Mapping Membrane Protein Backbone Dynamics: A Comparison of Site-Directed Spin Labeling with NMR ¹⁵N-Relaxation Measurements

Abstract

The ability to detect nanosecond backbone dynamics with site-directed spin labeling (SDSL) in soluble proteins has been well established. However, for membrane proteins, the nitroxide appears to have more interactions with the protein surface, potentially hindering the sensitivity to backbone motions. To determine whether membrane protein backbone dynamics could be mapped with SDSL, a nitroxide was introduced at 55 independent sites in a model polytopic membrane protein, TM0026. Electron paramagnetic resonance spectral parameters were compared with NMR ¹⁵N-relaxation data. Sequential scans revealed backbone dynamics with the same trends observed for the R₁ relaxation rate, suggesting that nitroxide dynamics remain coupled to the backbone on membrane proteins.

Introduction

Site-directed spin labeling (SDSL) can be used to investigate membrane proteins in many different environments (e.g., micelles and synthetic lipid bilayers) and thus provides an essential link between high-resolution methods that investigate membrane proteins in detergents (X-ray crystallography and NMR) and the more native-like lipid bilayer. SDSL has already been applied to investigate conformational switching in membrane proteins; however, the lack of investigation of the nitroxide electron paramagnetic resonance (EPR) lineshapes on membrane proteins has limited the quantification of side-chain and backbone dynamics in nanosecond (directly from the lineshape) to microsecond (interpretation of multiple spectral components) timescales. Although ¹⁵N NMR relaxation methods can be used to identify nanosecond backbone motions in proteins, only a few select polytopic membrane proteins have been investigated to date (1–5).

The approach that enabled the quantification of nanosecond backbone dynamics in soluble proteins involved crystal structures of spin labels on the model system T4 lysozyme (6–8), systematic perturbation of the internal dynamics of the nitroxide side chain (9), and comparison with NMR-determined dynamics (10). The combined data provided the rotameric preferences of the spin label and a nitroxide side-chain dynamic model, such that fast backbone modes in soluble protein could be deduced and quantified directly from the EPR lineshape (10). In membrane proteins, it is more difficult to use this three-pronged approach. The difficulty of obtaining high-resolution crystal structures, the reaction limitations of different nitroxide side-chain moieties, and the limited availability of NMR data for comparison are bottlenecks to obtaining quantitative SDSL dynamics data for membrane proteins. However, some progress has been made on both β-barrel and α-helical proteins. Crystal structures of the most common spin label, (1-oxyl-2,2,5,5-tetramethyl-3-pyrroline-3-methyl)-methanethiosfonate (MTSSL, or R1 when incorporated into a protein), on BtuB (12) and LeuT (13) have provided dynamic models and an understanding of the complex interactions the nitroxide moiety has with the membrane protein surface. The motional model proposed for membrane protein solvent-exposed α-helical sites does not preclude the quantification of backbone dynamics; however, the observed increased affinity of the nitroxide ring for the membrane protein surface could potentially reduce the sensitivity of the probe to backbone dynamics. To investigate the ability to quantify membrane protein backbone dynamics, EPR spectral measurements, scaled mobility and second moment, were compared to NMR 15N relaxation rates for a model polytopic membrane protein, TM0026 (Fig. 1) (14–16). TM0026 has two transmembrane segments and is folded, monomeric, and α-helical in decyl maltoside, dodecylphosphocholine, and several detergent mixtures, respectively (14–16). The trends in dynamics agreed well between the EPR and NMR measurements. An interesting observation, detected by both methods, was the relatively independent dynamics observed in the N- and C-terminal regions of TM2. The N-terminal region (TM2N), which precedes a proline residue and contains a GXXG motif, was more mobile than the C-terminal region of the helix (TM2C) and TM1. Overall, the data indicate that nitroxide scanning coupled with a method to evaluate the tertiary fold or function of spin-labeled mutants can provide meaningful information about the nanosecond backbone motions of membrane proteins.

Figure 1.

Topology of TM0026. Residues in gray were not spin labeled. Residues in which spin label incorporation disrupts the fold as assessed by the ¹⁵N,¹H-HSQC spectra are indicated as white with an outline. The transmembrane helices are labeled TM1 (defined by residues 5–27) and TM2, which is divided into two regions: TM2N (residues 34–45) and TM2C (residues 47–54).

