Probing Protein Secondary Structure using EPR: Investigating a Dynamic Region of Visual Arrestin

Abstract

One key application of site-directed spin labeling (SDSL) electron paramagnetic resonance (EPR) spectroscopy is the determination of sequence-specific secondary structure in proteins. Regular secondary structure leads to a periodic variation in both side chain motion and solvent accessibility, two properties easily monitored by EPR techniques. Specifically, saturation recovery (SR) EPR spectroscopy has proven to be useful for making accessibility measurements for multiple protein structure populations by determining individual accessibilities and is therefore well suited to study the structure of proteins exhibiting multiple conformations in equilibrium. Here we employ both continuous wave and SR EPR spectroscopy in combination to examine the secondary structure of a short sequence showing conformational heterogeneity in visual rod arrestin. The EPR data presented here clearly distinguish between the unstructured loop and the helical structure formed in the crystallographic tetramer of visual arrestin and show that this region is unstructured in solution.

1 INTRODUCTION

Site-directed spin labeling (SDSL) involves introducing a unique cysteine into a protein at a site of interest and reacting the sulfhydryl group with a cysteine-specific spin label, most often 2,2,5,5-tetramethylpyrroline-3-yl-methanethiosulfonate (MTSL). The resulting side chain, commonly referred to as R1, can now be monitored by EPR, providing a great deal of information about the structure of a protein (e.g., reviewed in [1, 2]). For example, the mobility of R1 can be determined from the shape of the CW EPR spectrum. The side chain mobility reflects the degree of local interactions and thus provides a means to differentiate sites where R1 is buried in the protein interior (low mobility) from those on the solvent exposed surface (high mobility). Thus a “scan” of R1 through a sequence of regular secondary structure will reveal an oscillatory behavior of the mobility, the period of which reveals the type of secondary structure. A complementary and independent means for identifying regular secondary structure is through solvent accessibility measurements, determined via the collision rate with diffusible paramagnetic reagents in solution [3–5]. Residues buried in the protein interior or on the solvent exposed surface will experience low and high collision rates, respectively. Thus both spin label mobility and solvent accessibility will be a periodic function of sequence position. Ideal α-helices are expected to display a periodicity of 3.6 [4, 6, 7] and β-sheets a periodicity of 2.0 [4, 8, 9], while regions without regular secondary structure are predicted to display no regular periodicity in mobility or solvent accessibility [10].

The principles underlying the determination of solvent accessibility in SDSL rely on the enhancement of the nitroxide spin-lattice relaxation rate (T₁⁻¹) due to collisions with fast-relaxing paramagnetic species in solution that lead to Heisenberg exchange [3, 7, 11]. Molecular oxygen and NiEDDA have been well-characterized and often employed as non-polar and polar reagents, respectively. Enhancement of spin lattice relaxation rates can be determined by power saturation (PS) or saturation recovery (SR) EPR; the methods have previously been compared [12, 13] and both are employed in the present study.

The most common EPR method for observing spin label solvent accessibility is power saturation [14–17]. Here, the intensity of the central line in the first derivative spectrum is observed as a function of the square root of incident microwave power. The point at which the saturation curve intersects a line with half of the initial slope is defined as the accessibility parameter, P_1/2. The change in P_1/2 (ΔP_1/2) upon addition of paramagnetic reagents has been shown to be correlated with the solvent accessibility of the nitroxide. Often, the accessibility parameter Π is reported, where Π = [ΔP_1/2/ΔH_pp]/[P_1/2/ΔH_ppDPPH [3, 5]. This parameter accounts for differences in resonator efficiencies by using the P_1/2 of a standard sample, DPPH, and also compensates for differences in the spin-spin relaxation time (T₂) by factoring in the central line width (ΔH_pp). A particularly useful parameter for revealing regular secondary structure is R, where R = [Π C_R⁻¹(O₂)/[Π C_R⁻¹(NiEDDA)] and C_R denotes the concentration of relaxation agent. This is similar to the contrast parameter Φ [18], but is normalized for concentration of the relaxation agent.

