Single-stranded-DNA-binding proteins (SSBs) are required for numerous genetic processes ranging from DNA synthesis to the repair of DNA damage, each of which requires binding with high affinity to ssDNA of variable base composition. To gain insight into the mechanism of sequence-nonspecific binding of ssDNA, NMR chemical shift and 15N relaxation experiments were performed on an isolated ssDNA-binding domain (RPA70A) from the human SSB replication protein A. The backbone 13C, 15N, and 1H resonances of RPA70A were assigned for the free protein and the d-CTTCA complex. The binding-induced changes in backbone chemical shifts were used to map out the ssDNA-binding site. Comparison to results obtained for the complex with d-C5 showed that the basic mode of binding is independent of the ssDNA sequence, but that there are differences in the binding surfaces. Amide nitrogen relaxation rates (R1 and R2) and 1H–15N NOE values were measured for RPA70A in the absence and presence of d-CTTCA. Analysis of the data using the Model-Free formalism and spectral density mapping approaches showed that the structural changes in the binding site are accompanied by some significant changes in flexibility of the primary DNA-binding loops on multiple timescales. On the basis of these results and comparisons to related proteins, we propose that the mechanism of sequence-nonspecific binding of ssDNA involves dynamic remodeling of the binding surface.
The replication, recombination, and repair of DNA are fundamental genetic processes carried out by large multiprotein assemblies. Separation of the DNA duplex into single strands is a key step in all of these pathways. Consequently, each assembly requires a protein to organize and protect single-strand DNA (ssDNA) during the time that it is exposed. This role is fulfilled by ssDNA-binding proteins (SSBs). One key aspect of the function of SSBs is their need to bind ssDNA regardless of sequence.
The major SSB in eukaryotes is replication protein A (RPA). RPA is a modular protein composed of three subunits (RPA70, RPA32, RPA14) containing seven structural domains. RPA binds tightly to ssDNA with an affinity of ∼10−10 M (Wold 1997). Binding of ssDNA occurs in three different stages, resulting in an excluded site size of first 8–10 nt, then 13–14 nt, and finally ∼30 nt (Blackwell and Borowiec 1994; Bastin-Shanower and Brill 2001). Although the exact mechanism is unknown (for review, see Mer et al. 2000), the initial stage of binding has been mapped to the region 181–422 of the RPA70 subunit (Gomes and Wold 1996; Gomes et al. 1996). The X-ray crystal structure of this region (termed RPA70AB) was determined in complex with d-C8, revealing a pair of oligonucleotide binding (OB)-fold domains (Bochkarev et al. 1997). One particularly striking result from the crystal structure of the RPA70AB complex was the observation of contacts between the protein and DNA that are base-specific. These findings are seemingly inconsistent with RPA's need to bind all ssDNA sequences, and require further investigation.
In the RPA70AB crystal structure, the L45 loop in the N-terminal OB-fold domain (RPA70A) curves around the DNA and is held in place by an aromatic ring, F269, stacking between the bases (Bochkarev et al. 1997). This closed conformation is reinforced by interactions between the DNA and strands β3 and β5′. A similar arrangement is observed for the C-terminal OB-fold domain (RPA70B), as well as in the crystal structures of Escherichia coli SSB bound to ssDNA (Raghunathan et al. 2000) and yeast aspartyl-tRNA synthetase bound to RNA (Ruff et al. 1995). In contrast, in the ssDNA-bound structure of the OB-fold domain of the α subunit of the telomere end-binding protein, the L45 loop appears to have a well-ordered extended conformation that does not make direct contact with the DNA (Classen et al. 2001).
Structures of SSBs in the absence of oligonucleotides show more open binding sites. Comparison of the ssDNA-bound and DNA-free structures of the B domain from the RPA70AB structure (Fig. 1B) shows that the L12 and L45 loops are most affected by the binding of ssDNA (Bochkarev et al. 1997; Bochkareva et al. 2001). These loops provide an open interaction surface and undergo large displacements (6.28 Å for L45 and 13.44 Å for L12) to close upon the DNA. In the A domain, L12 behaves similarly, but the conformation of the L45 loop is not significantly altered by the binding of ssDNA (Fig. 1A). The authors proposed that the change observed for RPA70B is more characteristic of the binding-induced changes in SSBs and that RPA70A was an anomaly that arose from crystal packing effects. However, smaller conformational changes on the order of 2 Å were observed for the L45 loop in E. coli SSB (Fig. 1C; Raghunathan et al. 1997, 2000). In summary, specific protein–DNA contacts in both loops result in a closed conformation of the OB-fold domain upon binding of ssDNA, but the extent to which the structure and dynamics of the loops are altered upon binding of DNA remains an unresolved issue.
