Solution NMR structure of an immunodominant epitope of myelin basic protein

Conformational dependence on environment of an intrinsically unstructured protein

Authors

  • Christophe Farès,

    1. Department of Molecular and Cellular Biology, and Biophysics Interdepartmental Group, University of Guelph, Canada
    Search for more papers by this author
    • Present address
      Max-Planck-Institut für Biophysikalische Chemie, NMR-Based Structural Biology, Göttingen, Germany.

  • David S. Libich,

    1. Department of Molecular and Cellular Biology, and Biophysics Interdepartmental Group, University of Guelph, Canada
    Search for more papers by this author
  • George Harauz

    1. Department of Molecular and Cellular Biology, and Biophysics Interdepartmental Group, University of Guelph, Canada
    Search for more papers by this author

  • Christophe Farès and David S. Libich contributed equally to this work.

G. Harauz, Department of Molecular and Cellular Biology, and Biophysics Interdepartmental Group, University of Guelph, 50 Stone Road East, Guelph, Ontario, Canada, N1G 2W1
Fax: +1 519 837 2075
Tel: +1 519 824 4120, ext. 52535
E-mail: gharauz@uoguelph.ca

Abstract

Using solution NMR spectroscopy, three-dimensional structures have been obtained for an 18-residue synthetic polypeptide fragment of 18.5 kDa myelin basic protein (MBP, human residues Q81–T98) under three conditions emulating the protein's natural environment in the myelin membrane to varying degrees: (a) an aqueous solution (100 mm KCl pH 6.5), (b) a mixture of trifluoroethanol (TFE-d2) and water (30 : 70% v/v), and (c) a dispersion of 100 mm dodecylphosphocholine (DPC-d38, 1 : 100 protein/lipid molar ratio) micelles. This polypeptide sequence is highly conserved in MBP from mammals, amphibians, and birds, and comprises a major immunodominant epitope (human residues N83–T92) in the autoimmune disease multiple sclerosis. In the polypeptide fragment, this epitope forms a stable, amphipathic, α helix under organic and membrane-mimetic conditions, but has only a partially helical conformation in aqueous solution. These results are consistent with recent molecular dynamics simulations that showed this segment to have a propensity to form a transient α helix in aqueous solution, and with electron paramagnetic resonance (EPR) experiments that suggested a α-helical structure when bound to a membrane [I. R. Bates, J. B. Feix, J. M. Boggs & G. Harauz (2004) J Biol Chem, 279, 5757–5764]. The high sensitivity of the epitope structure to its environment is characteristic of intrinsically unstructured proteins, like MBP, and reflects its association with diverse ligands such as lipids and other proteins.

Abbreviations
CNS

central nervous system

CSI

chemical shift index

DIPSI

decoupling in the presence of scalar interactions

DPC-d38

perdeuterated dodecylphosphatidylcholine

DSA

doxylstearic acid

EPR

electron paramagnetic resonance

Fmoc

9-fluorenylmethoxycarbonyl

gpMBP

guinea pig myelin basic protein

hMBP

human myelin basic protein

MAP

mitogen-activated protein

MBP

myelin basic protein

MHC

major histocompatibility complex

rmMBP

recombinant murine

RMSD

root mean squared deviation

SDSL

site-directed spin-labeling

SH3

Src homology domain 3

TFE-d2

deuterated 2,2,2-trifluoroethanol (CF3-CD2-OH)

TSP

3-(trimethylsilyl)-propionic acid

Multiple sclerosis is characterized by chronic inflammation of the myelin in the central nervous system (CNS), and major variants of the illness are considered to be primarily autoimmune in nature [1]. The 18.5 kDa isoform of myelin basic protein (MBP) is one of the most abundant proteins in CNS myelin; MBP maintains the compaction of the sheath by anchoring the cytoplasmic faces of the oligodendrocyte membranes [2], and is a candidate antigen for T cells and autoantibodies in multiple sclerosis [3]. The three-dimensional structure of MBP has not yet been elucidated to high resolution [4,5]. We recently used site-directed spin-labeling (SDSL) and electron paramagnetic resonance (EPR) spectroscopy to investigate the topology of MBP when bound to lipid bilayers of composition mimicking that of the cytoplasmic face of myelin [6,7]. In particular, the segment P85-VVHFFKNIVT-P96 (human sequence numbering, Fig. 1) was shown to be an amphipathic α helix lying on the surface of the membrane at a 9° tilt. The phenylalanyl residues in the middle of this segment penetrated deeply (up to 12 Å) into the bilayer, and the lysyl residue was in an ideal position for snorkeling [7]. There had been several previous, contradictory predictions of the kind of secondary structure of this segment of MBP, due to the plethora of experimental conditions, and the SDSL/EPR experiments demonstrated its α-helicity in situ. More recent crystallographic structures of an MBP polypeptide encompassing this segment, in a complex with human major histocompatibility complex (MHC) and autoimmune T-cell receptors [8,9], revealed an extended conformation, due to the structural requirements for MHC II binding [5,10].