Proteins in solution can adopt multiple conformations in slow exchange on the EPR time scale [19–24]. In such cases, the EPR spectrum will represent a weighted sum of the individual components, with the first derivative central line height being dominated by the most mobile component. Therefore, accessibility data derived from power saturation is strongly biased toward the most mobile and accessible population. Thus, accessibilities measured by power saturation might not reveal regular secondary structure in a sequence that adopts multiple conformations in solution.

SR EPR spectroscopy involves the direct measurement of T₁ [7, 11, 25, 26]. A high power microwave pulse is introduced in order to completely saturate the spin system. After the pulse, the spins relax to their original Boltzmann distribution, and the EPR signal returns with an exponential time course, which is fit to determine the T₁ value(s). One key advantage to this method is that the accessibilities of multiple populations of R1 in slow exchange on the T₁ time scale can be monitored individually [27]. Just as in power saturation, T₁ is measured in the presence and absence of O₂ and NiEDDA, and the change in T₁ is related to the Heisenberg exchange rate, W_exsuch that 2 W_ex = [T_1R⁻¹ – T₁⁻¹], where T_1R is the apparent spin-lattice relaxation time in the presence of a relaxation agent R, and T₁ is the spin-lattice relaxation time in the absence of any agent [13]. W_ex can also be presented in the concentration-independent form k_ex such that W_ex = k_exC_R [12]. The T₁⁻¹ in the absence of relaxation agent, which is equal to twice the electron relaxation rate W_ehas also been shown to be correlated with secondary structure [13].

Visual arrestin is a soluble protein expressed at high levels in the retinal rod cells and binds to light-activated, phosphorylated rhodopsin, thus quenching the signaling cascade in vision. A crystal structure of visual arrestin has been published, showing arrestin in a tetrameric form [28]. This crystal tetramer exists as a conformational heterodimer (α ₂β₂), showing heterogeneity in three regions. We have recently shown that at least one of these “plastic” regions (residues 68 – 79) is highly dynamic in solution and plays a significant role in binding to phosphorhodopsin [29]. The sequence 155 – 163 is a second plastic region not previously investigated. In two of the monomers of the crystal tetramer (α form) this region adopts a helical conformation, while in the other two (β form), this region is in an unstructured loop (Figure 1). The focus of the present study is to determine the secondary structure and dynamics of the 155 – 163 region in solution under physiological conditions using a combination of SDSL EPR spectroscopy techniques.

Fig. 1.

Crystal structures of visual arrestin (pdb 1CF1) in the α form (left), the β form (center), and of T4L (pbd 3LZM) (right). Regions being compared are highlighted in red (arrestin 155 – 163 and T4L 128 – 135) [28, 41]

2 EXPERIMENTAL

2.1 Sample Preparation

The bovine visual arrestin vector was obtained from Dr. V. Gurevich (Vanderbilt University), with the three native cysteine residues (C63, C128 and C143) mutated to either ASA or VSV, respectively. Unique cysteines were then individually introduced at positions 155 – 163 by PCR amplification of the arrestin plasmid in the presence of the appropriate forward and reverse primers (IDT DNA). Arrestin was expressed and purified as previously described [30]. The purified arrestin protein was labeled in 50 mM MOPS, 100 mM NaCl, pH 7.0 (MOPS/NaCl) with a ten-fold molar excess of 2,2,5,5-tetramethylpyrroline-3-yl-methanethiosulfonate spin label (MTSL, Toronto Research Chemicals) overnight at 4°C and excess label removed by extensive dialysis against MOPS/NaCl buffer. All proteins were concentrated to a final concentration of 180 µM as determined by the BCA assay (Pierce) using bovine serum albumin (BSA) as a standard.

2.2 Continuous wave (CW) EPR

CW EPR spectra were recorded on an X-band Bruker ELEXSYS 500 fitted with a super-high Q cavity. Samples (10 – 15 µl) were placed in a glass capillary and spectra were recorded at room temperature over 100 G with an incident microwave power of 10 mW, a modulation frequency of 100 kHz, and a modulation amplitude of 1 G. Typical scan times were between 5 and 40 minutes.

