Structural determinants of nitroxide motion in spin-labeled proteins: Solvent-exposed sites in helix B of T4 lysozyme

Abstract

Site-directed spin labeling provides a means for exploring structure and dynamics in proteins. To interpret the complex EPR spectra that often arise, it is necessary to characterize the rotamers of the spin-labeled side chain and the interactions they make with the local environment in proteins of known structure. For this purpose, crystal structures have been determined for T4 lysozyme bearing a nitroxide side chain (R1) at the solvent-exposed helical sites 41 and 44 in the B helix. These sites are of particular interest in that the corresponding EPR spectra reveal two dynamic states of R1, one of which is relatively immobilized suggesting interactions of the nitroxide with the environment. The crystal structures together with the effect of mutagenesis of nearest neighbors on the motion of R1 suggest intrahelical interactions of 41R1 with the i + 4 residue and of 44R1 with the i + 1 residue. Such interactions appear to be specific to particular rotamers of the R1 side chain.

Site-directed spin labeling (SDSL) has become a widely used tool to study protein structure, dynamics, and interactions (for reviews, see Hubbell et al. 1996, 1998, 2000; Columbus and Hubbell 2002; Klug and Feix 2004; Fanucci and Cafiso 2006). The SDSL approach requires a cysteine residue to be introduced at a selected site and subsequently modified with a nitroxide reagent to generate a disulfide-linked nitroxide side chain. The most commonly employed spin-label side chain is designated R1(Fig. 1). The EPR spectra of R1 directly encode information on the motion of the nitroxide, which in turn reflects structure and dynamics of the local protein environment at the labeling site. A primary goal in developing the SDSL method is to enable interpretation of EPR spectra in terms of protein structure and dynamics using well-characterized proteins as models. Success in this regard will allow SDSL to be used to infer these features in poorly characterized proteins that are challenging to other methods.

Figure 1.

Structure of the nitroxide side chain R1 showing designation of the various atoms and side-chain dihedral angles (χ). χ₅ is defined by S_δ−C_ε−C₃−C₄ as in Tombolato et al. (2006).

At the present time, simple one-component EPR spectra that arise from R1 on solvent-exposed surfaces of α-helices and in ordered turns are reasonably well understood (Guo et al. 2007). At these sites, the motion of R1 is determined by the internal motion of the side chain and local backbone dynamics (Columbus et al. 2001), and the EPR spectra provide a means for mapping backbone dynamics (Columbus and Hubbell 2004).

Interactions between the R1 side chain with nearby groups result in two-component EPR spectra that encode information about the tertiary fold of the protein (Mchaourab et al. 1996). However, the structural origins of the two-component spectra remain poorly defined, making it difficult to reliably relate such complex EPR spectra to protein structure and dynamics.

In previous studies using X-ray crystallography and site-directed mutagenesis, the two-component spectra of T4L 119R1 and T4L 115R1 were found to arise from two rotamers of the R1 side chain (Langen et al. 2000; Guo et al. 2007). In one rotamer, the nitroxide has hydrophobic interactions with a nonpolar pocket formed by side chains of an adjacent helix. This tertiary interaction gives rise to an immobilized state of the nitroxide. In the other rotamer, the nitroxide has no interactions and gives rise to a second spectral component corresponding to a mobile state. This information provides a basis for predicting or interpreting the outcome of an SDSL experiment. For example, using the rotamer library for R1 determined by crystallography (Langen et al. 2000; Guo et al. 2007), the side chain can be modeled at a selected site in a protein. The presence or absence of a nearby hydrophobic surface that could accommodate the nitroxide of one of the rotamers can be ascertained from the model and the likelihood of an immobilizing interaction evaluated.

While the presence of a hydrophobic surface may be sufficient to produce immobilization of the nitroxide, it is apparently not necessary. As will be discussed below, R1 residues at solvent-exposed sites in helix B (residues 40–50) in T4L also give rise to resolved two-component spectra with one component relatively immobilized. However, there is no hydrophobic surface in the immediate environment for interaction with the nitroxide. Not surprisingly, there are other interactions of the nitroxide that give rise to partial immobilization. The primary aim of this study is to use X-ray crystallography and site-directed mutagenesis to identify the interactions that give rise to two-component spectra of R1 in the B helix. This helix is well suited for an SDSL study because Matthews and coworkers have found that the protein stability and function are little perturbed by mutagenesis of any residue other than A42, K43, or L46 (Matthews 1995). To identify suitable sites for structural studies, R1 was introduced, one residue at a time, along the entire length of the helix. Surface sites 41 and 44, where R1 residues have two-component spectra, were selected for investigation. The results suggest that the more immobile states arise not from tertiary interactions but from interactions of the nitroxides with side chains in helix B itself.

Results

R1 motion in helix B of T4L

Figure 2A shows the location of residues 40–50 on helix B of T4L. The EPR spectra of R1 at each position are shown in Figure 2B. The spectra of R1 at positions 40, 44, and 48 were previously reported (Mchaourab et al. 1996) and are reproduced here for comparative purposes. EPR spectra of R1 at buried or partially buried sites 42, 43, 46, and 50 have, as expected, a dominant strongly immobilized component identified by well-resolved outer hyperfine extrema separated by a large 2A_zz′ (shown for 50R1, Fig. 2B). Residue 45R1, at a site of tertiary contact, also has a strongly immobilized state. Interestingly, the spectrum of 46R1 shows, in addition to a strongly immobilized state, a second (but minor) component corresponding to a highly mobile state (arrow, Fig. 2B). Residue L46 was found to give the largest destabilization upon mutation (Matthews 1995), and the mobile component likely arises from a locally unfolded state. This will be investigated in a future study.