Materials and Methods

Mutagenesis and protein expression

The TM0026 expression protocol was previously described (15). Briefly, all cysteine mutations were introduced using polymerase incomplete primer extension PCR (17). Plasmid containing the TM0026 gene was transformed into BL21(DE3)RIL cells (Stratagene, Santa Clara, CA) for expression in either Luria-Bertani or minimal media containing 50 μg/mL ampicillin. Cells were grown to an OD₆₀₀ of 0.6–0.8 and induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG, Research Products International; Mt. Prospect, IL) for 4 h at 310 K. ¹⁵N-labeled samples were expressed in minimal media supplemented with ¹⁵NH₄Cl. The cells were lysed in a buffer containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 1 mM tris(2-carboxyethyl)phosphine (TCEP), and pelleted for 60 min at 15,000 g. After the cell debris was removed, 10 mM n-decyl-β-D-maltopyranoside (DM; Anatrace, Maumee, OH) was added to the supernatant for 3 h at room temperature to solubilize TM0026. TM0026 was then purified by Co²⁺ immobilized metal affinity chromatography (IMAC) as previously described (14,18), with 5 mM n-dodecyl-β-D-maltopyranoside (DDM; Anatrace) and 15 mM n-decylphosphocholine (FC10; Anatrace) used as the detergent in the wash and elution buffers.

Spin labeling

Spin labeling of TM0026 mutants was performed as previously described (18). TM0026 was concentrated to ∼150–200 μM and passed through a PD-10 desalting column containing an elution buffer of 20 mM phosphate buffer (pH 6.2), 150 mM NaCl, 5 mM DDM, and 15 mM FC10 to remove TCEP and imidazole. The protein eluate was incubated with (1-oxyl-2,2,5,5-tetramethyl-3-pyrroline-3-methyl)-methanethiosfonate (MTSSL, R1; Toronto Research Chemicals, Toronto, Ontario) or the diamagnetic equivalent (1-acetoxy-2,2,5,5-tetramethyl-3-pyrroline-3-methyl)-methanethiosulfonate (R1′; Toronto Research Chemicals) at a 1:5 molar ratio of protein to spin label. The spin label was incubated with TM0026 overnight and then excess MTSSL was partially removed by passing the sample through a PD-10 desalting column. The final excess of MTSSL was removed after a 3-day incubation at room temperature using Co²⁺ IMAC. The elution fraction was concentrated and dialyzed against 4 L of 20 mM phosphate buffer (pH 6.2), 150 mM NaCl to remove imidazole for 1–2 h. Dialysis does not remove detergent, due to the low critical micelle concentration of the detergent mixture (¹H NMR was used to determine detergent concentrations (14)). The dialyzed protein was concentrated to 100 μM and spectra were recorded.

EPR spectroscopy

Protein samples (5 μL, ≈100 μM) were loaded into Pyrex capillaries (0.60 mm ID × 0.84 mm OD; Fiber Optic Center, New Bedford, MA). X-band EPR spectra of TM0026 cysteine mutants were recorded on a Bruker EMX spectrometer with an ER4123D dielectric resonator (Bruker Biospin, Billerica, MA) at room temperature. The spectra were normalized to a constant area and phase and were baseline corrected. Scaled mobility was calculated from the peak-to-peak central linewidth, δ_exp, using the expression M_s = (δ_exp⁻¹ − δ_i⁻¹)/(δ_m⁻¹ − δ_i⁻¹), where δ_i (8.4 G) and δ_m (2.1 G) are the corresponding values at the most immobilized and mobile sites observed in a protein under conditions in which the rotational diffusion of the protein does not contribute to the line width (19,20). The second moment is calculated as <H²> = ∫(B − <H>)² S(B)dB/∫ S(B)dB (spectral breadth), where <H> is the first moment (geometrical center of the spectrum), B is the magnetic field, and S(B) is the absorption spectrum.

NMR spectroscopy

Isotopically ¹⁵N, ¹H-labeled TM0026 for NMR experiments was prepared using M9 minimum medium containing ¹⁵NH₄Cl (99%; Cambridge Isotope Laboratories) as a nitrogen source. NMR samples included the addition of 10% D₂O for lock. Chemical shifts were obtained from the published TM0026 assignment (BMRB 18494) (16). NMR experiments were performed on Bruker AVANCE spectrometers operating at proton frequencies of 600 MHz and 800 MHz and equipped with Bruker 5 mm TXI cryoprobes, and recorded at 313 K. Spectra were processed with Topspin. Longitudinal ¹⁵N-relaxation, transverse ¹⁵N-relaxation experiments, and heteronuclear nuclear Overhauser effects were measured using two-dimensional ¹⁵N-¹H transverse relaxation optimized spectroscopy (TROSY)-based experiments at both 600 MHz and 800 MHz. R₁-relaxation experiments employed longitudinal delay times of 50, 100, 250, 500, and 1000 ms, and R₂ relaxation was measured with Carr-Purcell-Meiboom-Gill delays of 17, 51, 102, 204, and 492 ms. Relaxation measurements were performed at 313 K and data sets were processed and analyzed using NMRPipe (21). The D₂O exchange TM0026 sample was concentrated to 250 μL, diluted with 15 mL of D₂O, and then concentrated to a final volume of 600 μL. ¹⁵N-¹H TROSY-heteronuclear single quantum coherence (HSQC) spectra were recorded at 1, 9, and 19 h. Spectra were processed with NMRPipe utilizing relaxation data analysis features.

