|Home | About | Journals | Submit | Contact Us | Français|
DNA polymerase β (Pol β) is a 39 kDa enzyme that performs the vital cellular function of repairing damaged DNA. Mutations in Pol β have been linked to various cancers and these mutations further correlated with altered Pol β enzymatic activity. The fidelity of correct nucleotide incorporation into damaged DNA is essential for Pol β repair function and several studies have implicated conformational changes in Pol β as a determinant of this repair fidelity. In this work, the rate constants for domain motions in Pol β have been determined by solution NMR relaxation dispersion for the apo and substrate, binary forms of Pol β. In apo Pol β, molecular motions, primarily isolated to the DNA lyase domain, are observed to occur at 1400 s–1. Additional analysis suggests that these motions allow apo Pol β to sample a conformation similar to the gapped, DNA substrate bound form. Upon binding DNA, these lyase domain motions are significantly quenched whereas evidence for conformational motions in the polymerase domain become apparent. These NMR studies suggest an alteration in the dynamic landscape of Pol β due to substrate binding. Moreover, a number of the flexible residues identified in this work are also the location of residues, which upon mutation, lead to cancer phenotypes in vivo, which may be due to the intimate role of protein motions in Pol β fidelity.
DNA polymerase β (Pol β) is a 39 kDa, monomeric enzyme and member of the X-family of DNA polymerases. It functions in the base-excision repair (BER) pathway of damaged DNA in a template directed manner. Pol β fills in the nucleotide gaps in double-stranded DNA after DNA glycosylase removes the damaged base and AP endonuclease incises the DNA backbone. Pol β prefers short gapped or single base gapped DNA substrates.1In vivo, Pol β is confronted with the daunting task of repairing an estimated 20,000 DNA lesions per cell per day and must select the correct nucleotide from the cellular pool of dNTPs.2 In accord with its crucial role, mutations in Pol β have been implicated in prostate, cervical, gastric, and colon cancers.3-8 These mutants have altered DNA repair fidelity that is sequence and dNTP context dependent as well as having altered chemical steps and dNTP binding constants.9 Thus, nucleotide selectivity and the fidelity of nucleotide incorporation is an essential aspect of the enzymatic mechanism of Pol β.
In the search for the molecular determinants of dNTP incorporation fidelity, a number of biochemical studies have focused on identifying the rate-limiting step(s) in Pol β catalysis or on monitoring the rates of conformational changes. Nucleotide incorporation into gapped DNA follows an ordered reaction with DNA binding first, followed by Mg2+-dNTP binding. Upon binding the proper dNTP and formation of Watson-Crick pairing with the template strand of the DNA substrate, the phosphodiester bond is formed between the nucleotide and DNA, followed by subsequent release of pyrophosphate (PPi). It has been proposed that chemistry is rate limiting and therefore the step of nucleotidyl transfer could provide the ‘fidelity checkpoint’ for the overall reaction.10; 11 Other studies have suggested that conformational rearrangements (i.e. an ‘induced fit’ mechanism) in Pol β are also a mechanistic feature designed to ensure proper nucleotide selection and would thus contribute to the fidelity of this enzyme.12; 13 The conformational changes that occur during Pol β catalysis have mainly been resolved by X-ray crystallography (Scheme 1).12; 14; 15 The three-dimensional structure of Pol β consists of two domains, an 8 kDa amino terminal lyase domain and a 31 kDa polymerase domain. dRP lyase activity is contained within the 8 kDa domain whereas the polymerase domain possesses the catalytic residues for the nucleotidyl transferase function. Upon binding a gapped, double-stranded DNA substrate, the extended lyase domain closes around the DNA (Scheme 1A,B). Additional closure and conformational changes occur in both domains upon binding the dNTP (Scheme 1C,D). With these structures as a basis, numerous transient kinetics experiments have sought to characterize the timescale of these conformational changes.10; 16 These experiments have relied primarily upon intrinsic Trp fluorescence or that of an unnatural fluorophore (2-aminopurine), to relate changes in the DNA environment to protein conformational changes, and to subsequently inform on the important structural changes that occur along the enzymatic reaction coordinate.17; 18
Solution NMR methods are well-suited for monitoring conformational changes with atomic resolution, and do not require addition of bulky or potentially perturbing labels. A recent solution NMR study of full length Pol β investigated the sidechain dynamics of 13C-labeled methionine residues and allowed for identification and characterization of some aspects of ligand and metal-induced conformational activation.19; 20 We were prompted by those studies and by recently published backbone assignments of a Pol β/DNA complex21 to further investigate how motions in Pol β are altered at early stages along the enzyme reaction coordinate. Our results indicate extensive millisecond (ms) motions in apo Pol β, and that these motions are largely quenched in the lyase domain and enhanced in the polymerase domain upon formation of the binary Pol β/DNA/Mg2+ complex.
