|Home | About | Journals | Submit | Contact Us | Français|
Ubiquitin-like modifications, which are carried out by similar biochemical mechanisms, regulate nearly every aspect of cellular function. Despite the recent advancements in characterizing their enzymology, our knowledge about the dynamic processes of these modifications is still fragmentary. In this study, we have uncovered an intrinsic affinity between the SUMO E2 and the Cys domain of SUMO E1. NMR studies in combination with paramagnetic spin labeling demonstrate that this interaction is mediated by previously unknown interfaces on both E1 and E2, and places the two catalytic Cys residues of the two enzymes in close proximity. Site-directed mutagenesis and enzymatic assays indicate that the interaction is fundamentally important for the transfer of SUMO from E1 to E2. Results from this study suggest that the interaction between E2 and the Cys domain of E1 participates in guiding the E2's translocation to E1's enzymatic active site in ubiquitin-like modifications.
The discovery that ubiquitin-like proteins (Ublp) can conjugate to target proteins to regulate their cellular life spans and functions has revolutionized our understanding of eukaryotic regulation (Hershko and Ciechanover, 1998; Kerscher et al., 2006; Varshavsky, 1997). Despite their diverse biological functions, these Ublps all conjugate to other cellular target proteins using similar biochemical mechanisms (Haas and Siepmann, 1997; Pickart, 2001; Rose, 2005; Wilkinson, 1997). Conjugation of a Ublp to a target protein begins with the E1 enzyme, which catalyzes the adenylation of Ublp's C-terminal COOH group and then participates in subsequent formation of a thioester bond between the SH group of the catalytic Cys residue on E1 and the Ublp's C-terminal -COOH group. Ublp is then transferred to a conjugation enzyme, E2, where it forms a thioester bond with E2's catalytic Cys residue. In the final step, Ublp is attached to target proteins by the formation of an isopeptide bond between its C-terminal –COOH group and the ε-amino group of a Lys residue on the target protein. This step generally requires an E3 ligase.
Significant advances have been made in recent years on the molecular mechanism of ubiquitin-like modifications; however, structural studies do not always place the substrate proteins at the enzymatic active site of an enzyme. Recent structural studies have shown that an E1 contains at least three important functional domains: an adenylation domain that catalyzes the adenylation of a Ublp, a Cys domain containing a catalytic Cys residue which forms a covalent thioester bond with a Ublp, and a ubiquitin-like (Ubl) domain at the C-terminus (Lois and Lima, 2005; Walden et al., 2003). Studies on NEDD8 modifications indicate that the Ubl domain rotates 120 degrees during the E1-SUMO thioester formation step, which allows the binding of E2 near the catalytic Cys domain of the E1 (Huang et al., 2007; Huang et al., 2005). However, in this structure, the catalytic Cys residues of the E1 and E2 are 20 Å apart. Additional translocation of the E2 towards the catalytic Cys of E1 must occur, but the mechanism driving such a specific movement is still unclear.
In this study, we have uncovered an intrinsic affinity between the E2 and the E1 Cys domain of the small ubiquitin-like modifier (SUMO). This interaction orients the two catalytic Cys residues of the E1 and E2 as facing each other in close proximity. We have also demonstrated that this mode of interaction is important for the transfer of SUMO from E1 to E2. This study suggests that guided translocation of E2 to the catalytic Cys of E1 is mediated through multiple E2 binding sites on E1.
Proper positioning of an E2 for efficient receipt of a Ublp from E1 is likely to require the E2 to bind directly to an E1 surface near E1's catalytic Cys. We therefore investigated whether the SUMO E2 (also known as Ubc9) can bind directly to the E1's Cys domain. We expressed this domain, encompassing residues 166−382 of the SAE2 subunit. Using a 2H/13C/15N-enriched protein sample, we obtained nearly complete assignments (>96%) of the backbone resonances (Fig. 1A and 1B). The secondary structure of the Cys domain in solution was identified from chemical shifts of each residue using the program TALOS (Fig. 1A) (Cornilescu et al., 1999), which showed the isolated Cys domain's secondary structure to be nearly identical to that shown in the full-length E1 crystal structure (Lois and Lima, 2005). Three short extended strands were also predicted by TALOS. These short strands are located in surface loops and are not involved in forming β-sheets, but appear to favor backbone dihedral angles of β-strands. Overall, the isolated Cys domain maintains its structural integrity.