NMR chemical shift and relaxation studies enable a direct characterization of protein structure and dynamics. Heteronuclear relaxation has been used to show directly that the L12 loop in the bacteriophage Pf3 protein is highly flexible, but upon binding ssDNA becomes significantly less flexible (Horstink et al. 1999). In contrast, only small changes in dynamics were observed for the L45 loop. However, the L45 loop is particularly short in Pf3 and in the absence of DNA is located at the interface between two subunits of a Pf3 dimer; therefore, it is unclear if this feature is characteristic of all SSBs or unique to this protein (Folmer et al., 1995).
In this report we extend the analysis of SSB protein structure and dynamics, describing NMR chemical shift perturbation and 15N relaxation experiments on the isolated RPA70A domain. The chemical shift changes in RPA70A upon binding of d-CTTCA and d-C5 are compared, along with the changes in backbone flexibility of RPA70A induced by the binding of d-CTTCA. These results extend previous insights about the nature of the RPA recognition of ssDNA (Bochkarev et al. 1997; Bochkareva et al. 2001) and provide the foundation for a specific proposal to explain sequence-nonspecific binding of ssDNA that involves dynamic remodeling of the binding surface.
NMR spectroscopy was used to monitor the effects on RPA70A of the binding ssDNA. During ssDNA titrations, a single set of broadened peaks for the protein is observed at subsaturating levels of DNA. This behavior is typical of slow or intermediate exchange between the free and bound protein, consistent with the ∼5 μM binding affinity of RPA70A for ssDNA (A.I. Arunkumar and W.J. Chazin, unpubl.). To carry out the detailed analyses described below, the requisite backbone 13C, 15N, and 1H NMR resonance assignments for RPA70A were made in the absence and presence of d-CTTCA by standard triple resonance approaches (Cavanagh et al. 1996).
Binding-induced changes in RPA70A structure
The changes in NMR chemical shifts induced by binding of ssDNA were used as a means to monitor effects on the structure of RPA70A. The amide proton and nitrogen chemical shift perturbations are considered significant if >0.1 ppm. Figure 2A shows the significant chemical shift differences mapped on the three-dimensional structure of RPA70A. These results indicate that the ssDNA-binding surface includes strand β3 and the L45 loop on one side, and the short β-hairpin between β1′ and β2 on the other side. Indirect effects of binding appear to be present in three contiguous stretches of the backbone: W212–S223 including residues in the L12 loop and strand β2; R234–N239 in strand β3; and N266–E277 in the L45 loop. There are minimal perturbations of the chemical shifts at the distal side of the binding site (e.g., β1, α1, β4, β5). Thus, the chemical shift analysis is fully consistent with the crystal structure of the RPA70AB–ssDNA complex, in which the binding pocket for the DNA consists of loops L12 and L45 and the concave surface of the β-barrel at strand 3 (Fig. 2A).
To further test that the structure of RPA70A in the complex with d-CTTCA is fully consistent with the structure in the RPA70AB complex with d-C8, we have collected both 15N and 13C HSQC spectra under identical conditions for RPA70A in the presence of d-CTTCA and d-C5. Figure 2 shows overlays of selected regions from these spectra. The observed differences in chemical shifts are limited in number and small in magnitude, which implies that overall, the structure of RPA70A in these two complexes is very similar. However, these spectra also show the existence of some highly localized specific chemical shift differences, which can be attributed to the local structural adaptation of RP070A to the two different DNA sequences. The effect of differences in ring current shifts is not the cause of these differences in chemical shifts because the ring currents of the pyrimidine bases are fairly small (relative to purine bases) and similar in magnitude to each other (Case 1995).
