| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 280, Issue 25, 24043-24052, June 24, 2005
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
¶¶||**
From the
Department of Physiology and Biophysics, Case Western Reserve University, The Rainbow Center for Childhood PKD, Rainbow Babies & Children's Hospital, the ¶Case Western Reserve University Cancer Center, and the ||Cleveland Center for Structural Biology, Cleveland, Ohio 44106
Received for publication, March 11, 2005 , and in revised form, April 19, 2005.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The EGF receptor is an integral transmembrane glycoprotein of 1186 amino acids (reviewed by Wells et al. (9)). It comprises an extracellular N-terminal ligand binding domain, a single transmembrane helix, a 50-amino acid residue juxtamembrane domain (JX), a tyrosine kinase domain, and a long C-terminal tail, which contains five autophosphorylation motifs. Several intrinsic sorting signals and protein binding sites have now been mapped to the JX region, which regulates receptor trafficking and inactivation. These include basolateral sorting signals (10), a lysosomal sorting motif (11), a nuclear localization signal (12), as well as binding sites for calmodulin (13), 
-subunits of heterotrimeric Gs proteins (14), and phosphoinositide kinases (15). The JX region also includes post-translational modification sites; Thr654 is a known substrate for PKC (16), and Thr669 and Ser671 are substrates for MAPK (17, 18).
The sorting motifs in the JX region have recently been identified. The critical residues of a dominant basolateral sorting signal include a positively charged amino-terminal residue (Arg662) and a proline-rich core (667PXXP670) (10), a motif that bears some similarity with SH3-domain binding motifs. The signal is sufficient to target receptor released at the trans-Golgi network directly to the basolateral plasma membrane, and to recycle endosomal receptor back to basolateral membrane. A second, basolateral sorting motif has been localized to positions 658/659. This signal appears functionally redundant, being active only when the 667PXXP670 signal is deleted or mutated (10). Possibly, because phosphorylation of Thr654 of the activated receptor has been shown to enhance recycling, this feedback mechanism might also invoke the basolateral targeting signals (19). A lysosomal sorting signal formed by two leucine residues is present in the juxtamembrane domain at residues 679 and 680 (679LL680) (11, 20). As receptors with an inactive 679LL680 signal accumulate on the limiting membranes of multivesicular endosomes in ligand-stimulated cells, 679LL680 appears to be required for uptake to internal vesicles thereby enabling their delivery to lysosomes (11).
The presence of several sorting signals and protein interaction sites in one small protein domain is intriguing and prompts multiple questions regarding selection, as well as regulation of the comprised motifs. Structural information for this domain, however, remains sparse. This contrasts with the recent, exciting progress that has been made in the structural characterization of other EGF receptor domains. Crystal structures of the extracellular domains of EGFR (21-23), ErbB2 (24, 25), and ErbB3 (26), and the tyrosine kinase domain of EGFR (27) provided exciting insights into the ligand recognition and binding, and dimerization. Furthermore, NMR studies provided first characterizations of the transmembrane domains of EGFR (28) and ErbB2 (29).
The focus of this study is the structural characterization of the juxtamembrane domain of EGFR. This region has received less attention so far, partially due to its relatively small extent, the weak secondary structure predictions, and only the recent emergence of functional insight into this domain. Several regulatory protein-protein interaction motifs have now been located to this domain, and our interest focuses initially on the conformations of basolateral and lysosomal sorting signals, and the understanding of how sorting motifs may be regulated and activated. We previously studied peptides comprising selected regions of the JX domain of the EGF receptor, residues Arg645 and Ala674, by a combination of NMR and circular dichroism spectroscopic methods (30). In the present study, we extend that work and characterize the properties of the entire JX domain (residues 645-697). Utilizing isotopically labeled protein, heteronuclear NMR spectroscopy provided details on the conformational properties of this region in aqueous, and particularly in DPC-micellar solution, and facilitated the generation of a structural model of the micelle-bound domain. Functionally, the data strongly suggest that the role of these signals may be regulated by their membrane association and their accessibility in the intact receptor.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Plasmid DNA encoding the glutathione S-transferase fusion protein was used to transform the protease-deficient E. coli strain BL21. The 15N-labeled peptide and 15N/13C double-labeled peptide were grown in minimum media supplemented with trace elements as described by Oxenoid et al. (32). 1 g of [15N]ammonium chloride and 3 g of glucose were used for 15N-labeled peptide, whereas 1 g of [15N]ammonium chloride and 1.5 g of [13C]glucose were used to obtain 15N/13C double-labeled peptide.