2.3 Power Saturation EPR

Power saturation measurements were collected on a Varian E-102 Century series spectrometer fitted with a loop-gap resonator (Medical Advances, Milwaukee, WI). Samples were contained in a gas-permeable TPX capillary and equilibrated with a stream of either nitrogen or air at room temperature, or in the presence of 5 mM NiEDDA under nitrogen. Typically, data were collected at 16 powers ranging from 0.25 mW to 60 mW and P_1/2 values determined for each sample using a LabView program written by Dr. C. Altenbach (UCLA).

2.4 Saturation Recovery EPR

Long pulse saturation recovery data were collected on an X-band spectrometer built in-house based on the design of Huisjen and Hyde [31] fitted with a loop-gap resonator. The samples were placed in a gas permeable TPX capillary under the flow of either room temperature nitrogen or room temperature air regulated by a flowmeter (Matheson Gas Products). A 300 ns saturating pulse was applied to the center line, and the recovery of the signal was then measured at the same frequency with an incident microwave power of 0.5 mW. Typical acquisition times were 5 – 15 minutes. The rotational correlation times of the proteins studied here are ~50ns, the upper limit being set by the correlation time of the arrestin tetramer; in most cases, correlation times of R1 are on the order of a few ns. In this situation, both the rotational diffusion and nitrogen nuclear contributions to the relaxations are negligible and the relaxation rates reflect only Heisenberg exchange due to added paramagnetic species and the intrinsic spin lattice relaxation [13]. In some experiments, Heisenberg exchange was introduced by molecular oxygen in equilibrium with air or by the addition of 5 mM NiEDDA. The relaxations were fit to two components according to the general expression:

$$\mathrm{y}={\mathrm{A}}_{\mathrm{o}}+{\mathrm{A}}_1{\mathrm{e}}^{-(2\text{Wel}+2\text{Wexl})\mathrm{t}}+{\mathrm{A}}_2{\mathrm{e}}^{-(2\text{We2}+2\text{Wex2})\mathrm{t}}$$

where A₁ and A₂ are the amplitudes of each component, W_e is the intrinsic spin lattice relaxation rate, (2T₁)⁻¹and W_ex is the Heisenberg exchange rate with either oxygen or NiEDDA. For a single component fit, the second term is dropped.

3 RESULTS

3.1 Crystal Structure Analysis

The program Molmol (http://hugin.ethz.ch/wuthrich/software/molmol/index.html) was used to calculate the fractional solvent exposed surface area (fsa) of the native amino acids 155 – 163 of the four monomers in the visual arrestin crystal structure, as well as a solvent-exposed helix (residues 128 – 135) of T4 lysozyme (T4L) (Figure 2). A solvent radius of 1.4 Å and precision value of 3 were used for both. There is a clear periodicity of roughly 3.6 in the T4L plot. At least one turn of a helix is observed for the 155 – 163 region of the α form of arrestin, with residue D158 pointing inward. Therefore, if this region were helical in solution, we would expect to find decreased spin label mobility and solvent accessibility at this site. The fsa values calculated for solvent exposed and solvent shielded residues are quite similar between the compared regions of the α form of arrestin and T4L. The fsa values calculated for 155 – 163 of the β form of arrestin are generally high relative to T4L and show no obvious periodicity. Therefore, if this region of arrestin were unstructured in solution, spin labels would exhibit high mobility and solvent accessibility relative to T4L.

Fig. 2.

Fractional solvent accessibility of native amino acids 155 – 163 of visual arrestin in the α form (black filled circles), the β form (gray filled circles), and of T4L 128 – 135 (open circles)

