an upper distance limit of 0.01 nm and a weight 10 times that of other constraints. Covalent links declared in the sequence file cause the DYANA program to ignore van der Waals repulsion across the bridge. A flexible proline residue was built in a similar fashion by modification of the standard Pro residue, in which the ring is held flat. The CB–CG bond was removed and the ring closed by superimposition of three pseudoatoms with coordinates identical to CG, CD, QG with five new torsion angles. The contribution of these fixed upper limits to the target function was defined as DYANA type 2 [53] to avoid excessive weighting caused by the short distances; all other constraints were of the standard type 1. It is worth noting that the modified program also accepts fixed distances for interproton constraints and will therefore operate in the same manner as DYANA if volume constraints are not used. Stereospecific and pseudo-stereospecific assignments were made with respect to preliminary calculated structures using the program GLOMSA [52], modified to accept both upper and lower volume constraints. Structures were calculated from random starting points, following the standard annealing protocol defined in the program DYANA. Structure evaluation Cross validation of experimental constraints by random exclusion of subsets is an effective technique for evaluating structures, and a similar insight is provided by testing structures against alternative sources of information. Ramachandran plots are used widely, which is appropriate if the steric repulsions used in the structure calculation are soft, as in the quadratic term used in DYANA, rather than a force field that effectively constrains the structure to the most favoured backbone torsion angles. Predicted H-bonds in structures may also be compared with experimental evidence for the involvement of amide protons in H-bonds, but only if there were no such constraints used in the structure calculation. The detailed quality of agreement between the structure and the input constraints is also an important indication of consistency with experimental data if, as with the measurement of additional maximum NOE volumes used here, the comparison is made with the complete set of volumes and not merely with those which are measured in the first instance. Constraint violations were examined using the program DYANA. Superposition of the family of structures, calculation of the rmsd of atomic coordinates, and preparation of diagrams used MOLMOL [54]. Ramachandran plots were obtained using PRO- CHECK -NMR [55], and the optimal pattern of H-bonds was calculated using WHAT IF [56]. Finally, chemical shifts were calculated using the program TOTAL [57]. RESULTS AND DISCUSSION Assignment The complete sequence specific assignment of the chimera was carried out by identification of the spin systems using TOCSY and COSY experiments followed by the use of sequential CaH-NH NOEs [58]. The sequential assignment was interrupted by the presence of two proline residues, in which cases the following NOEs were observed: Ha/Hb of the previous residue to the Pro H* and NH of the following residue to the Pro Ha/Hb. The assignments have been deposited at the BioMagResBank, with accession number 5120. The chemical shifts are consistent with the 1 H q FEBS 2001 Solution structure of mEGF/TGFa44–50 (Eur. J. Biochem. 268) 6249 assignments reported for wild-type mEGF at pH 3.1 [20,21], pH 2.0 [22] and pH 6.8 [23] at 28 C. Structure calculations From 1174 assigned and integrated cross peaks input into INDYANA, 561 lower and 793 upper volume constraints were obtained. After adjustment for missing stereospecific assignments and elimination of redundant constraints, this yielded 955 lower and 1127 upper volume constraints, an average of 42.5 constraints per amino acid (19.5 lower volume limits and 23.0 upper volume limits), which is summarized in Fig. 1 and Table 1. In addition, 28 constraints for backbone torsion angles