Slow Dynamics around a Protein and Its Coupling to Solvent.

Solvent is essential for protein dynamics and function, but its role in regulating the dynamics remains debated. Here, we employ saturation transfer electron spin resonance (ST-ESR) to explore the issue and characterize the dynamics on a longer (from μs to s) time scale than has been extensively studied. We first demonstrate the reliability of ST-ESR by showing that the dynamical changeovers revealed in the spectra agree to liquid-liquid transition (LLT) in the state diagram of the glycerol/water system. Then, we utilize ST-ESR with four different probes to systematically map out the variation in local (site-specific) dynamics around a protein surface at subfreezing temperatures (180-240 K) in 10 mol % glycerol/water mixtures. At highly exposed sites, protein and solvent dynamics are coupled, whereas they deviate from each other when temperature is greater than LLT temperature (∼190 K) of the solvent. At less exposed sites, protein however exhibits a dynamic, which is distinct from the bulk solvent, throughout the temperature range studied. Dominant dynamic components are thus revealed, showing that (from low to high temperatures) the overall structural fluctuation, rotamer dynamics, and internal side-chain dynamics, in turn, dominate the temperature dependence of spin-label motions. The structural fluctuation component is relatively slow, collective, and independent of protein structural segments, which is thus inferred to a fundamental dynamic component intrinsic to protein. This study corroborates that bulk solvent plasticizes protein and facilitates rather than slaves protein dynamics.

Protein structures are dynamic rather than static. They fluctuate on many different time scales and continually switch among conformational states to execute a variety of functions. Protein dynamics can span over a wide time range from shorter than nanoseconds[1] to longer than milliseconds[2,3] even at subfreezing temperatures. It is well-accepted that protein functions are governed by their dynamic motions.[2−4] However, the role of solvent dynamics in the protein structure–dynamics–function relationship remains unclear. This study aims to explore, in a fully hydrated condition, the connection between the protein dynamics and the dynamics of the surrounding solvent.

Support for a strong coupling between protein and solvent dynamics mainly comes from studies using neutron scattering techniques[1,5−8] while some are from results using other tools such as dielectric, NMR, and femtosecond fluorescence spectroscopy.[9−12] Essentially, neutron scattering studies collected two scattering profiles of the hydrated protein, one with water and the other with deuterated water, both of which are partially hydrated (about 0.3 g of water per g of dry protein) to ensure that the surface hydration can be supercooled at low temperatures. As neutrons are much more sensitive to hydrogen atoms than deuterium, a comparison of the two measurements can therefore distinguish the dynamic components related to protein (which also contains hydrogen atoms) from those related to surface hydration. As a result, a dynamical changeover related to protein, which exhibits a steep increase in the amplitude of atomic motions of protein, was observed around 220 K, and more importantly, this changeover coincides with the changeover temperature observed for the dynamic crossover from low-T Arrhenius and high-T super-Arrhenius (the latter is known as the Vogel–Fulcher–Tammann, VFT) behavior of the surface hydration.[6] A possible connection between the crossover of the surface hydration and the liquid–liquid transition (LLT) in supercooled water was suggested.[13−15] Also, some studies reported that proteins only become biologically active above 220 K.[16] Taken together, these results suggest a picture in which the dynamics of protein and surface hydration are strongly coupled, and only at temperatures above 220 K is the structural flexibility restored, thus causing the protein to be biologically functional. Nevertheless, the coupling has remained debated. Some proteins were found to remain active below the protein dynamical transition temperature 220 K.[2,3,17,18] Other experimental works have questioned the existence of the fragile-to-strong crossover in protein surface hydration.[5,19−21] A unified model that protein motions are slaved by the bulk solvent (α fluctuation) and the hydration shell (β fluctuation) rather than the protein dynamic transition was proposed.[4] Taken as a whole, understanding the role of solvent dynamics in regulating the protein dynamics and functions remains rather limited. To resolve the controversy, it is desirable to explore the dynamics of individual sites around a protein and extend the studies of water–protein coupling to the hydration levels beyond those attainable in hydrated powders.[22]

Here, site-directed spin labeling (SDSL) in combination with saturation transfer electron spin resonance (ST-ESR) is introduced to explore the slow dynamics on the time range from microsecond to seconds in a fully hydrated protein over the temperature range 180–240 K. ST-ESR is developed to study the rotational dynamics on the very slow time scale (>μs), where conventional ESR is not sensitive.[23−26] A detailed comparison between the two ESR methods is provided (Figures S1 and S2). The basic principle of ST-ESR is to collect the spectra under saturation conditions and high modulation amplitudes to observe the response of the spin system on spectral diffusion of saturation by molecular motions.[27−30] SDSL in combination with ESR techniques is a powerful tool to explore local information on dynamics of molecular structure in an ensemble system.[31,32] Basically, change in the dynamic structure of solvent can be probed by a nitroxide radical doped in the solvent. As opposed to other techniques detecting the overall dynamic motions in the ensemble, the SDSL-ESR provides local information on the dynamics of spin probes and the corresponding potential energy associated with the local environment in the solvent. When the spin-label side chain is incorporated into a protein by the SDSL methods, the tethered probe is also sensitive to various dynamic components related to the protein, such as dynamics of the side chain, backbone fluctuations, and interactions between tertiary structures.[32−34] In this regard, the local environment reported on the ESR spectra of spin-labeled proteins is a composite of information from protein and nearby solvent molecules.