Binding-induced changes in RPA70A NMR relaxation parameters
To obtain insights into the effects of binding ssDNA on the dynamics of the RPA70A domain, we have measured 15N relaxation parameters. The 15N R1, R2, and NOE values for the backbone amides of RPA70A in the absence and presence of d-CTTCA are plotted in Figure 3. Variations observed in these parameters can be correlated with fast-timescale (picoseconds to nanoseconds) internal motions of the amide groups. Most informative are the comparisons of the trends in these values in the absence versus the presence of ssDNA. Each parameter provides a window into motions occurring in a specific range, thus it is quite usual to observe differences in one or two parameters while the others remain constant.
In the absence of DNA, the relaxation data show that the N-terminal residues (L190–W197) and C-terminal residues (E290–D291), as well as residues in the tip of L12 (W212–K220) and L45 (I264–N274), are flexible. In addition, residues F257, S258, and G260 are of note because they possess higher than average R2 values, which implies these residues possess significant flexibility on a timescale of milliseconds to microseconds.
Addition of ssDNA to RPA70A is seen to partially reduce the internal motions of some regions of the protein that have above-average flexibility in the absence of DNA. The most remarkable observation is the increase in 1H–15N NOE values in the L12 loop. A change in the NOE of residues A265–D275 in the L45 loop is also observed, but these are smaller in comparison. Several of these residues in the L45 loop also show significant changes in R1 and R2 values. Interestingly, the crystal structure indicates that the decrease in flexibility in the L45 loop is due to the direct involvement of only two residues (K263, F269) in binding ssDNA. Thus, the effect of constraining the backbone at these two points appears to be propagated to adjacent residues along the peptide chain.
Relating RPA70A NMR relaxation parameters to dynamics
Because the relaxation data themselves are difficult to interpret in terms of the dynamics of a molecule, they are usually fit to simplified motional models, such as that of the Model-Free formalism (Lipari and Szabo 1982a), to gain insight into the timescale and amplitude of internal motions. In this approach, flexibility is represented by the square of the order parameter (S2), which varies from values approaching unity in regions with restricted internal motions to lower values (typically S2 < 0.75) in more flexible regions of the protein.
We first performed a progressive series of analyses assuming isotropic and axially symmetric overall tumbling of the system (Mandel et al. 1995). Although the fits obtained using this approach were acceptable by standard criteria, unusually high values for the average order parameter (Savg2 > 0.90) were obtained (Table 1). This was of concern because S2 values are expected to be in the range of 0.85 ± 0.07 for a well-folded β-sheet protein like RPA70A (Goodman et al. 2000). The crux of the problem is that the relaxation data cannot be fit assuming that the protein tumbles as an isolated particle in solution because RPA70A seems to undergo a limited degree of nonspecific association at the high (>100 μM) concentration needed to carry out the experiments, particularly in the absence of DNA.
Incorporating the effect of nonspecific aggregation on the analysis of relaxation data is extremely difficult because the nature of the self-association is by definition not known. This results in an increased uncertainty and reduced accuracy in parameters derived from the relaxation data (Schurr et al. 1994; Fushman et al. 1997; Kelley et al. 1997; Fairbrother et al. 1998), although the relative value of these parameters should not be greatly affected. To make at least some compensation for the effect of self-association, we used the simplest possible model, that is, one that incorporates a single additional fitting parameter derived from an equilibrium distribution of an isolated monomer and an isotropically tumbling symmetric dimer in fast exchange (Schurr et al. 1994; Fushman et al. 1997). This analysis of the data for the free protein resulted in a shift to lower S2 values. Moreover, there was a considerably greater number of residues that could be fit with the simple one-parameter fitting model (M1), a trend associated with improved fitting of the data. In the ssDNA-bound state, the effects of using the more complex model were less straightforward to interpret. Although there is a decided shift toward lower-order parameters, a tendency toward simpler fitting models was not strictly observed, thus the results obtained without compensation for aggregation were retained. Figure 4A shows a plot of the order parameters versus residue number for both the free protein and the complex with d-CTTCA.
To ensure that our conclusions from the Model-Free analyses were accurate, the relaxation data were further analyzed using the reduced spectral density mapping approach (Farrow et al. 1995). The advantage of this method lies in the absence of assumptions made a priori regarding the internal and tumbling motion of the molecule. In the spectral density mapping approach, relative flexibility is reflected in the intensity of the spectral density function (J) sampled from the relaxation data at specific frequencies, J(ω0), J(ωN), and J(0.87ωH) (Farrow et al. 1995). In practical terms, picosecond N—H bond motions are approximately reflected in J(0.87ωH) and nanosecond motions by J(ωN). Consequently, a flexible segment of the backbone would be characterized by higher-than-average J(0.87ωH) values and lower values of J(ω0) and J(ωN) (Peng and Wagner 1992Peng and Wagner 1995). A plot of the value of the spectral density at specific frequencies versus residue number is provided in Figure 4B–D. The similarity in the trends of the S2 values and the spectral density functions increases confidence in the accuracy of our conclusions.