Figure 1.

Comparison of amino acid sequences of the primary immunodominant epitope from various species. The blastp/clustalw[56,57] alignment of sequences of 18.5 kDa MBP from mouse (Mus musculus), rat (Rattus norvegicus), chimpanzee (Pan troglodytes), human (Homo sapiens), bovine (Bos taurus), pig (Sus scrofa), horse (Equus caballus), rabbit (Oryctolagus cuniculus), guinea pig (Cavia porcellus), chicken (Gallus gallus), African clawed frog (Xenopus laevis), little skate (Raja erinacea), spiny dogfish (Squalus acanthias), and horn shark (Heterodontus francisci). Symbols mean that residues in that column are (*) identical in all sequences, (:) substitutions are conservative, and (.) substitutions are semiconservative. The sequence has been numbered 1′ to 18′, where 1′ corresponds to residues 81 and 78 in human and murine full-length 18.5 kDa MBPs, respectively. There is a high degree of conservation in this epitope, particularly in residues V6′ to F10′.

This segment of MBP is highly conserved in primary structure (Fig. 1), and is of biological and medical interest for several reasons. The human hMBP(P85–P96) region is a minimal B-cell epitope for HLA DR2b (DRB1*1501)-restricted T cells [3,11], and overlaps the DR2a-restricted epitope for T cells reactive to hMBP(V87–G106) [12]. There is evidence that segment hMBP(V86–P96) contributes to autoantibody binding, and also contains the T-cell receptor and MHC contact points [11,13]. Moreover, this portion of MBP is also a potential Ca2+–calmodulin binding site [14], and borders a potential SH3-ligand and two known mitogen activated protein (MAP) kinase sites [4].

Experimental treatments for multiple sclerosis based on polypeptide mimetics of MBP have focused on this and neighboring regions of the protein [11,13,15–28]. Several linear and cyclic analogs of hMBP(V87–P99) have been designed, analyzed structurally using NMR and molecular modeling, and evaluated for their ability to induce and/or inhibit experimental autoimmune encephalomyelitis in rats [22,23,25,28]. The cyclic analogs, in particular, showed promise as potential antagonist mimetics for treating multiple sclerosis as artificial regulators of the immune response. The linear polypeptide D82-ENPVVHFFKNIVTPR-T98 (human numbering) has been used to induce immunologic tolerance in patients with progressive multiple sclerosis [20], and clinical efficacy is under evaluation in a phase II/III clinical trial that is currently enrolling patients (http://www.biomsmedical.com) [29]. Thus, comparison of the tertiary structures of this epitope under various conditions is of interest to understand its pharmacokinetics.

We have initiated solution NMR studies of 18.5 kDa rmMBP to probe its three-dimensional conformation under structure-stabilizing conditions, namely 100 mm KCl, 30% trifluoroethanol (TFE-d2 by volume in water) [4,30], and 100 mm dodecylphosphatidylcholine (DPC-d38). Direct application of solution NMR to membrane-associated MBP is problematic because of the reduced mobility of the protein in a reconstituted protein–lipid assembly. The challenge is to find sample preparation conditions that would allow high-resolution NMR studies of MBP in an environment most closely mimicking the native myelin sheath. Although there have been previous NMR studies of other MBP-derived polypeptides [31–33], they could not, at the time, be compared with other structural analyses in environments representative of the in vivo situation. Here, we describe a solution NMR and CD spectroscopic investigation of a segment of MBP comprising the primary immunodominant epitope, to characterize further its conformational dependence on environment, and to complement and extend previous structural analyses that used SDSL/EPR and X-ray crystallographic techniques. The 18-residue polypeptide Q1′DENPVVHFFKNIVTPRT18′, was synthesized and is referred to here as FF2, because it comprises the second Phe–Phe pair (viz. F9′–F10′) within the classic 18.5 kDa MBP isoform.

A key consideration for solution NMR experiments on full-length MBP is the stabilization of secondary, and by extension, tertiary structural elements. Although there is no guarantee that the structure of FF2 will be representative of the intact protein, the conditions used here will help define solution conditions in which these criteria are met. Using chemical shift index (CSI) analysis of the resonances of the intact protein recorded in 30% TFE-d2, regions of secondary structure coincide very well with elements that were either predicted or shown to be transient in molecular dynamics simulations. Another major concern in studying IUPs in solution is their inherent flexibility and their extreme dependence on the global environment (as demonstrated below), necessitating novel NMR strategies [34,35]. A condition that creates a homogeneous population in solution allows for a ‘snapshot’ of the protein to be taken using solution NMR techniques. Thus, in addition to providing a complete characterization of the peptide per se, this work represents a step towards establishing and optimizing physiologically relevant and experimentally tractable solution NMR conditions that will eventually be applied to structural studies of the intact protein.