In the present study, we first demonstrate the capability of ST-ESR to reveal LLT temperatures in various glycerol/water mixtures. Various T4 lysozyme (T4L) mutants were prepared and labeled, one at a time, with three different spin labels. The mutation sites cover solvent-exposed and buried sites in T4L. Then, we utilize ST-ESR to systematically map out the variation in protein local (site-specific) dynamics around T4L surface at temperatures from 180 to 240 K. The use of different spin labels for individual sites allows us to discriminate between the dynamic components reported in the ESR spectra, hence distinguishing protein from solvent dynamics. New insights into the connection between protein and solvent dynamics are reported.

Results

To demonstrate that spin-label ST-ESR has sufficient sensitivity to the dynamic structure of bulk solvent at low temperatures, we investigate the properties of 10 mol % glycerol/water mixture (doped with TEMPOL, Figure A) at temperatures of 180–240 K. The ST-ESR data were analyzed to yield the rotational correlation time (τ) as a function of temperature (see the Materials and Methods section). The theoretical analysis indicates that (Figure A) the τ data can be approximated by the VFT relation at high temperatures and the Arrhenius-type relation at low temperatures (plotted by blue and red lines, respectively, in Figure A) with a distinct transition ∼190 K (denoted by TI). The observed transition TI coincides with the liquid–liquid transition (LLT) reported in the state diagram of glycerol/water mixtures (Figure S3).[35] LLT is a phenomenon in which a (supercooled) liquid transforms into another one via a first-order transition,[35−38] and its observation has been a challenging task. This result represents, to our knowledge, the first observation of LLT in the water/glycerol system by ST-ESR. The best-fit potential energy U, obtained from the fits of the Arrhenius-type relation (cf. Materials and Methods) within the temperature range (180–190 K), for the mixture is approximately 108 kJ/mol (red line in Figure A), which can be regarded as the characteristic potential energy for the rotational motions of spin probe in association with the second liquid state (liquid II) of the glycerol/water mixtures. When the characteristic energy U is greater, the constraint on the spin probe motions is greater, hence the slower rotational dynamics.

Figure 1

LLT temperatures and introduction to spin labels and T4L mutants. (A) TEMPOL spin label used to probe the dynamic structure of bulk solvent. Liquid–liquid transition in the 10 mol % glycerol/water mixture is observed ∼190 K (TI), demonstrating the ST-ESR sensitivity to solvent dynamics. (B) Temperature dependence of the τ data of the TEMPOL in various glycerol/water mixtures. A distinct τ changeover in each of the 6, 10, 12, and 14 mol % glycerol/water mixtures is observed. No changeover in τ occurs in 50 mol % glycerol/water mixture. Blue and red lines are the predictions by the VFT and Arrhenius-type relations, respectively. The average U over the Arrhenius-type fits is 108 kJ/mol with σ ∼ 2%. (C) Cartoon model of T4L (PDB code 3LZM) displaying the cysteine variants of T4L, one at a time. The mutation sites can be classified into the highly and less exposed groups, the latter of which comprise three subgroups, as indicated by colors. Each mutation site can be covalently conjugated with a spin probe to form either R1, K1, or RX side chain, as illustrated by site 80. The length and flexibility of the side chains are in the order K1 > R1 > RX. (See also Figures S3 and S4.)

We have performed further experiments to verify the observation of LLT in various glycerol/water mixtures (Figure B and Figure S3). We show that the dynamical changeover occurs approximately at 190 K in the solvents containing glycerol <11 mol %, whereas it shifts to 185 and 182 K in 12 and 14 mol % glycerol/water mixtures, respectively, and disappears in 50 mol % glycerol/water mixture. The trend of the changeover temperatures observed is consistent with the predictions of LLT temperature from the phase diagram (Figure S3).[35] Our results show strong ESR evidence for the connection between the dynamical changeover, observed in the ST-ESR data, and the liquid-I-to-liquid-II transformation. Importantly, we demonstrate that ST-ESR has sufficient sensitivity for molecular dynamics in the temperature range studied.