3.2 Spectral Line Shape

The shape of the CW EPR spectrum reflects the local motion of the spin label, and has three contributions: rotary diffusion of the protein, backbone motion, and R1 side chain flexibility. The spectra of arrestin individually labeled at residues 155 – 163 were recorded at both 20 and 200 µM, concentrations at which arrestin exists mainly (70%) as monomeric and tetrameric species, respectively [32]. At 20 µM, the spectra of 155R1 – 158R1 exhibit a single component with relatively fast motion, as indicated by the sharp spectral line shape. The spectra of 161R1 – 163R1 also reflect relatively fast motion, but have a second well-defined component corresponding to a more restricted motion (see arrows, Figure 3). The inverse central line width (ΔH_pp⁻¹) can serve as a semi-quantitative measure of spin label mobility, with increasing values of ΔH_pp⁻¹ implying increased mobility [1, 33]. The plot of ΔH_pp⁻¹ as a function of spin label location shows no helical periodicity (Figure 4), and residues 155 and 158 show no immobilized component which would be expected from the α form of arrestin shown in the crystal data. When compared to published ΔH_pp⁻¹ values of T4L residues 128 – 135 [13], it is obvious that the labeled arrestin residues are more mobile than even the most unrestricted helical sites in T4L. This is likely the result of the increase in backbone motion that is expected in the absence of significant secondary structure. The ΔH_pp⁻¹ values for arrestin slightly decrease when the protein concentration is increased to 200 µM, suggesting that this region is somewhat less dynamic when in the tetrameric state. This decrease in motion may be due to a decrease in rotary diffusion or an increase in backbone order upon tetramer formation. Solvent exposed arrestin residues possessing secondary structure have been shown to display similar concentration-dependent changes [32, 34], suggesting that a change in rotary diffusion is likely the reason for a drop in ΔH_pp⁻¹. None of the spectra recorded at 200 µM reflect any new immobilized component, and the ΔH_pp⁻¹ values are still larger than T4L values, suggesting that this region has an identical secondary structure in both the monomeric and tetrameric states. The lack of large spectral changes upon tetramer formation also confirms that the second component visible in some spectra is not the result of a heterogeneous mixture of oligomeric forms of arrestin.

Fig. 3.

X-band CW EPR spectra of visual arrestin spin labeled at residues 155 – 163 at 20 µM (black) and at 200 µM (gray) protein concentration. Arrows indicate a second well-defined motional component

Fig. 4.

Plot of the inverse central line width (ΔH_pp⁻¹) as a function of spin label location at both 20 µM (gray filled circles) and 200 µM (black filled circles) arrestin. Shown for reference is a plot of ΔH_pp⁻¹ values for T4L residues 128 – 125 (open circles) [13]

3.3 Power Saturation

The solvent accessibilities of spin-labeled arrestin residues 155 – 163 were determined using power saturation techniques and are reported in Figure 5 as the contrast parameter R (see Introduction) plotted as a function of residue number. This parameter describes the O₂ accessibility relative to that of NiEDDA based on measurements made on the same instrument. Any systematic differences in Π values from different laboratories are thus removed and R values can be directly compared. Because NiEDDA is strongly excluded from the protein interior while the smaller, non-polar O₂ has a finite accessibility, strong maxima occur in the R vs sequence plot for buried residues.

Fig. 5.

Plot of the concentration-adjusted contrast parameter R as a function of spin label location. R values for arrestin (filled diamonds) and T4L (open diamonds) were determined using power saturation (PS; black) and saturation recovery (SR; gray) EPR methods as described in the text

R values for the 155 – 163 sequence are compared to previously published data collected for T4L residues 128 – 135 in Figure 5. As anticipated, buried residues T4L129 and 133 correspond to the maxima in the plot that occur with a helical periodicity. All arrestin residues have R values ranging from ~6.5-9, consistent with the solvent exposed T4L residues (R=8-11; [12]). These data describe arrestin residues 155 – 163 as being both solvent exposed and without regular secondary structure.

3.4 Saturation Recovery

Representative SR decays in the presence and absence of oxygen are shown in Figure 6. The exponentials were fit to either one or two components, with the corresponding residuals (for nitrogen) shown below. In most cases two-component fits were superior to single-component fits. Previous studies of R1 in T4 lysozyme also found two-component relaxation curves and showed that the electron relaxation rate, W_edetermined either from the dominant component in a two-component fit or from the best fit to a single exponential, reveals regular secondary structure via periodic changes along a sequence [13]. The plot of W_e determined either way for arrestin 155 – 163 has no obvious periodicity, again supporting the idea of this region being unstructured (Figure 7). Arrestin residues display W_e values consistent with moderate to highly exposed residues in T4L.

