Assessment of Nanosecond Time Scale Motions in Native and Non

Sep 14, 2015 - The paramagnetic relaxation times of the aromatic and β protons of Tyr59 and His68 residues of the native ubiquitin and of Tyr59 resid...
0 downloads 5 Views 1MB Size
Download PDF
Article pubs.acs.org/JPCB

Assessment of Nanosecond Time Scale Motions in Native and NonNative States of Ubiquitin Olga B. Morozova and Alexandra V. Yurkovskaya* International Tomography Center of SB RAS, Institutskaya 3a, Novosibirsk, 630090, Russia Novosibirsk State University, Pirogova 2, Novosibirsk, 630090, Russia S Supporting Information *

ABSTRACT: The paramagnetic relaxation times of the aromatic and β protons of Tyr59 and His68 residues of the native ubiquitin and of Tyr59 residue of the non-native ubiquitin were determined from an analysis of chemically induced dynamic nuclear polarization (CIDNP) kinetics obtained during the photoreaction of the protein and 2,2′-dipyridyl excited in the triplet state. Using the paramagnetic relaxation times determined earlier for the radicals of free amino acids as an internal standard and assuming that the hyperfine interaction (HFI) anisotropy is very similar for the radicals of free amino acids and the corresponding radicals of amino acid residues in the proteins, we determined parameters that characterize the intramolecular mobility of different protons in native and two nonnative states of ubiquitin. The latter are denatured at pH 2 and 57 °C, and the A-state at pH 2 in a 60%/40% methanol/water mixture. The determination of the two parameters of intramolecular mobility (i.e., the correlation time of internal motion, τe, and the order parameter, S2) was only possible by analyzing paramagnetic relaxation data obtained at two magnetic fields (4.7 and 9.4 T) using nuclear magnetic resonance (NMR) spectrometry. Intramolecular correlation times fall into the submicrosecond−microsecond time scale. Longer correlation times and higher order parameters were found for the less accessible Tyr59 residue than for the His68 residue, as well as for the more buried β protons than for the aromatic protons for both of the protein residues in the native state. For Tyr59, intramolecular mobility increases following the loss of the tertiary structure of ubiquitin. These findings strongly support the reliability of the obtained data.

■

INTRODUCTION Proteins are inherently flexible, and their dynamics often play a central role in biological functions, influencing diverse processes, such as conformational selection in molecular recognition, catalysis, and allosteric regulation. Efficient and highly informative methods for studying protein dynamics are provided by the magnetic resonance spectroscopic techniques of nuclear magnetic resonance (NMR) and electron paramagnetic resonance (EPR). NMR methods are highly versatile; therefore, they allow the determination of the amplitude and correlation times of bond vector fluctuations.1 Conventional NMR relaxation studies can quantitatively evaluate the time scales and amplitudes of bond vector motions on time scales faster than the rotational correlation times of a system (i.e., 10−8 s).2 For example, rapid motions on the subnanosecond time scale can be examined by analyzing nuclear spin relaxation.3 Additionally, a combination of various NMR methods enables the study of the slower motions of a system. NMR can also identify conformational exchange processes on time scales slower than approximately 10−4 s through their effect on transverse relaxation rates or the appearance of separate resonances. However, over the biologically important 10−8−10−4 s range, NMR has a “blind spot”,2,4 indicating that it is difficult to evaluate the nanosecond time scale range using NMR methods, as demonstrated, for example, by Figure 1 of ref 5. Few NMR techniques are capable of examining motions © 2015 American Chemical Society

on this time scale. An important method of this type is based on studying the residual dipolar couplings (RDCs) in proteins placed in an aligning medium. RDCs offer insight into dynamic processes that is complementary to other approaches and provide access to mechanisms of molecular recognition.5,6 However, the technique is limited in the information it can provide; for example, an orientation-dependent dipolar interaction averaged over the time required to collect the NMR signal (i.e., tens of milliseconds) permits only a timeaveraged orientation of internuclear vectors. Such an approach permits structure refinement and, in some cases, provides information on the direction in which the motion occurs. However, quantitative information about the correlation times of the intramolecular motions is inaccessible for the RDC technique.6−8 Generally, fluctuations at a suitable frequency are required to access the time scale of interest. Recently, a new powerful approach toward addressing this problem was proposed by Ferrage et al.9 This group measured the magnetic field dependence of site-specific 15N nuclear spin relaxations in proteins over a wide field range (0.5−22.3 T). From the analyses of field-dependent relaxation data for 15N nuclei, it was concluded that, in general, nanosecond motions in proteins Received: July 29, 2015 Revised: September 14, 2015 Published: September 14, 2015 12644

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B

motional parametersthe order parameter and the correlation timefrom CIDNP data obtained at a single magnetic field strength. In this study, we revised the previous experimental approach to measure the time-resolved CIDNP at two magnetic fields (i.e., 4.7 and 9.4 T). We clearly demonstrate that this approach enables the determination of the motional parameters. As an object of our study, we selected bovine ubiquitin (UQ) in its native and non-native states. UQ is a small, 76-residue globular protein that has been used as a model system to study protein structure, stability, folding, and dynamics for the past 20 years.28 The spatial structure of ubiquitin taken from the Brookhaven Protein Database (PDB) is shown in Figure 1.