were obtained from 3 J HNHa coupling constants. Structures were calculated using the standard DYANA protocol for simulated annealing with torsion angle dynamics, with four additional dynamic variables for the conversion of volumes to distances. Preliminary structures were checked for possible stereo- specific assignments and also for unconstrained short interproton distances. The volumes at the positions of the cross peaks predicted for short distances were measured where possible. Out of 59 nondegenerate methylene and isopropyl groups, 44% were stereospecifically assigned, and pseudo-stereospecific assignments were made for 43% of the NOEs to fast flipping aromatic rings. Relaxation matrix calculations with preliminary structures gave corrected distances that violated the constraints obtained from INDYANA with an rmsd of 5.4%, hence, all converted distances were softened by 6% in the final calculations. The function r ¼ k/ 4 p V was used for converting volumes to distances was chosen after comparing results for r 24 and r 26 . The family of 10 structures with the lowest target functions obtained from 50 random starting structures was chosen to represent the chimera in solution, and is shown in Fig. 2. The atomic coordinates and constraints have been deposited at the RCSB Protein Data Bank, with accession code 1gk5. Structure evaluation The global rmsd per residue for backbone and heavy atoms with respect to the mean of the family of structures is plotted in Fig. 1. Backbone torsion angles were analysed using Ramachandran plots, excluding Gly, Pro, and terminal resi- dues, and the results are summarized in Table 2. Hydrogen bonds were identified using the program WHAT IF with default parameters [56]. A total of 14 H-bonds between backbone atoms were detected in 50% or more of the structures. These include all of the amide protons found by Montelione et al. to have exchange rates lower than 2.5 Â 10 24 : min 21 [20], with the exception of Val34 NH, for which no H-bond was found, and Leu15 NH, that formed H-bonds to Arg41 CO in four out of the 10 structures. In addition, a bifurcated H-bond was found involving Asp27 and Ser28 NH with Ile23 CO. An H-bond was also found between Asp46 NH and Gly36 CO in all structures. Significant H-bonds were found for amide protons in sidechains: in particular, bonding of Asn16 NdH to Cys42 CO, and Arg41 N1H to Tyr13 CO appeared in all 10 structures, and Arg41 NhH was bonded to Gly12 CO in eight out of 10 structures. The H-bonds were not used as constraints in the calculation, but these structures predict that the sidechain hydrogen bonds, together with Leu15 NH – Arg41 CO, stabilize the relative orientation of the C-terminal loop. This would explain the dramatic changes in structure and activity caused by mutating residue 41 [28]. The calculated structures also explain the large secondary structural shifts of Asn16 bCH 2 and Arg41 gCH 2 , which Fig. 1. NMR Data. Top: the number of meaningful NOE-derived constraints for each residue used in the calculations. White represents intraresidual constraints (Di ¼ 0), light grey sequential (Di ¼ 1), grey and black represent medium (Di , 5) and long-range (Di $ 5) constraints. Both lower and upper volume limits are included. Bottom: average rmsd (A ˚ ) for the backbone (A) and heavy atoms (B) with respect to the mean structure. The superimposition was performed for residues 5–47. Fig. 2. Stereo view of the of the 10 best chimera structures, superimposed using the backbones of residues 5–47. 6250 S. G. Chamberlin et al.(Eur. J. Biochem. 