The expressed Cys domain binds E2 in a specific manner. Titration of the 15N-labeled Cys domain with unlabeled Ubc9 induces chemical shift changes in specific residues of the Cys domain (Fig. 1B and Figure S1A). Similarly, specific chemical shift perturbations were observed in 15N-labeled Ubc9 when it was titrated with unlabeled Cys domain (Fig. 1C and Figure S1B). The estimated dissociation constant based on chemical shift perturbation is 87.4 ± 23.6 μM (Figure S2). These results demonstrate that the Cys domain of E1 can participate in specific interactions with E2.
Since the interaction is of low affinity and chemical shift perturbation is small, we further investigated the specific nature of the interaction using paramagnetic spin-labels. We used a nitroxide spin-label, (1-oxyl-2,2,5,5-tetramethylpyrroline-3-methyl)-methanethiosulfonate (MTSL), which can be covalently attached to Cys residues (Battiste and Wagner, 2000; Gillespie and Shortle, 1997). There are four Cys residues in Ubc9; two are exposed, and two are buried. In order to determine which Cys residues could be labeled under the conditions described in the Experimental Procedures section, wild-type 15N-labeled Ubc9 was covalently linked with MTSL. The spin label was reduced, and a 1H-15N HSQC spectrum was acquired. Both the Cys93 and Cys138 resonances were shifted so dramatically that their cross peaks could not be easily identified (data not shown). In contrast, Cys75 and Cys43 maintained their original chemical shifts. This data clearly shows that only Cys138 and Cys93 are modified under the experimental conditions. In order to obtain a sample with a single spin-label, two single mutants were produced, C93A and C138A. Spin labeling did not change the protein-protein interaction, because the 1H-15N HSQC spectrum of E1's Cys domain in complex with spin-labeled Ubc9 with a reduced proxyl group is similar to that of the Cys domain in complex with non-labeled Ubc9 (data not shown).
The unpaired electron of the proxyl group induces line-broadening effects on nearby nuclear resonances in a distance dependent manner (Krugh, 1976). Line-broadening effects on the 15N-labeled Cys domain were observed in the complexes with MTSL-labelled Cys93 (Fig. 2A) and Cys138 in Ubc9 (Fig. 2B). The superimposed spectra of the reduced (black) and oxidized (red) states show clear R2 relaxation enhancement effects. With the spin-label on Cys138 of Ubc9, the signals of residues 270−274 and 276 of SAE2 were severely attenuated (Fig. 2B and 2C). With MTSL attached to Cys93, two peaks corresponding to residues 172 and 202 were below noise level in the oxidized state (Fig. 2A and 2D). Residue 172 is next to the catalytic Cys173 of SAE2; therefore, this data indicates that the interaction brings the two catalytic Cys residues, Cys93 of Ubc9 and Cys173 of SAE2, close to each other. Although Cys173's resonance is still visible, the spin label significantly reduced its intensity. These well-defined paramagnetic relaxation effects unequivocally demonstrate the specific nature of the protein-protein interaction.
The distance constraints derived from the spin-labels were used for a structure calculation using the program HADDOCK (Dominguez et al., 2003). The small degree of chemical shift perturbation upon complex formation indicates that the proteins do not undergo global conformational changes upon complex formation. Therefore, docking calculations were performed with intermolecular distance constraints derived from the spin-labeling experiments as well as the crystal structures of the SAE2 Cys domain and Ubc9. The two loops of the Cys domain encompassing residues 219−239 and residues 291−304 that are missing in the crystal structure were built using the program Modeller (Sali and Blundell, 1993). In order to avoid potential bias of the loop conformation on docking results, an ensemble of structures with different loop conformations generated by Modeller was used as the input. The first two short extended strands identified from NMR chemical shifts are located in one of the flexible loop, based on information provided from the 1H-15N NOE experiment (Figure S3). Therefore, these loops and several adjacent residues were allowed to be flexible in the calculation. The third short strand adopts an extended conformation in the crystal structure, although it was not identified as a strand in the Protein Data Bank report. Other residues that show significant chemical shift changes upon complex formation, but are located in structurally well-defined regions, were allowed to have flexible sidechains. The calculation resulted in well-converged structures (Fig. 3A). The structures satisfy all spin-label derived distance constraints, and have favorable covalent geometries and Ramachandran statistics (Table 1). The internal consistency of the intermolecular distance constraints derived from two independent spin-labels further demonstrates the specific nature of this interaction.
