Introduction

For high-resolution NMR structure determinations, especially for using an automated NOE signal assignment procedure that is becoming popular, a sufficient number of NOE peaks with nearly complete and accurate assignments are required. These requirements cannot always be met, and depend on the nature of the target protein being examined. Widely used computational programs, such as CYANA (Güntert et al. 1997; Herrmann et al. 2002), ARIA (Nilges et al. 1997) and autoStructure (Huang et al. 2003), can determine using the automated NOE signal assignment procedure. In the case of CYANA, more than 8.4 NOE restraints per residue and more than 90 % of chemical shift assignments are necessary for a successful NMR structure determination using the automated NOE assignment procedure (Jee and Güntert 2003). For membrane proteins and loop-rich proteins, these requirements are hardly met.

Paramagnetic relaxation enhancement (PRE)-derived distance restraints are easily obtained using spin labels. Spin labels contain a stable lone-pair electron, and are attached to target proteins at specific sites. The lone-pair electron enhances R 2 relaxation rates of NMR active nuclei, and the distances between the spin-labeled moiety and the NMR active nuclei are determined from enhancements of the R 2 relaxation rates. PRE provides semi-quantitative distance information in the range of 15–24 Å. Although PRE can provide longer distance information compared with NOE, it is less precise due to the flexibility of the attached spin-labeled moiety and de-localization of the lone-pair electron.

PRE is effective in accurately determining the global fold of a protein with a limited NOE data set (Battiste and Wagner 2000). A number of comprehensive studies that use PREs for structural analysis of proteins have been performed (Liang et al. 2006; Volkov et al. 2006; Clore et al. 2007; Clore and Iwahara 2009; Simon et al. 2010; Madl et al. 2011). Recently, PRE has been utilized to determine inter-domain orientation and dimer interfaces (Madl et al. 2010; Yang et al. 2010), and the tertiary structure of membrane proteins in detergent micelles (Roosild et al. 2005; Zhou et al. 2008; Van Horn et al. 2009; Reckel et al. 2011). Detailed analyses of the influence of PRE-derived distance restraints on structure determinations have been reported for α-helical membrane proteins using simulated NMR data (Gottstein et al. 2012). It was reported that the number and location of spin labels affected the resulting protein structures. However, to our knowledge, the influence of PRE-derived distance restraints on high-resolution structure determinations of soluble proteins has not been investigated.

SAPK-interacting protein 1 (Sin1) is a subunit of the target of rapamycin (TOR) complex 2 (TORC2) (Frias et al. 2006; Jacinto et al. 2006; Yang et al. 2006). TOR is a serine/threonine-specific protein kinase, and the composition of TORC2 is conserved from yeast to humans (Cybulski and Hall 2009). Mammalian TORC2 (mTORC2) activates AGC-family protein kinases, such as Akt (protein kinase B), protein kinase Cα and SGK1, suggesting that mTORC2 may play a role in the regulation of cellular metabolism and proliferation (Hresko and Mueckler 2005; Sarbassov et al. 2005; Wullschleger et al. 2006; Facchinetti et al. 2008; García-Martínez and Alessi 2008; Cybulski and Hall 2009). In the fission yeast Schizosaccharomyces pombe, TORC2 activates the AGC-family kinase Gad8 (Matsuo et al. 2003; Ikeda et al. 2008). The conserved region in the middle domain of Sin1 (Sin1CRIM) binds specifically to Gad8 (Tatebe et al. in preparation), and recruits Gad8 to TORC2 for phosphorylation and subsequent initiation of downstream signal transduction. Based on NMR chemical shift values of Sin1CRIM reported by our group (Kataoka et al. 2014), more than half of the residues of Sin1CRIM protein were estimated to be located within loops.

In this report, the impact of PRE-derived distance restraints on the NMR structure determination of the loop-rich soluble protein Sin1CRIM is described. Following application of the conventional NOE-based structure determination procedure to determine the structure of Sin1CRIM, PRE-derived distance restrains were included in the NOE-based structure calculations. The impact of PRE-derived distance restraints on the structure determination was further investigated by changing the number of PRE- and NOE-derived distance restraints.