In the background of the cysteine-free T4L mutant, nitroxide-based side chains (R1, K1, and RX) were introduced (one at a time) using the SDSL methods (Figure S4). Single-cysteine residue is required to generate the side chain R1. In the cases with RX, a pair of cysteine residues at sites i and i + 4 in an α-helix is required to best reflect the backbone dynamics of α-helix.[39] In the cases with K1, an introduction of the genetically encoded unnatural amino acid p-acetyl-l-phenylalanine is required, as detailed in the Materials and Methods section.[40] ST-ESR is verified to have sufficient sensitivity to protein dynamics at low temperatures (Figure S2). The mutation sites studied can be classified into highly and less exposed groups, the latter of which contain three subgroups: helix surface, buried, and tertiary contact, as illustrated (Figure C). In terms of average distances from the distal nitroxide ring to protein backbone, K1 (ca. 12–13 Å) is greater than R1 (ca. 6–7 Å) and RX (ca. 5–6 Å). Side-chain flexibility (internal dynamics) is in the order K1 > R1 > RX.[40] With the replacement of R1 by RX, both the side-chain flexibility and the number of preferred side-chain conformational clusters (referred to as rotamer clusters herein; see Figure S4D) are decreased. A previous crystal study showed that the RX side chain in an α-helix of T4L exhibits a single, well-ordered, and energetically relaxed rotamer cluster,[39,41] demonstrating that the nitroxide of RX is more localized (i.e., characterized by single set of side-chain rotamers), and its internal side-chain motion is more restricted. As such, a replacement of R1 (or K1) by RX would certainly lead to a reduction in both the side-chain flexibility and the number of preferred rotamer clusters. Our strategy is, therefore, to probe the local environment of spin label in T4L (10 mol % glycerol/water mixture) and identify the dominant dynamic components in the ESR spectra via the modulation of spin-label sensitivity to the dynamics using the three probes.

The ST-ESR spectra (Figure S5) of the highly exposed sites (with R1 side chain) were analyzed to give the temperature dependence of the rotational correlation time (τ). The results of the highly exposed sites show a similar trend of variation in the plot of log(τ) versus 1/T (Figure A). Below TI (∼190 K), they overlap with the bulk solvent (i.e., 10 mol % glycerol/water mixture). Above TI, they are strikingly different from the bulk solvent. The data of the highly exposed sites can be fitted to Arrhenius-type relations (Figure B) to obtain potential energies U (Table S1; discussed later). In addition to the TI transition that corresponds to LLT of the bulk solvent ∼190 K, a transition around 220 K (denoted by TIII) is observed in all of the highly exposed sites.

Figure 2

Study of highly exposed sites in T4L. (A) Temperature dependence of the effective rotational correlation time τ, plotted in log(τ) vs 1/T, for the highly exposed sites (22, 80, 137, and 150). While the data are highly similar, they are distinctly different from the bulk solvent (colored in black) above TI (∼190 K). (B) Theoretical fits to the τ data. TI can be probed (in blue color) by the R1 side chain at all of the highly exposed sites. The data also exhibit TIII transition (∼220 K), which was not observed in the bulk solvent (10 mol % glycerol). (C) Results for site 80 carrying three different side chains. TI can only be observed in the results of R1 and K1, but not RX. As the length of the RX side chain is shorter, and its rigidity is greater than others, it becomes insensitive to the dynamics of bulk solvent. (D) Fits to the 80K1, and 76/80RX. TI transition is indicated by blue line. TII (∼205 K) is observed in the 76/80RX. (E) U of the dominant components. They are colored by temperature and local environment. Important U values are noted. Error bars correspond to experimental uncertainties. (See also Table S1 and Figure S5.)

To explore the significance of the dynamic components in the ESR spectra, we investigate site 80 with three different spin labels (Figure C,D). The results of 80R1 and 80K1 are similar, both of which reveal TI and TIII transitions. The similarity is due to the fact that, as site 80 is highly exposed, both K1 and R1 have sufficient sensitivity to probe the dynamic structure of bulk solvent, hence revealing the same transition TI. When RX is substituted for R1, the sensitivity of spin probe to TI is lost, but transitions TII (∼205 K) and TIII (∼220 K) are revealed in the data. While a reduction in the chain length and flexibility leads to a decrease in the sensitivity to bulk solvent, it causes RX to reflect another dynamical changeover which was detected by neither R1 nor K1. By using different lengths of spin probe, we are able to increase/decrease the spectroscopic sensitivity to bulk solvent.

U values obtained from the fits of Arrhenius-type relation are plotted in Figure E (see also Table S1), providing quantitative descriptions for the dynamic components. The dynamic states (H1, H2, H3, L1, L2, and L3) between the transitions (TI, TII, and TIII) are denoted by two sets of colors (Figure E) according to the local environment (highly versus less exposed) of spin probe. When R1 is replaced by K1 for site 80, the potential energies U change little in states H1 and H2, but it decreases clearly from 32 to 24 kJ/mol in state H3. When R1 is replaced by RX for site 80, U increases from 32 to 85 kJ/mol in state H3. These observations are consistent with the fact about the side-chain flexibility: K1 > R1 > RX. When the flexibility is greater, and there is less constraint to the spin probe, the U value is lower. This suggests that the dominant dynamic component reflected in the temperature range 220–240 K (H3 or L3, equivalently) is the internal motion of side chain. The other important effect expected by the substitution of RX for R1 is that the number of rotamer clusters is largely decreased, which would lead to a reduction in U. In a comparison of the U values for states H2 and L2, we observe a distinct reduction for the replacement of R1 (85 kJ/mol) by RX (53 kJ/mol) at site 80. The difference (ca. 32 kJ/mol) accounts for the energy barrier between the major rotamer clusters of R1. The dominant component reflected by spin labels in either H2 or L2 is the dynamics of spin-label rotamers.