Fusion protein synthesis was induced overnight at room temperature with 1 mM isopropyl 
-D-1-thiogalactopyranoside after the absorbance at 600 nm reached 0.6-0.8. Bacterial cells were harvested by centrifugation. The cells were lysed with B-PER reagent according to the manufacturer's instructions (Pierce), and the lysates were centrifuged at 15,000 rpm for 30 min. The pellet fraction included the fusion protein, and was solubilized with 70% formic acid. The EGF receptor-derived sequences were cleaved from glutathione S-transferase by adding cyanogen bromide and stirring at 60 °C for 2 h, followed by lyophilization. The dried powder was dissolved in 6 M guanidine hydrochloride at pH 8, and the solution was incubated with 15 ml of nickel-nitrilotriacetic acid beads (Qiagen) for 30 min. The beads were first washed with 6 M guanidine hydrochloride, 50 mM Tris-HCl, pH 8.0, until UV absorbance (280 nm) reached a baseline, then with 50 mM Tris-HCl, pH 8.0. The EGFR peptide was subsequently eluted from nickel-nitrilotriacetic acid beads with 0.1% formic acid and purified by reversed-phase high-performance liquid chromatography on a semi-preparative C8 column (10-mm diameter) (Vydac), using a linear gradient from 0.1% trifluoroacetic acid in water to 0.1% trifluoroacetic acid in acetonitrile in 50 min. Routinely, 5-8 mg of peptide per liter of cell culture were obtained. Peptide mass and sequence were analyzed by MALDI-TOF mass spectroscopy and NMR spectroscopy. The peptide was used for NMR studies without removal of the C-terminal His tag.
NMR Spectroscopy—NMR experiments were carried out using Varian INOVA 500, 600, and 800 MHz NMR spectrometers. Lyophilized peptide samples were dissolved in 500 µl of H2O containing 10% D2Oby volume to yield a sample concentration of 1.0 to 1.5 mM, and the solution was adjusted to pH 5.0. For micellar solutions, molar peptide: detergent ratios of 1:60 for DPC were used to achieve a final approximate peptide:micelle ratio of 1:1 (33). 1.5 mM peptide with 90 mM DPC detergent was dissolved in 500 µl of H2O containing 10% D2O by volume and adjusted to pH 5.0.
Series of two-dimensional and three-dimensional 15N-edited NMR experiments were recorded in both aqueous and micellar solutions on a 600-MHz spectrometer, including 2D-15N-HSQC spectra, 3D-15N-HSQC-TOCSY (60-ms mixing times), 3D-HNHA, heteronuclear 15N-1H NOE (using a saturation delay of 3.2 s), and 3D-15N-HSQC-NOESY experiments (150- and 300-ms mixing times in aqueous samples, and 80-ms mixing time in micellar samples), were recorded at 10, 25, and 35 °C for aqueous samples and at 25, 35, and 45 °C for DPC-micelle samples. The assignment of the backbone resonances of EGFR645-697 peptide in DPC micelles was achieved with standard triple-resonance methods and the following set of triple-resonance experiments at both 25 and 35 °C at 500 MHz: HNCO (34), HN(CA)CO (35), HNCA (34), HNCACB (36), and CBCA(CO)NH (37). The assignment of the side-chain resonances were accomplished with HCCH-TOCSY (38), HC-C(CO)NH, and CC(CO)NH (39) experiments at 35 °C on 500 MHz. Chemical shifts for prochiral protons were largely degenerate, therefore no stereospecific assignments were obtained.2
For the structure determination, deuterated dodecylphosphocholine (DPC-d38) was used to reduce the intensities of the resonances from the detergent. A series of 2D-15N-HSQC-NOESY analyses using mixing times from 40 to 150 ms was acquired and indicated that the NOE buildup was linear up to 100-ms mixing time. 3D-15N/13C-HSQC-NOESY experiments (80-ms mixing times) were recorded at both 25 °C on 600 MHz and 35 °C on 800 MHz. All NMR data were processed using software package NMRPipe (40) on RedHat 9.0 workstations and transferred into the program NMRView (41) for assignment and data evaluation. The chemical shift indices of H, C, C, and CO were calculated within NMRView based on random-coil chemical shift values from Wishart et al. (42) and using corrections for proline and temperature effects.