268) q FEBS 2001 imply relatively rigid sidechain conformations as well as the proximity of a strongly anisotropic or charged group. In fact both residues are close to aromatic groups, Tyr37 and Tyr13, respectively, as illustrated in Fig. 3. Smaller deviations in chemical shift are found for the Pro7 dCH 2 protons, at 3.4 and 2.7 p.p.m., that do not agree well with calculations based on the structure, 3.9 (0.3) and 3.9 (0.4) p.p.m., although they are within three standard deviations (given in parenthesis). In this case, the standard deviation is large because aromatic ring of Tyr29 is the main contributor to the shifts of Pro7, and its orientation is not well defined. The observed and calculated shifts, excluding amide protons, are compared in Fig. 4. Comparison with mEGF structures Two different groups have deposited solution structures of murine EGF at low pH in the RCSB Protein Data Bank (http://www.rcsb.com), 3egf [21] and 1eph [23] being the most recent. The family of structures in 3egf was computed using 644 distance constraints, 32 dihedral angle constraints and constraints for nine hydrogen bonds. Those of 1eph were computed using 355 distance constraints, 24 torsion angle constraints and constraints for eight hydrogen bonds. Unsurprisingly, PROCHECK-NMR [55] reports a higher effective resolution for the 3egf family than for 1eph, and the resolution of the chimeric structure presented here is similar to that of 3egf, despite the absence of H-bond constraints. Although this is a measure of precision rather than accuracy, it correlates with the comparison of observed and calculated shifts shown in Fig. 4. The standard deviations of the calculations are 0.36 p.p.m. for 1eph, 0.32 p.p.m. for 3egf, and 0.30 p.p.m. for the chimera, averaged over the 10 best structures in each case. Two structures from 3egf and 1eph, the first from each family, superimpose with a backbone rmsd of 0.53 nm for residues 1–53, which is uninformative. The mEGF structure has been presented previously as two structural domains, the N-terminal domain (residues Asn 1–Cys33) and the C-terminal domain (residues Asn 32–Leu 47). The last few residues form a poorly defined tail. The b sheet of the N-terminal domain for 10 EGF and EGF-like structures determined by solution NMR methods have been super- imposed by Tejero et al. [24]. These 10 structures exhibited a wide range of relative orientations of the two subdomains. It has also been concluded from 1 H linewidth studies, 15 N and 13 C relaxation rates, and molecular dynamics simulations that multiple orientations of the two subdomains may be in dynamic equilibrium in any one molecule [21,25–27]. Therefore, rmsd values have been presented separately for the entire molecule (residues 1–50), the entire molecule minus the C-terminal tail (residues 1–47), the N-terminal subdomain (residues 1–33), the C-terminal subdomain (residues 32–47) and the core (residues 2–6, 18–23, 26– 38 and 42 –45). Superimposition of the backbones of 3egf and 1eph gave an rmsd of 0.169 nm for residues 6–33 and 0.165 nm for residues 32–47. The overall best structure from the chimera family was superimposed with a structure from each of the previously reported families of structures for mEGF; the results are given in Table 3. The difference between the backbone of the chimera and the EGF structure is no greater than that found for all the EGF structures, leading to the conclusion that the increased activity of the chimera is not as a result of a structural change. Significantly, a bond between the NH of residue 15 and the carbonyl of residues 41 was identified in more than 40% of the structures in the family of the chimera (Table 4). This NH was reported to exchange slowly in hEGF (human EGF) and mEGF [20,29] and the formation of a Leu15 NH–Arg41 O-H-bond in calculations using a force field effectively defines the relative orientation of the N- and C-terminal domains [29]. As H-bond constraints were not used in the calculation of the chimera structure, this provides strong evidence that the overall structure is unchanged by the modification of the sequence. Given that the solution structure of mEGF/TGFa 44250 was found to be similar to that of other EGF structures, it is unlikely that the low affinity of the chimera for binding to the majority of cell surface EGFRs [18] results from gross structural changes in the unbound