Materials and methods

Sample preparation of Sin1CRIM protein

A pCold-GST expression vector (Hayashi and Kojima 2008) encoding the S. pombe Sin1 CRIM domain (amino acids 247–400), designated as Sin1CRIM in this paper, was constructed as previously reported (Kataoka et al. 2014). This was utilized for the expression, purification and preparation of non-labeled, 15N-labeled and 13C, 15N-labeled protein samples. Details of the protein sample preparation are provided in supplementary materials and methods.

Cysteine mutagenesis and spin-labeling of Sin1CRIM mutants

Single cysteine mutations of Sin1CRIM were introduced by the QuikChange site-directed mutagenesis method (Stratagene). Mutation sites were mostly selected from Ser/Thr residues, polar Arg/Lys/Asp/Glu/Gln residues expected to locate on the surface, and other Phe/Tyr/Leu/Val/Ala/Gly residues located on the edge of the secondary structure. Mutated 15N-labeled Sin1CRIM proteins were overexpressed and purified by the same procedure utilized for the wild-type protein, except that the buffer used in the final size-exclusion column chromatography (SEC) did not contain DTT. Following purification, the spin-labeled reagent MTSL [(1-oxyl-2,2,5,5-tetramethyl-·3-pyrroline-3-methyl) methanethiosulfonate] was attached to the thiol moiety of the introduced cysteine residues by mixing MTSL with mutant Sin1CRIM samples (0.1 mM) at a 10:1 (MTSL:protein) molar ratio and then incubating the mixture at 20 °C for 4 h. Unreacted MTSL was carefully removed by SEC. The conjugation of MTSL was confirmed by mass spectroscopy before and after the collection of both paramagnetic and diamagnetic NMR data. When the incomplete spin labeling was found, its NMR data were not used for the analysis. Sin1CRIM and spin-labeled Sin1CRIM mutants were monomer evaluated by SEC.

NMR sample preparation

NMR samples of Sin1CRIM for resonance assignments and NOESY measurements were prepared at a protein concentration of 0.5 mM in 90/10 % H2O/D2O containing 50 mM potassium phosphate (pH 6.8), 50 mM KCl and 1 mM DTT. For PRE measurements, NMR samples were prepared at a protein concentration of 0.1 mM to avoid contributions from additional undesired PREs arising from random “elastic” collisions between a molecule and the paramagnetic group of another molecule (Clore and Iwahara 2009). In fact, for spin-labeled K312C mutant, the peak intensity of 1H–15N HSQC at 0.2 mM was 10 % lower than the expected, indicating the presence of non-specific intermolecular interactions at 0.2 mM (Figure S1). Following collection of paramagnetic NMR data, MTSL was reduced by addition of a threefold molar excess of ascorbic acid into the NMR sample in order to collect diamagnetic NMR data. The reduction reaction was performed at 25 °C for more than an hour. NMR sample for residual dipolar coupling (RDC) measurements was prepared at a protein concentration of 70 μM in 90/10 % H2O/D2O containing 50 mM potassium phosphate (pH 6.8), 50 mM KCl, 1 mM DTT, and in the absence or presence of ~15 mg/ml Pf1 phage (ASLA Biotech Ltd). The NMR sample containing Pf1 phage was incubated at 4 °C for 12 h, and magnetically aligned in the NMR magnet at 30 °C for 5 h prior to the NMR measurement.