Results and Discussion

NMR spectroscopy

Resonance assignment

Standard ‘through-bond’ and ‘through-space’1H–1H homonuclear correlation experiments were employed to assign the resonances of the polypeptide FF2, and ultimately to provide the semiquantitative distance restraints for the calculation of its structure in aqueous (100 mm KCl, pH 6.5), organic (30% TFE-d2), and membrane-mimetic (DPC-d38 micelles, 1 : 100 polypeptide/lipid molar ratio) environments. The 1H spin systems for all of the 18 residues were revealed as frequency-connected peak families created by the isotropic mixing of the TOCSY experiments [36]. The sequence-specific assignment of these spin systems was deduced from the ‘fingerprint’ regions of the TOCSY and NOESY experiments, shown in Fig. 2 for all three conditions: aqueous solution (Fig. 2A,B), 30% TFE-d2 (Fig. 2C,D), and 100 mm DPC-d38 (Fig. 2E,F). The TOCSY spectra exhibit the J-correlated i HN to Hα frequencies of all residues except for the N-terminus and the two prolyl residues, whereas the NOESY spectra show the cross-relaxation peaks with frequencies corresponding to the HN of residue i and Hα of residue (i−1) in close proximity. Despite the small size of the polypeptide, some degree of overlap was present, especially for the consecutive residues H8′, F9′, and F10′ with similar spin systems (Fig. 2A,C,E), and additional correlations from both experiments were needed to lift the ambiguity. However, no secondary set of cross-peaks was observed, which suggested that FF2 formed a single, dominant, fast-averaging structure in the three solution conditions investigated. The complete resonance assignments for the three conditions are given in the Supplementary Material (Table S1).

Figure 2.

Results of NMR correlation experiments of the FF2 polypeptide in (A, B) aqueous solution (100 mm KCl, pH 6.5), (C, D) 30% TFE-d2, (E, F) 100 mm DPC-d38 micelles, pH 6.5. Panels present 1HN1Hα fingerprint regions of (A, C, E) a two-dimensional TOCSY (DIPSI-2) spectrum with mixing time of 120 ms, and (B, D, F) a two-dimensional NOESY spectrum with mixing time of 300 ms. Labels were added showing the relevant peak assignments, by residue number.

To strengthen further the relevance of FF2 as a polypeptide model for the immunodominant epitope of MBP, the 13C frequencies of the backbone spins of FF2 were also assigned and compared with those previously published for full-length MBP under the same 30% TFE-d2 conditions [30]. Assignments were carried out on the standard heteronuclear single-quantum (HSQC) experiment and were based on the 1H assignment presented above. Because of the low abundance of the 13C nuclei, the sample concentration was raised to 20 mm, for which excellent solubility was still achievable in 100 mm KCl and 30% TFE-d2. At this concentration, only minor 1H chemical shift differences were observed relative to the low concentration samples (data not shown), which implied that polypeptide aggregation was minimal.

Secondary structure analysis

For those residues of the full-length rmMBP (recorded in 30% TFE-d2) with definite peak identification (refer to values described previously [30], Accession No. 6100 in the BioMagRes Bank database, http://www.bmrb.wisc.edu), there generally is very good agreement with the chemical shifts identified in FF2 recorded under the same conditions. The HN and Cα atoms were identified in 15 residues in the Q78T95 sequence of rmMBP and differ on average by 0.2 and 1.2 p.p.m., respectively, with the corresponding primed residues of FF2. However, in each case there is one outlying larger difference: residues F9′ (ΔδCα = 5.4 p.p.m. versus F86) and F10′ (ΔδHN = 0.46 p.p.m. versus F87), possibly due to steric effects in the local environment. These overall small deviations suggest similar Φ and Ψ angles in both structures throughout the central segment of the polypeptide, with an exception perhaps in the vicinity of the Phe–Phe pair. Observed differences in the Cα chemical shifts may be due to changes in local environment because of tertiary interactions present in the intact protein and absent in FF2. Small deviations in the pH of the two samples may also account for the chemical shift differences.

The secondary fold of FF2 in all three conditions was assessed using the chemical shifts of the Hα and Cα atoms. A database of chemical shift indices was compiled by Wishart et al. [37] to identify residues involved in ordered secondary structures. Typically, α-helical structures are identified by an uninterrupted segment of four or more residues that have a positive Cα chemical shift difference (downfield displacement) and a negative Hα chemical shift difference (upfield displacement) relative to the random coil chemical shift values for the same residue dissolved in water [37]. The CSI analyses of our assignments, shown in Fig. 3, indicate a noticeable tendency of a central 10-residue segment of the polypeptide to adopt a helical conformation from residues 5′ to 14′, for samples in TFE-d2 (Fig. 3B) and in DPC-d38 (Fig. 3C), but not in KCl (Fig. 3A). This tendency is shown by the uninterrupted downfield Cα and upfield Hα shifts for that stretch of amino acids. Based on the CSI of FF2 in KCl, there is conflicting evidence of secondary structure formation (Fig. 3A). The Hα shifts seem to indicate weak α helix formation, which is unsubstantiated by the Cα chemical shifts.

Figure 3.