|
For the measurement of the residual dipolar coupling constants a TROSY-HSQC, and a phase cycle-modified TROSY selecting the up-field 15N-{1H} multiplet component were carried out at 35 °C (44). Both spectra were processed identically by NMRPipe and evaluated by NMRView. The 1H-15N residue dipolar coupling values (RDC) were calculated for each residue by subtracting the difference between 15N frequencies in these spectra of the isotropic sample from the gel aligned sample.
Structure Calculation—1146 unambiguously assigned NOEs were obtained for use in structural calculations (Table I). 15N-NOE cross-peaks were separately calibrated for each residue using intraresidue dN(i,i) cross-peaks (2.70-3.05 Å). For residues where dN(i,i) cross peaks were absent or overlapped, the intensity of dN(i,i+1) was used instead (1.70-3.60 Å). If both these reference cross-peaks were not quantifiable, we used the largest calibration factors found for any other
|
3JHN-N coupling constants were derived from HNHA experiments (45). The coupling constant values of <6 Hz were converted into angle restraints of 60 ± 30° error. The H, C, C, and CO chemical shifts were used as input for the TALOS program (46) to generate additional torsion angle restraints with an error of limit of ±30°.
N-H residual dipolar coupling (RDC) restraints were used only for residues found to be located in any of the three helical domains established by initial structural calculations. RDC restraints were grouped by helical segment, and estimates of Da and rhombicity R (Da/Dr) for individual alignment tensors were derived by using the program PALES (47). For this, structures calculated without RDC restraints were used. Also, CNS structure calculations were performed, in which tensor variables were systematically varied. These showed that large ranges for the variables were compatible with the data of 27 N-H RDC, without affecting the final energies of the structures, the alignment, or the convergence of the calculations. Although PALES suggested values of Da and R of 12-12.6 and 0.42 to 0.61, respectively, the test calculations indicated that wide ranges of Da and R were equally applicable to all three groups. Thus, for the three groups of RDC restraints, uniform values of Da = 12 and R = 0.55 to a single common alignment tensor were introduced in the second step of the structure calculations.
1,146 distance restraints (396 intraresidue, 453 sequential, and 297 medium range), 55 dihedral angle restraints from both TALOS and 3JHN-N coupling constants were used as experimental input for simulated annealing calculations using the software package CNS (version 1.1) (48). A simulated annealing protocol comprising two major steps was used. First, a high temperature conformational search phase in torsion angle space (1,000 steps at 50,000 K with a 15-fs time step) was followed by a cooling phase with torsion angle dynamics (2,000 steps from 50,000 to 1,000 K, 15-fs time step). The resulting structures were further refined in the second step in three stages, including 27 N-H RDC restraints: a high temperature dynamics phase in Cartesian space (5,000 steps from 1,000 K to 0 K in 5-fs time step), followed by a Cartesian dynamics annealing phase with (10,000 steps from 1,000 K in 5-fs time step), and a final minimization phase. 46 of 100 calculated structures were taken to represent the NMR structural ensemble, with non-converged structures being identified using final force-field energies and significant violations of covalent restraints as selection criteria. The structures were analyzed by PROCHECK-NMR (49), visualized with either SPDBV (50) or MOLMOL (51).