NMR measurements and analyses

NMR experiments were performed using an AVANCE I 800 MHz spectrometer or an AVANCE III 950 MHz spectrometer (Bruker). Spectra measured for resonance assignments were as previously described (Kataoka et al. 2014). To obtain distance restraints, 13C-edited NOESY (Muhandiram et al. 1993) and 15N-edited NOESY spectra (Zhang et al. 1994) were recorded. To obtain PRE distance restraints, 1H–15N HSQC spectra of MTSL-conjugated proteins in the paramagnetic and diamagnetic state were recorded. To obtain χ1 angle restraints, three bond JC·C·and JNC· coupling constants (Hu and Bax 1997) were measured. {1H–15N} heteronuclear NOE spectra were measured with and without 3 s of proton pre-saturation in an interleaved fashion (Farrow et al. 1994). 1H–15N RDCs were measured using in-phase and anti-phase 1H–15N HSQC pulse sequences under both isotropic and anisotropic conditions (Ottiger et al. 1998).

Uniformly sampled NMR spectra were processed using NMRPipe (Delaglio et al. 1995), while non-uniformly sampled (NUS) NMR spectra were processed using the Rowland NMR tool kit (http://rnmrtk.uchc.edu/rnmrtk/RNMRTK.html). Spectra were analyzed using Magro NMRView (http://bmrbdep.protein.osaka-u.ac.jp/en/nmrtoolbox/magro_nmrview.html) (Johnson and Blevins 1994; Kobayashi et al. 2007) and Sparky 3.115 (Goddard and Kneller 2008).

Chemical shift assignments of Sin1CRIM have been previously described in detail (Kataoka et al. 2014; BMRB accession code 11546). Resonances for 94 % of the backbone 1HN and 15N, 92 % of 13C’, 95 % of 13Cα, 91 % of 1Hα, 72 % of side-chain 1H and 69 % of side-chain 13C were assigned. Backbone torsion angle restraints were estimated using the program TALOS+ (Shen et al. 2009). Chemical shift assignments of spin-labeled Sin1CRIM were from the wild-type Sin1CRIM comparing the HSQC spectra, simply the nearest cross peak.

PRE-derived distance restraints

PRE-derived distance restraints were calculated from intensity ratios of 1H–15N HSQC spectra in the paramagnetic and diamagnetic states (Figures S2 and S3) by the method introduced by (Battiste and Wagner 2000). The 1H–15N HSQC spectra were recorded within 90 min for each, since the spin labeled samples were not stable (Figure S4). Details of the PRE-derived distance calculations are provided in supplementary materials and methods.

PRE-derived distance restraints were classified into three types; (1) Peaks with an intensity ratio < 0.8 and detectable in the oxidized spectra, (2) severely broadened peaks and not detectable in the oxidized spectra, and (3) peaks with an intensity ratio >0.8. Peaks in class (1) were restrained as the calculated distance. Peaks in class (2) were restrained with no lower distance limit and upper distance limits of distances estimated from the noise level. Peaks in class (3) were restrained with no upper distance limit and lower distance limits of distances (11 Å for R291C and K321C and 10 Å for others) calculated from intensity ratio of 0.8 considering experimental errors. PRE-derived distance restraints were introduced between amide protons and Cβ atoms of residues which were replaced by cysteine for the paramagnetic labeling (Reckel et al. 2011; Gottstein et al. 2012), with an error of ±7 Å. The flexibility was not considered here as a first approach. The concentration dependence of the PREs of the spin-labeled K312C mutant did not show the significant intermolecular PREs (Figure S5).

Structure calculation

Structure calculations and automated NOE assignments were performed using CYANA 3.95 (Güntert et al. 1997; Herrmann et al. 2002). Fifty structures were calculated, and ten structures with the lowest target function were selected for evaluation of the structural quality and treated as representative. When structure calculations were combined with automated NOE assignments, the NOE peaks were automatically assigned in seven cycles of the structure calculations, and NOE assignment tables were utilized for the final structure calculation. Fifty structures were calculated and ten structures with the lowest target function were selected in each cycle. Details of the structure calculations are provided in supplementary materials and methods. The atomic coordinates of the refined structures of Sin1CRIM have been deposited in the Protein Data Bank with accession code 2RUJ.