Figure A,B shows plots of log(τ) versus 1/T for the helix surface sites and the result of theoretical fits, in which two transitions TII (∼205 K) and TIII (∼220 K) can be clearly observed. As these helix surface sites are relatively less exposed than those shown in Figure A, TI is thus not revealed in the data. To verify the above statement, we compare the results of 131R1, 131K1, and 131RX (Figure C,D). Neither R1 nor RX at site 131 are able to report transition TI. When K1 is substituted for R1 (which leads to an increase in the side-chain length and flexibility), TI (the signature transition of the bulk solvent) is restored in the data (Figure D). A distinct reduction in U (from 41 to 25 kJ/mol) occurs in the temperature range 220–240 K (i.e., L3 or H3) upon the substitution (Figure E), consistent with the finding for site 80 as discussed earlier. When R1 is replaced by RX at site 131, we observe a distinct increase in U (from 41 to 77 kJ/mol) in state L3 (220–240 K) due to the decreased side-chain flexibility, and a reduction in U (from 72 to 49 kJ/mol) in state L2 due to the decreased number of side-chain rotamer clusters. These observations are consistent with the findings of the highly exposed sites (discussed in Figure ). Taken together, the results (Figures and 3) support that the dominant components in states H2 and L2 are the rotamer dynamics. The average U over the highly exposed sites in state H2 is approximately 82 kJ/mol (Table S1). Upon substitution of RX for R1, the average U becomes 51 kJ/mol in state L2: specifically, 53 kJ/mol (76/80RX) and 49 kJ/mol (127/131RX). A reduction of 31 kJ/mol (approximately) happens because of the decreased number of preferred rotamers in the RX as compared to the R1.

Figure 3

Study of helix surface sites. (A) Temperature dependence of τ, plotted in log(τ) vs 1/T, for the helix surface sites (which belong to the less exposed category). They all display a similar trend of variation. (B) Theoretical fits (red lines) to the data of the helix surface sites (44, 61, 65, 72, 109, 115, and 131). Transitions TII (∼205 K) and TIII (∼220 K) are revealed in the data. TI is absent. (C) Experimentally determined log(τ) of the site 131 carrying three different side chains, R1, K1, and RX. (D) Fits to the results of site 131 carrying K1 and RX labels. With the replacement of R1 by K1 (which is among the longest side chain), TI transition is restored. Transitions TII and TIII are observed in the RX data. (E) U of the dominant components for the helix surface sites. They are colored by temperature and local environment. Important U values are noted. Error bars correspond to experimental uncertainties. (See also Table S1 and Figure S5.)

Figure shows the results for the buried sites (Figure A,B), the tertiary contact sites (Figure C,D), and the corresponding U values (Figure E and Table S1). As displayed, the τ data of the buried sites are clearly different from the result of the highly exposed site 80R1 (Figure A). The τ data of the tertiary contact sites are within the two extreme conditions (Figure C), i.e., the highly exposed (80R1) and the buried (99R1). Transitions TII (∼205 K) and TIII (∼220 K) are observed in the data, in line with the finding for all of the less buried sites.

Figure 4

Study of buried and tertiary contact sites. (A) Temperature dependence of τ, plotted in log(τ) vs 1/T, for the buried sites (99, 129, 133, and 153). The data show a highly similar trend of variation and are clearly different from the highly exposed site 80 (blue). (B) Theoretical fits to the data of the buried sites. Transitions TII (∼205 K) and TIII (∼220 K) are identified in the data. (C) Plot of log(τ) versus 1/T for the tertiary contact sites (4 and 74). The data are between the two extreme cases, 99R1 (a buried site) and 80R1 (a highly exposed site). (D) Theoretical fits to the results of 4R1 and 74R1. Transitions TII and TIII are revealed. (E) U values of the dominant components for the buried and tertiary contact sites. States L1, L2, and L3 are colored according to Figure E. (See also Table S1 and Figure S5.)

The assignment of the dynamic components is described below. In state L1, the average U over the less exposed sites is 148 kJ/mol with an average deviation (σ) of 7% (Table S1). As the variations in U are reasonably small and consistent over different labeling sites (including the helix surface, buried, and tertiary sites), the component in L1 can be assigned to overall structural fluctuations related to an intrinsic dynamic component in T4L. In state L2, the average U (for all of the less exposed sites with R1) is 73 kJ/mol (σ ∼ 8%); specifically, it is 76 kJ/mol over the helix surface sites and 73 kJ/mol over the buried sites. The average U in state L2 is representative of the characteristic energy describing the dynamics of jumps between the preferred rotamers of R1. In state L3, the average U over all sites in the less exposed group is 69 kJ/mol with a relatively large deviation (σ ∼ 31%). The large deviation is due to the fact that an additional local constraint to the side-chain motions is present for the buried sites but absent for the helix surface sites. We, therefore, consider the U of state H3 (33 kJ/mol with σ ∼ 10%) to be representative of the internal dynamics (of small-amplitude motions) of R1 side chain. The difference in U between states H3 (33 kJ/mol with σ ∼ 10%) and L3 (91 kJ/mol with σ ∼ 4%, for the buried sites only) is attributed to the local confinement effect in the buried sites; namely, the contribution from the local steric clashes to the internal dynamics of the R1 side chain is approximately 58 kJ/mol. In addition, we also measure ST-ESR spectra of lyophilized T4L mutants spin-labeled, one at a time, at solvent-exposed and buried sites (Figure S6). The temperature dependence of the τ data decreases monotonically with increasing T, independent of the labeling site. Importantly, all of the transitions (TI, TII, TIII) are absent from the τ data of the dehydrated samples. This result suggests that the presence of solvent molecules is necessary for a protein to exhibit the dynamical diversity observed.