Figure 2.

Sites in T4L where R1 was placed along helix B and the corresponding EPR spectra. (A) Ribbon diagram of the T4L structure. Spheres at the C_α atoms identify residues where R1 has been introduced. The C_α atoms of the most solvent-exposed sites (fractional side-chain solvent accessibility >0.4) are shown in magenta. (B) EPR spectra of R1 at the indicated positions. Arrows “i” and “m” indicate spectral components that arise from immobile and mobile states of R1, respectively. Resolved hyperfine extrema (2A_zz′) are identified for 50R1.

The most solvent-exposed sites (fractional solvent accessibility of the side chain >0.4) are 40, 41, 44, and 48 (magenta spheres, Fig. 2A). Of these, only the spectrum of 48 appears to have a single component while the other three clearly consist of mobile and relatively immobile components (m and i arrows, respectively, Fig. 2B). To identify the origins of the more immobile component, sites 41 and 44 were selected for crystallographic and mutagenic analyses.

Structure and interactions of 41R1

The structure of 41R1 was determined and refined to 1.4 Å with R _work = 15.2% and R _free = 20.2%. The data collection and refinement statistics are given in Tables 1 and 2. The electron density of the 41R1 side chain is shown in Figure 3A. Residue 41R1 is located at a crystal contact interface. The electron density for the 41R1 side chain is well-resolved from the backbone to S_δ of the disulfide group (Fig. 3A); electron density for the nitroxide ring is partially resolved. Rotamers of R1 will be specified using the notation of Lovell et al. (2000), m (minus), p (plus), and t (trans), where m is nominally −60°, p = +60°, and t = 180°. For χ₁, the angle is defined such that χ₁ = 0 when S_γ eclipses the N backbone atom of the side chain. This notation is retained for variations of ∼±30° about the nominal value. Because electron density is not resolved beyond the S_δ atom in all structures, rotamers of R1 will be designated according to χ₁ and χ₂ dihedrals.

Table 1.

Summary of data collection statistics^a

Table 2.

Summary of refinement statistics

Figure 3.

Structure of 41R1. (A) Electron density superimposed on a stick model of 41R1. The 2F _o−F _c map is contoured at 1.3σ. (B) Distance of the S_δ atom of the disulfide group from the H atom modeled on 41C_α. (C) Distances of atoms in the nitroxide ring from nearest neighbors in the lattice. A symmetry-related molecule is colored cyan, and residues belonging to this molecule are indicated with a prime (′). (D) Space-filling model of 41R1 and neighboring groups.

The 41R1 side chain adopts a {t,p} state (χ₁ = −175°, and χ₂ = 57°). This rotamer was also observed in the structures of 119R1 and 65R1 (Langen et al. 2000) and is apparently one of the preferred rotamers of R1 in helices. In the {t,p} rotamer, the S_δ atom lies close to the hydrogen atom modeled on C_α with a distance of 3.2 Å (Fig. 3B), suggesting a possible C_αH···S_δ interaction. A close apposition of S_δ to the C_αH is also characteristic of {m,m} rotamers previously reported (Langen et al. 2000; Guo et al. 2007). Sufficient electron density is resolved for the nitroxide ring to determine χ₃ as ≈86°, which is consistent with values observed in previous structures (Langen et al. 2000; Guo et al. 2007). This value of χ₃ places the nitroxide ring of R1 in a position to contact side chains in a symmetry-related molecule (Fig. 3C,D). Interestingly, the interaction with R119′ of the symmetry-related molecule is similar to the intrahelical interaction of 115R1 with R119 at i + 4 (Guo et al. 2007).

Figure 4A (left panel) shows the EPR spectrum of 41R1 in solution (solid trace) and a fit to a two-component MOMD model (dashed trace). A satisfactory fit is obtained with a z-axis anisotropic motion for the mobile component, characteristic of R1 on other helix surface sites (Columbus et al. 2001), and isotropic slow motion for the immobilized state. Figure 4A (right panel) shows the individual components derived from the simulation.

Figure 4.

Effect of mutations of neighbor residues on the EPR spectra of 41R1 in solution. (A−E) EPR spectra of 41R1 in the context of indicated mutations (solid traces) and fits to 41R1 and 41R1/E45A (dashed traces). The individual components determined from the fits are shown to the right. For 41R1, i and m components for the fit have τ = 10.5 ns (isotropic) and τ = 1.5 ns, S = 0.37, and β_D = 36, respectively. For 41R1/E45A, i and m components for the fit have τ = 10.5 ns (isotropic) and τ = 1.6 ns, S = 0.31, and β_D = 36, respectively. Arrowheads on the mobile component of 41R1 indicate resolved parallel and perpendicular orientations with respect to the magnetic field. (Inset) A diagram showing the relationship between the nitroxide molecular frame (x _M, y _M, and z _M) and the z-axis of the diffusion tensor (z _R) for z-axis anisotropic motion. For simplicity, the methyl substituents at positions 2 and 5 of the nitroxide ring are not shown. As is customary, z _M lies along the nitroxide p orbital, x _M lies along the N−O bond, and y _M is selected for a right-handed coordinate system.