Validation of calculated structure

Determined structures were validated by examining Pearson’s linear correlation coefficient between experimental RDC values and back-calculated RDC values obtained from the structures using the program PALES (Zweckstetter and Bax 2000). Experimental RDC values were measured using the nmrDraw program (Delaglio et al. 1995). Only RDC values of residues that were classified as being located in helix or extended regions by the TALOS+ program were employed for the validation procedure.

Results

Secondary structure and dynamics of Sin1CRIM

The relationship between the secondary structure and the dynamics was investigated for Sin1CRIM. The 1H–15N heteronuclear NOE values of both N- and C-terminal regions of Sin1CRIM (amino acids 247–273 and 396–400) were below 0.6, indicating that both N- and C-terminal regions are flexible (Fig. 1) (Kay et al. 1989). For the central region (amino acids 274–395), the 1H–15N heteronuclear NOE values were above 0.6, indicating that the central region is structurally ordered. The secondary structure was estimated by the backbone dihedral angles evaluated by the TALOS+ program (Shen et al. 2009). The central region had a low content of secondary structure elements, and more than half of the residues were located in loop regions (Fig. 1). That is, Sin1CRIM is a highly loop-rich protein, although these loops are inflexible and structurally ordered.

Fig. 1
figure 1

{1H–15N} heteronuclear NOE values of Sin1CRIM. Residue numbers are denoted on the horizontal axis. The secondary structure elements of Sin1CRIM determined by the program TALOS+ are shown above the graph. Errors for NOE values were estimated by a Monte Carlo procedure

Full size image

Structure calculation without PRE-derived distance restraints

The structure of Sin1CRIM was calculated using a conventional NOE-based procedure, and combined with automated NOE assignments using the program CYANA. However, the calculated structures were not converged sufficiently, mainly due to the insufficient number (>1,000) of NOE restraints (Fig. 2; Table 1). The backbone root mean square deviation (RMSD) values of the ordered region (amino acids 275–395) and the correlation coefficient between experimental and back-calculated RDC values using the calculated structures were 3.06 ± 0.89 and 0.56 ± 0.12 Å, respectively (Table 1).

Fig. 2
figure 2

Superimposition of the final ten structures calculated by the program CYANA combined with automated NOE assignments in the absence of PRE distance restraints

Full size image
Table 1 Structural statistics of Sin1CRIM
Full size table

Although most 1H–15N HSQC peaks of Sin1CRIM were well resolved with high signal intensities (Fig. 3), many signals in the 3D NMR spectra were weakened or disappeared. For example, in the H(CCO)NH and C(CO)NH spectra, many signals were not observed, which imposed a limit on the signal assignments of side-chain chemical shifts (~70 %), and the 13C- and 15N-edited NOESY spectra could not be measured with high signal-to-noise (S/N) ratio. Additionally, the NMR sample of Sin1CRIM precipitated within 3 days of the NMR measurements at 30 °C. These experimental limitations made the structure determination difficult.

Fig. 3
figure 3

Site-specific spin labeling of Sin1CRIM protein. a Spin-labeled sites on Sin1CRIM (amino acid 272–397). Red spheres indicate sites of spin labels from which PRE distance restraints were obtained and used for structure determination (T280, S282, R291, S301, K312, L332, S371, T384 and A394). Yellow spheres indicate sites of spin labels from which PRE distance restraints were obtained and not used for structure determination (S317, F361 and A386). Blue spheres indicate sites of spin labels from which PRE distance restraints could not be obtained (G321, Q331, Q341, G355 and R366). b An overlay of 1H–15N HSQC spectra of Sin1CRIM (WT) (blue), MTSL-conjugated Sin1CRIM (K312C) in the diamagnetic (green) and paramagnetic (red) states. c Intensity ratio of peaks calculated from 1H–15N HSQC spectra of Sin1CRIM (K312C) in the paramagnetic and diamagnetic states. Error bars indicate experimental uncertainties based on the noise level in the NMR spectra