might have been underestimated by traditional high-field NMR approaches, which might have led to an overestimation of the slower “supra-τc” motions.9 This study highlights the ability of NMR methods to evaluate motions on the nanosecond time scale when the field strength is varied to match the appropriate frequency range.9−13 This is because magnetic field variation enables variation of the nuclear Zeeman interaction strength, thereby enabling access to a different frequency window.3 Another option to reach the same goal (i.e., to change the quantifiable frequency range) is given by the paramagnetic labeling of proteins with stable radicals and measuring EPR relaxation data or analyzing paramagnetic nuclear relaxation enhancements (PREs).14,15 Paramagnetic spin labeling enables a stronger sensitivity of paramagnetic systems to motions on time scales on which the relaxation of diamagnetic molecules is less sensitive. Furthermore, this approach can be combined with dynamic nuclear polarization (DNP), which provides additional structural and dynamical information on proteins and enhances protein NMR signals.16,17 The primary contribution to PRE-NMR is provided by electron−nuclear dipole−dipole interactions resulting in a 1/r6 distance dependence of the PREs. Consequently, higher PREs are expected for nuclei located close to the electron; however, their NMR signals are significantly broadened and shifted, often to such an extent that they become undetectable, thus limiting the range of accessible PREs. Thus, although the permanent presence of radicals in the system allows the acquisition of important structural and dynamical information, it is not without difficulties. In this study, we also exploit paramagnetic relaxation enhancements to gain site-specific information about protein mobility on the nanosecond time scale. However, the present novel approach is unconventional. Specifically, we propose the transient generation of radicals to address the problems originating from the presence of radicals in the sample. To obtain nuclear relaxation rates in radicals with lifetimes of only several microseconds, we harness a spin hyperpolarization technique termed chemically induced dynamic nuclear polarization (CIDNP), which detects radicals indirectly via spin polarization of their recombination products.18,19 The timeresolved version of CIDNP enables a highly precise characterization and evaluation of radicals, as well as their reactivity and spin relaxation properties.20,21 In addition, CIDNP can increase the inherently low sensitivity of NMR up to several orders of magnitude.22 Time-resolved CIDNP (TR CIDNP) has been used in our previous studies on proteins in different states.23−27 It has been demonstrated that this method provides site-specific information about the mobility. Indeed, motional correlation times can be obtained for the individual nuclei of CIDNP-active amino acids that are accessible to a probing dye molecule. Typical nuclear T1-relaxation times in the paramagnetic state fall in the range of tens of microseconds; this allows the evaluation of motional correlation times on the nanosecond time scale. We have established a relationship between the intramolecular mobility of CIDNP-active residues in native bovine αlactalbumin and their side-chain accessibility;23 for non-native states it was shown that the successive loss of tertiary structure has a different influence on the variation of paramagnetic nuclear relaxation times of tryptophan and tyrosine residues, which is caused by their different mobilities. However, quantitative analysis of the CIDNP data encountered difficulties, arising from the ambiguity in extracting two

Figure 1. Ribbon representation of the three-dimensional structure of ubiquitin showing the side chains of the CIDNP-active residues His68 and Tyr59, and the structures of histidine and tyrosine with proton numbering. The structure of ubiquitin was generated using the PDB coordinates for ubiquitin (1UBQ.pdb) and the program YASARA Scene.

This structure is well documented,29 and it shows that there are only two polarizable residues on its surface that are accessible to the dye molecule, namely, tyrosine in position 59 and histidine in position 68 (Tyr59 and His68, respectively). The total side chain accessibility (TSA) values for both of the residues were determined according to the standard procedure 30 for three structures (taken from 1UBQ.pdb, 29 1D3Zb.pdb,31 and 1OGW.pdb32); the average TSA values for the three structures are 18% and 57% for Tyr59 and His68, respectively. In non-native states of UQ, the signals of only a single residue (i.e., Tyr59) can be obtained because, under acidic conditions, histidine is not competitive with tyrosine for the reaction with the dye 2,2′-dipyridyl excited in its triplet state.25 In the present study, we apply CIDNP to the native and two non-native states of UQ at two magnetic field strengths. These non-native states are a denatured state at pH 2 and 57 °C33 and a partially folded A-state in a 60%/40% methanol/ water mixture at pH 2.34,35 Using the TR CIDNP method, we determine the nuclear relaxation times of different protons in the transient paramagnetic state; these data enable a robust determination of motional parameters, with correlation times falling on the nanosecond time scale, and order parameters, which are currently beyond the scope of other spectroscopic techniques.

■

EXPERIMENTAL SECTION A detailed description of the time-resolved CIDNP setup operating at proton frequency of 200 MHz has been given in our previous publication.36 The samples, sealed in a standard 5 12645

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B mm NMR pyrex ampule, were irradiated by a COMPEX Lambda Physik excimer laser (wavelength 308 nm, pulse energy up to 150 mJ) in the probe of a 200 MHz Bruker DPX-200 NMR spectrometer. The optical system consisted of a quartz lens, a prism, and a cylindrical light-guide (5 mm diameter). The pulse sequence for time-resolved CIDNP experiments was as follows: radiofrequency (rf) saturation pulses−laser pulse− evolution time, τ−rf detection pulse−free induction decay. The newly built setup is based on a Bruker Avance III 400 MHz NMR spectrometer. The optical system is identical to that of a DPX-200 NMR spectrometer, providing the same irradiation conditions. The comparison of data obtained at two magnetic fields opens an opportunity to improve the accuracy of our relaxation data evaluation by utilizing the theory of magnetic field dependent effects on paramagnetic nuclear relaxation times T1. As the background signals in the spectrum originating from Boltzmann polarization are suppressed, only resonances from the polarized products formed during the variable delay τ appear in the CIDNP spectra. In all kinetic measurements, a rf pulse with duration of 4 μs was used. The timing corresponds to the center of the rf pulse (i.e., 2 μs for τ = 0) on all experimental CIDNP kinetics shown herein. Conditions were as follows: for native state, ambient temperature, pH 5.9; for A-state, pH 2, 60%/40% methanol/ water mixture, ambient temperature; for the denatured state, pH 2, t = 57 °C. Ubiquitin, 2,2′-dipyridyl (DP), methanol-d4 and D2O were used as received from Sigma-Aldrich. Concentrations were 1.2 mM for ubiquitin, 0.6 mM for DP at pH 2 and 5 mM at pH 5.9. The pH of the NMR samples was adjusted by addition of DCl. No correction was made for the deuterium isotope effect on the pH.

Figure 2. 200 MHz 1H CIDNP spectra obtained in the photoreaction of 2,2′-dipyridyl and ubiquitin (a) at pH 5.9 and at an ambient temperature (upper spectrum, native state); (b) ambient temperature, 60% methanol-d4 in D2O, pH 2 (middle spectrum, A-state); (c) pH 2, 57 °C (lower spectrum, denatured state). The spectra were obtained immediately after the laser pulse with a detecting rf-pulse length of 4 μs.