Amino acid sequence of the FF2 polypeptide, and survey of sequential and medium-range NOEs, and conformation-dependent chemical shifts of FF2 dissolved in (A) aqueous solution (100 mm KCl, pH 6.5), (B) 30% TFE-d2, and (C) 100 mm DPC-d38 micelles, pH 6.5. Thick, medium, and thin bars indicate strong, intermediate, and weak NOE intensities, respectively, linking the residues involved in sequential (dαN, dβN and dNN) and medium-range (dαβ and dαN/dβN) NOE connectivities. The 13Cα and 1Hα chemical shifts are plotted relative to the random coil values available from Wishart et al. [37], calibrated to TSP.

In order to explain this apparent ambiguity, the global conformation of the FF2 polypeptide was examined by CD spectroscopy under various conditions (Fig. 4). In aqueous solution (pure water, and 100 mm KCl, pH 6.5), the spectra indicated that the polypeptide had little or no regular secondary structure. In organic and membrane-mimetic conditions (30% TFE and 20 mm DPC, respectively), the spectra clearly indicated an α-helical conformation. These results are consistent with previous CD spectroscopic studies of MBP and MBP fragments [38–40] and support the inclusion of loose dihedral angle restraints in the structure calculations of FF2 in TFE-d2 and DPC-d38 (see below).

Figure 4.

CD spectroscopy of the FF2 polypeptide in various solution conditions. The solid line represents FF2 in 100 mm KCl, pH 6.5; the dotted line represents FF2 in 20 mm DPC; the dashed line represents FF2 in 30% TFE; the dot-dash line represents FF2 in water. The spectra of FF2 in TFE and DPC show the characteristic double minima at 207 nm and 222 nm of an α helix. In contrast, the spectra of FF2 in 100 mm KCl and pure H2O are indicative of a primarily random coil conformation.

NOE analysis

The pattern and size of NOE connectivities extracted from the NOESY experiment also provide an independent indication of the secondary structure of FF2. The diagrams in Fig. 3 show the classification of NOE connectivities into either sequential (i, i+1) or medium range (i, i+2) (i, i+3), and (i, i+4) categories. The extremities of each line connect the cross-relaxing residues, whereas the thicknesses relate to the magnitude of the interaction (weak, medium, strong). The characteristic types of NOE connectivities for an α helix were observed throughout the sequence, but were particularly consistent for a segment of residues between positions 5′ and 15′. These included the sequential dNN(i, i+1) and dαN(i, i+1), and medium-range dαβ(i, i+3), dαN(i, i+2), dβN(i, i+2), dαN(i, i+3), and dβN(i, i+3). Numerous other (i, i+3) and (i, i+4) connectivities were also observed between side-chain protons over this same sequence. This pattern reinforces the α-helical model for the stretch of residues between P5′ and P16′.

The structures of the FF2 polypeptide presented here are largely based on intramolecular NOE connectivities. The monomeric medium-sized FF2 (2.2 kDa) is predicted to have a rotational correlation time just above the critical value for which NOE cross-peaks vanish, owing to the equal contribution of the cross-relaxation through the zero- and double-quantum transitions. Correspondingly, in the 100 mm KCl and 30% TFE-d2 samples, the NOESY cross-peaks are small but have the same sign as the diagonal peak. In the 100 mm DPC-d38 sample, cross-peaks are larger because the FF2 polypeptides in association with the micelles have a longer correlation time.

Sufficient NOE cross-peaks were compiled, partially assigned, and measured to calculate the structure of FF2 in 100 mm KCl (pH 6.5), 30% TFE-d2, and in 100 mm DPC-d38 micelles. Although two-dimensional NOESY spectra were measured for several mixing times (100, 200, and 300 ms) and were all used to assign connectivities, the magnitude of NOEs was based on the Gaussian-function fitted volume of cross-peaks from the two-dimensional spectrum recorded at 300 ms. For the correlation time regime of FF2 under all conditions, the NOE build-up curves are expected to vary quasi-linearly over the time range covered by these mixing times. The NOE cross-peaks with heavy overlap were fit using the sum-over-box algorithm in the sparky package.

As described in Experimental Procedures, the aria/cns calculations were provided with: (a) chemical shift assignments; (b) a list of NOE cross-peak volumes that were tentatively assigned; and (c) for the TFE-d2 and DPC-d38 structures, loose initial backbone dihedral restraints (−180° < Φ < 0, −90° < Ψ < 30°). Additional loose H-bond distance restraints (2.5 < Oi… Ni+4 < 3.5) did not improve the quality of the 10 best structures, but reduced the occurrence of NOE-violated structures over the ensemble of 100 structures. Approximately 200 NOE distance restraints were used for each condition, of which ≈ 50% were interresidual (Table 1). These NOE connectivities were either sequential, and/or short-ranged (connecting 1H separated by 2–4 residues in the primary sequence).