Discussion

Highly Exposed Sites Are Decoupled from Bulk Solvent above TI

In the result of the highly exposed sites (with R1 side chain; Figure ), we observe the transitions from (low to high temperatures) H1 to H2 and then H3 states, each of which are characterized by distinct U values. The temperature dependence of spin probe dynamics in H1 (180–190 K) is well consistent with the result of the bulk solvent (Figure A), indicating that below TI the molecular dynamics of the highly exposed sites is strongly coupled to the bulk solvent. The bulk solvent is so dominant in H1 such that the nitroxide probes, whether attached to T4L or not (i.e., R1 side chain versus free TEMPOL probe), experience the same characteristic potential energy (ca. 108 kJ/mol); moreover, the U is very close to 110 kJ/mol reported as a characteristic of the glass-forming solvent around 200 K.[42] This indicates that, in H1, the internal dynamics of side chain contributes insignificantly to the spectra and is largely dominated by bulk solvent. Above TI, the τ data of the R1 on T4L behave differently from the bulk solvent (Figure A), supporting a view that the bulk solvent no longer maintains a commanding influence on the protein dynamic components. We show (Figures E and 3E, Table S1) that the dominant component in H2 is correlated with the dynamics of side-chain rotamer clusters. Upon the transition from H2 to H3 at higher temperatures, the internal dynamics (characterized by relatively smaller-amplitude motions) of side chain is confirmed to be dominant in the spectra. In addition, we note the following about the generality of the result. It is noteworthy that site 22 is on an unstructured loop of T4L, but it exhibits the same transitions as those observed for the highly exposed sites on structured helices of T4L. This observation suggests that the ST-ESR may be applicable to a more flexible protein with disordered regions.

Protein Slow Dynamics and Its Coupling to Bulk Solvent

In the result of the less exposed sites, the bulk-solvent-related transition TI is barely observed. The average U over the less exposed sites (13 sites, totally) in L1 (180–205 K) is approximately 148 kJ/mol with σ ∼ 7%. The slow dynamics (featured by longer τ values) is characterized by small variations in U and appears to have weak correlation with the local environments (e.g., helix surface or buried environments) as well as the type of spin label (R1 versus RX). We, therefore, assign the observed dynamic component for the overall structural fluctuations rather than the dynamics coupled to the protein surface hydration as the latter is expected to be largely different in the energy U between the hydrations of helix surface and buried sites. (Note that the protein surface hydration shell is suggested by other studies such as femtosecond spectroscopy and Overhauser DNP,[43−46] but it is not present as a dominant component in our ST-ESR spectra sensitive to a slower time scale.) Strikingly, the temperature-dependent behavior of the τ data in L1 is distinctly different from that of the bulk solvent in the lower temperature range studied: the corresponding U is 148 kJ/mol (Table S1) for the overall structural fluctuations as opposed to 108 kJ/mol for the bulk solvent. This finding demonstrates that the overall structural dynamics of T4L is not dominated by the dynamics of bulk solvent. An important implication is that the overall structural dynamics is not slaved to the bulk solvent dynamics, contrary to the finding of the highly expose sites in H1. Given that the influence of the bulk solvent is largely excluded from the spectra of the less exposed sites, the dynamic components reported in L1 are more relevant to protein dynamics in the very slow time scale (from μs to s).

In state L2 (205–220 K, which overlaps a part of H2), the dominant component is assigned to the dynamics describing the jumps between the preferred rotamers in the less exposed sites (U ∼ 73 kJ/mol), as discussed earlier. The average U over the less exposed sites is approximately 9 kJ/mol less than that over the highly exposed sites. This reduction suggests that the number of side-chain rotamers is generally less for the less exposed sites than the highly exposed sites, in good agreement with the MD simulations.[47] The same observation can be made from the results of site 131 with different spin labels as follows: in L2, U is 88 kJ/mol (131K1), 72 kJ/mol (131R1), and 49 kJ/mol (127/131RX). In this regard, we conclude that, in L2, the reduction in U is caused by a decrease in the number of preferred rotamer clusters (cf. Figure S4D).

In L3, the dynamics of side chain is dominant. The U values for L3 include not only the internal dynamics of side chain (as observed for H3), but also the interactions with nearby residues (e.g., the helix surface sites) and the confinement effect due to the restricted local space (e.g., the buried sites). The U values in L3, therefore, can be used to yield the energies associated with the restricted local space and environments.