■

RESULTS AND DISCUSSION CIDNP Spectra. CIDNP spectra obtained from the photoreaction of 2,2′-dipyridyl (DP) with UQ at pH 5.9 and at an ambient temperature (native state), in 60% methanol in water at pH 2 (A-state), as well as at 57 °C and pH 2 (denatured state), are shown in Figure 2. In the native state, the His68 and Tyr59 protons are polarized. Enhancements are observed for H2, H4 of histidine; H3,5, H2,6 of tyrosine; and the β protons of both of the residues. In accordance with our expectations, only the Tyr59 residue is polarized in the nonnative states, and no reaction occurs under acidic conditions between the triplet dipyridyl cation (TDPH+, pKa = 5.8) and histidine carrying a positive charge on the imidazole ring (HisH+, pKa = 6.1).37 With a high degree of accuracy, the CIDNP signal intensities linearly correlate with the corresponding hyperfine coupling (HFC) values. For His68, the ratio of I(H2):I(H4):I(β) = 1:0.75:(−1) (native state) corresponds to the relative hyperfine coupling constants (HFCCs) 1:0.8: (−1).21 For Tyr59, the I(H3,5):I(β) ratios are (−1):1.2 (native state) and (−1):1.3 (A-state), and the I(H,3,5):I(β) ratio is (−1):1.4 for the denatured state. The ratio of the HFCC’s (−1):1.221 indicates that the spin density distribution in the radicals of the amino acid residues of UQ is similar to that in the radicals of the corresponding free amino acids. This finding presents an opportunity to use the paramagnetic nuclear relaxation times found for the protons of tyrosine and histidine as an internal standard to determine the correlation times for the intramolecular motion of the His and Tyr protons in UQ. CIDNP Kinetics. CIDNP kinetic data for the native state are shown in Figure 3, while the data for the non-native states are

Figure 3. 200 MHz (solid symbols) and 400 MHz (open symbols) 1H CIDNP kinetics obtained during the photoreaction of 2,2′-dipyridyl and native ubiquitin at pH 5.9, ambient temperature: (a) for the H4 proton (circles) and β protons (triangles) of His68; (b) for the H3,5 protons (circles) and β protons (triangles) of Tyr59. The amplitude of polarization (enhanced absorption or emission) is plotted.

shown in Figure 4. The details of the CIDNP formation in reversible photoinduced radical reactions were described previously.24,25 A spin-correlated radical pair is formed upon quenching the reaction of the triplet excited dye by the amino acid. In the geminate stage of the reaction, nuclear-spin dependent intersystem crossing and spin-selective recombination result in the geminate CIDNP formation. This stage is not resolved in our experiment, and geminate polarization is detected as polarization with no delay after the laser pulse. Radicals that escape geminate recombination carry polarization that is opposite in sign to the geminate polarization. Thus, radical termination in the bulk leads to the transfer of CIDNP of the opposite sign into the diamagnetic products, resulting in the so-called CIDNP cancellation effect. The cancellation is, however, incomplete because of the paramagnetic relaxation of 12646

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B

the nuclei j (aromatic or β protons) in the radical i Pj(Ri), and the corresponding polarization in the product Pj(Pri), as follows: 2

dR D = −∑ kiR iRD dt i=1

(1)

dR i = −kiR iRD dt

(2)

dPj(R i) dt Figure 4. 200 MHz (solid symbols) and 400 MHz (open symbols) 1H CIDNP kinetics obtained during the photoreaction of 2,2′-dipyridyl and ubiquitin for the H3,5 protons (circles) and β protons (triangles) of Tyr59 at (a) ambient temperature, 60% methanol-d4 in D2O, pH 2 (A-state); (b) pH 2, t = 57 °C (denatured state). The amplitude of polarization (enhanced absorption or emission) is plotted.

dPj(Pr) i dt

= −kiPj(R i)RD − kiβi R iRD −

Pj(R i) T1(ij)

(3)

= kiPj(R i)RD + kiβi R iRD

(4)

Equation 2 refers to two equations with indexes 1 and 2, and eqs 3 and 4 refer to four equations each for all possible combinations of indexes i and j. Thus, the total system contains 11 equations. Here, ki is the radical termination rate constant for the reaction between the dye radical and the radical of the residue i, T1(ij) is the paramagnetic spin−lattice relaxation time of the nuclei j in the radical i, and the parameter βi represents the polarization per pair, which is created following the recombination of pairs of radicals (dye radical and radical i) that escape geminate recombination and instead react in the solvent bulk; these are typically termed F-pairs. In most cases, β ≈ 3PG/R0, where PG is the geminate polarization at t = 0. The first terms on the right side of eqs 3 and 4 describe the transfer of polarization from the radicals to the diamagnetic molecules in the termination reaction; the second terms represent the formation of polarization in F-pairs. The third term in eq 3 corresponds to the loss of polarization in the radicals caused by paramagnetic nuclear relaxation. We assumed that the yield of radicals that escape from the triplet geminate radical pair is much greater than the yield of geminate recombination. This is true for the triplet radical pair precursor and the slow intersystem crossing compared with the diffusional radical pair separation. In addition, we assumed that the radicals disappear only by recombining with one another (with the rate

the nuclei in radicals. As the relaxation time decreases, the detected CIDNP grows with increasing delay times after the laser pulse. In addition, polarization is formed in F-pairs, which in the case of a triplet precursor has the same sign as the geminate polarization. In some cases, namely, when the nuclear T1-relaxation in radicals is rapid, CIDNP formation in F-pairs results in the growth of polarization compared to geminate CIDNP. Thus, the behavior of time-resolved CIDNP provides a precise tool to measure the nuclear spin relaxation times in the paramagnetic states. The simulation method used for modeling the CIDNP kinetic data is described in detail below. CIDNP Kinetic Data Analysis. The observed CIDNP kinetic data were interpreted using the approach of Fischer et al.20,38 Because only two residues are polarized in the native state, it was possible to find the parameter set for the simultaneous treatments of the four kinetics for the aromatic and β protons of each residue. The system of differential equations includes the equations for the concentration of the dye radical RD, the radicals of the residues Ri (where the index i refers to the histidine or tyrosine residue), the polarization for