In this structure, the two catalytic Cys residues of E1 and E2 approach each other and are in close proximity (Figure 3B). The distance between the two sulfur atoms is approximately 14 Å, which is consistent with the spin-labeling result discussed above. Since there are no protein groups located in between the two catalytic Cys residues, they can be brought within bonding distances with a slight structural adjustment of the complex, for example, upon binding the thioester-bonded SUMO. The buried surface area is approximately 1091Å2, a value consistent with the low affinity of the complex.
The structure is supported by chemical shift perturbation data and paramagnetic relaxation effects. Segments of the proteins where chemical shift perturbation was greater than at least 1.5 times the spectral digital resolution are indicated in the structures of Ubc9 and the Cys domain of SUMO E1 (Fig. 3B). The Ubc9 residues include or are within the segments of 14−19, 26, 36, 92, 102, 129−134, and 146 (Fig. S2). Similarly, the SAE2 residues include or are within the segments of 198, 200−206, 211, 218−220, 236−237, 290, 308, 342−343, 352, and 381−382 (Fig. S1). Most of the residues that show chemical shift perturbation are located in the protein surfaces that face each other in the complex. The large surface areas and the low interaction affinity indicated by these chemical shift perturbation studies also suggest that more than one interaction mode may occur between Ubc9 and the Cys domain without significant changes in their relative orientation. Figure 3C shows the Ubc9 residues Cys93 and Cys138 where the spin-label is attached as well as the E1 residues that are most affected by the spin-labels. The structure illustrates how the two spin-labels provided intermolecular distance information.
The structure as well as chemical shift perturbation indicates that the flexible loop encompassing residues 218−240 of SAE2 is a critical component in binding Ubc9. This segment is disordered and missing in the crystal structure. The structure of this loop cannot be defined in semi-rigid body docking calculations; however, the structure implies that the loop of E1 binds residues 129−134 of Ubc9, due to their close proximity and significant chemical shift changes upon complex formation. Notably, significant chemical shift perturbations were also observed for residues at the N-terminus of Ubc9, which were previously shown to be important in binding the Ubl domain and a surface of SUMO that is conserved among all SUMO paralogues (Bencsath et al., 2002; Huang et al., 2005; Tatham et al., 2003). This result suggests that the Cys domain also binds to this multi-protein binding surface of Ubc9.
We tested whether the newly identified E1-E2 interaction, which brings the catalytic Cys residues of the SUMO E1 and E2 into proximity, is important for the transfer of SUMO from E1 to E2 during SUMO modification processes. Since the flexible loop at the interface could not be well defined, we relied on chemical shift perturbation to identify potentially important residues for the interaction. Residues 129−136 of Ubc9 showed the largest chemical shift perturbations, but were not previously shown to bind the Ubl domain or other regions of E1. We therefore tested their roles. Four Ubc9 mutations, Q130A, A131D, E132A, and Y134A had been created previously (Lin et al., 2002), and had been shown not to perturb the three-dimensional structure of Ubc9. To examine the transfer of SUMO from E1 to E2, the SUMO•E1 thioester complex was formed first, and then several different concentrations of wild-type and mutant Ubc9, ranging from 0.2 μM to 0.8 μM, were added to initiate the transfer of SUMO from E1 to E2 (Fig. 4A and 4B). The amounts of transferred SUMO were monitored at different time points. The transfer rates of all mutants were slower than that of the wild type. In particular, mutants A131D and Y134A had the slowest rates in this transfer step, indicating that they play more important roles in this process. In order to correlate the loss of function of these mutants to the E1-E2 interaction, quantitative enzyme kinetic analyses were performed for all Ubc9 mutants (Figure 4C), using a previously developed approach (Bohnsack and Haas, 2003; Siepmann et al., 2003). The apparent Km value for wild-type Ubc9 (111 ± 57 nM) is very similar to that previously determined for ubiquitin (102 ± 13 nM). The apparent Km values of Ubc9 mutants increased from six to more than twenty fold in comparison to that of the wild-type Ubc9. In contrast, the apparent Vmax values remain similar within the experimental uncertainties (Figure D), except for that of A131D, which was reduced by approximately four fold. These results confirm that the mutations significantly reduced the affinity of Ubc9 for E1.