The z-axis anisotropy arises from a constrained motion of the z _R axis of the diffusion tensor in a cone whose symmetry axis lies along (or at a tilt angle of β_D with respect to) the nitroxide z _M molecular axis (Fig. 4, inset). The z-axis anisotropy is recognized in the EPR spectrum by resolved features that correspond to parallel and perpendicular orientations of z _R with respect to the magnetic field (arrowheads, Fig. 4A, right panel). In the MOMD model, the amplitude of z _R motion is constrained by a restoring potential from which an order parameter (S) is derived. It has been suggested that the physical origin of the restoring potential is limited torsional oscillations about χ₅ which together with oscillations about χ₄ give rise to the anisotropic motion (the “χ₄/χ₅ model”) (Columbus et al. 2001); motions about χ₁ through χ₃ are strongly constrained due to the apparent C_αH···S_δ interaction described above for both {t,p} and {m,m} rotamers and the high-energy barrier to rotate about χ₃ (Fraser et al. 1971). Thus, the {t,p} rotamer in the crystal structure of 41R1 can account for the observed z-axis anisotropic motion of the mobile component in solution.

The immobile state in solution must arise from interactions of the ring with nearby groups. The identity of the groups cannot be determined from the crystal structure due to the intermolecular contacts of R1 in the lattice; so an approach was taken in which potentially interacting side chains were mutated. Replacement of the E45 residue (i + 4) with alanine dramatically reduced the immobile component (Fig. 4B, left panel, solid trace). The spectrum of 41R1/E45A is again well-fit (dashed trace) with two components similar to those in 41R1 (Fig. 4B, right panel), but with the immobile component decreased to ≈10% compared to ≈40% in 41R1. This result suggests an intrahelical interaction of 41R1 with E45; a model for this putative interaction is considered in the Discussion. The mutant 41R1/E45D gives a single-component spectrum corresponding to highly ordered z-axis anisotropy (Fig. 4C), while 41R1/E45Q results in a spectrum similar to 41R1 but with an enhanced population of the immobile component (Fig. 4D). These results clearly show that the motion of R1 is modulated by the identity of the i + 4 side chain.

NMR evidence suggests that isomerization of P37, four residues from 41, may result in slow conformational exchange involving the B helix (Heinz et al. 1994). If such exchange exists, it does not contribute to the complex motion of 41R1 as shown by the essentially identical spectra of 41R1 and 41R1/P37A (Fig. 4E).

Structure of 44R1

The structure of 44R1 was determined and refined to 2.1 Å with R _work = 20.2% and R _free = 25.0%. The data collection and refinement statistics are given in Tables 1 and 2. The electron density maps of the 44R1 side chains are shown in Figure 5A. There are three molecules in the asymmetric unit, and the chains are designated A, B, and C; residues within the structures are designated with a corresponding subscript. Both 44R1_A and 44R1_B side chains are located at crystal contact sites, and the electron densities of the nitroxide rings are completely resolved (Fig. 5A). The side chains adopt {m,m} states (Table 3) in which the distances from the S_δ to a hydrogen modeled on C_α are 2.8 and 2.7 Å for 44R1_A and 44R1_B, respectively (Fig. 5B), similar to the {m,m} state for R1 at noncontact sites (Langen et al. 2000; Guo et al. 2007).

Figure 5.

Structures of 44R1. (A) Electron density superimposed on a stick model of 44R1 corresponding to the A, B, and C chains in the lattice; residue E45 is also shown. The 2F _o−F _c maps are contoured at 1.3σ, 1.2σ, and 1.1σ for 44R1_A, 44R1_B, and 44R1_C, respectively. (B) Distances from the S_δ of the disulfide to the H atom modeled on 44C_α are shown. Contacts of the nitroxide ring of 44R1_A and 44R1_B with symmetry-related molecules are too numerous to show clearly.

Table 3.

Summary of side-chain dihedral angles and rotamer designation

The 44R1_C side chain is not at a lattice contact site; the nitroxide is 14 Å from the closest point of a symmetry-related molecule. The electron density of the nitroxide ring in 44R1_C is not resolved, but the disulfide group has strong density (Fig. 5A) defining a {t,m} rotamer (χ₁ = 173° and χ₂ = −96°). This rotamer, observed only in one other structure (Guo et al. 2007), may be stabilized in part by interactions of the disulfide S_δ atom with the backbone carbonyl oxygen of 44R1_C (3.8 Å) and the backbone amide nitrogen of Glu 45 (3.7 Å) (Fig. 5B). The large absolute value of χ₂ places S_δ relatively far from the backbone, suggesting a relatively weak interaction.

Structural comparison of the A, B, and C chains in the asymmetric unit shows that these three chains have different conformations. The structures of the N-terminal subdomains of the different conformations are virtually superimposable, as are the C-terminal subdomains. The difference between the three 44R1 chains is the hinge-bending angle between the N- and C-terminal subdomains. Differences in the hinge-bending angle that moves one subdomain relative to the other and opens or closes the active site cleft are commonly observed in T4L crystal structures (Zhang et al. 1995). Relative to wild-type, the hinge-bending angles for the A, B, and C structures are calculated to be 3.2°, −4.5°, and 15.8°, respectively, using the program DynDom (Hayward and Berendsen 1998).