Table 1. Values of Paramagnetic Nuclear Relaxation Times T1 for the Aromatic Protons and β Protons of the Tyr59 and His68 Residues of Ubiquitin Obtained from a Simulation of the CIDNP Kinetics of the Native, Denatured State and A-State, the Corresponding T1(pr)/T1(aa) Values, the Parameters of Intramolecular Motion (S2 and τe) Determined from an Analysis of the Relaxation Data, and T1 Values for the Free Amino Acids T1, μs AA residue His68, native state Tyr59, native state Tyr59, A-state Tyr59, denatured state Hisb Tyrc

proton(s)

200 MHz

H4 β H3,5 β H3,5 β H3,5 β H4 β H3,5 β

± ± ± ± ± ± ± ±

15 60 24 160 19 71 21 67 16 200 63 200

5 10 8 40 5 14 5 15

T1(pr)/T1(aa) 400 MHz

30 140 90 440 45 250 29 92

± ± ± ± ± ± ± ±

a

20 (15 ) 30 (110a) 30 (70a) 100 10 50 6 20

200 MHz

400 MHz

S2

τe, ns

0.94 0.30 0.38 0.60 0.30 0.36 0.33

0.94 0.55 1.11 2.20 0.71 1.25 0.46

0 0.15 0.55 >0.55 0.2

0.11 0.57 1.26 >1.26 0.8 1.8 0.38

0

The value of T1 used in the relaxation data analysis in cases when a solution for S2 and τe could not be found using T1 value corresponding to the center of the best-fit range. bData taken from ref 37. cData taken from ref 42. a

12647

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B

motion within a protein, τe; and (iii) the square of the generalized order parameter, S2, which characterizes the amplitude of the intramolecular motion in the molecular reference frame. The primary mechanism of paramagnetic nuclear relaxation in short-lived radicals is caused by the intramolecular dipole−dipole interaction of the nuclei with unpaired electron spins. The anisotropic hyperfine interactions are stochastically modulated by the motion of amino acid residues in a molecular reference frame, as well as by the overall rotational Brownian motion of the protein. The order parameter satisfies the inequality 0 ≤ S2 ≤ 1, in which lower values indicate larger amplitudes of internal motions. For a protein with isotropic motions, the theory predicts the following:

constant ki). The initial polarizations for each pair [ij] were taken as Pj(Pri) = PGij = −Pj(Ri), which is consistent with the spin-sorting nature of the S−T0 radical pair mechanism.39 The initial concentrations of the radicals of the dye, histidine residue, and tyrosine residue were R0, αR0, and (1 − α)R0. Finally, there were four fitting parameters (consisting of the values T1(ij), k1/k2, and R0 × k1) and four scaling factors. In our previous publication,27 from an analysis of the geminate CIDNP spectra of a mixture of histidine and tyrosine and of the UQ in the native state, we estimated the value of α as 0.7. The ratio of k1/k2 = 0.7 was also found from fitting the kinetic data. A deviation of the simulated curve from the experimental data caused by a change in one of the parameters can be compensated by a change in the other parameters. However, each parameter can be varied only in a limited range to obtain a satisfactory fit. Notably, agreement of the CIDNP kinetics for the H4 proton of His68 was obtained at the two magnetic fields. This is a clear indication that the parameters of the second-order radical termination are very close at both fields, which is achieved by providing the same irradiation conditions for the two experimental designs. Additionally, similar parameters for the radical termination were obtained at the two fields for the denatured UQ (see below). That is the reason that the simulation procedure was performed with the constraint that identical k1 × R0 values be used for the experimental kinetic data sets at the two magnetic fields. The best-fit value for this parameter is 8.3 × 104 s−1. The values of T1(ij) obtained are given in Table 1. For the aromatic protons of His68 and Tyr59, the CIDNP value increases with time; consequently, the range of T1 that provides a good fit is wider than that for aromatic protons. For the β protons, the CIDNP kinetics has a maximum, which narrows the range of acceptable T1 values. In the non-native states, the presence of only one type of protein radical with the radical center at the tyrosine residue reduces the number of equations to five: one equation describes radical termination in the second-order reaction of the dye radical and the protein radical, while the other four equations describe the polarization of the two groups of protons (i.e., H3,5 and β). Because the signal-to-noise ratio is higher for the denatured proteins, paramagnetic nuclear relaxation times are determined more precisely. The parameters of the second-order radical termination were 5.1 × 104 s−1 (200 MHz) and 5.4 × 104 s−1 (400 MHz) for the A-state and 1.3 × 105 s−1 (200 MHz) and 1.5 × 105 s−1 (400 MHz) for the denatured state. The similar values for 200 and 400 MHz, along with the higher recombination rates for the denatured state, corroborate the reliability of the method. The values of the paramagnetic nuclear relaxation times are given in Table 1 for the free amino acids and for UQ. Relaxation Data Analysis. The ultimate goal of our study is to determine the correlation times, τe, of intramolecular motions in the native state, in the molten globule state and in the denatured state of the protein using the accessible sitespecific information about paramagnetic nuclear relaxation times. To reach our goal, we utilized the so-called “model free approach” developed by Lipari and Szabo, which incorporates intramolecular motions, in addition to the overall rotational motion of a protein.40 According to this theory, the spectral density function of the square of a fluctuating field, (ΔB)2, which is responsible for relaxation, is determined by the following: (i) the correlation time of the motions of a protein as a whole; (ii) the effective correlation time for the internal

⎡ S2τM (1 − S2)τ ⎤ 1 ⎥ = (ΔB)2 ⎢ + T1(pr) 1 + [ω Nτ ]2 ⎦ ⎣ 1 + [ω NτM]2

(5)

where τM is the correlation time of the protein tumbling as a whole; τ−1 = τM−1 + τe−1, and in most cases, τ ≈ τe. The value of τM is relatively easy to obtain here; the value of 4.03 ns for native ubiquitin is taken from the available literature data.41 For free amino acids, assuming that the rotational correlation time of the corresponding radical τc is approximately 100 ps, we obtain that the condition (ωNτc)2 ≪ 1 is met for both of the magnetic fields used in this study, which correspond to the NMR frequencies of ωN = 2π × 200 MHz and ωN = 2π × 400 MHz. Thus, the paramagnetic relaxation times of a proton in a free amino acid, T1(aa), do not depend on ωN; instead, they only depend on the motional correlation time τc(aa) according to the following relationship: 1 = (ΔB)2 τc(aa) T1(aa)