growth factor. The importance of the ligand C-tail for EGFR binding was demonstrated in mutagenesis studies on the conserved Table 1. Summary of relevant constraints used for calculating the structure of the chimera. The number of individual NOEs, before adjustment for nonstereospecifically assigned protons, is given in parenthesis. Note that the lower limit of the NOE volume determines the upper distance limit (upl ) and the upper volume determines the lower distance (lol ). Constraint type Lower volume (upl ) Upper volume (lol ) Intraresidue (Di ¼ 0) 355 (222) 310 (236) Sequential (Di ¼ 1) 215 (140) 254 (192) Medium-range (2 # Di # 4) 153 (82) 206 (134) Long-range (5 # Di) 232 (117) 357 (231) Total per residue 19.5 (11.4) 23.0 (16.2) Torsion angles 28 Table 2. Statistics for the family of 10 chimera structures. Note that violations are calculated after conversion of NOE volumes into distance limits, i.e. from lov to upl and from upv to lol. Target function range 0.47–0.77 A ˚ 2 (62%) Scaling factors (standard deviation) Proton–proton 89.3 (0.3) Amide proton–proton 97.0 (0.3) Proton–methyl 106.0 (0.5) Methyl–methyl 122.5 (3.0) Backbone rmsd (6–47 N, Ca, CO) 0.47 A ˚ Heavy atom rmsd 0.81 A ˚ Average sum (maximum) of upl violations 3.4 (0.17) A ˚ Average sum (maximum) of lol violations 2.8 (0.26) A ˚ Average maximum van der Waals violation 0.09 A ˚ Consistent violations . 0.2 A ˚ 0 Residues in Ramachandran regions (%) Most favoured 62.1 Allowed 35.1 Generously allowed 2.8 Disallowed 0.0 q FEBS 2001 Solution structure of mEGF/TGFa44–50 (Eur. J. Biochem. 268) 6251 leucine residue in both EGF (L47) [59] and TGFa (L48) [60] and has been confirmed using transferred NOE enhancement data for titration of TGFa with the EGFR [61]. As the conserved leucine residue was not changed in the chimera, this suggests that nonconserved residues in the C-tail of the growth factor also make important contri- butions to receptor binding and that these are context dependent, i.e. the altered environment of the TGFa C-tail relative to the main murine EGF structural motif may disrupt interactions required to stabilize the receptor bound form of the ligand. Consistent with this proposal is the report that the C-tail, which is very flexible in the nonbound state and poorly defined by NMR, has restricted mobility upon receptor binding [62]. Interestingly, the secondary structural shifts of the bridging sidechains, Asn16 and Arg41, are larger in the chimera than in EGF, suggesting that the C-loop of the unbound chimera has reduced mobility with respect to the rest of the structure. Further studies of the receptor bound forms of the chimera and other related ligands will be necessary to define the nature of the interactions leading to receptor recognition and dimerization. CONCLUSIONS Because of the limited supply of the chimeric growth factor, the protein concentration used in this work was about one half of that used in determining the structures 1eph and 3egf [21,23]. Apart from that, the instrumentation and experi- mental methods were similar. The difference in approach lies in the methodology for structure calculation: hydrogen bond constraints were not used in this work and NOE Fig. 3. Ribbon diagram of the secondary structure in residues 5–47 of the overall best structure of the chimera. The three disulfide bridges are also shown, together with the sidechains of Asn16 and Arg41, which form H-bonds to backbone CO groups. The stability of these bonds is implied by large secondary structural shifts of the bCH 2 protons, shown as spheres, which are generated by the rings of Tyr13 and Tyr37. Table 3. Rmsd (A ˚ ) for superposition of backbone atoms in the structures 1eph, 3egf, and the chimera. Values above the diagonal are for residues 6 –33 and those below for 32–47. 