Binding-induced change in RPA70A dynamics
The variation of the generalized order parameter and spectral densities for the backbone of RPA70A in the absence and presence of d-CTTCA is shown in Figure 4. In the absence of DNA, systematically lower-order parameters are observed in the L12 and L45 loops, which indicate the presence of above-average internal motions on the picosecond–nanosecond timescale for these residues. In addition to these fast internal motions, certain slow conformational exchange processes on the microsecond–millisecond timescale appear to affect part of the L45 loop because Rex terms (data not shown) are required for residues in the regions L248–V254 and S258–T261, respectively.
In comparing the DNA-free to the DNA-bound state of the protein, we find a modest trend toward increased order parameters in the presence of d-CTTCA. The spectral density mapping analysis shows that the most significant effect of binding ssDNA is a decrease in the J(0.87ωH) values (reduction in flexibility) for the L12 and L45 loops. The most significant changes in dynamics upon binding are observed in the L12 loop for residues N214–E218, where the order parameters increase ∼0.2 units and J(ωH) decreases. The response to binding ssDNA in the long L45 loop appears to be more subtle. To obtain an overview of the effects on the dynamics of RPA70A induced by the binding of d-CTTCA, the changes in the order parameters from the Model-Free analysis are mapped onto the structure of RPA70A in Figure 2B.
One rather unexpected effect associated with the binding of ssDNA is evident in the observation of exchange terms (Rex > 1 Hz) for 31 residues in the presence of ssDNA. These sites appear to be localized over a contiguous stretch of the protein, involving strand β3 (T236–N239) and the loop connecting helix α1 and strand β4 (L248–V254). A significant part of L45 (between residues S258 and T284), including G260–K263 in β4′, Y276–F280, and T284 in β5′, also shows slow conformational exchange on the microsecond–milliseconds timescale.
The observed exchange terms for residues in the binding site are not a trivial artifact of the ssDNA coming on and off the protein, but represent slower-timescale motions in the protein complex. The binding affinity of RPA70A for d-CTTCA is sufficiently high (Kd ∼ 5 μM) to ensure >99% saturation in the presence of a sufficient excess of ligand (A.I. Arunkumar and W.J. Chazin, unpubl.). Assuming the on-rate of ssDNA approaches the diffusion controlled limit of 108 M−1 sec−1, the upper limit for the off-rate is in the range of 103 sec−1, which is a reasonable approximation based on values for E. coli SSB of kon = 109 M−1 sec−1 and koff = 700 sec−1 (Kozlov and Lohman 2002). With an off-rate of ∼103 sec−1, the 1-msec refocusing delay used in the CPMG sequence in this study is too short to be significantly affected, thus the Rex terms must arise from a different process. The same conclusion can be drawn by considering that the magnitude of the Rex terms depends on the product of the populations and the square of the chemical shift difference. Given that the protein is saturated (papb |LL 0.0099), there would have to be an unreasonably large chemical shift difference between the bound and free proteins for an exchange term to be detected (Palmer et al. 1996; Evenäs et al. 1999). If the Rex terms did, indeed, arise from the ssDNA coming on and off the protein, a correlation between the magnitude of the binding-induced change in chemical shift and the presence of an exchange term would be expected, but this is not observed.
Direct contacts with the d-C8 oligonucleotide involving residues in L12, L45, β3′, β4′, and β5′ are found in the X-ray structure of the RPA70AB complex. These contacts are anticipated to result in a reduction in flexibility in the ssDNA-binding site of RPA70AB. L12, which is flexible in the absence of DNA, is found to undergo a substantial reduction in flexibility on the picosecond timescale. However, the response to DNA binding in L45 is more complex, as we detect changes in internal dynamics occurring on two different timescales and in opposite directions.