Full size image

Site-directed spin labeling and PRE-derived distance restraints

Since the high-resolution structure of Sin1CRIM could not be determined by a conventional NOE-based procedure, PRE-derived distance restrains were employed. In order to obtain the PRE-derived distance restraints, 29 single cysteine mutants were designed for site-directed spin labeling using MTSL (Table S1). Fourteen of these mutants were successfully introduced into the Sin1CRIM gene by genetic engineering, and the respective recombinant proteins overexpressed and purified. Following conjugation of MTSL onto the thiol moiety of the cysteine residue, 1H–15N HSQC spectra were measured for each mutant under both paramagnetic and diamagnetic conditions. Following the NMR measurements, the conjunction of MTSL was confirmed by mass spectroscopy for all samples. Of the mutants generated, mutants S269C, Q331C, Q341C, R366C and S399C were not used to collect PRE-derived distance restraints (Fig. 3a). For S269C and S399C, the substituted cysteines were located in flexible regions (Fig. 1). For Q341C, marked chemical shift changes were observed for many peaks compared with wild-type (Figure S2). For Q331C, a 1H–15N HSQC spectrum with sufficient signal-to-noise ratio was not obtained (Figure S2). For R366C, the 1H–15N HSQC spectra of the paramagnetic and diamagnetic states differed significantly (Figure S2). For the residual 9 mutants (T280C, S282C, R291C, S301C, K312C, L332C, S371C, T384C and A394C), the distances between the lone spin of MTSL and amide protons were evaluated based on a comparison of the signal intensity ratios of the 1H–15N HSQC spectra of the paramagnetic and diamagnetic states (Figures 3 and S3). It is noted that PRE-derived distances for flexible residues, where the 1H–15N heteronuclear NOE values were below 0.6 (Kay et al. 1989) (Fig. 1), were not used for the structure calculations.

Structure calculation with PRE-derived distance restraints

About 100 PRE-derived distance restraints obtained from each mutant were used ‘independently’ for the structure calculation combined with the automated NOE assignments. A total of nine sets comprising about 100 PRE-derived distance restraints were obtained from nine mutants (Fig. 3a), however, no sufficient improvement in the RMSD values or RDC correlation coefficients of the calculated structures was achieved when each set of about 100 PRE-derived distance restraints was used ‘independently’ (Table S2). The location of the mutations and sets of PRE-derived distance restraints were widely dispersed over the Sin1CRIM molecule (Figs. 3a , S6). This result clearly suggests that about 100 PRE-derived distance restraints obtained from one single-site spin label were insufficient to improve the quality of the calculated protein structure.

All data obtained from the nine mutants were then used ‘simultaneously’ for the structure calculations combined with automated NOE assignments since each data set of about 100 PRE-derived distance restraints failed to improve the quality of the calculated structure. The total number of PRE-derived distance restraints employed was 867, comprising 163 upper and 704 lower distance limits. The accuracy and convergence of the calculated structures improved dramatically (Fig. 4; Table 1). The backbone RMSD values and the correlation coefficient between experimental and back-calculated RDC values using the calculated structures were 0.91 ± 0.17 and 0.86 ± 0.05 Å, respectively (Table 1 ; Figure S7).

Fig. 4
figure 4

Superimposition of the final ten structures calculated by the program CYANA combined with automated NOE assignments in the presence of PRE distance restraints

Full size image

The structure was further refined by Xplor-NIH (Schwieters et al. 2003, 2006). A superimposed picture of the refined structures and a ribbon representation of the lowest energy structure are shown in Figure S8, and the structural statistics is presented in Table S3. The structure of Sin1CRIM and its homologs have not been reported. The lowest energy structure was submitted to Dali server (Holm and Rosenström 2010) to find the similar structure. The most similar structures were Ubiquitin-fold modifier 1 (PDBID 1L7Y, Z-score = 5.4) and Ras binding domains from RalGDS (PDBID 1RAX, Z-score = 5.3 and PDBID 1LXD, Z-score = 5.1). These structures have ubiquitin-like fold and belong to the ubiquitin-like superfamily on the SCOP database (Murzin et al. 1995). Since the structures possessing the Z-score above 2 have similar fold (Holm et al. 2008), Sin1CRIM seems to have ubiquitin-like fold.