The average C_α B-factor for the entire N subdomain in chain C is 39 Ų, substantially higher than that for the A and B chains (23 Ų and 28 Ų, respectively) due to the reduced contact between subdomains in the open 44R1_C structure and the lack of lattice contact in the B helix. Nevertheless, the side-chain conformations of most residues are identical in chains A, B, and C and the WT structure (PDB ID: 3LZM). Exceptions are residues Phe 4, Phe 67, Asp 72, and Phe 104 that adopt unique conformations in chain C. Conformational changes of these residues have previously been associated with hinge-bending motion in T4L (Zhang et al. 1995).

Interactions of 44R1 in solution

The single {t,m} rotamer for 44R1_C and lack of density for the nitroxide ring leave unresolved the origin of the two spectral components for R1 at this site. To explore possible interactions of 44R1 in solution, side chains within reach of 44R1 were mutated to alanine. The locations of these mutated sites are shown in Figure 6A and include the i ± 1, i ± 4, and i + 3 residues located on the same helix and residues located on a nearby loop (N53, N55). Because NMR studies showed that cis–trans isomerization of Pro 37 results in a slow conformational isomerization of T4L (Heinz et al. 1994), proline 37 was also mutated to alanine.

Figure 6.

Effect of mutations of neighbor residues on the EPR spectra of 44R1 in solution. (A) Ribbon diagram of T4L showing the position of 44 (sphere) and neighboring residues (stick models). (B) EPR spectra of 44R1 in the context of indicated mutations. For comparison, the 44R1 spectrum (gray trace) is superimposed on all other spectra.

The EPR spectra of 44R1 in the context of these mutations are shown in Figure 6B, where the spectrum of the mutant (black trace) is compared with the spectrum of 44R1 alone (gray trace). As can be seen, all mutations except E45A leave the populations of mobile and immobile components relatively unchanged; in 44R1/E45A the immobile component is substantially reduced suggesting an interaction with the i+1 glutamate. As shown in Figure 5A,B, the {t,m} rotamer projects the R1 side chain in a direction to allow interactions with E45. Although the electron density for E45 is weak in the 44R1_C structure (Fig. 5A), the density that is resolved suggests that E45 has changed conformation from the {m,m} rotamer in the WT protein (PDB ID: 3LZM) to an {m,t} rotamer, further supporting the existence of an interaction with R1. Glutamate 45 is the same side chain with which 41R1 is proposed to interact. The possibility that the changes in both 41R1 and 44R1 are due to a general alteration in protein structure due to the 45 mutation will be considered in the Discussion, along with models for potential interactions of 41R1 and 44R1.

In contrast to 41R1, the EPR spectrum of 44R1 is significantly better fit with an X-axis anisotropic motion for the mobile component Figure 7A (dotted trace); a comparison of the best fits to X- and z-axis anisotropic motions is given in Supplemental Figure S1. The individual components derived from the fit are shown in the right-hand panel. The spectrum of the 44R1/E45A mutant is also fit well with two components similar to those in 44R1 (Fig. 7B) but with the immobile population amounting to only ≈10% compared to ≈30% in 44R1 alone.

Figure 7.

Analysis of the EPR spectra of 44R1 and 80R1 with an X-axis anisotropy model. (A−C) EPR spectra of 44R1, 44R1/E45A, and 80R1 (solid traces) together with fits based on X-axis anisotropy for the mobile component (dashed traces). Individual components derived from the fits are shown to the right. For 44R1, i and m components for the fit have τ = 13 ns (isotropic) and τ = 1.9 ns, S = 0.44, β_D = 90, γ_D = 36, respectively. For 44R1/E45A, i and m components for the fit have τ = 13 ns (isotropic) and τ = 2.0 ns, S = 0.34, β_D = 90, γ_D = 36, respectively. The fit for 80 R1 has τ = 1.5 ns, S = 0.37, β_D = 90, γ_D = 35. (Inset) Diagram showing the relationship between the nitroxide molecular frame (x _M, y _M, and z _M) and the z-axis of the diffusion tensor (z _R) in X-axis anisotropy. For simplicity, the methyl substituents at positions 2 and 5 of the nitroxide ring are not shown.

In the case of X-axis anisotropy, the z _R axis of the diffusion tensor is constrained to move in a cone whose symmetry axis is aligned along (or at a tilt of γ_D with respect to) the x _M molecular axis of the nitroxide, rather than the z _M axis of the nitroxide as for z-axis anisotropy (Fig. 7, inset). Although the order parameter and correlation time for the mobile components are similar to those of 41R1, the spectral line shapes are clearly distinct (cf. Figs. 4A and 7A). In particular, spectral features corresponding to parallel and perpendicular orientations of the nitroxide z _M axis with respect to the magnetic field are not resolved as they are in z-axis anisotropic motion.

The apparent X-axis motion is not unique to the {t,m} rotamer. For example, one rotamer of T4L115R1 is a {t,m} state, but the motion is consistent with z-axis anisotropy (Guo et al. 2007). In addition, the structure of 80R1 in T4L shows R1 to be in an {m,m} state (Langen et al. 2000), and the spectrum is best fit by a single component with X-axis anisotropy (Fig. 7C and Supplemental Fig. S1).