(6) 2

Combining eqs 5 and 6, and assuming that (ΔB) is very similar for an amino acid residue in a protein and the corresponding free amino acid, we obtain the following equation: −1 ⎡ T1(pr) S2τM (1 − S2)τ ⎤ ⎥ = τc ⎢ + T1(aa) 1 + (ω Nτ )2 ⎦ ⎣ 1 + (ω NτM)2

(7)

Thus, the free amino acids provide a standard for relaxation measurements. Obtaining this simple formula is possible because the dominant contribution to nuclear relaxation in the paramagnetic species is given by the anisotropic components, Aaniso, of the HFCCs, which are identical for free amino acid radicals and for the corresponding residue radicals in a protein. To prove that this is the case, we plotted a correlation between the 1/T1(aa) obtained in our studies versus the square of Aaniso for histidyl and tyrosyl radicals (see the Supporting Information). A good linear correlation was found, providing a clear indication that anisotropic HFCCs are responsible for the measured relaxation times and corroborating our analysis. As we show below, isotropic HFCCs remain nearly identical in free amino acids and in their residues in UQ; we assume that the same also is true for the anisotropic components of the same interactions. To perform the relaxation analysis, the ratio T1(pr)/T1(aa) was determined at two magnetic fields, providing two equations for calculating the two unknowns, τe and S2. To facilitate the analysis, we calculated the dependence of the T1(pr)/T1(aa) 12648

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B ratio on the effective motional correlation time τe according to eq 7, using S2 as a parameter. An important aspect of this study is that the ratio T1(pr)/ T1(aa) is measured at two magnetic fields. Previously, corresponding data analysis performed for a single field has encountered two problems. First, two parameters of interest cannot be obtained from a single T1(pr)/T1(aa) value. Second, even when the S2 value is known, there is an ambiguity in determining τe because the T1(pr)/T1(aa) dependence on the correlation time has two branches: a descending and an ascending branch (Figure 5). Therefore, the assignment of τe to

Figure 6. Dependencies of ratio of paramagnetic nuclear relaxation times of a protein and of free amino acid, T1(pr)/T1(aa), on the effective correlation time, τe, for internal motion calculated for UQ in its native state at 200 MHz (lower curves) and 400 MHz (upper curves) for (a) S2 = 0 (solid line) and S2 = 0.15 (dashed line); (b) S2 = 0.55 (solid line) and S2 = 0.9 (dashed line). The vertical positions of the symbols correspond to T1(pr)/T1(aa) resulting from CIDNP kinetics obtained at 200 MHz (solid symbols) and 400 MHz (open symbols) for (a) the H4 proton (circles) and β protons (triangles) of His68; (b) the H3,5 protons (circles) and β protons (triangles) of Tyr59.

Figure 5. Dependencies of ratio of paramagnetic nuclear relaxation times of a protein and of free amino acid, T1(pr)/T1(aa), on the effective correlation time, τe, for internal motion calculated with τM of native ubiquitin (4.03 ns) at 200 MHz (lower curves), at 400 MHz (middle curves), and at 800 MHz (upper curves) for different values of S2, as indicated above the curves.

a particular branch also requires additional knowledge about the system.23,24 As shown in Figure 5, measurements performed at two NMR frequencies (e.g., 200 and 400 MHz, as shown here) allow both problems to be circumvented. By first plotting T1(pr)/T1(aa) as a function of τe, the order parameter, S2, can be determined, which agrees with the relaxation data for both fields. Subsequently, the branch of the τe dependence is immediately identified, and the correlation time is obtained. Notably, the method functions best for the combination of fields that we have used. An improvement is not expected by going to higher fields (i.e., where the NMR sensitivity and spectral resolution are increased (e.g., at 800 MHz)). This is because most of the predicted values of T1(pr) are too long with respect to the kinetic window of our experiment, and the determination of T1(pr) in this case becomes problematic. With the aim to determine τe and S2, eq 7 was solved numerically, and the solution is shown in Figures 6 and 7. First, for each group of protons in each residue, we find an S2 value that is consistent with the data obtained at both of the fields. Subsequently, the correlation time can be determined with a known S2 value. Possible ambiguities are removed using the data for the two fields. For example, τe could not be assigned to the descending or ascending branch of the τe dependence if only one field was used. An additional constraint used for fitting the data is related to relationship between the surface accessibility of the residues and their mobility in which the more accessible residues have a higher mobility. It is reasonable to expect that τe and S2 are higher for the more buried Tyr59

Figure 7. Dependencies of ratio of paramagnetic nuclear relaxation times of a protein and of free amino acid, T1(pr)/T1(aa), on the effective correlation time, τe, for internal motion calculated at 200 MHz (lower curves) and 400 MHz (upper curves) for (a) UQ in the A-state, S2 = 0.2; (b) UQ in the denatured state, S2 = 0. The vertical positions of the symbols correspond to T1(pr)/T1(aa) resulting from CIDNP kinetics obtained at 200 MHz (solid symbols) and 400 MHz (open symbols) for the H3,5 protons (circles) and β protons (triangles) of Tyr59.

residue than for the His68 residue and for β protons than for aromatic protons. It was found that not all the best-fit values of T1 given in Table 1 provide a solution for the two motional parameters, and a different value within the best-fit range was taken. These values (in parentheses), along with the resulting τe and S2 values, are given in Table 1. The exact coincidence of the kinetics for the H4 proton of His68 at the two NMR frequencies suggests identical T1 values in these two cases. This clearly suggests that S2 is close to zero and that the solution for τe is at the left branch of the dependency of T1(pr)/T1(aa) on τe. If T1(pr) = 15 ns is considered, then S2 and τe = 0.11 ns provide a solution. In contrast, for Tyr59, the order parameter is approximately 0.5 for the aromatic protons and is close to unity for the more buried β protons; their correlation times significantly increase compared with His68 because the latter have much more freedom of movement. Specifically, the values of T1(pr)/T1(aa) for more buried β protons are equal to 0.3 at 12649