Spin-label Experiments—The DPC-bound and water-accessible residues of EGFR645-697 in DPC micelles were determined by measuring the effect of the MnCl2 and 5-doxylstearic acid on the intensities of cross-peaks in 1H-15N-HSQC spectra. 2 mg of peptide and 15 mg of DPC were solubilized in 450 µl of 10% D2O, 100 mM sodium acetate, pH 5.0. 5-µl aliquots of a 60 mM stock solution of MnCl2 (1.5 mg of MnCl2 in 200 µl of water) were titrated into the peptide sample, followed by the acquisition of HSQC spectra at 25 °C. The titration was halted after a total of 3 mmol of MnCl2 had been added, which caused about half the number of the original cross-peaks to disappear. For 5-doxylstearic acid, 1.5 mg of 5-doxylstearic acid were solubilized in 200 µl of 30% DPC solution. Each time, 5 µl of 5-doxylstearic acid solution, 0.2 mmol of 5-doxylstearic acid, was added. The HSQC spectra in the presence of 0.8 mM MnCl2 and of 1 mM 5-doxylstearic acid, respectively, were selected to represent the data and used for evaluation. The reduction intensities were calculated as the ratio of peak intensities in spectra from the sample containing the respective spin-labeled compounds over the peak intensities derived from the same sample before adding the probes.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
The secondary structure of EGFR645-697 was deduced from the chemical shift index of H protons, from 3JHNHA couplings, and from medium range NOEs (Fig. 2C). All data suggested a void of defined structure in the N-terminal half of the peptide. For the C-terminal half of JX, both H chemical shift index (CSI) and 3JHNHA indicated the presence of two adjacent helical regions between Asn676 and Ile691. Nevertheless, only few medium range NOE, d(i,i+3), and d(i, i+4) were observed, and the experiments generally lacked continuous medium range NOEs that distinguish a stable helical backbone conformation from the indicated, more dynamic nascent helices in the C-terminal half. Heteronuclear NOE relaxation data collected at identical conditions (Fig. 3A) revealed two regions with larger than average values: residues Val650 to Thr669 had NOE ratios around 0, and the region between residues Leu679 and Lys689 exhibited NOE ratios of around 0.2. These relatively low values corroborated the highly dynamic nature of the backbone, as stably folded regions in proteins generally exhibit NOE ratios of above 0.5.
EGFR645-697 in DPC Micelles—Because the EGF receptor JX region is in close proximity to biomembranes, we investigated how the lipid microenvironment affected the structure of this domain. As a solution NMR-compatible membrane mimic we used DPC, a micelle-forming detergent that has been effectively and widely used in similar studies (32, 52), including our previous assessment of the JX domain peptide EGFR645-674 (30). The protein was reassigned in the presence of DPC, with backbone assignments of EGFR645-697 being obtained using 15N-labeled peptide and 2D-15N-HSQC, 3D-15N-HSQC-TOCSY, 3D-15N-HSQC-NOESY, and HNHA experiments at 25, 35, and 45 °C. The 13C/15N-doubly-labeled peptide and a selection of triple-resonance experiments at 35 and 45 °C were further used for assignment confirmation and to achieve essentially complete side-chain assignments.