1eph 3egf Chimera 1eph – 1.69 2.01 3egf 1.65 – 1.69 Chimera 1.28 1.99 – Fig. 4. Calculated vs. observed secondary structural shifts for murine EGF (1eph [23] and 3egf [21]) and the EGF/TGFa chimera. In each case, chemical shifts were calculated using the program TOTAL [57] and averaged over the 10 best structures. The limits of the estimated accuracy of the calculation are indicated by dashed lines. 6252 S. G. Chamberlin et al.(Eur. J. Biochem. 268) q FEBS 2001 volumes (intensities), not precalibrated distances, were used as input. This work made use of a simple extension of the DYANA procedure [38] for calculating structures from NMR data that is based firmly on experiment, including error bars, and minimizes the possibilities for subjective influence. The protocol ensures that the maximum amount of information is extracted from the spectra and therefore it is possible to account for the effects of spin diffusion simply by loosening constraints, without significant loss of precision. The struc- tures of the chimera were calculated with no electrostatic energy terms. Hence, the accuracy of the solutions obtained here is indicated by the well defined hydrogen bonds found. Several of these were identified in previous studies of EGF, and the backbone fold of the chimera is clearly similar to those of the EGF structures. The presence of a backbone hydrogen bond from the N- to C- terminal domain, which was identified in the hEGF structure [29], together with those of the sidechains, is particularly significant. The chemical shifts calculated for the chimera clearly support the sidechain orientations, and the pattern of shifts is similar to that found in native EGF. This shows that the relative orientation of the domains is unchanged and the modified activity of the chimera does not result from any major structural alteration. The properties of this chimeric growth factor should therefore help to elucidate the importance of heterodimerization and homodimerization of the EGF receptors. REFERENCES 1. Cohen, S. (1962) Isolation of mouse submaxillary gland protein accelerating incisor eruption and eyelid opening in the new born animal. J. Biol. Chem. 237, 1555–1562. 2. Gregory, H. (1975) Isolation and structure of urogastrone and its relationship to epidermal growth factor. Nature 257, 325–327. 3. DeLarco, J.E., Reynolds, R., Carlberg, K., Engle, C. & Todaro, G.J. (1980) Sarcoma growth factor from mouse sarcoma virus- transformed cells. Purification by binding and elution from epidermal growth factor receptor rich cells. J. Biol. Chem. 255, 3685–3690. 4. Higashiyama, S., Lau, K., Besner, G., Abraham, J.A. & Klagsbrun, M. (1991) A heparin-binding EGF-like growth factor secreted by macrophage-like cells is related to EGF. Science 251, 936–939. 5. Shoyab, M., McDonald, V.L., Bradley, J.G. & Todaro, G.J. (1988) Amphiregulin: a bifunctional growth-modulating glycoprotein produced by the phorbol 12-myristate 13-acetate-treated human breast adenocarcinoma cell line MCF-7. Proc. Natl Acad. Sci. USA 85, 6528– 6532. 6. Shing, Y., Christofori, G., Hanahan, D., Ono, Y., Sasada, R., Igarashi, K. & Folkman, J. (1993) Betacellulin: a mitogen from pancreatic b-cell tumors. Science 259, 1604–1607. 7. Toyoda, H., Komurasaki, T., Uchida, D., Takayama, Y., Isobe, T., Okuyama, T. & Hanada, K. (1995) Epiregulin. A novel epidermal growth factor with mitogenic activity for rat primary hepatocytes. J. Biol. Chem. 270, 7495 –7500. 8. Wen, D., Peles, E., Cupples, R., Suggs, S.V., Bacus, S.S., Luo, Y., Trail, G., Hu, S., Silbiger, S.M., Levy, R.B., Koski, R.A., Lu, H.S. & Yarden, Y. (1992) Neu differentiation factor: a transmembrane glycoprotein containing an EGF domain and an immunoglobulin homology unit. Cell 69, 559–572. 9. Marchionni, M.A., Goodearl, A.D.J., Chen, M.S., Bermingham McDonogh, O., Kirk, C., Hendricks, M., Danehy, F., Misumi, D., Sudhalter, J., Kobayashi, K., Wroblewski, D., Lynch, C., Baldassare, M., Hiles, I., Davis, J.B., Hsuan, J.J., Totty, N.F., Otsu, M., McBurney, R.N., Waterfield, M.D., Stroobant, P. & Gwynne, D. (1993) Glial growth factors are alternatively spliced erbB2 ligands expressed in the nervous system. Nature 362, 312–318. 