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B 200 MHz and 0.55 at 400 MHz, resulting in S2 = 0.15 and τe = 0.57 ns. The less accessible Tyr59 residue with the two values of T1(pr)/T1(aa) of 0.38 and 1.11 gives S2 = 0.55 and τe = 1.26 ns. For the β protons of Tyr, the values of T1(pr)/T1(aa) = 0.6 (200 MHz) and T1(pr)/T1(aa) = 2.2 (400 MHz) which are nearly asymptotic (Figure 5) were found, which suggests a restricted motion. No definite solution could be obtained in this case. However, it is reasonable to assume that the values of S2 and τe are higher than are those for the H3,5 protons. In Figure 6, the examples with S2 = 0.55 and S2 = 0.9 are shown. We calculated the dependency of T1(pr)/T1(aa) on τe with τM = 13 ns for UQ in the A-state.34 The value of the correlation time for the denatured protein was calculated from its hydrodynamic radius, RS, using Stokes’s law, τM = 4πηRS3/ 3kBT, in which η is the viscosity of the solution, kB is Boltzmann’s constant, and T is the temperature. The value of RS for denatured UQ is 26 Å.43 Thus, the calculated τM value for the denatured UQ is 16 ns. Since this value is relatively high, it did not have mich influence on the result obtained. Because of the better signal-to-noise ratio of the CIDNP kinetic data, T1(pr) was determined with relatively high accuracy, and it was possible to obtain values for S2 and τe using the best-fit values listed in Table 1. For the A-state, S2 = 0.2 was found for both the H3,5 and β protons, and the β protons were found to have a higher τe value of 1.8 ns than the H3,5 protons of 0.8 ns. For the denatured state, T1(pr)/T1(aa) was identical for the H3,5 and β protons at both 200 and 400 MHz, resulting in identical τe = 0.38 ns values at S2 = 0. Thus, upon denaturation, the amplitude of intramolecular motion gradually increased; in the denatured state the motion is nearly isotropic as S2 → 0 and the speed of the motions increases.

denatured state, both of the parameters for the two groups of protons coincide. Notably, the unambiguous determination of the correlation times and order parameters for internal motions was only possible via a comparison of the paramagnetic nuclear relaxation times obtained at the 200 and 400 MHz NMR frequencies. The use of two magnetic fields is a crucial improvement of the method, which reduces ambiguities in determining the motional parameters. Thus, our method provides the ability to analyze motions of biomolecules on the nanosecond time scale which is related to the condition that the product of NMR frequency and correlation time is close to or greater than unity. The technique developed in this study enables the determination of both motional parameters, the correlation time and the order parameter; furthermore, differences in these parameters for different protons of the same amino acid residue can be reliably determined. This is in contrast to the existing methods for studying such motions, i.e., an analysis of RDCs, which provides only the order parameter, and the previously reported field-dependent relaxation measurements, which study the motions of N−H fragments of a protein. Both of these methods are more generally applicable than CIDNP and enable the analysis not only of a few accessible CIDNP-active residues on the protein surface but also of other residues. However, the selectivity of CIDNP with respect to a few specific residues can also be regarded as an advantage of the method, depending on the particular application. The NMR spectrum is modified such that the number of NMR lines is strongly reduced, and only a few residues are highlighted by hyperpolarization and are used as local probes for protein motions. The combination of our method with the CIDNP pulse labeling method44−46 can provide additional information about the structural parameters in the denatured states of proteins, further increasing the potential of time-resolved CIDNP in biomolecular applications.

■

CONCLUSION In this study, we obtained and analyzed site-specific intramolecular motions of amino acid residues on the nanosecond time scale. To this end, we used a combination of the controllable formation of short-lived radicals and the nuclear hyperpolarization of proteins on the microsecond time scale. A detailed analysis of CIDNP time evolution provides information on the intramolecular mobility of polarized residues in the native and non-native states of proteins. To accomplish this challenging task, we refined a nondestructive approach to instantly generate short-lived radicals using a laser pulse in reversible photochemical reactions involving electron (hydrogen) transfer between a water-soluble photosensitizer and the surface accessible residues of histidine and tyrosine. These reactions are accompanied by CIDNP, which is a type of nuclear hyperpolarization that is formed due to the spin selectivity of radical recombination and the sensitivity of the singlet−triplet conversion in transient radical pairs to the nuclear spin.18 The high potential of time-resolved CIDNP enabled the determination of intramolecular correlation times and order parameters for the aromatic and β protons of the Tyr59 and His68 residues in the native state of ubiquitin, of the Tyr59 residue in the A-state, and of the denatured state of the protein. The intramolecular correlation times and order parameters increase for the less accessible Tyr59 residue and for the more buried β protons relative to the aromatic protons within the individual residue. For Tyr59, these parameters decrease in the expected sequence: native state → A-state → denatured state. In the A-state, the order parameters were found to be identical for the H3,5 and β protons, although the correlation times were found to differ significantly. In the

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcb.5b07333. HFI constants, paramagnetic nuclear relaxation times, and correlation between paramagnetic nuclear relaxation times and HFI anisotropy (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Tel: +7 383 3331333. Fax: +7 383 3331399. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the Russian Science Foundation (Grant No. 15-13-20035). We are thankful to Dr. K. L. Ivanov (International Tomography Center) for valuable comments on the text of the paper and for motivating discussions.

■

ABBREVIATIONS UQ, ubiquitin; CIDNP, chemically induced dynamic nuclear polarization; TR CIDNP, time-resolved chemically induced dynamic nuclear polarization; HFCCs, hyperfine coupling constants; RDC, residual dipolar coupling; PREs, paramagnetic relaxation enhancements; DNP, dynamic nuclear polarization 12650