In the absence of ssDNA, the central region of loop L45 (residues I264–Y276) is characterized by a high degree of flexibility on the picosecond timescale, and the N terminus of this loop (S258–T261) shows excess slower (microsecond–millisecond) timescale motions reflected in Rex terms. In the presence of ssDNA, the excess fast-timescale motions of the central region of the loop are partly quenched, but the slower-timescale motions in the N-terminal region are not greatly affected. In contrast to the two loops, we find that strand β3′ is not affected on the faster timescales, but shows increased slow-timescale motions in the DNA-bound state. Increases in slower-timescale motions were also detected for residues in β4′, L12, and L45, which together constitute a significant portion of the DNA-binding site. Thus, in addition to quenching of fast-timescale motions, there is a clear trend toward increased microsecond-to-millisecond motions of RPA70A upon binding to ssDNA.
The changes in structural dynamics have important implications with respect to understanding the response of RPA70A to the binding of ssDNA. The majority of the observed effects appear to be localized to the direct binding surface, where a clamping action of the L12 loop appears to assist in holding the DNA in the cavity. The trend in the fast-timescale motions of the L45 loop is similar. Thus, the consensus view of the binding site is that it becomes less open and flexible upon binding DNA. Decreased motions upon binding of ssDNA for residues in the binding site were also observed for the OB-fold domain of bacteriophage Pf3, where a purportedly open binding site closes in on the DNA through the capping action of the DNA binding wing or the L12 loop (Horstink et al. 1999). Loss of flexibility in the binding site has also been reported for a variety of dsDNA-binding proteins (e.g., Baber et al. 2000; Cave et al. 2000).
The apparent increase in the slower-timescale motions upon binding of ssDNA for residues in the RPA70A binding site is much less intuitive. Nonetheless, similar observations were made for an OB-fold domain of topoisomerase I (Yu et al. 1996). These effects presumably correspond to a significant entropic contribution to the free energy of binding. However, before any definitive statements can be made about the importance of the ssDNA-binding-induced changes in dynamics at this timescale, experiments to directly characterize these millisecond–microsecond motions are required.
Structural flexibility is clearly a key element in the recognition of DNA by SSBs. The plasticity of the SSB binding scaffold is complemented by the flexibility of ssDNA; these factors presumably work together to facilitate the optimization of the interaction surfaces. Flexibility has also been invoked as a key element of recognition of dsDNA. For example, it has been proposed that dynamics in the binding site are essential for recognition of a range of promoter sequences by E. coli transcription factor MarA (Dangi et al. 2001). Thus, structural flexibility appears to be an important element of sequence-nonspecific recognition of both ssDNA and dsDNA.
Hypothesis to explain sequence-nonspecific binding of ssDNA by SSBs
The crystal structure of RPA70AB with d-C8 reveals interactions with the DNA through hydrogen bonds to both the bases and the phosphate backbone, as well as through stacking interactions of aromatic amino acids with the bases (Bochkarev et al. 1997). The stacking interactions as well as the hydrogen bonds to the phosphates had been predicted from previous biophysical studies. In contrast, the hydrogen bonds to the bases were entirely unexpected because RPA, like all SSBs, is a sequence-nonspecific DNA-binding protein.
How can base-specific hydrogen bonds (H-bonds) be rationalized with the functional necessity for binding ssDNA of any sequence? We propose that base-specific H-bonds contribute to the overall affinity of RPA to ssDNA, but they do not mediate a sequence-specific readout of sequence. Implicit in this assumption is that an array of RPA side chains at the binding interface enable similar H-bonds to form for the other DNA bases.
In our model, base-specific contacts are formed for all sequences because the binding surface is flexible in the absence of DNA and can be induced to fit whatever sequence is presented for binding: the ssDNA-binding interface dynamically adapts to the DNA sequence. Thus, one set of base-specific contacts for a given DNA sequence is replaced by a different set of base-specific contacts for another DNA sequence. This hypothesis is fully consistent with the conclusions of Bochkarev, Edwards, and coworkers based on their comparative analyses of the crystal structures of RPA70AB free and bound to d-C8 and with their proposal that their structure of RPA70AB captures a single complex from a family of isoenergetic structures (Bochkarev et al. 1997; Bochkareva et al. 2001). The concept of structural flexibility is strongly supported by the intrinsically high degree of flexibility of the ssDNA-binding sites of RPA70A (this study) and Pf3 (Horstink et al. 1999), as well as by the lack of order in the DNA-binding site of T4 gp32 in complex with ssDNA (Shamoo et al. 1995). Flexibility is viewed as essential for attaining a best fit to any given sequence; a dynamically remodeled binding site is not possible with a rigid preformed binding surface. Clearly, structural characterization of additional ssDNA complexes is needed to validate our hypothesis.