|
, C, C, and CO secondary chemical shifts (CSI) suggested the presence of three helical regions, between residues Lys652 and Arg662, residues Asn676 and Glu685, and residues Phe688 and Leu694 (Fig. 2D). The helical conformations were supported by 3JHNHA-coupling constants, measured by an HNHA experiment, as all residues in the three regions exhibited 3JHNHA < 6 Hz. The only exception is Leu659, with a coupling constant of 6.6 Hz. Among the large number of NOEs numerous contiguous helix-indicative NOEs (dN(i,i+3) and dN(i, i+4)) were observed for these regions. The separation between helices 2 and 3 was suggested despite the observation of several NOEs between these helices by a non-helical CSI for Thr686 and Glu687, the 3JHNHA-coupling constant of 6.6 Hz for Glu687, and the sequential NOE pattern between Glu687 and Phe688. A heteronuclear NOE relaxation experiment was performed to assess the conformational flexibility (Fig. 3A). Micelle-bound JX domain showed significantly higher NOE ratios than in aqueous solution, which indicated the higher conformational order of JX bound to micelles. Two sequential regions exhibit NOE ratios of 
0.6, indicating only limited internal motions. These regions corresponded well to the helical domains established by the structural data discussed above, and suggested the presence of conformationally stable helices in the micelle-associated JX. Residues Glu663 to Pro675, the sequence between the first and second helix, exhibited only small to moderate chemical shifts changes in DPC when compared with aqueous solution. Consistently, the consensus CSI of this region indicated that this region remained in random-coil conformation. The 3JHNHA-coupling constants for all residues but Val665 and Thr669 were in the range of 6-7 Hz. Moreover, the values of heteronuclear NOE ratios for this region were 0.0-0.2, further substantiating the conformational averaging and flexibility of this region. This indicated the absence of interactions between this loop and the detergent micelles.
Topologies of EGFR645-697 with DPC Micelles—To investigate details of the JX association with DPC micelles, the distance-dependent broadening effect of paramagnetic agents was utilized. The micelle-integrating spin-label 5-doxylstearic acid (5-DSA) reports on the proximity of residues to the spin label and thus the micelle interior, whereas Mn2+ as solely water-soluble agent assesses the water accessibility of the peptide residues (53). The broadening was detected as signal intensity changes of amide cross-peaks in 15N-HSQC spectra upon stepwise addition of either DSA or MnCl2 solutions. In combination, these results allowed us to distinguish micelle surface adsorption versus insertion, and to identify the helical surfaces that face the micelles. The relative ratios of cross-peaks intensities in the presence and absence of the broadening agents versus the amino acid sequence of EGFR645-697 are shown in Fig. 3 (B and C). The flexible region between Glu663 and Ala674, the PXXP basolateral signal, was water-exposed, because all of the observable residues showed a large reduction of intensities by Mn2+, and only small changes of intensities by 5-DSA. A similar environment was detected for the two-residue kink between helices 2 and 3, because residues 684-686 were largely water/and Mn2+-accessible. In contrast, the seven residues in the flexible region at the N terminus showed significant reductions in intensities with 5-DSA, and only a small reduction in intensities with Mn2+. This indicated a close proximity of these charged residues to the micelle surface, corroborating the previously proposed micelle binding of this region. Furthermore, the large effect of 5-DSA suggested a deep insertion of the backbone of these residues into the interface, similar to the snorkeling model suggested for the interaction of positively charged residues in amphipathic, membrane-adsorbed helices (54).
For the helical regions, complementary and partially periodic patterns of intensity reduction were observed for both 5-DSA and Mn2+. The pattern for 5-DSA largely matched the 3- to 4-residue periodicity of helices. For example, in the first helix residues Lys651, Leu655, and Leu658 showed the largest decrease in the intensities by 5-DSA, whereas these (and neighboring residues) showed only small changes in intensities by Mn2+. Similarly, 5-DSA more strongly affected the intensities of residues Asn676, Leu679, Leu680, and Leu683 in helix 2, whereas no clear pattern was observed for helix 3. Again, the peak intensities in these regions were only minimally affected by Mn2+. Both the fact that these regions were differently altered by micelle and water based agent and the periodicity of these effects with minima for primarily hydrophobic residues supported the surface adsorption of the peptide, with the amphipathic helices being placed at the micelle-water interface with their hydrophobic surfaces toward the micelle interior.
Structure Calculation for Micelle-bound EGFR645-697—Distance restraints for structure calculations were obtained from 13C- and/or 15N-HSQC-NOESY spectra recorded with a mixing time of 80 ms. A total of 3782 NOEs was identified, of which 70% were unambiguously assigned, and 15% were assigned in several rounds of iterative structure calculations and NOE assignment. No long range NOEs could be unambiguously identified, partially due to the severe overlap of side-chain 13C and 1H chemical shifts, resulting in about 15% of NOES being ambiguous or unassigned. After removal of redundant and ambiguous NOEs, the final structure calculations used a total of 1146 NOE cross-peaks (396 intraresidue, 453 sequential, and 297 medium range), 55 dihedral angle restraints from both 3JHN-HA coupling constants and TALOS, and 27 N-H RDC (detail under "Materials and Methods").