The 39 kDa monomeric Pol β was studied by solution NMR spectroscopy. Using TROSY-based triple resonance experiments, 86% of Cα, and Cβ, and 82% of NH for non-proline residues were assigned in the apo complex, which corresponds to approximately 95% of the observable resonances in a 1H-15N correlation spectrum. The two-dimensional 1H-15N TROSY spectrum22 for the apo enzyme is shown in Figure 1. Backbone assignments were published previously for the isolated lyase subdomain and a portion of the polymerase domain23; 24 of the apo enzyme as well as for full-length Pol β bound to a 22-mer gapped DNA substrate.21 Many of the assignments presented here for the full-length apo protein are similar to those of the isolated domains, with significant differences near the domain boundaries.
Two magnesium ions (Mg2+) are required for catalytic activity of Pol β. Addition of millimolar quantities of MgCl2 to our NMR samples did not result in changes in the TROSY spectra. This result was consistent with observations from the 13Cε-Methionine NMR experiments.20 In contrast, formation of the binary complex by addition of a double-stranded, single nucleotide gapped DNA substrate (see Methods) resulted in numerous changes in the TROSY spectrum (SI Figure 1). The response of the amide resonances to binding of the DNA substrate was monitored by NMR titration. Most resonances that shift upon substrate binding appear to be in slow to intermediate chemical exchange (SI Figure 2). In agreement with the nanomolar affinity determined for this substrate,25 saturation of the NMR sample occurred at approximately equimolar concentrations of Pol β and substrate (SI Figure 2). The chemical shift changes observed upon substrate binding are shown in Figure 2. Figure 2C shows the composite 1H/15N chemical shift changes after saturation by substrate. Residues with chemical shift changes greater than 1.5 standard deviations from the 10% trimmed mean (0.17 ± 0.07 ppm) are indicated above the horizontal line in Figure 2C and are plotted onto the Pol β/DNA structure in Figure 2D,E. In total there are 27 residues that demonstrate a significant response to DNA binding (Q8, E21, L22, N24, V29, S30, K60, G64, I69, F76, G80, E86, N98, L100, T121, D130, H134, G139, F146, N164, E165, T176, D192, F223, V238, L259, and G268).
The TROSY spectrum of Pol β bound to our DNA substrate is similar to the spectrum of Pol β bound to a 22-mer gapped substrate.21 Backbone assignments for substrate-bound Pol β were determined by comparison to the published data, with ambiguities resolved using TROSY-based triple resonance experiments. Six residues are assigned in the apo enzyme but unassigned in the binary complex, presumably due to exchange broadening in the presence of DNA (T10, G14, F25, G105, K141, and K234). In addition, 12 residues are assigned in the binary complex but not in the apo enzyme (D17, K41, G66, R89, T104, R137, Q159, S202, L211, C239, V303, and I323), again likely due to exchange broadening.21
15N transverse relaxation rate constants (R2,avg) were measured using the relaxation-compensated Carr-Purcell-Meiboom-Gill dispersion experiment.26; 27 This experiment measures the average relaxation rate constant (R2,avg) of 15N in-phase and anti-phase coherence. At the shortest τcp time (0.625 ms), the overall mean R2,avg values are 16.4 ± 5.8 s– and 20.2 ± 7.1 s– for the apo and binary Pol β enzymes. The increase in R2,avg observed for the binary complex is consistent with a mass increase of 9.6 kDa upon complex formation. Mean R2,avg values were also compared for the individual domains of the enzyme because in the apo enzyme the lyase domain is somewhat isolated from the main body of the protein (Scheme 1) and could potentially have a different rotational correlation time (and therefore a different transverse relaxation rate constant) than the rest of the enzyme. For apo Pol β, R2,avg = 15.5 ± 8.1 s–1 and 17.2 ± 5.9 s–1 for the 8 kDa lyase domain (res. 1 – 90) and 31 kDa DNA binding domain (res. 91 – 335) respectively, and therefore do not appear to be significantly different from each other. However, the distribution of R2,avg for the lyase domain is skewed with a significant number of residues with R2,avg lower than expected (SI Figure 3). This suggests that some independent rotational diffusion of the lyase domain may occur in the apo enzyme. In the binary complex the lyase and polymerase domains have similar R2,avg values (21.4 ± 10.2 s- and 19.9 ± 6.5 s–, respectively) reflecting the more compact structure in the presence of DNA substrate.