Discussion

The first study to correlate R1 side-chain motion with protein structure in T4L showed that the motion of R1 at several solvent-exposed helical sites, where tertiary interactions are unlikely, was independent of the nearest neighbor residues at positions i ± 3 and i ± 4 (Mchaourab et al. 1996). The results presented here show that this is not general and apparently depends on the rotameric state of R1 and, for the cases studied, the identity of the residue at i + 4 (41R1) or i + 1 (44R1) in the helix. In the following paragraphs, the rotamers of R1 based on crystallographic data obtained so far are summarized, and how they may relate to local interactions is discussed.

Rotamers of R1 in helices

The {χ₁,χ₂} rotamers of R1 in helices determined from previously reported crystal structures (Langen et al. 2000; Guo et al. 2007) and those reported here are shown in Figure 8. The {m,m}, {t,p}, and {t,m} rotamers are apparently preferred for R1 at such sites. Rotamers with χ₁ in the p configuration are not allowed in helices due to steric clashes with the backbone. Although allowed on steric grounds, rotamers of R1 with χ₂ in the t configuration are so far rare and represented by only a single example, an {m,t} rotamer for T4L 75R1 located at a tightly packed crystal contact site (Langen et al. 2000). In a total of 19 structures determined so far (summarized in the thesis of Fleissner [2007]), the {m,t} rotamer has not made another appearance and a {t,t} rotamer has yet to be observed. In the {m,t} and {t,t} rotamers, the S_δ atom is moved away from the backbone, eliminating the presumed attractive interactions of S_δ with the backbone C_αH atom (as for {m,m} and {t,p}) or with the backbone peptide group (as for {t,m}), accounting for their rarity (or absence). In the T4L 75R1 {m,t} rotamer, the S_δ has surrogate contacts with other atoms in a symmetry-related molecule.

Figure 8.

Rotamer preferences of R1 at solvent-exposed helical sites. Side-chain dihedral angles χ₁ and χ₂ from the present work are shown by circles, and those from Langen et al. (2000) and Guo et al. (2007) are shown by triangles. Dashed boxes identify ±30° about the average values for the dihedral angles.

For native side chains, it is generally the case that the set of preferred rotamers is the same at lattice contact and noncontact sites. It also appears that this will be the case for R1, at least with respect to {χ₁,χ₂}, because the data in Figure 8 were derived from structures in which R1 is at both lattice contact and noncontact sites. Thus, it is likely that the rotamers shown in Figure 8 also represent those found for R1 on helices in solution.

In quantum mechanical calculations aimed at enumerating the sterically allowed rotamers of R1 at solvent-exposed sites in helices, Tombolato et al. (2006) identified the {m,m}, {t,p}, {m,t}, and {t,t} configurations for {χ₁,χ₂}. However, the experimental data and the above arguments suggest that the {m,t} and {t,t} rotamers will not likely be important for R1 at such sites. The fact that the experimentally observed {t,m} rotamer was not predicted may be due to a requirement of intramolecular interactions involving the nitroxide ring for stabilization (as in 44R1_C); such interactions were not included in the calculations. For dihedral angles further along the side chain, the calculations predicted (nominally) χ₃ = ±90° and χ₄ = ±75° or 180°. For χ₄ = ±75°, χ₅ was predicted to be near 0° or ±100°; for χ₄ = 180° the potential energy for χ₅ was found to be rather similar for all values between ±77°. As will be discussed in future publications, these values for χ₃–χ₅ are not in conflict with available experimental data on R1 and will be employed where needed in the models presented below.

Potential interactions of R1 rotamers in helices

The structures presented here provide examples for each of the three commonly observed {χ₁,χ₂} rotamers of R1 shown in Figure 8. These structures provide a basis for constructing models of the R1 side chain to reveal potential rotamer-specific interactions of the nitroxide ring that could account for immobilized states of R1. As discussed below, the results of modeling add support to conclusions drawn from mutagenesis studies. To construct models of R1 for each of the three {χ₁,χ₂}rotamers, values of χ₃–χ₅ are taken either from experimental data where available or from Tombolato et al. (2006).

Figure 9A (left panel) shows the {m,m} rotamer of the 44R1_A structure (Table 3) with the i ± 3, i ± 4, and i + 1 side chains shown in space-filling form. In this case, the entire side chain was resolved in the crystal structure, and the model is based on the complete structure (Fig. 5). Although 44R1_A is located at a lattice contact site, the values of χ₃, χ₄, and χ₅ (−95°, 76°, −86°) are reasonably close to one set of allowed values predicted by Tombolato et al. (2006) for R1 in the absence of interactions (−90°, 75°,−100°). In this conformation, the nitroxide ring lies between the neighboring side chains, and is not sufficiently close to any of them for a stable interaction at room temperature; allowed rotations about χ₄ to −75° or 180° or any rotation about χ₅ does not alter this conclusion. The ring is closest to K48 at the i + 4 position but can make only limited contacts with that side chain. Because the disulfide dihedral (χ₃) is nominally ±90°, another possible conformation that could exist in solution is that formed by isomerization of the χ₃ bond to 90° (from −95°) (Fig. 9A, right panel). Again, the ring is too far from any side chains to establish immobilizing interactions for any allowed combinations of χ₄ and χ₅. Although the models of Figure 9A are based on the specific structure of 44R1_A, the conclusions are general for the {m,m} rotamer at helix sites, and this rotamer may account for the experimental result that the motion of R1 at some sites in T4L is not influenced by steric interactions with nearest neighbors (Mchaourab et al. 1996).