The significant set of the weak unassigned and ambiguous NOEs were primarily observed in the 13C-edited NOESY experiment and appeared to originate from other than short and medium range NOEs. In an attempt to assign these, we turned to both manual as well as automatic NOE assignment approaches with the programs ARIA (55) and CANDID (56). In the presence of manually assigned short and medium range NOEs, dihedral angles, RDC restraints, and individual well defined folds were identifiable that explained subsets of the observed unassigned presumed long range NOEs. However, it was not possible to identify a single conformation compatible with the majority, or even more than a quarter of these NOEs (data not shown). To us, these results suggested the presence of significant conformational averaging, and the absence of a stable fold (in which the N-terminal half would consistently interact with the C-terminal half of the peptide). Nevertheless, the observation of these NOEs was a reflection of temporary proximities and dynamic, intermittent interactions between these helices. The conformational averaging, however, precluded the direct use of these NOEs in the calculations, and they were therefore omitted from the experimental data.
|
N(i,i+3) or dN(i,i+4)) bridge this region (Fig. 2D) and accurately defined the degree of bending in conjunction with the RDC restraints.
Due to the absence of long range NOEs, the spatial arrangement between helices 1 and 2/3 remains undefined. Fig. 4 thus represents an arbitrary, possible conformation, and spacing between the helices in coordinate space. The precision of the structure ensemble is presented using a spline drawn through the C position of a selected conformer, with the line-width of the spline encoding the angular order parameters of phi-torsion angles for the respective residue (generated by the program MOLMOL (51)). The high definition of all three helical regions (in torsion angle space) was apparent in the narrow line width for these structure elements. At the same time, the termini and more importantly, the PXXP basolateral sorting motif comprising loop (Glu663 to Pro675) were highly flexible and disordered, giving rise to the uncertainty of the relative positions between helix 1 and helices 2 and 3.
|
The parameters for the alignment were estimated separately for each helix with the program PALES (47). Additionally, test calculations with CNS, in which the scaling factor Da and the rhombicity R were systematically altered, indicated that RDC data for all three helices were compatible with the presence of a single alignment tensor. Despite the significant conformational flexibility within regions of JX, this is reasonable as the helices are covalently linked, and foremost, because the entire JX domain was aligned as a mixed-micelle complex. The RDC data were therefore included in the above calculations using a single tensor, and Da and R values of 12 Hz and 0.55, respectively. This resulted in a lowering of the backbone r.m.s.d. values for all helical regions compared with calculations without RDCs (data not shown). All back-calculated RDCs for each helical segment in all structures matched the experimental data very well (Fig. 5B). Furthermore, the structural precision of helices 2 and 3 was improved, because these helices were separated only by a two-residue kink that was well characterized by a number of medium range distance observations. Although the RDC restraints are compatible with four, mathematically equivalent relative orientations between these two structural elements, due to the covalent and local conformational restraints only the presented 126° kink conformation was observed in the accepted structural ensemble (albeit one other orientation was infrequently observed in the calculations). In contrast, the absence of long range NOEs between helices 1 and 2 (and 1 and 3) had the direct consequence that, as mentioned above, the distance between these helices was unknown, and that the identification of the correct average orientation out of the four symmetry-related RDC solutions was not possible.
A Model for the Average Micelle-bound Structure—A distinction between the equivalent, symmetry-related RDC solutions can be made by using a second set of RDC data, obtained from a significantly different residual alignment condition. Only the real, physical average orientation and conformation is shared between the symmetry-related solutions calculated for two independent, individual alignment conditions. Obtaining such a second independent RDC set was however unsuccessful in the case of JX, because attempts to use different methods and conditions to generate partially aligned, DPC-bound JX either failed or did not provide independent alignments. Nevertheless, the symmetry-related solutions characterize biochemically different structures, and thus are in principle also distinguishable by other criteria or restraints. For example, in this case the JX domain adsorption to the micelle is not equally possible through a common hydrophobic helical surface in all symmetry-related solutions. We were thus interested in exploring, whether the translational and orientational restrictions resulting from the micelle adsorption of the protein and determined in the spin-label study would effectively reduce the symmetry related uncertainties in the above calculations.