Site-directed mutations were similarly introduced on full-length E1 in order to further examine the interaction. Structural calculations, as well as chemical shift perturbation, indicate that the loop encompassing residues 218−240 is important for the E1-E2 interaction. In order to test this hypothesis, I238 was chosen for site-directed mutagenesis, because this residue showed significant chemical shift perturbation upon complex formation. The neighboring residue I235 was also included in the mutation. We introduced an I235A/I238A double mutation in the full-length SUMO E1. Both residues have small 1H-15N NOE values (0.17 and 0.45 for I235 and I238, respectively) indicating that these residues are flexible, and thus the amino acid substitutions should not affect the structural integrity. Overall conjugation assays showed that the E1 mutant had a significantly reduced activity in overall conjugation assays in comparison to the wild-type E1 (data not shown). In order to confirm that the mutation specifically affects the transfer of SUMO from E1 to E2, the E1•SUMO thioester complex was formed first and then Ubc9 was added to initiate the transfer of SUMO from E1 to E2. The mutation did not affect the amount of SUMO•E1 thioester complex formed (Fig. 4E), but strongly affected the transfer of SUMO from E1 to E2 (Fig. 4F). A combination of site-directed mutagenesis and biochemical assays indicated that the E1-E2 interaction described here is important for the ubiquitin-like modification.
This study has revealed, for the first time, that E2 has an intrinsic affinity for a site near the catalytic Cys residue of E1. Site-directed mutagenesis and biochemical analyses have shown that the newly uncovered E1-E2 interaction plays an important role in the transfer of SUMO from E1 to E2. The specificity of the interaction is demonstrated by chemical shift perturbation as well as definitive paramagnetic line-broadening effects. The relative orientation of the two proteins has been well-defined through the use of two independent proxyl spin labels without including any ambiguous structural restraints. This approach is well suited for this and other low affinity protein-protein complexes whose structures cannot be easily determined using conventional approaches. The internal consistency in the distance constraints obtained from two different spin-labels offers cross-validation of the structure.
This study has unexpectedly revealed that the E2 region previously identified as the binding site for the ΨKxE motif (Ψ is a bulky hydrophobic residue, K is the Lys where modification occurs, x is any residue, E is a Glu) of target proteins (Bernier-Villamor et al., 2002; Lin et al., 2002) is also important for binding the E1 Cys domain for the transfer of SUMO from E1 to E2, although the E1 does not contain the ΨKxE motif. Residues Q130 and A131 appear to be most important for substrate recognition, as indicated by biochemical data (Lin et al., 2002). Residues Q130 and E132 are the most critical for the Cys domain binding (Fig. 4). This study has revealed the dual function of this important site of E2 adjacent to the catalytic Cys93 residue. This segment was not identified in a previous Ala scanning mutagenesis (Huang et al., 2007), possibly because the enzymatic assay used could be misled by artifacts, such as the E2's inhibitory effects (Siepmann et al., 2003; Wee et al., 2000). Another possibility for the difference is that the E2 enzyme for NEDD8 modification, Ubc12, contains an additional amino terminal peptide extension that contributes significantly to binding its cognate E1, and thus may not rely on an additional interaction analogous to that observed in this study. The Cys domains of SUMO and NEDD8's E1s vary widely in their lengths and sequences. However, a large surface loop in NEDD8 E1's Cys domain can also be found near the catalytic Cys residue in a position similar to that of the SUMO E1 loop that is important in binding Ubc9. Thus, the interaction between E2 and the Cys domain of E1 is likely to occur in other ubiquitin-like modifications.
The present study offers a critical missing piece to the puzzle of the E1-E2 transthiolation mechanism. Recent studies have suggested that thioester formation of an Ublp with the catalytic Cys of E1 triggers the rotation of the Ubl domain of E1, which allows the binding of E2 to the Ubl domain as well as the adenylation domain (Huang et al., 2007; Lois and Lima, 2005; Walden et al., 2003). Our studies indicate that the intrinsic affinity between E2 and the Cys domain of E1 drives the further translocation of E2 to the catalytic Cys residue of E1. The interaction between E1 and E2 revealed here can be accommodated without dramatic conformational changes of E1 when it is “charged” with an Ublp.