Table 1.  Structural statistics of the FF2 polypeptide structures under various solution conditions: 100 mm KCl, pH 6.5; 30% (vol) TFE-d2; 100 mm DPC-d38, pH 6.5.
 100 mm KCl30% TFE-d2100 mm DPC-d38
Restraint for calculation
 Total number of NOE restraints182266199
 Unambiguous168246183
 Intraresidue122116107
 Sequential356840
 Short-range (long range)11(0)60(2)36(0)
 Dihedral angle03030
Restraint violations
 Distance restraints = 0.3 Å (= 0.5 Å)0(0)0(0.65)0(0)
 Dihedral angle restraints of = 500.90
Deviations from idealized geometry
 Bonds (Å)0.0015 ± 0.000090.00296 ± 0.000290.00185 ± 0.00007
 Angles (°)0.26 ± 0.010.46 ± 0.030.29 ± 0.01
 Impropers (°)0.13 ± 0.010.43 ± 0.150.12 ± 0.01
 Dihedral (°)44.23 ± 0.6340.66 ± 0.7039.63 ± 0.38
 VdW (Å)12.07 ± 0.7825.12 ± 2.7110.92 ± 0.57
Energies (kcal·mol−1)
 NOE restraint energy1.88 ± 0.879.69 ± 4.982.17 ± 0.39
 Total energy− 512.4 ± 35.5− 501.4 ± 39.9− 544.9 ± 28.8
Ramachandran statistics (%)
 Most allowed37.574.182.4
 Additionally allowed43.913.510.0
 Generously allowed9.35.91.2
 Disallowed9.36.46.4
RMSD from mean structure
 Backbone atoms (overall)3.69 ± 1.251.07 ± 0.260.86 ± 0.37
 All heavy atoms (overall)4.34 ± 1.281.55 ± 0.311.41 ± 0.41
 Backbone atoms (2° structure)0.29 ± 0.120.67 ± 0.180.42 ± 0.17
 All heavy atoms (2° structure)0.84 ± 0.351.09 ± 0.270.92 ± 0.19

For each solution condition, the 10 lowest energy structures were overlaid and represented from two different orthogonal perspectives as line-connected heavy atoms (backbone), as secondary structure schematics (ribbons), and as space-filling models (Fig. 5). As summarized in Table 1, these structures have low energies (both for the restraint potentials and overall potentials), small distance and angular deviations from idealized molecular geometries, and few NOE violations. The root mean square deviations (RMSD), calculated from atom positions of the 10 best structures relative to the mean structure, are reasonably low for all heavy nuclei (i.e. excluding hydrogens) and for backbone nuclei. For the organic and membrane-mimetic conditions, considering only residues 5′ to 16′, these RMSD values are further reduced by ≈ 0.5 Å. This segment is a well-defined helix, with Φ and Ψ torsion angle pairs falling within the allowed α-helical region of the Ramachandran plot [41]. Under aqueous conditions, deviations from the allowed regions of the Ramachandran plot are greater than observed under the other two conditions, suggesting the incomplete formation of an α-helical structure. It should be noted that the majority of residues (81.4%) fall into the allowed or generously allowed regions, which suggests that the peptide adopts a structure (in the core region) similar to a helix. There is extreme flexibility of the polypeptide near the termini, particularly residues D2′, E3′, R17′ and T18′ which have the largest deviations from the most highly populated regions of the Ramachandran plot, and which contribute to the proportion of residues in the disallowed space.

Figure 5.

Structure of the FF2 polypeptide in (A) 100 mm KCl, pH 6.5, (B) 30% TFE-d2, (C) DPC-d38 micelles, pH 6.5. To provide two different perspectives, a 90° rotation along the horizontal axis was used to convert the left structure to the right structure. The N-terminus is at the left for every structure. The best-fit overlays of the 10 lowest overall energy structures obtained with the aria protocol, described in Experimental Procedures, are illustrated as a line-model of the covalent bonds between heavy atoms, or as ribbons (A only). In (B) and (C), the means of the 10 lowest energy structures are presented as schematic representations of α-helical structure, and as space-filling models. The surfaces in the latter representations are colored with a red-to-white-to-blue gradient indicating the electrostatic partial charge distribution (red = positive, white = neutral, blue = negative).

In aqueous solution (100 mm KCl, pH 6.5), the polypeptide forms a relatively stable core, and suggests a weakly helical conformation in the most highly conserved region (V6′ to F10′). These results are consistent with the CD data (Fig. 4) and with our recent molecular dynamics simulations that showed this segment to have a propensity to form transient α helices in aqueous solution [7]. The NMR structures obtained under such conditions would thus be consistent with a compendium of conformers in fast exchange.

In the organic and membrane-mimetic environments, the helical segment stretches over 10 residues, forms three loops, and exhibits little curvature. As expected, the helix also delineates the discrete amphipathic nature of the polypeptide. To illustrate this segregation of hydrophobic/hydrophilic residues around the helical conformation, Fig. 5 shows the electrostatic surface charge of residues P5′ to P16′ of the proposed structures of FF2 from two orthogonal view angles. The partitioning of charges onto opposing faces of the helix further reinforces the amphipathic nature of this peptide. A noticeable difference between the two structures is seen, however, in both N- and C-termini. In TFE-d2, the ends bend abruptly at the site of the two prolyl residues, and fold back towards the hydrophilic side of the helix. In DPC-d38, the helix is more elongated despite similar interruptions of the helix at P5′ and P16′.