There are several possibilities for including additional restraints on helices in structural calculations (see for example Zdunek et al. (60)). We decided to utilize distance restraints to reflect the approximate spherical size of DPC micelles of 22 Å, with generous error bounds accounting for the uncertainty in size or deviation from spherical shape. Furthermore, the orientation of the surface-adsorbed helices were defined via shorter distance (restraints) for the experimentally determined non-polar surface-bound face, and longer distances for the water-accessible site on the opposing helical face. This resulted in 21 generic distance restraints that define the average distance between every amide in any of the three helices and the center of the detergent micelle as either 21 ± 1 or 23 ± 2 Å, depending on its location in the polar or hydrophobic face of the helix. The addition of these restraints to the above calculations had profound effects and reduced possible conformations in coordinate space. Accepted structures evenly clustered into three distinct groups (Fig. 6A), indicating that of the four symmetry-related relative helical orientations only two were compatible with orienting the correct interface for binding toward the micelle. In the first cluster (Fig. 6B), helices 1 and 2 were nearly orthogonal, and the size of the sphere further broadly defines and limits the spatial separation and translational freedom between the helices. The second cluster (Fig. 6C) represents an alternative symmetry-related solution with helix 1 being anti-parallel to helices 2/3. Due to the broad definition of the adsorption face of the helices, two translational variants were observed, in which helix 1 was accommodated on either side of helices 2/3 (Fig. 6A). Overall, the clusters were consistent with all data and appeared to be equivalent average structures of a dynamic, but diffusionally restricted scenario. Although from these two a reliable identification of the correct average structure cannot be made, differences in the properties of these solutions should be noted. Fig. 6 (B and C) shows the dimensions of the alignment tensor. The source of its high rhombicity under the present alignment conditions (with compressed acrylamide gel) would be expected to be of steric origin, and the dimension of the model with parallel helices appeared more suitable to generate a mixed micelle with such oblate dimensions. Second, the variable but in principle relatively close separation of helices in this group of structures is consistent with the observation of averaged long range NOEs discussed above.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Although principally an independent domain, the deduced close proximity of the JX domain to the plasma membrane brought us to evaluate the effects of membrane interactions on the JX peptide structure by using DPC micelles. As shown, the structural properties of JX are strongly influenced by mimics of the plasma membrane. Our results provided evidence for the binding of JX domain peptide to micelles, causing the formation of three helical domains; from residues Lys652 to Arg662, Asn676 to Glu685, and Phe688 to Leu694. All helices were shown to be amphipathic and to use their hydrophobic sides for micelle-surface binding (Fig. 7).
The C-terminal half of EGFR645-697 peptide comprises two of these helical regions separated by a kink. The position of helix 2 (residues 676-685) agrees well with secondary structure predictions, and the nascent helix in this region in aqueous solution. It also conforms with former observations on peptides comprising this region of the EGFR receptor and its ErbB2 homologues showing helical backbone conformations when bound to detergent micelles.3 This region is included in the previously solved crystal structure of the EGFR kinase domain (27). There, Asn676 to Glu685 were found in an extended conformation, packed adjacent to the following first -sheet of the tyrosine kinase domain. Because the truncated tyrosine kinase domain of EGFR starting at residue 683 was shown to be active without the JX part (64) and no other biochemical data supported the direct interaction between JX part and tyrosine kinase domain, the binding of this JX part to the tyrosine kinase domain may be caused by crystal packing rather than representing the physiological structure of this region. However, in conjunction with our findings the extended structure in the crystal demonstrates the structural flexibility of this region, which may be essential for or a reflection of its function ability to interact with variable partners such as the plasma membrane, the kinase domain, as well as the established interactions with other proteins such as calmodulin (13), and foremost the unknown protein recognizing the comprised dileucine sorting signal.