Relation to Previous Studies

As proposed in the unified model, proteins are subject to the α and β fluctuations, both of which originate from solvent.[4] The α fluctuations control the overall shape of the protein, exhibiting a VFT-like behavior at higher temperatures. The β fluctuations slave protein internal motions, following an Arrhenius-type relation at lower temperatures. While the unified model suggests the complete slaving of protein to solvent, other studies (as described in the Introduction) support a mutual coupling between protein and solvent. Our results are in line with many of the previous studies showing the existence of dynamical changeovers in the temperature range 180–240 K. Moreover, we reveal new insights into the protein local dynamics and its coupling to the bulk solvent dynamics. The site-specific labeling approach allows a clear separation between the dynamics of the protein buried and exposed sites, hence identifying the contribution from the bulk solvent. We therefore identify the dynamic components in the spectra that are coupled to protein (Figure ). The dominant components include the overall structural dynamics (in L1), the dynamics of the preferred rotamer clusters (in either L2 or H2), and the internal dynamics of side chain (in H3). While one of the observed transitions coincides with the conventional protein dynamical transition temperature (∼220 K; TIII), the other (∼205 K; TII) has not been reported in the slow time range. Both the transitions are clearly different from the bulk-solvent-related transition TI (∼190 K). If the literature protein dynamical transition around 220 K was due to an intrinsic property of solvent, then the choice of the solvent (pure water versus glycerol/water mixtures), which results in different LLT temperatures, should matter. This study thus suggests that the dynamical behaviors of protein and solvent are not tightly related. This observation is in line with our earlier discussion that protein dynamics in state L1 is not slaved to the bulk solvent dynamics. Furthermore, the protein-related transitions TII and TIII can be consistently observed at the labeling sites around T4L exposed and buried surfaces, adding yet more evidence that the slow dynamics reported (U ∼ 148 kJ/mol) is indeed related to protein-intrinsic dynamics.

Figure 5

Hierarchy of protein dynamics and bulk solvent dynamics as observed by ST-ESR. The approximate temperature for the transitions between dominant dynamic components is indicated. The left panel represents the observed liquid–liquid transition in the 10 mol % glycerol/water mixture, in which the U of liquid II is determined in this study. The right panel shows the protein-related dynamic components identified and the respective potential energy U describing the characteristic energy of the R1 dynamics in association with the dominant dynamic component.

Proteins must move between conformational energy landscapes to perform physiological functions. To discriminate the contributions of individual dynamic components to protein motions, we performed investigations at temperatures sufficiently low to cease (or reduce) individual components. The dynamic components identified here are potentially important for constructing dynamic models required for understanding how proteins function on a molecular level at physiological temperatures.

Conclusions

This study has employed SDSL-ST-ESR to map out the variation in protein local dynamics in the time range from microseconds to seconds, hence providing new insights into the dynamics on a longer time scale than has been extensively explored. We demonstrate the capability of ST-ESR spectroscopy to discriminate differences in dynamics between the liquid I and liquid II states in the glycerol/water mixtures, accordingly establishing the connection between the dynamical changeover and the LLT of bulk solvent. We show that the bulk solvent (10 mol % glycerol/water) dynamics can only dominate the dynamics of the highly exposed sites in T4L below transition TI (∼190 K) in liquid II state. For other sites that are relatively less exposed, protein maintains control over the dynamics itself throughout the temperature range studied. The temperature-dependent behaviors of the protein-related dynamic components are not dominated by the solvent dynamics. The dynamics of spin labels is shown to reflect the overall structural dynamics in T4L (180–205 K), the dynamics of rotamer clusters (205–220 K), and the internal side-chain dynamics (220–240 K). The overall structural dynamics is collective and independent of protein structural segments, providing information for understanding the fundamental dynamic component of a protein that has not been reported. More than one protein-related dynamical transition is revealed. However, these dynamic components are arrested in the dehydration state. This study not only reveals the hierarchy of the protein dynamics associated with side-chain motions, but also provides quantitative descriptions for the dynamic components observed in the ST-ESR results of the fully hydrated T4L. The presence of hydration is required for protein to exhibit its dynamics, whereas it does not dominate the protein dynamics. The studies presented here support that bulk solvent plasticizes protein and facilitates rather than slaves protein dynamics.