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B

■

(21) Morozova, O. B.; Ivanov, K. L.; Kiryutin, A. S.; Sagdeev, R. Z.; Köchling, T.; Vieth, H.-M.; Yurkovskaya, A. V. Time-Resolved CIDNP: An NMR Way to Determine the EPR Parameters of Elusive Radicals. Phys. Chem. Chem. Phys. 2011, 13 (14), 6619−6627. (22) Lee, J. H.; Okuno, Y.; Cavagnero, S. Sensitivity Enhancement in Solution NMR: Emerging Ideas and New Frontiers. J. Magn. Reson. 2014, 241, 18−31. (23) Morozova, O. B.; Hore, P. J.; Bychkova, V. E.; Sagdeev, R. Z.; Yurkovskaya, A. V. Time-Resolved CIDNP Study of Non-Native States of Bovine and Human a-Lactalbumins. J. Phys. Chem. B 2005, 109 (12), 5912−5918. (24) Morozova, O. B.; Yurkovskaya, A. V.; Sagdeev, R. Z.; Mok, K. H.; Hore, P. J. Time-Resolved CIDNP Study of Native-State Bovine and Human a-Lactalbumins. J. Phys. Chem. B 2004, 108 (39), 15355− 15363. (25) Morozova, O. B.; Yurkovskaya, A. V.; Tsentalovich, Y. P.; Forbes, M. D. E.; Hore, P. J.; Sagdeev, R. Z. Time Resolved CIDNP Study of Electron Transfer Reactions in Proteins and Model Compounds. Mol. Phys. 2002, 100 (8), 1187−1195. (26) Ivanov, K. L.; Yurkovskaya, A. V.; Hore, P. J.; Lukzen, N. N. Theoretical Study of the CIDNP Kinetics of the Reaction of a PhotoExcited Dye with a Reactive Spot on the Surface of a Sphere: Application to CIDNP of Proteins. Mol. Phys. 2006, 104 (10−11), 1687−1702. (27) Ivanov, K. L.; Lukzen, N. N.; Morozova, O. B.; Mok, K. H.; Yurkovskaya, A. V. Quantitative Approach to CIDNP in Proteins with Several Polarizable Residues on the Surface. Appl. Magn. Reson. 2011, 41 (2−4), 251−266. (28) Jackson, S. E. Ubiquitin: A Small Protein Folding Paradigm. Org. Biomol. Chem. 2006, 4 (10), 1845−1853. (29) Vijay-Kumar, S.; Bugg, C. E.; Cook, W. J. Structure of Ubiquitin Refined at 1.8 a Resolution. J. Mol. Biol. 1987, 194 (3), 531−544. (30) Koradi, R.; Billeter, M.; Wuethrich, K. Molmol: A Program for Display and Analysis of Macromolecular Structures. J. Mol. Graphics 1996, 14, 51−55. (31) Cornilescu, G.; Marquardt, J. L.; Ottiger, M.; Bax, A. Validation of Protein Structure from Anisotropic Carbonyl Chemical Shifts in a Dilute Liquid Crystalline Phase. J. Am. Chem. Soc. 1998, 120 (27), 6836−6837. (32) Alexeev, D.; Barlow, P. N.; Bury, S. M.; et al. Synthesis, Structural and Biological Studies of Ubiquitin Mutants Containing (2S, 4S)-5-Fluoroleucine Residues Strategically Placed in the Hydrophobic Core. ChemBioChem 2003, 4, 894−896. (33) Ibarra-Molero, B.; Loladze, V. V.; Makhatadze, G. I.; SanchezRuiz, J. M. Thermal Versus Guanidine-Induced Unfolding of Ubiquitin. An Analysis in Terms of the Contributions from ChargeCharge Interactions to Protein Stability. Biochemistry 1999, 38 (25), 8138−8149. (34) Wirmer, J.; Berk, H.; Ugolini, R.; Redfield, C.; Schwalbe, H. Characterization of the Unfolded State of Bovine Alpha-Lactalbumin and Comparison with Unfolded States of Homologous Proteins. Protein Sci. 2006, 15 (6), 1397−1407. (35) Brutscher, B.; Brüschweiler, R.; Ernst, R. R. Backbone Dynamics and Structural Characterization of the Partially Folded a State of Ubiquitin by 1H, 13C, and 15N Nuclear Magnetic Resonance Spectroscopy. Biochemistry 1997, 36, 13043−13053. (36) Morozova, O. B.; Kiryutin, A. S.; Sagdeev, R. Z.; Yurkovskaya, A. V. Electron Transfer between Guanosine Radical and Amino Acids in Aqueous Solution. 1. Reduction of Guanosine Radical by Tyrosine. J. Phys. Chem. B 2007, 111 (25), 7439−7448. (37) Tsentalovich, Y. P.; Morozova, O. B.; Yurkovskaya, A. V.; Hore, P. J.; Sagdeev, R. Z. Time-Resolved CIDNP and Laser Flash Photolysis Study of the Photoreactions of N-Acetyl Histidine with 2,2′-Dipyridyl in Aqueous Solution. J. Phys. Chem. A 2000, 104 (30), 6912−6916. (38) Vollenweider, J. K.; Fischer, H. Absolute Chemically Induced Nuclear Polarizations and Yields from Geminate Radical-Pair Reactions. A Test of High-Field Radical-Pair Theories. Chem. Phys. 1988, 124 (3), 333−45.