A comparison of R2,avg at τcp = 0.625 ms (R2(1/τcp) versus amino acid sequence shows a rather uniform range of values across the entire protein with several exceptions of elevated transverse relaxation rates noted in Figure 3. For both the apo and binary enzymes a significant number of residues have R2,avg values higher than the mean protein value. These elevated transverse relaxation rates suggest that conformational exchange motions may exist in both the apo enzyme and the binary complex. The details of the elevated transverse relaxation rate constants were explored in more detail by TROSY-CPMG relaxation dispersion experiments, at 600 and 800 MHz on both forms of the enzyme.
TROSY-based CPMG relaxation dispersion experiments27 were performed at 296 K on apo Pol β at both 14.1 and 18.8 T. Overall, 30 resonances show dispersion profiles characteristic of millisecond (ms) motions (Figure 4). Of these, 8 pairs of resonances (I33,I73; I38,L22; I174,Y271; F275,K141; A70,W325; L19,Y322, A23,I277; and R83,D145) were overlapped in the 1H-15N two-dimensional spectrum and therefore the observed motions cannot be assigned to a single residue. Of the resolved resonances, 12 (E21, Y36, A43, I46, D74, F76, A78, E86, K87, F99, E123, D130) dispersion curves were of sufficient quality for fitting only at 800 MHz. For the remaining residues (K27, N28, A47, E75, R102, D116, L122, R126, H135, and F146), relaxation dispersion data at both static fields was further analyzed (Figure 4 and SI Figure 4). The kinetics of the observed motions varied considerably across the enzyme sequence, with kex values ranging from 600 s–1 to > 4000 s–1. Four residues (N28, A47, H135, and F146) had kex values ≥ 4000 s–1 and could not be adequately quantitated using the CPMG dispersion experiment in the absence of other data.28 Thus, for these residues the kex value represents a lower limit to the kinetics of the exchange process. Individual fits to the dispersion data for the remaining residues with kex < 4000 s–1 were all within a factor of two of each other, suggesting that a global model in which kex is shared by all remaining residues may be a more appropriate description of the data. In the individual fitting procedure, each residue is assumed to have its own kex and Rex values. In the global model all residues share a single kex value while Rex remains a residue-specific value. The individual and global models were compared and the P-value from F-tests is 0.77 thus favoring the mathematically simpler, global model. This conclusion is also supported by AIC differences of 44 when comparing the global versus individual model. Therefore, the data for these 18 residues included in the global fitting support a single conformational exchange process with an exchange rate constant equal to 1410 ± 130 s–1 (Figure 4 and SI Figure 4). For the four residues with kex values exceeding 4000 s–1, notably, H135 and F146 are in close proximity to each other in the structure of Pol β and are somewhat isolated from the other residues undergoing ms motions (Figure 5). However, N28 and A47 are spatially separate residing on the ends of α-A and α-B, respectively.
For the binary Pol β enzyme (containing 5 mM Mg2+ and the single-gapped DNA substrate), significantly fewer residues exhibit dispersion profiles. Only E21 shows dispersion (Figure 6). Analysis of the dispersion data for E21 in the binary enzyme gives kex = 400 ± 200 s–1 and Rex = 16.9 ± 7.8 s–1, compared to values of 1410 ± 130 s–1 and 3.3 ± 1.5 s–1 for the apo enzyme respectively. Of the residues that possessed ms motions in the apo enzyme based on the CPMG dispersion experiment, only E21, N28, and D130 have elevated R2(1/τcp) values in the binary enzyme. In addition, a number of other residues in the polymerase domain have elevated R2(1/τcp) values in the binary complex (Figure 3B) but do not show dispersion in the CPMG experiment (SI Figure 5). This suggests that although conformational motions appear to be present in the substrate-bound enzyme, the kinetics of the motions reside in a time regime to which the CPMG dispersion experiment is insensitive. Several flexible residues in apo lyase domain (T10, G14, F25) were unassigned in the DNA bound enzyme, therefore whether they remain flexible in the bound enzyme is not known. However, other lyase residues including Y36, A43, I46, A47, D74, E75, and F76 are assigned in the DNA bound form and do not show relaxation dispersion profiles nor elevated R2 values suggesting the motions in the lyase domain is reduced in the binary DNA complex.