Figure 9.

Possible intrahelical interactions of R1. (A) Left panel: R1 in the {m,m} rotamer of the 44R1_A crystal structure (Table 3). Right panel: 44R1_A in a conformation in which χ₃ has isomerized from −95° to 90°. (B) Left panel: R1 in the {t,p} rotamer of the 41R1 crystal structure (Table 3). The nitroxide ring was modeled using χ₄ = −75° and χ₅ = 100°. Right panel: 41R1 in a conformation in which χ₃ is isomerized from 86° to −96° and χ₄ and χ₅ are −80° and 0°, respectively. (C) R1 in a {t,m} rotamer. Dihedral angles χ₃, χ₄, and χ₅ are modeled as −90°, 180°, and 28°, respectively.

Figure 9B shows a similar representation of the {t,p} rotamer of the 41R1 structure (Table 3). The nitroxide ring was only partially resolved in the structure (Fig. 3). In the left panel of Figure 9B the ring is positioned using χ₄ = −75° and χ₅ = 100°, compatible with the partially resolved ring density. In this configuration the ring does not make significant contacts with neighboring side chains; selection of other allowed values of χ₄ or χ₅ does not alter this conclusion. Again, an alternative conformation in solution is formed by isomerization of the disulfide along with some changes in the other dihedrals to minimize steric overlaps with E45 (Fig. 9B, right panel). In this conformation, the nitroxide ring lies in van der Waals contact with E45 at the i + 4 position. This could account for the immobilized state of 41R1 in solution and the dramatic drop in population of the immobilized state in E45A; slow isomerization of the disulfide could give rise to equilibrium between mobile and immobile states represented by the structures in the figure.

Figure 9C shows a model of the {t,m} rotamer of 44R1_C. In the structure of 44R1_C ring density was not resolved. The value of χ₃ is modeled as −90°; isomerization of the disulfide is not possible due to steric clashes with the helical backbone. Dihedrals χ₄ and χ₅ are modeled as 180° and 28°, respectively, compatible with quantum mechanical predictions (Tombolato et al. 2006) and selected to minimize steric overlap with E45. In this configuration, the nitroxide ring makes van der Waals contact with E45 that could account for the immobile state of 44R1 in solution and the reduction of the immobilized population in the E45A mutant. Interestingly, residue E45 in the 44R1_C structure appears to have isomerized to the {m,t} state from the {m,m} state in the WT structure, a change which may optimize the interaction with 44R1. The origin of the more mobile state of 44R1 in solution is unclear. It may arise from isomerization of χ₄ which would move the ring away from E45, or from isomerization of E45 itself. Alternatively, in solution at room temperature {m,m} or {t,p} states may be populated that are not observed in the low-temperature crystal structure. At 100K, enthalpic interactions between 44R1 and E45 may select the {t,m} state. Low temperature rotamer selection was previously observed for 115R1 (Guo et al. 2007). Unfortunately, the already high B-factor in helix B of the 44R1_C structure at 100K (average backbone B-factor of 50 Ų) precludes determination of a sufficiently high-resolution room-temperature structure to check this point.

The above models provide reasonable explanations for the existence of immobilized states of 41R1 and 44R1 in solution. Moreover, they suggest in general that {t,p} and {t,m} rotamers of R1 with allowed values of χ₃– χ₅ can position the nitroxide ring to enable interactions with i + 4 and i + 1 residues, respectively. The strength of the interaction will depend on the identity and configuration of the neighboring residue, and the data presented here suggest that glutamic acid is one that can have substantial attractive interactions with the nitroxide ring. On the other hand, the {m,m} rotamer of R1 allows little or no interaction with neighboring residues in a helix, emphasizing the rotamer-specific nature of the interactions. The nature of the interactions that determine particular rotamers of R1 at a given site are not yet fully understood and will be the subject of a future study. However, interactions of the nitroxide with neighboring side chains will clearly be one of the determinants.

The above discussion has focused on the simplest model for the origin of the two-component EPR spectra of 41R1 and 44R1, namely, slow physical exchange between two (or more) rotamers of the side chain, one of which has an interaction with a neighboring side chain. It should be noted that some slow conformational exchange process involving the B helix could also be involved, either in modulating the conformation of E45 or the R1 rotamer populations themselves.

For both 41R1 and 44R1, E45 is a common site of the suggested interactions that determine the motion of the respective nitroxides. Thus, it is possible that the mutation E45A produces a global change in the B helix structure that is reflected in changes in mobility of both 41R1 and 44R1; indeed, E45 appears to make a salt bridge with K48 in the WT structure. Two arguments suggest that this is not the case. First, mutations E45A and K48A separately and together produce essentially no change in the thermal stability of the protein (Heinz et al. 1994). Second, the mutation E45A in the context of the 41R1 and 44R1 proteins do not result in significant line shape changes in the individual components of the complex EPR spectra but only of the relative populations of the components. Moreover, the mutation K48A, which disrupts the putative salt bridge, has no effect on the 44R1 spectrum. Because the EPR spectra are very sensitive indicators of the local structure and dynamics, this argues against global structural changes due to the E45A mutation. One must also consider the possibility that the A41 and S44 residues are themselves important for protein structure and that mutation to the spin label produces changes in structure. This is unlikely since mutations of these residues to several others produced essentially no change in protein stability or function (Matthews 1995). Indeed, extensive mutational analysis of the B helix revealed that mutation of only the buried or partially buried sites 43 and 46 produced significant destabilization. Note that these are the sites at which R1 is strongly immobilized and produces partial unfolding at 46 (Fig. 2).