10. Carpenter, G. & Wahl, M. (1991) The epidermal growth factor family. In Peptide Growth Factors and Their Receptors (Sporn, M.B. &Roberts, A.B., eds), pp. 69–171. Springer-Verlag, New York. 11. Alroy, I. & Yarden, Y. (1997) The erbB signaling network in embryogenesis and oncogenesis: signal diversification through combinatorial ligand–receptor interactions. FEBS Lett. 410, 83–86. 12. Campion, S.R. & Niyogi, S.K. (1994) Interaction of the epidermal growth factor with its receptor. Prog. Nucleic Acid Res. Mol. Biol. 49, 353– 383. 13. Groenen, L.C., Nice, E.C. & Burgess, A.W. (1994) Structure– function relationships for the EGF/TGFa family of mitogens. Growth Factors 11, 235–257. 14. Lemmon, M.A. & Schlessinger, J. (1994) Regulation of signal transduction and signal diversity by receptor oligomerization. Trends Biochem. Sci. 19, 459– 463. 15. Beerli, R.R. & Hynes, N.E. (1996) Epidermal growth factor related peptides activate distinct subsets of ErbB receptors and differ in their biological activities. J. Biol. Chem. 271, 6071–6076. 16. Tzahar, E., Waterman, H., Chen, X., Levkowitz, G., Karunagaran, D., Lavi, S., Ratzkin, B.J. & Yarden, Y. (1996) A hierarchical network of interreceptor interactions determines signal transduction by neu differentiation factor/neuregulin and epidermal growth factor. Mol. Cell. Biol. 16, 5276–5287. 17. Puddicombe, S.M., Wood, L., Chamberlin, S.G. & Davies, D.E. (1996) The interaction of an epidermal growth factor/transforming growth factor a tail chimera with the human epidermal growth Table 4. Optimal hydrogen bonds in the family of chimera structures, found using the program WHAT-IF [56]. The total score for the 10 structures is given, which should have a value of 10.0 for a ‘perfect’ hydrogen bond. Leu15 NH Arg41 O 0.796 Asn16 Nd2 Cys42 O 6.013 Val19 NH Asn32 O 4.177 Met21 NH Thr30 O 3.958 His22 Nd1 Ser28 O 2.625 Ile23 NH Ser28 O 2.262 Leu26 NH Ser25 Og 2.560 Asp27 NH Ile23 O 2.594 Ser28 NH Ile23 O 2.901 Tyr29 NH Ser28 Og 2.939 Thr30 NH Met21 O 3.027 Asn32 NH Val19 O 6.594 Cys33 NH Cys31 O 0.247 Tyr37 NH Val34 O 5.443 Ser38 NH His44 O 3.853 Gly39 NH Ser38 Og 0.377 Gly39 NH His44 O 0.793 Arg41 Nh2 Gly12 O 4.174 Arg41 N1 Tyr13 O 8.107 Arg41 Nh2 Tyr13 O 1.646 Cys42 NH Gly39 O 3.777 Glu43 NH Ser38 O 0.789 Glu43 NH Gly39 O 0.731 Glu43 NH Asp40 O 2.947 His44 NH Ser38 O 1.774 His44 NH Gly39 O 0.557 Ala45 NH Glu43 O 2.645 Asp46 NH Gly36 O 3.909 q FEBS 2001 Solution structure of mEGF/TGFa44–50 (Eur. J. Biochem. 268) 6253 factor receptor reveals unexpected complexities. J. Biol. Chem. 271, 30392–30397. 18. Neelam, B., Richter, A., Chamberlin, S.G., Puddicombe, S.M., Wood, L., Murray, M.B., Nandagopal, K. & Davies, D.E. (1998) Structure–function studies of ligand-induced epidermal growth factor receptor dimerization. Biochemistry 37, 4884–4891. 19. Chamberlin, S.G. & Davies, D.E. (1998) A unified model of c-erbB receptor homo- and heterodimerisation. Biochim. Biophys. Acta 1384, 223–232. 20. Montelione, G.T., Wu ¨ thrich,K.&Scheraga,H.A.(1988) Sequence-specific 1 H NMR assignments and identification of slowly exchanging amide protons in murine epidermal growth factor. Biochemistry 27, 2235–2243. 21. Montelione, G.T., Wu ¨ thrich, K., Burgess, A.W., Nice, E.C., Wagner, G., Gibson, K.D. & Scheraga, H.A. (1992) Solution structure of the Murine epidermal growth factor determined by NMR spectroscopy and refined by energy minimization with restraints. Biochemistry 31, 236– 249. 22. Kohda, D. & Inagaki, F. (1988) Complete sequence-specific 1 H Nuclear Magnetic resonance assignments for mouse epidermal growth factor. J. Biochem. (Tokyo) 103, 554–571. 23. Kohda, D. & Inagaki, F. (1992) Three-dimensional nuclear magnetic resonance structures of mouse epidermal growth factor in acidic and physiological pH solutions. Biochemistry 31, 11928–11939. 