There have been several NMR dynamics investigations of protein-DNA interactions. These studies, like this one, observed conformational fluctuations in the apo enzyme or DNA binding protein that were suggestive of inherent flexibility that was important for interacting with the DNA. In DNA binding protein PBX homeodomain the C-terminal extension transiently forms, in the absence of DNA, an a-helical segment,29 whereas cAMP binding to CAP facilitates reorientation of the DNA binding domains in this protein. 30; 31 Similar phenomena were also observed in the DNA repair enzyme, Uracil DNA Glycosylase in which NMR studies suggested the dynamic sampling of alternate enzyme conformations.32 More recently, backbone flexibility in DNA ligase D was suggested to be involved in both activation and inhibition of this enzyme. 33
The apo Pol β enzyme exhibits conformational exchange motions at 22 resolved amide positions. These sites are located primarily in the dRP lyase domain and at the base of the DNA binding site in the polymerase domain (Figure 5). Many of these residues directly contact or are in close proximity to the substrate in the binary complex. All but four of these residues (N28, A47, H135 and F146) have the same kex value (1400 s–1). These four residues have R2 values that vary with τcp in an almost linear fashion indicating significantly elevated exchange rate constants (SI Figure 4). Interestingly, H135 and F146 are somewhat isolated structurally from those with kex = 1400 s–1. In contrast, A47 and N28 are both located at opposite ends of the lyase domain and near residues with kex = 1400 s–1. Therefore it is not clear why the CPMG dispersion data for these two resonances suggest greatly elevated exchange rate constants.
For the 18 residues with kex = 1400 s–1, most appear to be in the fast exchange limit, precluding estimation of equilibrium site populations for this exchange process. For the residues that may be in intermediate to slow exchange, the dispersion data is not of sufficient quality to allow for extraction of populations and Δω values from non-linear least squares fitting. However, for groups of residues moving in a concerted fashion, the Rex values obtained from dispersion fitting should be linearly related to the 15N chemical shift difference between apo and substrate-bound Pol β (Δδ2) 34 if the apo enzyme is in equilibrium with a conformation similar to that when Pol β is bound to DNA. These data are shown in Figure 7. This analysis clearly identifies two clusters of residues. The slope of the line (papbΔω2/kexΔδ2) for the first cluster of residues (shown in blue) is 3.55 × 10–5. Assuming that Δω = Δδ, this slope is equivalent to papb/kex, allowing estimation of the equilibrium populations, which are 95%:5% for the major and minor conformations. This further suggests that in the apo enzyme the closing rate constant for dRP lyase domain is 70 s–1. This value is similar to prior estimates of the rate constant for domain closure upon nucleotide binding,35 which based on existing models for Pol β function should be distinct from the motion observed by our NMR experiments since dNTP is not included in our samples. This correlation, which suggests the apo enzyme samples the DNA substrate-bound form argues against an induced-fit mechanism, at least with respect to DNA binding. The data points shown in red in Figure 7 (E21, N28, D74, F76, and E86) comprise the second cluster and the equilibrium populations determined from the slope of the fitted line are unrealistic (99.7%:0.3%). It is unlikely that such a highly skewed population distribution would show upward curving relaxation dispersion data. This observation could be due to additional ring current effects from the DNA that would alter Δδ but not Δω thus invalidating the assumption that Δω = Δδ. Alternatively, these residues could be part of a separate motional process that coincidentally also has an exchange rate constant equal to 1400 s–1.