Although the interaction described here is weak, our studies indicate that it contributes synergistically in a “guided” translocation of a substrate to the active site. Weak protein-protein interactions have also been shown to be functionally critical in many other processes involving ubiquitin and ubiquitin-like modifications (Pornillos et al., 2002; Reverter and Lima, 2005; Shen et al., 2006; Song et al., 2004). These weak interactions can play determining roles in the presence of other binding sites.
DNA coding for amino acid residues 166−382 of the SAE2 subunit was amplified through polymerase chain reaction, and inserted into the vector pET-28a for expression of the E1 Cys domain protein with a His6-tag on the C-terminus. For all mutants, site-directed mutagenesis were carried out with the QuikChange PCR kit (Stratagene), using designed primers to create the desired amino acid residue changes. The sequences of all new plasmids were confirmed by DNA sequencing before protein expression and purification. All the recombinant proteins were expressed and purified using a Ni-NTA column as described previously (Lin et al., 2002). For NMR studies, proteins were labeled with 15N, 13C or 2H as described previously (Liu et al., 1999). The E. coli strain for expressing full-length E1 was a generous gift from Dr. Christopher Lima (Sloan-Kettering Institute, NY).
All SUMO conjugation assays were conducted at 37°C in the presence of ATP and its regeneration system (50 mM Tris, pH 7.5, 5 mM MgCl2, 2 mM ATP, 10 mM creatine phosphate, 3.5 Units/ml creatine kinase, and 0.6 Units/ml inorganic pyrophosphatase). For E1•SUMO thioester formation, 20 μM SUMO and 5 μM E1 were incubated for 10 minutes. For conjugation to substrate RanGAP1, 15 μM SUMO and 15 μM RanGAP1418−587 were incubated together with 0.5 μM E1 and 0.5 μM Ubc9 for 3 minutes. The reactions were terminated by adding SDS-gel loading buffer, and the mixtures were resolved via SDS-polyacryamide gel electrophoresis.
In order to examine the transfer of SUMO from E1 to E2, SUMO transfer assays were performed as follows. E1•SUMO thioester complex was formed first in a mixture containing ATP regeneration system (see above), 1 μM Alexa680 (Invitrogen) fluorescent dye-labeled SUMO, and 1 μM E1 protein. After 10 minutes of incubation at 37 °C, the reaction was halted by adding EDTA to a final concentration of 30 mM to remove Mg2+. Due to fast transfer rates, all the solutions for the following steps were pre-cooled on ice, and the assay was carried out in the cold room. Various concentrations of Ubc9 were added to the E1•SUMO/EDTA mixture, and samples were taken at different time points after the start of the assay and stopped by mixing with non-reducing 4M urea/SDS gel-loading buffer; the samples were fractionated by SDS-PAGE. The gel was briefly rinsed with water before the fluorescent image of the gel was taken with a Li-Cor infrared fluorescence scanner in combination with the software Odyssey. The software ImageQuant 5.2 was used for quantification.
For the steady-state kinetic experiment, preliminary assays were carried out to determine the appropriate E1 concentration range in which the rate of RanGAP1-SUMO conjugation is linearly dependent on the E1 concentration. In addition, the optimal assay duration was also determined for the examination of initial rates of the product formation. In the final kinetic assays, 10 μM RanGAP1 was mixed with 5μM SUMO1, 100 nM E1, an ATP regeneration system, and serial concentrations of E2 proteins (25, 40 and 62.5 nM for wild-type Ubc9; 100, 160 and 250 nM for Q130A mutant; 800, 1250 and 2000 nM for A131D mutant; 160, 250, 400 nM for E132A mutant; and 500, 800, 1250 nM for Y134A mutant). The reactions were carried out for 10 minutes at 37 °C before it was stopped by the addition of an equal volume of SDS-gel loading buffer containing DTT. The samples were resolved by SDS-polyacrylamide gel electrophoresis, and the bands were detected by Western blot with mouse monoclonal antibody against SUMO-1 (Abgent) and IRDye 680-labeled secondary antibody (Li-Cor Biosciences), in combination with the Odyssey infrared imaging system. SUMO protein alone was also loaded onto the gel in a series of known amount in parallel to the assay samples to ensure that the quantification of the protein bands is in the linear range of the detection. Each data point was repeated three times, and the amount of assay product formed in each reaction was calculated as the percentage of the band intensity of RanGAP1-SUMO over that of free SUMO and RanGAP1-SUMO combined. For the Lineweaver-Burk plot, the reciprocal of the amount of RanGAP1-SUMO formed per minute was plotted against the reciprocal of Ubc9 concentration, and the Km and Vmax values were extracted from linear fittings of each set of data.