Materials and Methods

Sample Preparations and Spin-Labeling Reactions of the R1 and RX Side Chains

The pseudo-wild-type T4L construct containing the substitutions C54T and C97A was subcloned into NdeI/HindIII site of pET28a vector (New England Biolabs), and then used to prepare constructs with single-cysteine mutants (for R1) and double-cysteine mutants (for RX) (Figure S4). All mutants were generated via the Quikchange site-directed mutagenesis kit (Stratagene) and verified by DNA sequencing. The recombinant pET28a vector was transformed into the E. coli BL21(DE3) expression strain (Novagen). Recombinant proteins fused with six histidines at the N-terminal of T4L were expressed and purified by affinity Ni column as previously described.[48] Briefly, bacterial cultures were grown at 37 °C in LB medium containing kanamycin (50 μg/mL) until OD600 reached 0.8–1.0. T4L overexpression was induced upon addition of 1 mM of IPTG (isopropyl-β-D-thiogalactopyranoside) at 30 °C for 1–2 h. The cells were then harvested by centrifugation, and the supernatant was discarded. The cell pellet was collected and resuspended in ice-cold lysis buffer (25 mM Tris-HCl, 25 mM Mops, pH 7.6, 40 mM imidazole, and 1 mM PMSF). The resuspended pellet was sonicated on ice for 5 min, followed by centrifugation at 13 000g for 50 min. The supernatant was filtered through 0.22 μm filter and then loaded onto an affinity Ni column using HisTrap HP (GE Healthcare) at a flow rate about 0.5–1 mL/min. The column was washed with 3 column volumes of wash buffer (200 mM NaH2PO4, pH 7.6, 40 mM imidazole, 0.5 M NaCl). The T4L fraction was eluted with elution buffer (200 mM NaH2PO4, pH 7.6, 0.5 M imidazole, 0.5 M NaCl). Imidazole was removed by using a PD-10 desalting column (GE Healthcare) equilibrated with buffer A (50 mM MOPS, 25 mM NaCl, pH 6.8). Purified protein was confirmed by SDS-PAGE with Coomassie blue staining, and protein concentration was estimated via absorption spectroscopy at 280 nm.

Spin labeling of the cysteine variants of T4L mutants was performed in buffer A. Single-cysteine mutant of T4L was labeled with a 10-fold molar excess of (1-oxy-2,2,5,5-tetramethyl-3-pyrroline-3-methyl) methanethiosulfonate spin label (MTSL) (Alexis Biochemicals) per cysteine residue in the dark. The reaction was allowed to proceed at 4 °C for at least 6 h for solvent-accessible sites, and at room temperature overnight for buried sites. Double-cysteine mutant of T4L was designed to locate at the i and i + 4 positions of an α-helix and reacted with a 10-fold molar excess of HO-1944 (Toronto Research Chemicals) at 4 °C overnight[39] (Figure S4). Excess reagent was removed by desalting using a PD-10 column equilibrated with the same buffer. MALDI-TOF experiments were conducted to confirm the identity of proteins carrying spin labels.

Preparation of T4L Mutants Containing Unnatural Amino Acid To Form K1

To generate the nitroxide side chain K1, the genetically encoded unnatural amino acid p-acetyl-l-phenylalanine (p-AcPhe) was reacted with a hydroxylamine-functionalized nitroxide (HO-4120) (Figure S4), as previously reported.[40,49] Briefly, the suppressor tRNA and aminoacyl-tRNA synthetase (aaRS) pair was employed to incorporate the unnatural amino acid p-acetyl-l-phenylalanine (p-AcPhe) into the recombinant protein in response to the amber (TAG) stop codon. Amber mutants of T4L at the site of interest were generated using the Quikchange site-directed mutagenesis method in the pseudo-wild-type T4L background and were confirmed by DNA sequencing. T4L plasmid containing mutated gene and pEVOL plasmid (kindly given by Professor P.G. Schultz) containing the evolved tRNA/aaRS pair were introduced into the E. coli BL21(DE3) expression strain. LB medium containing kanamycin (50 μg/mL), chloramphenicol (34 μg/mL), and p-acetyl-l-phenylalanine (0.3 g/L) (Amatek Chemical) was prepared for inoculation. The culture was incubated at 37 °C until it reached an OD600 of 0.6–0.8. Expression of two plasmids was induced by addition of 1 mM IPTG and 0.02% l-arabinose at 37 °C for 3–4 h. Cells were harvested by centrifugation. The purification of T4L containing p-AcPhe is the same as the aforementioned procedure for the cysteine variants of T4L. The complete incorporation of p-AcPhe into T4L was verified by MALDI-TOF experiments.

The genetically encoded p-AcPhe was then reacted with a hydroxylamine reagent (HO-4120) to generate a ketoxime-linked (K1) side chain (Figure S4). Labeling of the T4L mutants containing p-AcPhe was performed in mildly acidic buffer (50 mM sodium phosphate and 25 mM NaCl at pH 4.0). Purified T4L mutants were reacted with a 10-fold excess of HO-4120 (Toronto Research Chemicals) at 37 °C overnight. Unreacted reagents were removed by PD-10 desalting column, eluting with buffer A.