Motion of 44R1

The best MOMD fit to the X-band EPR spectrum of 44R1 is obtained with X-axis anisotropy for the mobile component. A multifrequency EPR study of 44R1 under the present conditions may be necessary to convincingly resolve the relatively subtle differences between X- and z-axis motions for R1 at this site (Supplemental Fig. S1) (Barnes et al. 1999). Nevertheless, there is little doubt that X-axis motions can occur for R1 in proteins as illustrated by T4L 80R1, where the distinction is more clearly reflected in the spectral line shapes (Fig. 7; Supplemental Fig. S1).

The X-axis anisotropic motions could arise from relatively unrestricted motion about χ₅. In this case, the z _R diffusion axis would lie along the χ₅ bond (Fig. 7, inset), approximately the direction chosen for the simulations. As suggested by quantum-mechanical calculations (Tombolato et al. 2006), this could occur if χ₄ adopts a trans configuration (180°) wherein χ₅ has a flat potential-energy surface between ±77°. This origin of X-axis anisotropy was previously suggested for the extreme X-axis anisotropic motion of R1 at sites 214 and 300 in the G_α subunit of transducin (Van Eps et al. 2006). Alternatively, backbone fluctuations could contribute; indeed, the B helix has high backbone B-factor in the 44R1_C structure (50 Ų compared to 30 Ų for the whole C chain). It should be pointed out that the two-component fits to the EPR spectra assume the two states are in slow exchange on the EPR timescale. No information is currently available on the rates of R1 rotamer exchange, and it is possible that details of the line shapes may be influenced by an exchange process.

In summary, the data presented above add to the rotamer library being accumulated for the R1 side chain, and the simple models presented provide reasonable explanations for the existence of the immobilized states of the R1 side chain at 41 and 44 in terms of intrahelical interactions of particular R1 rotamers with neighboring side chains. Knowledge of the R1 rotamers provides a basis for evaluating computational methods aimed at modeling R1 (Tombolato et al. 2006) and a foundation for estimating the contribution of R1 “flexibility” to inter-spin distance distributions determined by the increasingly popular DEER and DQC methods (Berliner et al. 2000). Elucidation of the possible interactions of the nitroxide in this and a previous study (Guo et al. 2007) provides a beginning toward interpreting the multicomponent EPR spectra frequently found for R1 at helical sites.

Materials and Methods

Construction, expression, purification, and spin labeling of T4L mutants

The template T4L gene used for mutagenesis was that coding for the cysteine-free “pseudo-wild-type” protein containing the substitutions C54T and C97A (Matsumura and Matthews 1989; Nicholson et al. 1991) kindly provided by F.W. Dahlquist (University of California, Santa Barbara). Cysteine mutants at positions 40, 44, 48, and 80 have been previously reported (Mchaourab et al. 1996, 1999). Other mutants were constructed using the overlap extension PCR method as described by Ho et al. (1989). Genes for all mutants were verified by sequencing.

T4L proteins were expressed and purified as previously described (Mchaourab et al. 1996; Columbus et al. 2001). Protein purity was at least 95% judged by SDS-PAGE. Spin labeling of single-cysteine substitution mutants in labeling buffer (50 mM MOPS, 25 mM NaCl, pH 6.8) was carried out at 4°C overnight with a 10-fold molar excess of 1-oxyl-3-methanesulfonythiomethyl-2,5-dihydro-2,2,5,5-tetramethyl-1H-pyrrole to generate the R1 side chain. Excess spin label was removed using a HiTrap (Amersham) desalting column, and the labeled proteins were concentrated using YM-10 Microcon filter concentrators (Millipore). The spin-labeling reagent was synthesized as reported (Hankovszky et al. 1980; Berliner et al. 1982)

Crystallization of T4L containing nitroxide side chains

Crystallization of spin-labeled T4L proteins was carried out under the nonreducing conditions described previously for a T4L mutant with an engineered disulfide bond using the hanging-drop vapor diffusion method with a mixture of sodium and potassium phosphate as the precipitant (Jacobson et al. 1992). Crystallization trials were set up at room temperature (∼21°C) by mixing 2 μL of protein solution (10–20 mg/mL in 50 mM MOPS, 25 mM NaCl, pH 6.8) with an equal volume of well buffer (2.0 M Na/KPO₄, 0.24 M NaCl, 40 mM 2-hydroxyethyl disulfide) of four different pH values (pH 6.2, 6.6, 6.9, 7.2). Drops of the protein solution were suspended over 1 mL of the same buffer. The crystals of 41R1 appeared within a week, and crystals of 44R1 appeared in about 8 weeks.