Fig. 6.

Representative saturation recovery curves (top) for spin labeled arrestin under nitrogen (black) and air (20% O₂gray), as well as the magnified residual from fitting the nitrogen exponential to a single component (middle) and to two components (bottom)

Fig. 7.

Plot of the electron relaxation rate (W_e) as a function of arrestin residue number fit to a single component (filled black circles) and to two components (filled gray circles). Also shown are the published W_e values of T4L residues 128 – 135 (open circles) [13]

Data were also collected in the presence of either 0.26 mM (20%) oxygen or 5 mM NiEDDA as exchange reagents. As expected for solvent exposed sites, the recovery time is noticeably shortened in the presence of these paramagnetic agents (e.g. Figure 6). The exchange rate constant is defined as k_ex= W_ex/C_R where C_R is the concentration of the exchange reagent. Just as in the plots of ΔH_pplog R, and W_ethe plots of k_ex (NiEDDA) and k_ex (O₂) have no clear oscillation for arrestin residues 155 – 163, again suggesting a lack of regular secondary structure (Figure 8). Notably, there is no dip in accessibility for residues 155 and 158, which would be expected if this region existed in the crystal-like helix. Arrestin residues and highly exposed T4L residues have similar k_ex (O₂) values (~1.5-2 MHz/mM). As for power saturation data, k_ex values obtained using SR can be used to calculate the contrast parameter R = k_ex (O₂)/k_ex (NiEDDA). These SR data are in excellent agreement with the R values obtained using PS data, and again define this region as solvent exposed relative to T4L data (Figure 5).

Fig. 8.

Plot of k_ex values in the presence of 0.26 mM O₂ (black) and 5 mM NiEDDA (gray) for arrestin residues 155 – 163 as determined by SR EPR. Shown for reference are the published values of k_ex for T4L residues 128 – 135 [12]

3.5 Correlation Factor α

The correlation between Π and solvent accessibility has been established in several protein systems. Recently, a correlation constant α, where α = W_ex/Π, was determined using T4L 44R1 [12]. When this value of α (1.8 MHz) is applied to Π values obtained for arrestin, the values of W_ex (power saturation) are in good agreement with W_ex (SR) values (Figure 9). This highlights the correlation between accessibility determined by power saturation and SR and validates the value of α determined in a previous study.

Fig. 9.

Correlation of Heisenberg exchange rates (W_ex) between spin labeled arrestin in the presence of oxygen (grey) or NiEDDA (black), as determined by power saturation and saturation recovery EPR and an α value of 1.8 MHz. Solid line represents a slope of 1

4 DISCUSSION

The crystal structure of visual arrestin suggests that residues 155 – 163 may be in equilibrium between a helix and an unstructured loop. Because we have recently shown that another structurally dynamic region of arrestin (residues 68 – 79) is of considerable functional significance [29], the structure and dynamics of the 155 – 163 sequence warrants investigation. In addition, it is important to resolve whether this conformational heterogeneity is the result of a functionally important equilibrium or is simply an artifact of the crystalline state of the protein. For this purpose, the mobility and accessibility of R1 were determined at each site along the sequence under physiologically relevant buffer conditions.

Each of the arrestin residues tested display faster motion than even the most mobile T4L residue reported, and no periodic variations in spin label motion were detected along the sequence. The presence of a more immobile component at some of the sites, most notably at 161R1-163R1, suggests some restriction of R1 motion. Even with this additional degree of order, the results in Figures 3 and 4 are consistent with the backbone of this sequence being dynamically disordered on the ns time scale [35] and are not consistent with helical structure, even a helix completely exposed to the solution.

When arrestin concentrations were increased in order to promote tetramer formation, no large spectral changes were observed. This suggests that the secondary structure and dynamics of this region do not change as a function of oligomeric state. Moreover, the secondary structure of this region does not appear to be sensitive to rhodopsin binding, as 157R1, 160R1, and 162R1 are able to bind to light activated phosphorhodopsin but show no changes in spectral motion [29]. This region may instead be important in binding to another signaling protein or for general protein stability upon binding to its receptor.