REFERENCES

(1) Luginbühl, P.; Wuthrich, K. Semi-Classical Nuclear Spin Relaxation Theory Revisited for Use with Biological Macromolecules. Prog. Nucl. Magn. Reson. Spectrosc. 2002, 40 (3), 199−247. (2) Bax, A.; Grishaev, A. Weak Alignment NMR: A Hawk-Eyed View of Biological Stucture. Curr. Opin. Struct. Biol. 2005, 15, 563−570. (3) Kowalewski, J.; Mäler, L. Nuclear Spin Relaxation in Liquids: Theory, Experiments, and Applications. In Series in Chemical Physics; Moore, H. J., Spencer, N. D., Eds.; CRC Press Taylor & Francis Group: Boca Raton, FL, 2006; Vol. 2, p 426. (4) Palmer, A. G., III NMR Probes of Molecular Dynamics: Overview and Comparison with Other Techniques. Annu. Rev. Biophys. Biomol. Struct. 2001, 30, 129−155. (5) Lakomek, N. A.; Carlomagno, T.; Becker, S.; Griesinger, C.; Meiler, J. A Thorough Dynamic Interpretation of Residual Dipolar Couplings in Ubiquitin. J. Biomol. NMR 2006, 34 (2), 101−115. (6) Lange, O. F.; Lakomek, N. A.; Farés, C.; Schröder, G. F.; Walter, K. F. A.; Becker, S.; Meiler, J.; Grubmuller, H.; Griesinger, C.; de Groot, B. L. Recognition Dynamics up to Microseconds Revealed from an RDC-Derived Ubiquitin Ensemble in Solution. Science 2008, 320 (5882), 1471−1475. (7) Peti, W.; Meiler, J.; Bruschweiler, R.; Griesinger, C. Model-Free Analysis of Protein Backbone Motion from Residual Dipolar Couplings. J. Am. Chem. Soc. 2002, 124 (20), 5822−5833. (8) Meier, S.; Grzesiek, S.; Blackledge, M. Mapping the Conformational Landscape of Urea-Denatured Ubiquitin Using Residual Dipolar Couplings. J. Am. Chem. Soc. 2007, 129 (31), 9799−9807. (9) Charlier, C.; Khan, S. N.; Marquardsen, T.; Pelupessy, P.; Reiss, V.; Sakellariou, D.; Bodenhausen, G.; Engelke, F.; Ferrage, F. Nanosecond Time Scale Motions in Proteins Revealed by HighResolution NMR Relaxometry. J. Am. Chem. Soc. 2013, 135, 18665− 18672. (10) Redfield, A. G. High-Resolution NMR Field-Cycling Device for Full-Range Relaxation and Structural Studies of Biopolymers on a Shared Commercial Instrument. J. Biomol. NMR 2012, 52 (2), 159− 177. (11) Roberts, M. F.; Redfield, A. G. High-Resolution 31P Field Cycling NMR as a Probe of Phospholipid Dynamics. J. Am. Chem. Soc. 2004, 126 (42), 13765−13777. (12) Roberts, M. F.; Redfield, A. G. Phospholipid Bilayer Surface Configuration Probed Quantitatively by 31P Field-Cycling NMR. Proc. Natl. Acad. Sci. U. S. A. 2004, 101 (49), 17066−17071. (13) Roberts, M. F.; Cui, Q.; Turner, C. J.; Case, D. A.; Redfield, A. G. High-Resolution Field-Cycling NMR Studies of a DNA Octamer as a Probe of Phosphodiester Dynamics and Comparison with Computer Simulation. Biochemistry 2004, 43 (12), 3637−3650. (14) Sharp, R.; Lohr, L.; Miller, J. Paramagnetic NMR Relaxation Enhancement. Recent Advances in Theory. Prog. Nucl. Magn. Reson. Spectrosc. 2001, 38 (2), 115−158. (15) Clore, G. M.; Iwahara, J. Theory, Practice, and Applications of Paramagnetic Relaxation Enhancement for the Characterization of Transient Low-Population States of Biological Macromolecules and Their Complexes. Chem. Rev. 2009, 109, 4108−4139. (16) Prisner, T.; Köckenberger, W. Dynamic Nuclear Polarization: New Experimental and Methodology Approaches and Applications in Physics, Chemistry, Biology and Medicine. Appl. Magn. Reson. 2008, 34 (3−4), 213−218. (17) Griesinger, C.; Bennati, M.; Vieth, H.-M.; Luchinat, C.; Parigi, G.; Höfer, P.; Engelke, F.; Glaser, S. J.; Denysenkov, V.; Prisner, T. F. Dynamic Nuclear Polarization at High Magnetic Fields in Liquids. Prog. Nucl. Magn. Reson. Spectrosc. 2012, 64, 4−28. (18) Kaptein, R.; Dijkstra, K.; Nicolay, K. Laser Photo-CIDNP as a Surface Probe for Proteins in Solution. Nature (London, U. K.) 1978, 274 (5668), 293−4. (19) Hore, P. J.; Broadhurst, R. W. Photo-CIDNP of Biopolymers. Prog. Nucl. Magn. Reson. Spectrosc. 1993, 25 (4), 345−402. (20) Vollenweider, J. K.; Fischer, H.; Hennig, J.; Leuschner, R. TimeResolved CIDNP in Laser Flash Photolysis of Aliphatic Ketones. A Quantitative Analysis. Chem. Phys. 1985, 97 (2−3), 217−34. 12651

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652

Article

The Journal of Physical Chemistry B (39) Salikhov, K. M.; Molin, Y. N.; Sagdeev, R. Z.; Buchachenko, A. L. Spin Polarization and Magnetic Effects in Chemical Reactions; Elsevier: Amsterdam, 1984. (40) Lipari, G.; Szabo, A. Model-Free Approach to the Interpretation of Nuclear Magnetic Resonance Relaxation in Macromolecules. 1. Theory and Range of Validity. J. Am. Chem. Soc. 1982, 104 (17), 4546−59. (41) Lee, A. L.; Wand, A. J. Assessing Potential Bias in the Determination of Rotational Correlation Times of Proteins by NMR Relaxation. J. Biomol. NMR 1999, 13, 101−112. (42) Tsentalovich, Y. P.; Morozova, O. B. Laser Flash Photolysis and Time Resolved CIDNP Study of Photoreaction of 2,2′-Dipyridyl with N-Acetyl Tyrosine in Aqueous Solutions. J. Photochem. Photobiol., A 2000, 131 (1−3), 33−40. (43) Kamatari, Y. O.; Ohji, S.; Konno, T.; Seki, Y.; Soda, K.; Kataoka, M.; Akasaka, K. The Compact and Expanded Denatured Conformations of Apomyoglobin in the Methanol-Water Solvent. Protein Sci. 1999, 8 (4), 873−882. (44) Mok, K. H.; Hore, P. J. Photo-CIDNP NMR Methods for Studying Protein Folding. Methods 2004, 34 (1), 75−87. (45) Mok, K. H.; Nagashima, T.; Day, I. J.; Jones, J. A.; Jones, C. J. V.; Dobson, C. M.; Hore, P. J. Rapid Sample-Mixing Technique for Transient NMR and Photo-CIDNP Spectroscopy: Applications to Real-Time Protein Folding. J. Am. Chem. Soc. 2003, 125 (41), 12484− 12492. (46) Lyon, C. E.; Suh, E.-S.; Dobson, C. M.; Hore, P. J. Probing the Exposure of Tyrosine and Tryptophan Residues in Partially Folded Proteins and Folding Intermediates by CIDNP Pulse-Labeling. J. Am. Chem. Soc. 2002, 124 (44), 13018−13024.

12652

DOI: 10.1021/acs.jpcb.5b07333 J. Phys. Chem. B 2015, 119, 12644−12652