Impact of PRE-derived distance restraints on structure determination

As described above, use of the PRE-derived distance restraints dramatically improved the determined NMR structure. In an effort to gain insight into the impact of PRE-derived distance restraints on the structure determination, the number of NOE-derived and PRE-derived distance restraints was depicted against residue number (Figs. 5 , S9). As shown by the black bars in Fig. 5a, the NOE-derived long-range distance restraints were found in the limited regions, around residues 320 and 380. Additionally, as shown by the gray bars in Fig. 5a, the number of NOE-derived distance restraints was relatively small in both the N-terminal (residues 247–300) and C-terminal (residues 385–400) regions. On the other hand, as shown by the red bars in Fig. 5a, the PRE-derived upper distance limits were evenly found over the structured region, and complemented the lack of NOE-derived long-range distance restraints, especially in the N-terminal region (Fig. 5).

Fig. 5
figure 5

NOE- and PRE-derived distance restraints. a The number of distance restraints for each residue that was used for the structure determination of Sin1CRIM. PRE, Long-range NOE and other NOE are shown in red, black and gray, respectively. b The NOEs used for the structure determination are shown by lines on the lowest target function structure of Sin1CRIM. c PREs used for the structure determination are shown by lines on the lowest target function structure of Sin1CRIM

Full size image

For the evaluation of the contribution of PRE-derived distance restraints on the structure determination, the structure calculations were repeated in the absence or presence of PRE-derived distance restraints using the same NOE-based upper distance limits. The distance restraints obtained from the final cycle and used to determine the structure are shown in Fig. 4. By employing the PRE-derived distance restraints, the RMSD values improved from 1.51 ± 0.59 to 0.98 ± 0.20 Å, and the correlation coefficient between the experimental RDC values and back-calculated RDC values obtained from the calculated structures improved from 0.59 ± 0.04 to 0.87 ± 0.05 (Fig. 6). This result indicates that PRE-derived distance restraints are reliable and can complement the insufficient NOE-based distance restraints, although the distance restraint is weak due to the large error range of ±7 Å.

Fig. 6
figure 6

Superimposition of the final ten structures calculated in the absence (a) or presence (b) of PRE-derived distance restraints. NOE upper distance limits, created through automated NOE assignments in the presence of PRE-derived distance restraints, were applied to both calculations

Full size image

Minimum number of PRE-derived distance restraints

For the evaluation of the minimum number of PRE-derived distance restraints, the number of PRE-derived distance restraints (163 and 704 of upper and lower distance limits, respectively) was randomly decreased in a stepwise manner (87.5, 75.0, 62.5, 50.0, 37.5, 25.0 and 12.5 %) and structure calculations were performed in each case. The convergence of the calculated structures and the correlation between the experimental and back-calculated RDC values decreased exponentially with decreasing number of PRE-derived distance restraints (Fig. 7). When decreasing the number of PRE-derived distance restraints by more than 50 %, the structure suddenly became dramatically worse.

Fig. 7
figure 7

Evaluation of effective minimum number of PRE. Structure calculations were performed using 12.5, 25.0, 37.5, 50.0, 62.5, 75.0, 87.5 and 100 % of the PRE-derived distance restraints. The RMSD values of backbone atoms and correlation coefficients between experimental and back-calculated RDC values using the calculate structures

Full size image

In an effort to improve the structure further, an additional five single cysteine mutants (S317C, G321C, G355C, F361C and A386C) were designed based on the determined structure to increase the number of PRE-derived distance restraints. Of these mutants, G321C and G355C were not used to collect PRE-derived distance restraints since marked chemical shift changes were observed (Figure S2). Thus the PRE-derived distance restraints collected from the residual three mutants (S317C, F361C or A386C) were used in structure calculations to improve the structure (Figs. 3a , S6). However, the structure did not improve significantly by employing these additional constraints (Figure S10).