This important difference can be rationalized from the nature of the solvent. Previously, Bates et al. [7] performed molecular dynamics simulations of the central immunodominant segment in water, with an added chlorine (Cl) counterion, and demonstrated that it had a propensity to form an α helix. However, this structure was transient in the absence of stabilizing factors. In general, the organic solvent TFE is electrically neutral and preferentially aggregates around the polypeptide, displacing water, and thereby forming a low dielectric environment that favors the formation of intrapeptide hydrogen bonds [42]. Hence, in this instance, the terminal and side chain charges must come into close contact at the expense of bending energies. The zwitterionic DPC, by contrast, provides not only a hydrophobic surface from its acyl chain, but both positive- and negative-charge contacts to the polypeptide chain, allowing it to adopt a much more relaxed conformation. The notion that the solvent environment can elicit structural changes in this polypeptide, and by extension to the whole rmMBP structure, is a major concern in the choice of membrane-mimetic environment [4]. However, despite the slight bend in the termini, the overall secondary structure is preserved by the presence of TFE-d2, while avoiding possible aggregation and precipitation at the high concentrations necessary for NMR.

Paramagnetic relaxation effects

The position of FF2 in DPC-d38 micelles was also investigated using two paramagnetic agents, 5-doxylstearic acid (5-DSA) and FeCl3, which, respectively, partition inside or outside the hydrophobic interior of the micelles. These molecules act locally as strong signal-relaxing agents, causing a broadening proportional to the inverse of the average of the distance to the sixth power (<r−6>), between the unpaired electron of the paramagnetic agent and the interacting nucleus. Thus, these agents can report the positioning of individual residues, and on the orientation of the whole helix relative to the micellar core. The effects of these agents were measured on an ensemble of cross-peaks belonging to the same residue spin system in TOCSY spectra measured with a 40 ms mixing time, and are summarized in Fig. 6. For the 5-DSA titration data, there are three short regions of strong relaxation effect (V6′–V7′, F9′–F10′–K11′ and I13′–V14′), separated by regions of lower effect (H8′, N12′). The termini of the polypeptide are generally not affected by the presence of the 5-DSA in the micelles. A reverse trend is apparent when the experiment is repeated on the same FF2/DPC-d38 sample to which FeCl3 was added, although the effect seems less pronounced. Here, the regions of larger broadening are located near positions V7′ and N12′, as well as in the vicinity of the C-terminus. However, the regions of high relaxation with 5-DSA have relatively lower relaxation because of the presence of Fe3+. The apparent fast relaxation of V7′ in the presence of both paramagnetic agents suggests that the residue may lie at the micellar interface where it would be exposed to both Fe3+ and 5-DSA. A residue that shows slow relaxation under both conditions is H8′, although this may be due to unfavorable electrostatic interaction between its side chain and the Fe3+ ions. Although the Fe3+ ion is soluble in aqueous solution, its location is also dictated by electrostatic interactions which are unfavorable in the vicinity of the partially positively charged side chain of histidine. These results demonstrate that the polypeptide α helix forms distinct hydrophobic and electrostatic contacts with the DPC micelles, and are in agreement with the SDSL/EPR mapping and positioning of the α-helical model of this epitope of MBP on the surface of a lipid bilayer [6,7].

Figure 6.

Paramagnetic relaxation effects of 5-DSA, and of FeCl3, on the FF2 polypeptide in DPC-d38 micelles. Normalized signal amplitude of TOCSY (mixing time = 40 ms) spin system cross-peaks is displayed as a function of residue position for FF2 dispersed in DPC-d38 micelles for each step of the titration of (A) 5-DSA (0.5–2 mm), and (B) FeCl3 (0–1.5 mm). The residual amplitudes were measured for the ensemble of resolvable peaks of each spin system at the ω2 frequency of the HN.

Biological significance

MBP is an ‘intrinsically unstructured’ (or ‘conformationally adaptive’) protein [4]. Such proteins constitute roughly one third of the eukaryotic proteome, and are generally involved in signaling and/or cytoskeletal assembly [43,44]. Although seemingly unstructured in isolation, their large effective volume facilitates rapid and specific interaction with a variety of ligands, the association of which, in turn, effects a conformational change. Often, defined segments of these proteins have a propensity to form an α helix, and represent a binding target for some other protein [44]. The classic 18.5 kDa MBP isoform fits well into this paradigm, because it is membrane-associated in vivo, but also interacts with a plethora of other proteins, such as calmodulin, actin, tubulin, clathrin, and SH3-domain containing proteins [4]. Here, we focused on a conserved segment of MBP which is known to be α-helical when bound to a membrane, is a potential calmodulin-binding site, and also a primary immunodominant epitope in multiple sclerosis. The helicity of this epitope when associated with calmodulin is probable but not yet proven [14], but it is extended when bound to the MHC [8,9]. Thus, it exhibits a conformational adaptability depending on its environment and binding partners.