Data collection, structure determination, and refinement

For data collection, crystals were flash-frozen in a cryostream of N₂ gas at 100K using mineral oil as cryoprotectant. Diffraction data were collected at 100K at Brookhaven National Laboratory using beamline X8C of the National Synchrotron Light Source. Data reduction and scaling were done with the programs DENZO and SCALEPACK (Otwinowski and Minor 1996). Data processing statistics are listed in Table 1. The crystals of 41R1 grew in the space group P3₂21 and were isomorphous with WT T4L crystals with one molecule per asymmetric unit. The crystals of 44R1 grew in the space group P2₁2₁2₁ and were non-isomorphous with WT crystals with three molecules per asymmetric unit.

The structure of 44R1 was determined by molecular replacement using the program EPMR (Vagin and Teplyakov 1999). The WT T4L coordinates (PDB ID: 3LZM) were used as a search model with residues 44, 54, and 97 replaced by glycine. The structure of 41R1 was determined by rigid-body fitting of the WT T4L structure using EPMR (Kissinger et al. 1999) because the crystals of 41R1 were isomorphous with WT T4L crystals. The structures of 41R1 and 44R1 were refined using the programs SHELXL (Sheldrick and Schneider 1997) and CNS (Brunger et al. 1998), respectively. The refinement was monitored using the free R factor. A subset (5%) of the diffraction data was chosen randomly and omitted from refinement for calculation of the free R factor. The model was rebuilt manually during refinement using the program O (Jones et al. 1991). The spin-label side chains were manually built when the R factor dropped to ∼20%. The final structures were validated with the programs ProCheck (Laskowski et al. 1993), SFCHECK (Vaguine et al. 1999), Errat (Colovos and Yeates 1993), and WHAT_CHECK (Hooft et al. 1996). The final refinement statistics are reported in Table 2. The final coordinates and structure factors have been deposited in the Protein Data Bank under entries 2Q9D and 2Q9E for 41R1 and 44R1, respectively.

EPR spectroscopy and spectral simulations

EPR spectra of spin-labeled proteins (from 100 to 500 μM) were recorded on a Varian E-109 X-band spectrometer fitted with a loop-gap resonator at room temperature using a microwave power of 2 mW and modulation amplitude optimized to the natural line width of each individual spectrum. All protein samples contained 30% (w/w) sucrose in order to increase solution viscosity and thereby minimize the contribution of protein rotation to the EPR spectral line shape (Mchaourab et al. 1996, 1999).

Experimental EPR spectra were fit to a one- or two-component MOMD model using the program NLSL developed by Freed and co-workers (Schneider and Freed 1989; Budil et al. 1996). Each spectrum to be fit arises from R1 at a solvent-exposed helix site, and appropriate starting values for the elements of the A and g magnetic tensors were taken as g_xx = 2.0078, g_yy = 2.0058, g_zz = 2.0022 and A_xx = 6.2, A_yy = 5.9, A_zz = 37.0 (Kusnetzow et al. 2006). Axially symmetric motion is assumed for simplicity; in terms of the principle values of the rotational diffusion tensor, R_z ≡ R _par, R_x = R_y ≡ R _perp. In the modified spherical tensor representation the motion is described by the geometric mean rotational diffusion constant <R> = (R _par R _perp ²)^1/3 and the rotational anisotropy parameter N = R _par/R _perp (Budil et al. 1996). The effective correlation time is computed as τ = 1/6<R>. To constrain the number of fitting parameters, immobilized states are taken to have a simple isotropic motion (N = 1), as would be the case if R1 motion is determined by rotational diffusion of the roughly spherical T4L molecule. The results justify this assumption. The more mobile state in two-component spectra is fit by the MOMD model (Budil et al. 1996) that describes anisotropic motion. Spatial ordering of the diffusion tensor that gives rise to the anisotropic motion is accounted for by an order parameter, S, computed from the single C ₂₀ coefficient of the ordering potential varied in the simulations (Budil et al. 1996); the resulting motion is that of diffusion of the z _R axis of the diffusion tensor in a cone. The tilt of the diffusion tensor with respect to the molecular axis of the nitroxide is specified by the Euler angles (α_D, β_D, γ_D). For axially symmetric motion (R_x = R_y), only β_D and γ_D need be specified. For z-axis anisotropic motion, the diffusion tilt was fixed at β_D = = 36°, γ_D = 0° as previously reported (Columbus et al. 2001). For X-axis anisotropic motion, the diffusion tilt was fixed at β_D = 90°, γ_D = 36°, placing the z _R axis of the diffusion tensor approximately along the N−O (X-axis) bond of the nitroxide. For the immobile component with S = 0, the tilt angles have no effect on the spectra.

Fitting of the two-component spectra thus involved variation of <R>, N, and C ₂₀ as dynamic parameters to describe the motion of the mobile component and only <R> for the immobile component (N = 1). The program NLSL does not allow for independent variation of the populations in two-component fits, and the relative populations are an output of the fitting. After initial fits were obtained manually by exploring parameter space, final best fits in a least-squares sense were obtained by allowing small variations of the A and g tensor values as well as the above parameters.

Electronic supplemental material

Supplemental Figure S1 shows the comparison of the MOMO fits with X- and z-axis anisotropic motions to the EPR spectra of T4L 44R1 and 80R1. This figure is included in a PDF file named Guo_Supplemental_Material.pdf.