Discussion

PRE-derived distance restraints are useful for structure refinement of the soluble ordered loop-rich protein

PRE-derived distance restraints have been evaluated and utilized for the structure determination of soluble and membrane proteins (Battiste et al. 2000; Gottstein et al. 2012). In the case of soluble proteins, Battiste et al. (2000) have shown that PRE-derived distance restraints, in combination with HN–HN NOEs without hydrogen bond restraints, can determine the global fold of a soluble protein, where the protein structure was determined with a backbone RMSD of 2.3 Å. In the case of membrane proteins, Gottstein et al. (2012) have shown that PRE-derived distance restraints, in combination with limited NOEs and hydrogen bond restraints, could provide sufficient structural information for the accurate structure determination of an α-helical membrane protein. Here, we have shown that PRE-derived distance restraints, in combination with an insufficient number of NOEs without hydrogen bond restraints, can dramatically improve the structure of a soluble protein in terms of both accuracy and convergence, and resulting in a backbone RMSD of 0.91 Å (Fig. 4). Therefore, at least in the case of soluble proteins, the employment of PRE-distance restraints is useful in the determination of an accurate high-resolution structure.

In general, secondary structure elements and related hydrogen bond restraints are key factors in NMR structure determination of proteins. For example, for most membrane proteins, the secondary structure is fixed by dihedral angles and hydrogen bond restraints (Gottstein et al. 2012). Our target, Sin1CRIM, is a structurally ordered loop-rich protein, and the content of the secondary structure elements is much lower than that found in other proteins. The hydrogen bond restraint is not used in the structure calculation since the hydrogen bond pair is not identified. In this sense, Sin1CRIM is not a good target for NMR structure determination using conventional procedures. Here, employment of a sufficient number of PRE-derived distance restraints dramatically improved the structure of Sin1CRIM in terms of both accuracy and convergence. Therefore, the use of PRE-derived distance restraints can facilitate the structure determination of structurally ordered loop-rich proteins.

Impact of PRE-derived distance restraints on automated NOE assignments

In the absence of PRE-derived distance restraints, with automated NOE assignments, the backbone RMSD value was 3.06 ± 0.89 Å and the correlation coefficient between experimental and back-calculated RDC values was 0.56 ± 0.12 (Fig. 2; Table 1). On the other hand, in the absence of PRE-derived distance restraints, with fixed NOE upper distance limits that were created by the CYANA automated assignment procedure in the presence of PRE-derived distance restrains, the RMSD value was 1.51 ± 0.59 Å and the correlation coefficient was 0.59 ± 0.04 (Fig. 6a). Although both calculations were performed in the absence of PRE-derived distance restraints, the RMSD values varied greatly. This difference is derived from the improvement in the automated NOE assignments by the PRE-derived distance restraints. The number of final NOE distance restraints increased from 924 to 967 by employing PRE-derived distance restraints, and the increased distance restraints were mostly long-range (Table 1 ; Figure S11). Additionally, the NOE distance restraints created without the PRE-derived distance restraints contained inappropriate restraints that were absent in those restraints created using the PRE-derived distance restraints. Therefore, in the automated NOE assignment process, the use of PRE-derived distance restraints led to an increase in the number of NOE-derived upper distance limits, especially long-range distance restraints, and a decrease in the number of inappropriate restraints.

Conclusions

The present work shows that employment of PRE-derived distance restraints can contribute to the high-resolution NMR structure determination of proteins. Combined with conventional NOE-based procedures, the use of PRE-derived distance restraints significantly improved the convergence and accuracy of the determined structure. PRE-derived distance restraints could also be used in the automated NOE assignment process of CYANA. The PRE-assisted structure calculation procedure presented here can be utilized as a powerful option to determine the high-resolution NMR structure of proteins, especially in cases where chemical shift assignments and/or NOE data are insufficient.