24. Tejero, R., Bassolinp-Klimas, D., Bruccoleri, R.E. & Montelione, G. (1996) Simulated annealing with restrained molecular dynamics using COGEN: Energy refinement of the NMR solution structures of epidermal and type-a transforming growth factors. Protein Sci. 5, 578– 592. 25. Li, Y.C. & Montelione, G.T. (1995) Human type-a transforming growth factor undergoes slow conformational exchange between multiple backbone conformations as characterized by nitrogen-15 relaxation measurements. Biochemistry 34, 2408– 2423. 26. Celda, B., Biamonti, C., Arnau, M.J., Tejero, R. & Montelione, G.T. (1995) Combined use of 13 C chemical shift and 1 H a - 13 C a heteronuclear NOE data in monitoring a protein NMR structure refinement. J. Biomol. NMR 5, 161–172. 27. Fadel, A.R., Jin, D.Q., Montelione, G.T. & Levy, R.M. (1995) Crankshaft motions of the polypeptide backbone in molecular dynamics simulations of human type-a transforming growth-factor. J. Biomol. NMR 6, 221 –226. 28. Hommel, U., Dudgeon, T.J., Fallon, A., Edwards, R.M. & Campbell, I.D. (1991) Structure–function-relationships in human epidermal growth-factor studied by site-directed mutagenesis and 1 H-NMR. Biochemistry 30, 8891– 8898. 29. Campbell, I.D., Baron, M., Cooke, R.M., Dudgeon, T.J., Fallon, A., Harvey, T.S. & Tappin, M.J. (1990) Structure–function relation- ships in epidermal growth factor (EGF) and transforming growth factor-alpha (TGFa). Biochem. Pharmacol. 40, 35– 40. 30. Brennan, L. (1998) NMR of peptides and proteins. PhD Thesis, University of Southampton, UK. 31. Nilges, M. & O’Donoghue, S.I. (1998) Ambiguous NOEs and automated NOE assignment. Prog. NMR Spectrosc. 32, 107–139. 32. Turner, D.L. (1994) A method for obtaining precise initial structures from rotating frame nuclear Overhauser enhancements and its application to Cyclosporin A. J. Magn. Reson. A107, 239–242. 33. Rees, N.H., Penfold, D.J., Rowe, M.E., Chowdhry, B.Z., Cole, S.C.J., Samuels, R.I. & Turner, D.L. (1996) NMR studies of the conformation of destruxin A in water and in acetonitrile. Magnetic Resonance Chem 34, 237–241. 34. Turner, D.L. (1995) The conformation of the monensin A-sodium complex in solution determined from self-consistent NOE distance constraints. J. Magn. Reson. B108, 137–142. 35. Messias, A.C., Kastrau, D.H.W., Costa, H.S., LeGall, J., Turner, D.L., Santos, H. & Xavier, A.V. (1998) Solution structure of Desulfovibrio vulgaris (Hildenborough) ferrocytochrome c 3 : Structural basis for functional cooperativity. J. Mol. Biol. 281, 719–739. 36. Brennan, L., Turner, D.L., Messias, A.C., Teodoro, M.L., LeGall, J., Santos, H. & Xavier, A.V. (2000) Structural basis for the network of functional cooperativities in cytochrome c 3 from Desulfovibrio gigas: solution structures of the oxidised and reduced states. J. Mol. Biol. 298, 61 –82. 37. Brennan, L., Turner, D.L., Fareleira, P. & Santos, H. (2001) Solution structure of cytochrome c: insights into the structural basis for ligand detachment. J. Mol. Biol. 308, 353 –365. 38. Gu ¨ ntert, P., Mumenthaler, C. & Wu ¨ thrich, K. (1997) Torsion angle dynamics for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273, 283–298. 39. Kumar, A., Wagner, G., Ernst, R.R. & Wu ¨ thrich, K. (1980) A two-dimensional nuclear Overhauser enhancement (2D NOE) experiment for the elucidation of complete proton – proton cross relaxation networks in biological macromolecules. Biochem. Biophys. Res. Commun. 95, 1–6. 40. Marion, D. & Wu ¨ thrich, K. (1983) Application of phase sensitive two-dimensional correlated spectroscopy (COSY) for measure- ments of 1 H- 1 H spin coupling constants in proteins. Biochem. Biophys. Res. Commun. 113, 967–974. 41. Braunschweiler, L. & Ernst, R.R. (1983) Coherence transfer by isotropic mixing: application of proton correlation spectroscopy. J. Magn. Reson. 53, 521 –528. 42. Bartels, C., Xia, T.H., Billeter, M., Gu ¨ ntert, P. & Wu ¨ thrich, K. (1995) The program XEASY for computer-supported NMR spectral- analysis of biological macromolecules. J. Biomol. NMR 6,1–10. 