The tethering of four relatively weak binding domains in RPA results in the high overall affinity for ssDNA, while maintaining a significant off-rate within each binding module. Within this context, the entropic penalty associated with loss of flexibility upon binding may contribute in a positive sense to keeping the affinity of any single binding element low and in the biologically relevant micromolar range. Moreover, damping of motions of the ssDNA by binding of the protein will reduce the conformational entropy of the DNA, thereby facilitating processing in subsequent steps. The linkage of multiple weak interactions is being increasingly viewed as a highly effective mechanism for multistep processing in biology (Lohman et al. 1998). A dynamic binding scaffold is an attractive approach to the organization of multiprotein assemblies that participate in multistep genetic processes such as DNA replication, recombination, and repair.
Materials and methods
Uniformly 15N and 13C, 15N-enriched hRPA70182–291 (RPA70A) was expressed in E. coli and purified as described elsewhere (A.I. Arunkumar and W.J. Chazin, in prep.). The concentration of protein samples used for NMR ranged from 0.5 mM to 1.0 mM in an aqueous buffer (90% H2O/10% 2H2O) containing 25 mM Na2HPO4, 50 mM NaCl, and 5 mM d10-DTT with the pH adjusted to 7.5. The ssDNA complex was prepared by titrating a fourfold excess of DNA oligomer (d-CTTCA, d-C5) into a solution of free protein in the same buffer. This results in >99% saturation of the protein with ligand. The HPLC purified oligonucleotides were purchased from Midland Certified Reagent and used without further purification.
Standard 3D triple resonance experiments for backbone and side-chain assignments of RPA70A were acquired on Varian and Bruker spectrometers operating at 600 MHz and 800 MHz 1H frequencies. The transfer of these assignments from the DNA-free to the DNA-bound state was straightforward with a few exceptions. Of the 110 residues, the backbone resonances of 103 were unambiguously assigned in both the free protein and the d-CTTCA complex in the oxidized state of the protein. All relaxation experiments (R1, R2, and 1H–15N NOE) were performed at 298 K on Bruker AVANCE-500 and DRX-600 spectrometers using standard inverse detected pulse sequences (Kay et al. 1989; Skelton et al. 1993), modified to include a gradient-enhanced water suppression scheme (Sklenár et al. 1993). The 1H–15N NOE experiment was acquired with a 3-sec presaturation period preceded by a 2-sec relaxation delay. The recovery times in the R1 experiment were set to 10 (×2), 50, 60 (×2), 80, 100, 150 200, 250, 300, 400, 500, 600, 700 (×2), 800, and 1000 msec, and those for the R2 experiment to 4, 12, 20 (×2), 32, 40, 48, 60, 70, 80, 88, 100, 112, and 120 msec. Typical acquisition parameters were 160 (15N) by 2048 (1H) real points in the ω1 and ω2 dimensions, respectively, with 32 transients collected for each t1 increment and the phase-sensitive TPPI method for quadrature detection in the indirect dimension (Marion and Wüthrich 1983). The spectral widths for nitrogen and proton dimensions were 30 ppm and 14 ppm, respectively. The center of the 15N spectral width was set to 119 ppm, and the 1H carrier was placed on the water signal at 4.7 ppm from the respective base spectrometer frequencies. All spectra were processed with FELIX 97.0 (Molecular Simulations Inc.).
Analysis of relaxation data
The relaxation rates and NOEs were analyzed within the standard Model-Free formalism (Lipari and Szabo 1982a,b) extended to include both fast (τf) and slow (τs) internal motion (Clore et al. 1990), using Model-Free software (version 4.01; Palmer et al. 1991; Mandel et al. 1995). The anisotropic motional model involved the consideration of molecular asymmetry in the calculation of the relaxation rates (Woessner 1962). The program TENSOR 1.1 (Dosset et al. 2000) was used to perform a rigorous analysis of the R2/R1 ratios for residues in the rigid parts of the protein to obtain initial estimates of the parameters describing the axially symmetric diffusion tensor. The N—H vector orientation with respect to the diffusion frame of axis was calculated from the X-ray structure of the DNA bound protein. The reduced spectral density mapping was performed using the approach of Farrow et al. (1995).