ESR Measurements

The final concentration of the spin-labeled T4L was 0.3–0.5 mM in buffer A containing 30% (v/v) glycerol (which is equivalent to the 10 mol % glycerol/water condition, approximately). Oxygen was removed from the sample by using a freeze–thaw method. Approximately, 40 μL of sample volume was loaded into two capillaries (Kimble, 34507-99) prior to being placed in 4 mm quartz ESR tube. All ESR measurements were carried out on a Bruker ELEXSYS E580 spectrometer equipped with X-band microwave bridge (microwave frequency, 9.45 GHz), an ER 4122 SHQE cavity, and an ER 4131 VT unit for temperature control. For cw-ESR measurements, spectra were acquired with microwave power of 1.5 mW, 100 kHz modulation frequency, 1 G modulation amplitude, and 200 G sweep width. For ST-ESR measurements, spectra were acquired according to the reported procedure,[23] with 90 mW incident microwave power, 50 kHz modulation frequency, 5 G modulation amplitude, and 150 G sweep width. Modulation phase was set at 90° with phase-sensitive detection at second harmonic. Null phase method was used to minimize the signal at unsaturating microwave power.[29] For the measurements at increasing temperatures, the ESR probe-head was precooled to 180 K prior to the transfer of ESR sample tube into the cavity. Spectra were recorded stepwise from 180 to 240 K with an equilibrium time >10 min for each temperature.

Analysis of ST-ESR Measurements

The rotational correlation time (τ) of nitroxide spin labels can be estimated by analyzing the ST-ESR spectrum of TEMPOL (4-hydroxy-2,2,6,6-tetramethylpiperidine-1-oxyl) spin labels according to a previously published method.[23] Basically, the method is based on the fact that the integral of ST-ESR spectra is sensitive to the spin–lattice relaxation time of the electron of the spin label, which in turn is directly dependent on the rotational correlation time. From the known viscosity data and the related rotational correlation times of the TEMPOL spin label in pure glycerol, the rotational correlation times of unknown samples can therefore be determined from the ST-ESR spectra. The reliability of this method has been previously demonstrated in glycerol/water mixtures containing TEMPOL.[23]

The present study followed the above method. We first calculated the τ(T) data of the spin label (Figure S1A) using a modified Stokes–Einstein equation[23]where η is the solvent viscosity, V is the volume of molecule, k is a dimensionless interaction parameter, τ0 is the zero viscosity rotational correlation time, kB is the Boltzmann constant, and T is absolute temperature. For the case of TEMPOL, the volume V is 0.180 nm3; k is 0.09, and τ0 is negligibly small.[23] The viscosity of anhydrous glycerol in the low-temperature region was obtained using the Williams–Landel–Ferry (WLF) equation.[23] We then collected the ST-ESR spectra of the TEMPOL spin label in anhydrous glycerol as a function of temperature from 180 to 294 K (Figure S1B). TEMPOL spin label was purchased from Sigma-Aldrich. The final concentration of the TEMPOL spin probes was 0.3–0.5 mM. The spectra were analyzed to yield the two spectral characteristics, L ratio (L′/L) and relative integrated intensity, as a function of T, followed by the substitution of T by τ using the calculated τ(T) data displayed in Figure S1A. Thus, we obtained plots of the L ratio (L′/L) and relative integrated intensity (see Figures S1C and S1D for details) of the ST-ESR spectra as a function of the rotational correlation time τ. As shown (Figure S1), this approach by ST-ESR enables us to measure rotational correlation times of spin labels up to values of 102 s. These data (Figure S1) were then used as a calibration for the rotational correlation times of the samples in the present study. Note that although the dynamics of protein-attached spin label is derived from the reference curves of the spectra of free TEMPOL probe, the temperature-dependent τ values are considered to reflect largely the anisotropy of local spin-label motions. The validity of this analysis approach for extracting local restricted motions from the ST-ESR spectra of spin-labeled protein was previously reported.[50] For a macroscopically disordered system (e.g., spin-labeled proteins in solution), the L ratio parameter (used in the present study) was proven to have sufficient sensitivity to the local restricted motions of spin labels.[51] From these studies, it follows that the parameters derived from the ST-ESR spectra of spin-labeled proteins are sufficiently sensitive to local anisotropic motions. We therefore consider the correlation time τ obtained as an effective τ corresponding to the anisotropic motions of spin label.

Fitting of the Temperature Dependence of Rotational Correlation Time

The obtained τ data of the TEMPOL in the glycerol/water mixtures (Figure A,B) exhibit a clear dynamical transition TI, consistent with the state diagram of glycerol/water mixtures (Figure S3). As demonstrated previously, the data are characterized by the VFT relation at temperatures above TI and the Arrhenius-type relation at temperatures below TI. The VFT relation[37]has been widely used for understanding the nature of the transitions in glassy materials, where D is a constant related fragility, and T0 is the ideal glass transition temperature at which correlation time appears to diverge. The Arrhenius-type relation is given as[37]where R is gas constant, and U is the potential energy describing the rotational dynamics of the incorporated spin labels in the host material. The temperature dependence of the τ data from the spin-labeled T4L studies was fit to the Arrhenius-type relation. The random sample consensus (RANSAC) algorithm was used to choose the inliers for linear regression in the respective temperature ranges.[52] In this algorithm, two random points were selected for each iteration, and the number of iteration steps is set to 500 to generate a reproducible and reliable result. The threshold was set to 0.032 to rule out the outliers. The probability of choosing inliers in the selected temperature range was above 90% as criteria to define the range in which the chosen inliers follows the same Arrhenius-type behavior.