43. Bru ¨ nger, A.T. (1992) X-PLOR. A System for X-Ray Crystal- lography and NMR. Yale University Press, New Haven, CT, USA. 44. Wareham, R.S., Kilburn, J.D., Rees, N.H., Turner, D.L., Leach, A.R. & Holmes, D.S. (1995) Synthesis and solution conformation of a C 2 symmetric macrobicycle. Tetrahedron Lett. 36, 3047– 3050. 45. Wareham, R.S., Kilburn, J.D., Turner, D.L., Rees, N.H. & Holmes, D.S. (1995) Homeomorphic isomerism in a peptidic macrobicycle. Angewandte Chemie Int. 34, 2660–2662. 46. Kalk, A. & Berendsen, H.J.C. (1976) Proton magnetic relaxation and spin diffusion in proteins. J. Magn. Reson. 24, 343 –366. 47. Boelens, R., Koning, T.M.G. & Kaptein, R. (1988) Determination of biomolecular structures from proton-proton NOEs using a relaxation matrix approach. J. Mol. Struc. 173, 299 –311. 48. Boelens, R., Koning, T.M.G., van der Marel, G.A., van Boom, J.H. & Kaptein, R. (1989) Iterative procedure for structure determi- nation from proton-proton NOEs using a full relaxation matrix approach. Application to a DNA octamer. J. Magn. Reson. 82, 290–308. 49. Tropp, J. (1980) Dipolar relaxation and nuclear Overhauser effects in nonrigid molecules: The effect of fluctuating internuclear distances. J. Chem. Phys. 72, 6035–6043. 50. Olejniczak, E.T. (1989) Including methyl rotation in simulations of spin-lattice relaxation. J. Magn. Reson. 81, 392–394. 51. Koning, T.M.G., Boelens, R. & Kaptein, R. (1990) Calculation of the nuclear Overhauser effect and the determination of proton- proton distances in the presence of internal motions. J. Magn. Reson. 90, 111–123. 52. Gu ¨ ntert, P., Braun, W. & Wu ¨ thrich, K. (1991) Efficient computation of 3-dimensional protein structures in solution from nuclear- magnetic-resonance data using the program DIANA and the supporting programs CALIBA, HABAS and GLOMSA. J. Mol. Biol. 217, 517–530. 53. Gu ¨ ntert, P. & Mumenthaler, C. (1997) DYANA User’s Manual, ETH, Zu ¨ rich, Switzerland. 54. Koradi, R., Billeter, M. & Wu ¨ thrich, K. (1996) MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graphics. 14, 51–55. 55. Laskowski, R.A., Rullmann, J.A.C., MacArthur, M.W., Kaptein, R. & Thornton, J.M. (1996) AQUA and PROCHECK-NMR: 6254 S. G. Chamberlin et al.(Eur. J. Biochem. 268) q FEBS 2001 Programs for checking the quality of protein structures solved by NMR. J. Biomol. NMR 8, 477–486. 56. Hooft, R.W.W., Sander, C. & Vriend, G. (1996) Positioning hydrogen atoms by optimizing hydrogen-bond networks in protein structures. Proteins 26, 363–376. 57. Williamson, M.P. & Asakura, T. (1993) Empirical comparisons of models for chemical-shift calculation in proteins. J. Magn. Reson. B101, 63–71. 58. Wu ¨ thrich, K. (1986) NMR of Proteins and Nucleic Acids. John Wiley & Sons, New York, USA. 59. Matsunami, R.K., Yette, M.L., Stevens, A. & Niyogi, S.K. (1991) Mutational analysis of leucine 47 in human epidermal growth factor. J. Cell. Biochem. 46, 242–249. 60. Lazar, E., Watanabe, S., Dalton, S. & Sporn, M. (1988) Transforming growth factor a: mutation of aspartic acid 47 and leucine 48 results in different biological activities. Mol. Cell Biol. 8, 1247–1252. 61. McInnes, C., Hoyt, D., Harkins, R., Paglia, R.N., Debanne, M.T., O’Connor-McCourt, M. & Sykes, B.D. (1996) NMR study of the transforming growth factor-a (TGF-a) -epidermal growth factor receptor complex. J. Biol. Chem. 271, 32204–32211. 62. Hoyt, D.W., Harkins, R.N., Debanne, M.T., O’Connor-McCourt, M. & Sykes, B.D. (1994) Interaction of transforming growth factor a with the epidermal growth factor receptor: binding kinetics and differential mobility within the bound TGFa. Biochemistry 33, 15283–15292. q FEBS 2001 Solution structure of mEGF/TGFa44–50 (Eur. J. Biochem. 268) 6255 . 2001 geometry, the set of solutions reflects any uncertainty in the calibration as well as alternative fits to the set of distances. The main advantages of the procedure. change. Significantly, a bond between the NH of residue 15 and the carbonyl of residues 41 was identified in more than 40% of the structures in the family of the chimera (Table