Numerous epitopes of MBP have antigenic properties (13–32, 83–99, 111–129, 145–170, human sequence numbering) [45]. Their structural characterization is necessary to gain insight into their behavior as therapeutic agents, conditions under which a large variety of environments are encountered. Recently, Tzakos et al. determined the structure of the guinea pig myelin basic polypeptide gpMBP(Q74–V85), using solution NMR of the polypeptide dissolved in dimethylsulfoxide, and modeled its interaction with an MHC receptor site [27]. The segment QKSQRSQDENPV from the guinea-pig sequence, corresponds to the 13-residue segment hMBP(Q74–V86) of the human sequence, which is N-terminal to our 18-residue FF2 polypeptide. Thus, the overlap region between gpMBP(Q74–V85) and FF2 is only six residues (QDENPV), of which QDE were least well-defined conformationally in both studies, due to being at the termini of both constructs. Similarly, minimal direct comparison can be made with previous studies of other MBP segments [31,33,46] or the cyclic analogs [28].

The FF2 sequence is highly conserved evolutionarily compared with the rest of the protein (Fig. 1), and there are several post-translational modifications within it: Q1′ can be deamidated, R17′ can be deiminated, and T15′-and T18′ can both be phosphorylated by MAP kinases [4]. In all species except fish, this sequence is followed by a triproline repeat (P19′P20′P21′) and comprises a potential SH3-ligand (P16′R17′T18′P19′), which could be expected to form a polyproline type II helix [47]. Thus, the MBP segment that we have studied may be critical in the protein's interaction with the myelin membrane, potentially in proper positioning of this putative SH3-ligand and the two known MAP kinase sites for functional roles beyond membrane adhesion.

The structures of this segment have been well-characterized under a variety of conditions and using different biophysical approaches, here and elsewhere [7]. This investigation serves to guide ongoing solution NMR investigations of the full-length protein. The problems faced here are similar to those in NMR structural studies of other membrane-associated and/or intrinsically unstructured proteins such as α-synuclein [48–50], and similar strategies are thus suggested to probe MBP's conformational ensemble. Whereas the study of Bates et al. [7] indicated that the MBP segment PVVHFFKNIVTP was α-helical in situ in a membrane, this high-resolution NMR structural study proved its α-helicity in a stabilizing solution environment, and supports the use of DPC-d38 or TFE-d2[30] as a structure-stabilizing condition for solution NMR studies of the full-length protein.

Experimental procedures

Peptide synthesis

The 18-residue polypeptide hMBP(Q81–T98) (Q1′DENPVVHFFKNIVTPRT18′), encompassing the immunodominant epitope region matching a membrane surface-interacting α helix (V86 to T95), was synthesized via 9-fluorenylmethoxycarbonyl (Fmoc) chemistry at the Advanced Protein Technology Centre (Hospital for Sick Children, Toronto, Canada). The polypeptide was purified by reversed-phase HPLC on a C18 column (7.8 × 300 mm, Phenomenex, Torrance, CA). As determined spectroscopically at 230 nm, the polypeptide eluted after 30 min from a linear gradient binary solvent system (0–60% CH3CN in H2O with 0.1% trifluoroacetic acid, in 60 min) at a flow rate of 1 mL·min−1. This method yielded 200 mg of polypeptide; purity and identity were confirmed by ESI-MS (not shown). The polypeptide, here referred to as FF2 (because it comprises the second of two Phe–Phe pairs within 18.5 kDa MBP, viz., F9′–F10′), required no further purification and was used directly.

Sample preparation for NMR spectroscopy

FF2/KCl

The FF2 polypeptide was dissolved in 100 mm KCl, pH 6.5, to a final concentration of 2 mm. The 550 µL sample was transferred to a standard 5 mm high-precision microcell tube (528 pp, Wilmad-Labglass, Buena, NJ). For the measurements of natural abundance 13C, the polypeptide concentration was increased to 20 mm. The sample temperature was maintained at 298 K.

FF2/30% TFE-d2

Homonuclear 1H experiments were performed on a 600 µL FF2 NMR sample prepared by dissolving the polypeptide to a concentration of 5 mm in 30% TFE-d2 (Cambridge Isotope Laboratories, Andover, MA) in H2O. As for the aqueous solution, the sample was transferred to a standard 5 mm high-precision microcell tube. The polypeptide concentration was increased to 20 mm for experiments involving natural abundance 13C. The sample temperature was maintained at 300 K.

FF2/DPC-d38

All experiments were performed on a 550 µL sample comprising 1 mm FF2 polypeptide and 100 mm perdeuterated DPC-d38 (Cambridge Isotope Laboratories) in a 50 mm phosphate buffer, adjusted to pH 6.5 and containing 10% D2O. After dissolving the detergent and the polypeptide in the buffer, the sample was transferred to a standard 5 mm high-precision microcell tube and left to anneal for 30 min at 60 °C before use. The sample temperature was maintained at 318 K during measurements. This sample was also titrated with 5-DSA (55 mm solution in CD3OH) to obtain final concentrations in the range of 0–2 mm, and FeCl3 (55 mm aqueous solution) to obtain final concentrations ranging from 0 to 1.5 mm.