Results and Discussion

TM0026 structure

The resonance assignments of TM0026 were previously published (16), and the transmembrane α-helical regions (TM1 (residues 5–27) and TM2 (residues 35–54)) were identified based on carbon chemical shifts (Fig. 2 A). The deuterium exchange rates correlate with these regions (Fig. 2 B); however, overall, TM2 has faster deuterium exchange rates than TM1. A gradient in the deuterium exchange rate is observed for TM2, with the C-terminal region (TM2C) exchanging more slowly than the N-terminal region (TM2N). The deuterium exchange rate is modulated by solvent accessibility and hydrogen-bond strength. Since the accessibility should be similar for the two transmembrane helices, the data suggest that TM2 (especially TM2N) has weaker hydrogen bonding than TM2C and TM1.

Figure 2.

Secondary structure of TM0026. (A) Differences between carbon chemical shifts and random coil values are plotted versus residue number (ΔC^α − ΔC^β) = 1/3 (ΔC^α_i−1 + ΔC^α_i + ΔC^α_i+1 − ΔC^β_i−1 − ΔC^β_i − ΔC^β_i+1). (B) Deuterium exchange rates indicate that the transmembrane α-helical regions are between residues 5–27 (TM1) and 35–54 (TM2). Secondary structure is indicated with gray cartoon with waves indicating helices and the line indicating the loop.

Nitroxide scan of TM0026

The nitroxide side chain R1 was introduced at residues 1–27, 29–30, 31–38, 40–45, 47–54, and 57–61 (see Figs. S1–S3 in the Supporting Material) to investigate the dynamics captured in the EPR lineshape (e.g., side-chain versus backbone dynamics). The secondary structural elements are clearly defined by the lineshapes and lineshape measurements such as scaled mobility and second moment (Fig. 2). Narrow, sharp spectra are observed for residues 59–61, indicating a disordered C-terminus. The N-terminus spectra have multiple components and are mobile, but not to the extent of a completely disordered sequence. These residues may interact with the micelle surface. Labeled residues within the short loop that connects the two transmembrane helices have lineshapes similar to those of the N-terminus, indicating that the loop is unstructured but likely has interactions with the transmembrane helices or the micelle. The lineshapes throughout the transmembrane regions vary in terms of dynamics and the number of components. A helical periodicity in both scaled mobility and second moment in membrane protein helices is not always observed, due to 1), interactions of the spin label with the surface of the membrane protein (13); and 2), a lack of tertiary contacts to modulate the nitroxide side-chain dynamics. For a two-transmembrane protein, there is a small contact region if the helices are tilted (which is preferred (22)); thus, tertiary contacts throughout the helical segments do not exist. In addition, when helical periodicity is observed in membrane proteins, the differences between the surface and contact residues are less dramatic than those observed for soluble proteins (19). Typical lineshapes for membrane or detergent facing nitroxide side chains are observed (e.g., L7R1, S11R1, V20R1, T44R1, and F40R1). Some spectra indicate a nitroxide with highly restricted motion (e.g., A13R1, F34R1, L40R1, and L48R1) and other spectra indicate mobile nitroxide side chains (e.g., I9R1, W12R1, I19R1, V45R1, and V49R1). Before analyzing the dynamics further, we assessed the overall fold for selected spin-labeled mutants using ¹⁵N, ¹H-HSQC spectra and the diamagnetic label R1′.

Spin-label perturbations of structure and dynamics

The mobile nitroxides within the transmembrane helices suggested that some of the spin-labeled proteins may have perturbed structures. Therefore, we recorded ¹⁵N, ¹H-HSQC spectra for M1R1′–A5R1′, L7R1′, S8R1′, F10R1′, W12R1′, A13R1′, V15R1′–E17R1′, Y23R1′, and V45R1′, and compared them with the spectrum of the wild-type (Fig. S4). Of these mutants, only K4R1′, E17R1′, Y23R1′, and V45R1′ had significant line broadening and were missing backbone resonances greater than 8.2 ppm, indicating that the overall fold and/or dynamics was significantly perturbed. K4 may be important for positioning the helix at the headgroup region of the micelle (T3R1′ and A5R1′ are not perturbed). E17 is most likely protonated, since it is localized to the middle of TM1 and a counter positive charge does not appear to be in proximity. The E17 polar side chain may be important for hydrogen bonding with main-chain or side-chain (e.g., Y23, Y24, or T44) atoms. Precipitation was observed for Y23R1′ and V45R1′, indicating that these mutants destabilize the fold, which is supported by the NMR spectra. K4R1′, E17R1′, Y23R1′, and V45R1′ were eliminated from further analysis and not included in the scaled-mobility and second-moment plots (Fig. 3).