The first residues of the adjacent tyrosine kinase domain, residues Phe688 to Gly697, were included in our JX peptide to avoid possible truncation effects on the structure of JX. Surprisingly, this part of the peptide was found to be structured and form helix 3. This two-turn helix contrasts the conserved -strand conformation of this region in the crystal structure of EGFR (27), and known structures of the tyrosine kinase domains (65), but is likely only a direct consequence of removing this region from its native structural and folding context.
|
EGFR is found as monomers, homodimers, and heterodimers depending on its activation state. The majority of unactivated receptor is monomeric, only slowly internalizing and recycling to the basolateral surface in polarized cells. Upon binding of ligands to extracellular domain, the EGFR dimerizes either into homo- or heterodimers and is rapidly internalized to early endosome. Because the JX domain in all of our experimental condition was always a monomer, our model is most likely to represent the structure of JX domain in the unactivated monomer EGFR.
Both basolateral and lysosomal sorting signals determine the fate of endosomal EGFR were identified in the JX domains. Each signal was located in different structural units. The dominant basolateral signal, 667PXXP670, was located in the long flexible loop and not associated with micelles, whereas the recessive basolateral signal (658LL659) and lysosomal signal (679LL680) were located in the first helix and the second helix, respectively, facing the surface of micelles. The constitutively active recycling of the unactivated EGFR to the basolateral surface correlates with our model, in that only the PXXP-basolateral signal was accessible, whereas other signals were sterically obstructed from their sorting machineries. This suggests that membrane binding and steric accessibility modulate the activity of the sorting signals.
|
|
FOOTNOTES |
|---|
* This work were supported by National Institutes of Health Grant DK54178 (to C. R. C. and F. D. S.) and by a grant from the March of Dimes Birth Defects Foundation (to C. R. C.). Parts of this research were performed in the Environmental Molecular Sciences Laboratory (a national scientific user facility sponsored by the U.S. DOE Office of Biological and Environmental Research) located at Pacific Northwest National Laboratory, operated by Battelle for the DOE. 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. 
** To whom correspondence should be addressed: Dept. of Physiology and Biophysics, School of Medicine, 10900 Euclid Ave., Cleveland, OH 44106-4970. Tel.: 216-368-5405; Fax: 216-368-1693; E-mail: fds{at}case.edu.
1 The abbreviations used are: EGFR, epidermal growth factor receptor; ErbB2 and ErbB3, erythroblastic leukemia viral oncogene homolog 2 and 3, respectively; JX, juxtamembrane domain of EGFR; DPC, dodecylphosphocholine; 5-DSA, 5-doxylstearic acid; PKC, protein kinase C; MALDI-TOF, matrix-assisted laser desorption ionization time-of-flight; SH3, Src homology domain 3; r.m.s.d., root mean square deviation; MAPK, mitogen-activated protein kinase; 2D, two-dimensional; 3D, three-dimensional; NOE, nuclear Overhauser effect; NOESY, nuclear Overhauser effect spectroscopy; TOCSY, total correlation spectroscopy; HSQC, heteronuclear single quantum coherence; RDC, residual dipolar coupling; TROSY, transverse relaxation optimized spectroscopy.
2 Complete chemical shift data and tables of J coupling constants and RDC constants have been deposited in the BioMagResBank databank under accession number 6579
[BMRB]
.
3 L. Ma, personal communication.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
) and DPC micelles, pH 5.0, 35 °C (
). Error bars indicate the standard deviation of the NOE intensity ratio determined from measured background noise levels. Paramagnetic broadening effects obtained for JX in DPC micelles using Mn2+ and 5-doxyl stearic acid. B, intensity retention for JX in presence of 0.8 mM MnCl2. C, intensity retention plot for JX in the presence of 1 mM of 5-doxyl stearic acid.
), second helix (
), and third helix (
) are shown separately.