Solution NMR spectroscopy

The high-resolution 1H, 13C, and 15N NMR spectra were recorded on a Bruker Avance (Bruker BioSpin, Milton, ON, USA), spectrometer operating at a field of 14.1 T (corresponding to the resonance frequency of 600.1 MHz for 1H) and implemented with a triple resonance gradient inverse probe. The 90° pulses were typically 12 and 15 µs, and the spectral widths were set to 12 and 165 p.p.m. for 1H and 13C, respectively. Solvent (water) signal purging was achieved using a 2 s presaturation pulse with the carrier frequency set on the water 1H signal. The phase-sensitive two-dimensional TOCSY [36] (with DIPSI-2 [51] isotropic mixing times: 50–120 ms) and two-dimensional NOESY [52] (mixing times: 100–300 ms) experiments were typically acquired using a recycling delay of 2 s, 128 increments, and 96 scans per increment, for a total experimental time of ≈ 5.12 h. The natural abundance 1H–13C HSQC [53] spectra were acquired using gradient pulses for coherence selection recording: 112 increments × 1024 scans, and 144 increments × 160 scans, respectively. The 1H and 13C chemical shifts were referenced indirectly to 3-(trimethylsilyl)-propionic acid (TSP). The resonance assignments are reported in Table S1 (Supplementary Material). All spectra were processed using the xwinnmr package (Bruker BioSpin), and analyzed using sparky 3 (TD Goddard & DG Kneller, University of California, San Francisco).

Structure calculation and molecular modeling

All interhydrogen distance restraints were derived from the NOE cross-peak volumes measured on the 300 ms two-dimensional NOESY spectra and were used towards calculating a family of structures determined using cns v1.1 [54], operating under aria v2.0 [55] for partially automated NOE assignments; both programs were installed on a personal computer running the Intel/Linux operating system. The chemical shift assignment was based on the identification of 1H spin systems on the two-dimensional TOCSY spectra, and the connectivities were identified mainly from 1HN1HN and 1HN1Hα fingerprint regions. The 1H–1H NOE cross-peaks were fit to a Gaussian curve and integrated using sparky. The peak volumes and chemical shifts were used as the distance restraint input for aria, along with their most probable assignment. Stereochemical and overlap assignment ambiguities were automatically resolved by the aria protocol. Additional restraining input parameters included loose H-bonding (C = Oi… HNi+4) and backbone torsion angles (−180° < Φ < 0, −90° < Ψ < 30°), based on the CD and chemical shift analyses (see Results and Discussion). One hundred structures were generated at the end of 14 iterated simulated annealing steps, with gradual decreases in the NOE violation tolerance and in the partial assignment threshold. In each simulated annealing step, the polypeptide underwent molecular dynamics simulations with 3 fs time steps, during which it was submitted to 10 000 heating steps from 0 to 2000 K, and 16 000 cooling steps back to 50 K. Each step used square-well distance and torsion restraints, and the standard protein topology and parameters defining the bonded and nonbonded geometrical energy functions provided by the cns package [54]. Structural analyses and generation of structure figures were carried out using molmol 2.6 (ETH, Zürich, Switzerland), also running on an Intel/Linux operating system.

Data deposition

The 1H and 13C chemical shifts (Supplementary Material, Table S1) were deposited in the BioMagResBank (http://www.bmrb.wisc.edu) with Accession No. 6857.

CD spectroscopy

CD spectroscopy of samples of FF2 (0.2 mg·mL−1) in aqueous solution (100 mm KCl, pH 6.5), in 30% TFE, and in 20 mm DPC (pH 6.5), was performed using a Jasco J600 spectrapolarimeter (Japan Scientific, Tokyo, Japan). Samples had volumes of 1 mL, and were measured in a 0.1 cm path-length quartz cuvette. Measurements were taken at a 100 nm·min−1 rate, at 0.1 nm intervals, over a range of 190 to 250 nm. All measurements were recorded at ambient room temperature. Four successive scans were recorded, the sample blank was subtracted, and the scans were averaged. The data averaging and smoothing (using an inverse square algorithm) operations were accomplished with the sigmaplot (SPSS, Chicago, IL) computer program.

Acknowledgements

This work was supported by the Canadian Institutes of Health Research (Operating Grant MOP 43982 to GH), and by the Natural Sciences and Engineering Research Council of Canada (to GH). DSL was the recipient of an Ontario Graduate Scholarship. The authors are grateful to Dr Nam-Chiang Wang of the Peptide Synthesis Facility (Hospital for Sick Children, Toronto) for synthesizing the FF2 polypeptide, and to Ms Valerie Robertson, Dr Martine Monette, and Dr Vladimir Ladizhansky for insightful discussion and advice with the NMR experiments.

Ancillary