Figure 3.

TM0026 dynamics mapped with EPR spectral parameters. The scaled mobility (circles) and second moment (squares) are plotted versus residue number. Average values are indicated as black solid bars for the protein regions spanned by the bar. Secondary structure is indicated by the gray cartoon, with waves indicating helices and the line indicating the loop.

SDSL captures membrane protein backbone dynamics

To investigate whether the EPR spectral parameters (Fig. 3) indeed represent backbone dynamics (rather than purely side-chain dynamics), we measured backbone ¹⁵N R₁, ¹⁵N R₂, and heteronuclear ¹H-¹⁵N nuclear Overhauser effect (NOE) values at 800 MHz (Fig. 4). R₁ values are highly sensitive to backbone nanosecond motions, and in comparison, R₂ values are much more difficult to interpret since values decrease with nanosecond motions and increase with microsecond-to-millisecond motions. TM0026 R₁ values are lower in the transmembrane regions of TM0026 and increase at the termini. The R₁ values of the periplasmic loop are low except for residue R30. V20 in TM1 has a high R₁ value compared with the rest of the helix. TM0026 R₂ values are more variable per residue and a gradient is observed such that R₂ increases from the termini toward the loop in both transmembrane helices. The same trend is observed in the deuterium exchange rates and carbon chemical shifts for TM2, but not TM1. Since this gradient is not observed in the R₁ or ¹H-¹⁵N NOE data, the dynamics gradient is likely due to microsecond-to-millisecond backbone dynamics. An overall comparison of the EPR and R1 NMR data indicates a striking correlation.

Figure 4.

¹⁵N R₁, ¹⁵N R₂, and heteronuclear ¹H-¹⁵N NOE values for TM0026 measured at 800 MHz. Secondary structure is indicated by the gray cartoon, with waves indicating helices and the line indicating the loop. Average values are indicated solid black bars for the protein regions spanned by the bar.

Structural origins of TM0026 nanosecond backbone dynamics

The loop and C-terminal regions of TM0026 have the highest scaled mobility, indicating they are more dynamic than the transmembrane helices; however, as noted from the lineshapes, the loop is not as dynamic as the C-terminus. The loop is predominantly rigid, which is expected based on its length (six residues) and amino acid composition. G28 likely provides the backbone flexibility to break the α-helix and induces a turn. P31 provides rigidity to the linker backbone and induces the turn back toward the micelle to orient the second transmembrane α-helix. In addition, the remaining residues have bulky side chains that may restrict the backbone motion.

The EPR lineshapes of A13 on TM1, and F47 and L48 on TM2 (Figs. S2 and S3) are distinct from all other TM0026-labeled sites and reflect a highly immobilized spin label (23,24), indicating that these residues likely form the tertiary contacts between TM1 and TM2. Previously, A13R1 was used to assess the loss of this tertiary interaction in different detergents (14). Most transmembrane helices in polytopic membrane proteins are packed at an angle with respect to each other (22) and the EPR lineshapes of these three nitroxides are consistent with a single cross point at a tertiary contact among these three residues. The two helices do not move independently of each other. By calculating the overall correlation time using R₂/R₁ for TM1, the loop, TM2, or all three regions, we estimated the correlation time to be ∼13 ± 2 ns (molecular weight of the protein-detergent complex ≈ 22 ± 3 kD). Thus, the differences in nanosecond dynamics are due to local backbone oscillations. The difference in dynamics between TM1 and TM2C is puzzling. TM1 contains E17, which may form bifurcated hydrogen bonds with the backbone and thereby weaken the backbone hydrogen bonds (25,26). A comparison of the average EPR spectral parameter (Fig. 3) and R₁ NMR relaxation rates (Fig. 4) for TM2N and TM2C suggests that the dynamics of TM2N and TM2C are relatively independent, with TM2C being less dynamic than TM2N. The differences observed in the backbone dynamics of TM2 may be due to a kink formed by P46. However, a proline kink is not necessary to induce the observed dynamics, since Metcalfe et al. (27) observed a similar trend for a monomeric version of phospholamban in which a proline was not present. TM2N has two glycine residues that form a GXXG motif (28), which may contribute to the higher mobility of this region. These dynamic observations complement structural studies that suggested that transmembrane helices are flexible and hydrogen-bond shifts facilitate transmembrane dynamics (29).

Conclusions

Nanosecond backbone dynamics are reflected in membrane protein EPR lineshapes and, coupled with sequential scanning, match the trends observed in ¹⁵N R₁ relaxation experiments. As with any method involving chemical probes, care must be taken to ensure that labeling does not alter the tertiary fold and function. Despite this drawback, SDSL has distinct advantages over other methods, including small sample requirements and the ability to investigate membrane proteins (regardless of molecular weight) in a lipid bilayer.