Ubc9 was first treated with excess tris(2-carboxylethyl)phosphine (TCEP) for approximately two hours at room temperature to ensure that all Cys residues were reduced, followed by dialysis to remove excess TCEP. A 15-fold excess of MTSL dissolved in acetylnitrile was added to the Ubc9 solution, and the reaction was carried out at 4°C overnight. Excess MTSL was removed by extensive dialysis. The complex was prepared with a molar ratio of 1:1.1 Cys domain:Ubc9 and had pH of 7.2.
2H/13C/15N-enriched Cys domain was used to acquire the following spectra for resonance assignments: HSQC, TROSY, TROSY-HNCA, TROSY-HN(CO)CACB and TROSY-HNCACB. All spectra were acquired at 25°C on a 600 MHz Bruker Avance instrument equipped with a cryo-probe. For each complex of spin-labeled mutant Ubc9 and 15N-labeled SAE, two HSQC experiments were acquired; one with MTSL in the oxidized state (paramagnetic state) and one with MTSL reduced by addition of 5-fold excess concentrated ascorbic acid to the NMR tube, which only increased the volume of NMR sample by 1%. The latter 15N-HSQC was acquired two hours after the addition of ascorbic acid. All spectra were processed and analyzed with the programs nmrPipe, nmrDraw (Delaglio et al., 1995), and nmrView. The intensity ratios of oxidized versus reduced spectra were obtained from the peak heights, and further normalized to the largest ratio value before distance constraints were calculated.
Intermolecular distance constraints between the spin label on Ubc9 and amide protons of the SAE2 domain were calculated from paramagnetic line broadening effects using a protein rotational correlation time of 13 ns, based on the molecular weight of the complex (Battiste and Wagner, 2000). Because the complex is of low affinity, the rotational correlation time could be influenced significantly by the unbound states. However, theoretical simulation shows that when the overall correlation time changes from 6.5 ns to 26 ns, the calculated distances change about 3 Å in comparison to the distance calculated with correlation time of 13 ns. This range is covered by the 6 Å upper and lower bounds. The distance restraints were defined as follows. Residues with oxidized:reduced peak intensity ratios of 0 were set with upper limits of 20 and the lower limits were set to 1.8 Å. Residues with oxidized:reduced peak intensity ratios between 0 and 0.65 have both lower and upper limits defined by calculations. For residues with peak intensity ratios greater than 0.75, their lower limits were calculated, but upper limits were set at an arbitrary 62 Å to ensure that they are sufficiently large and that the exact value does not influence the docking results. The distance constraints used in the structural calculation are between the sulfur atom of Ubc9's cysteine and the protons of SAE2. Therefore, in order to compensate for the difference in the distances between the sulfur group and the unpaired electron as well as the flexibility of the proxyl group, 6 Å bounds were used for the constraints. A total of 107 distance constraints was used in the structural calculation.
The structure of the Cys domain-Ubc9 complex was calculated using the program HADDOCK (Dominguez et al., 2003). The X-ray structures of the Cys domain and Ubc9 were used as the starting structures. An ensemble of five Cys domain structures with diverse loop region structures, as generated by the program Modeller, was used as an input file. The two loop regions of SAE2 (residues 212−238 and 291−304) were defined as “fully flexible”. Other residues of the Cys domain and Ubc9 that showed significant chemical shift perturbation as well as significant solvent exposure (>50% for Cys domain and >30% for Ubc9) were defined as “semi-flexible”. All other residues on both proteins were fixed in the calculation. The chemical shift changes were added as ambiguous constraints in the final stages of structural calculation, which did not significantly alter the structure of the complex, but slightly improved the convergence of the structure. A total of 1000 structures were initially generated. The top 250 structures were subjected to simulated annealing calculations, and the best 20 structures were analyzed and shown here.
The work is supported by NIH grant CA094595 and GM074748 to Yuan Chen. We thank Dr. Yeon-Kyun Shin for information on MTSL spin-labeling.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Accession Numbers The coordinates have been deposited in the Protein Data Bank under PDB ID code 2PX9.