Intensities of the individual peaks were measured for the various time points for each relaxation data set using FELIX macros. The relaxation rate constants, R1 and R2, were obtained by fitting the resultant time profiles of the intensities to a monoexponential decay function (Palmer et al. 1991; Skelton et al. 1993). The uncertainty associated with the measurement of peak heights was determined as the standard deviation of the difference in peak heights obtained for all the resolved peaks from duplicate spectra divided by 2½ (Palmer et al. 1991). The data were fit to a two-parameter model by using the curve-fit module of the Model-Free 4.01 software (Palmer et al. 1991; Mandel et al. 1995). The quality of the fit was judged by the 95% confidence limits of the χ2 goodness-of-fit test. The reported uncertainties in the relaxation rate constants are those obtained from the fit. The 1H–15N NOE values were calculated from the ratio of the cross-peak intensities measured with and without saturation of the amide protons.
The Model-Free approach for a monomeric system was extended to include the presence of monomer/dimer equilibrium in solution, and the data analysis followed a previous study (Fushman et al. 1997). For relaxation curves that decay monoexponentially, it is reasonable to assume that the measured relaxation parameters R1 and R2 represent a weighted average of the values for the monomer and dimer, respectively (Marshall 1970). The analysis is simplified further by assuming nonspecific dimerization, which eliminates the need for any special treatment of residues at the dimer interface. This also leads to a single set of values for the order parameter (S2) and exchange term (Rex), and the largest effect of the equilibrium is restricted to the correlation time. Within this framework the overall tumbling of the dimer is treated differently from that of the monomer. For an isotropic dimer the global correlation time is set to twice the value of the monomer tumbling time (Grzesiek et al. 1997). A second special case was also considered, where the dimer is modeled as a dumbbell composed of two monomer units that sample all possible orientations with respect to each other. The resultant effect on the N—H vector orientations with respect to the axially symmetric diffusion tensor (D∥/D⟂ = 1.8) is to average the angular component of the spectral density function (Fushman et al. 1997). The resultant spectral density function used to calculate the relaxation rates for the dimer is given by the following expression:
where, A1 = 1/5, A2 = 2/5, and A3 = 2/5. The relaxation times τiD for the dimer are expressed in terms of the monomer tumbling time τcM using Perrin's expressions for the frictional coefficient of a prolate ellipsoid with semiaxis ratio 2:1 (Fushman et al. 1997). These analyses were performed using the Model-Free software rewritten to incorporate the above changes.
Table Table 1.. Summary of fitting various models to the relaxation data
Degrees of freedom (f)
M2 (S2, τe)
M3 (S2, Rex)
M4 (S2, τe, Rex)
M5 (S2, Sf2, τs)
Anisotropic dimer refers to a dumbbell-shaped molecule whose long axis is twice the short axis. The degrees of freedom (f) are calculated from the difference between the total data points and the number of variables used for the model-free fit. χ2/f is the chi-squared per degrees of freedom. τC is the global correlation time. RMD is the monomer/dimer ratio. D∥/D⟂ is the ratio of the diffusion constants for an axially symmetric molecule. The five dynamic models were obtained by simplifying the expression for the spectral density function described by the extended Model-free approach (Clore et al. 1990). The average order parameter (S2) is calculated for the rigid parts of the protein backbone that do not participate in either fast internal motion or slow exchange.
0.92 ± 0.03
0.93 ± 0.03
Isotropic monomer/isotropic dimer
0.91 ± 0.03
Isotropic monomer/anisotropic dimer
0.90 ± 0.04
0.94 ± 0.03
0.94 ± 0.03
Isotropic monomer/isotropic dimer
0.90 ± 0.03
Isotropic monomer/anisotropic dimer
0.94 ± 0.03
This research was supported in part by the Vanderbilt Center for Molecular Toxicology (NIH ES00267), the Vanderbilt Ingram Cancer Center, the Canadian Institutes for Health Research (A.M.E., C.H.A.) and the National Cancer Institute of Canada (A.M.E., C.H.A.). We thank Markus Voehler for assistance in setting up NMR experiments, Arthur G. Palmer III for providing access to the Model-Free code, and one reviewer for exceptionally insightful suggestions.
The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.