The solvent accessibility of R1 along the 155 – 163 sequence was examined by both saturation recovery and power saturation methods. Previous studies in T4L have validated the use of saturation recovery as a measure of accessibility [13], but the data presented here represent the first application of the strategy to a complex system of unknown solution conformation. Both methods reveal a high solvent accessibility of each residue along the sequence, on the same order as that for the most exposed residues in T4 lysozyme. The lack of periodicity is again consistent with this region being natively unfolded. An alternative interpretation of the data is that the 155 – 163 sequence is in fact helical, but is of marginal stability. In this case, the presence of R1 at buried sites could destabilize the structure and fail to reveal the expected periodic behavior. Destabilization of a marginally stable helix by R1 has been previously reported [27, 36, 37]. However, in that case R1 at solvent-exposed sites had sequence dependent motional restrictions that revealed an ordered structure, unlike the uniformly mobile and exposed residues along 155 – 163.

The SR data are best fit to two exponentials with both the major and minor components showing a lack of periodicity in accessibility. The CW spectra of R1 at sites 161 – 163 exhibit two motional components that can account for the two relaxation times, but the origin of the minor component of the SR relaxation for 155 – 160 will require further study. Whether or not it arises from a second unresolved population would not influence the conclusions drawn here.

Arrestin is known to exist primarily as a tetramer at high concentrations (>100µM). As mentioned above, a crystallographic tetramer has the 155 – 163 sequence in two different structures; a flexible loop (β form) and a regular helix (α form), the latter stabilized by contacts with an adjacent molecule in the tetramer. The very small changes in nitroxide mobility at all sites in going from monomer to tetramer (Figure 3) argues strongly against a tetramer in solution similar to that in the crystal where the 155 – 163 sequence is in two distinctly different structures with different contact interactions. Recently a topological model for the solution tetramer of visual arrestin was proposed in which each monomer has the same environment [32]. The model was generated by rigid body docking of a single chain from the crystal structure (the A chain) in which 155 – 163 is in a helical conformation. Thus, the model has this sequence helical but solvent-exposed at the surface of the structure with very limited contacts on the neighboring monomer. The results presented here are generally compatible with this model in that the 155 – 163 sequence resides in a similar solvent-exposed environment in each monomer of the tetramer, and offers a refinement of the model in showing the sequence to be disordered rather than helical. Disorder in this region is compatible with the stability of the proposed tetramer, considering the small contact area of the helix with its neighbor. However, the contact is between the 155 – 163 helices in adjacent monomers, and the model predicts that residues 155 – 163 may be within ~20 Å of their counterparts in an adjacent monomer; indeed, inter-molecular disulfide cross-linking has been observed for both 157C and 162C mutants (32). Thus, one might expect an observable dipolar interaction between nitroxides in the tetramer that are detectable as line broadening at distances < 20Å. However, covalent cross-linking reports even transient low-probability encounters, and dipolar interactions could be very weak in disordered sequences. It is possible that some of the small broadening observed upon tetramer formation has contributions from dipolar interactions (Figures 3 and 4). Long-range distance and distributions measurements will be required to decide whether spin interactions are detected between residues of the disordered sequences in the tetramer.

5 CONCLUSION

Several EPR techniques were utilized to characterize the secondary structure of arrestin residues 155 – 163 in solution. Spin label motion, as well as solvent accessibility studies using two distinct techniques, describe this region of visual arrestin as an unstructured loop. One of these techniques, SR, was applied for the first time to a protein region of ambiguous secondary structure. Apparently, the helix formed in the crystal structure is simply an artifact of the crystallization conditions or the environment in the crystal lattice. Regions of proteins that are highly dynamic, such as unstructured loops, have previously been shown to be forced into a more compact, ordered state by solutes commonly used in crystallography [38–40], and we have recently shown that the entire oligomeric arrangement of visual arrestin is different in solution than in crystalline form [32]. Taken together with the data presented here identifying 155 – 163 as a loop in solution, it is evident that protein structure needs to be evaluated in a setting as physiologically relevant as possible and that the CW and SR methodologies are in fact interchangeable for protein structure studies.