In the binary enzyme form only E21 shows a measurable CPMG relaxation dispersion profile with kex = 400 s–1 though there is significant uncertainty in this value. Nonetheless this data suggests a similar time scale for motion as in the apo enzyme. Other residues with conformational exchange motions in the binary enzyme (Figure 3) are largely located in the polymerase domain (A70, D130, G139, Q169, S204, L211, H212, S229, Y296, and T297). This suggests that DNA binding reduces ms motions in the lyase domain relative to the apo enzyme but enhances motions in the polymerase domain (Figure 8). For residues such as G139, S229, Y296, and T297, which are very close to the DNA in the bound form we cannot determine whether the observed flexibility is only present in the bound form or that DNA ring-current effects unmask motions through an increase in the Δω value. Resolution of this issue will require further characterization. For the remainder of residues distant from the DNA, ring current effects are likely to not be as significant. The motions in the polymerase domain detected only in the binary complex may represent the flexibility in this region necessary for dNTP binding. DNA binding appears to alter the Pol β energy landscape and increase flexibility in additional regions near the dNTP binding site. Whether this increase in motions is necessary to proper selection of dNTP requires additional investigation.
Strikingly, a significant number of the flexible residues identified in these NMR studies correspond to mutations that have been identified in colon cancer tumors and are known to alter Pol β enzymatic activity.36 Mutations of 10 residues in Pol β (N24, G80, D116, R126, H135, G139, E145, S204, S229, and S275) were identified in colorectal tumors and enzymes bearing these mutations were found to be faulty in the DNA BER pathway. Five of these residues show evidence for motions in the binary complex, whereas the remaining five have ms motions in the apo enzyme only. In addition, five other cancer-associated mutants (D74, T79, T121, H134, and E147) reside adjacent to the flexible residues identified in this NMR work. While the data presented here do not demonstrate a causal link between changes in Pol β flexibility and cell transformation, these data suggest that mutation residues that give rise to aberrant Pol β function reside at sites that are undergoing ms motions. Furthermore, the millisecond motions of these cancer-associated residues are ligand dependent, suggesting an important role for ms motions in this enzyme's function.
DNA oligonucleotides were purchased from Integrated DNA Technologies, Inc. (Coralville, IA). The DNA sequence was designed as described previously. 12 The DNA sequences corresponding to the template, primer, and downstream oligonucleotide of the gap are 5‘-CCGACGGCGCATCAGC-3‘, 5‘-GCTGATGCGC-3‘, and 5‘-pGTCGG-3‘. The downstream oligonucleotide was 5‘-phosphorylated. The template, primer, and phosphorylated downstream oligonucleotides (1:1:1) were mixed in 50 mM HEPES, pH 7.4, and 100 mM KCl. The mixture was incubated sequentially at 95°C for 5 min, slowly cooled to 25°C for 2 hours, and immediately transferred to ice. The quality and extent of the annealing was assessed by an 18% native polyacrylamide gel, followed by methylene blue staining.
The tagless DNA construct of rat Pol β was transformed into BL21 DE3 cells. Protein was expressed in the presence of 1 mM IPTG at 20°C overnight. The cells were resuspended in Buffer A (50 mM HEPES, pH 7.4, 100 mM NaCl, 1mM EDTA, 2mM DTT) in the presence of Roche EDTA Free Protease Inhibitor Cocktail and the sample was sonicated for 5 × 30 seconds on ice, followed by centrifugation at 10,000 RPM at 4°C. The supernatant was filtered through 0.45 micron filter before being loaded onto a HiTrap Heparin column (GE) and separated by an NaCl gradient [Buffer A: 50 mM HEPES pH 7.4, 100 mM NaCl, 1mM EDTA, 2mM DTT; Buffer B: 50 mM HEPES pH 7.4, 2 M NaCl, 1mM EDTA, 2mM DTT] using a fast protein liquid chromatography system (Äkta, GE Healthcare). Fractions containing pure Pol β were identified by electrophoresis on a 10% SDS PAGE gel followed by Coomassie staining, and were pooled, then diluted to 1:10 with Buffer C [50 mM HEPES pH 7.4, 75 mM NaCl, 1mM EDTA, 2mM DTT]. The protein sample was loaded onto a SP Sepharose column (GE), and again separated by a NaCl gradient [Buffer C: 50 mM HEPES pH 7.4, 75 mM NaCl, 1mM EDTA, 2mM DTT; Buffer B: 50 mM HEPES pH 7.4, 2 M NaCl, 1mM EDTA, 2mM DTT]. Fractions containing pure Pol β were identified by 10% SDS PAGE followed by Coomassie blue staining, and were pooled, then concentrated to <1 mL. Finally, the protein was exchanged into buffer D [50 mM HEPES pH 7.4, 100mM KCl, 2mM DTT, 0.02% sodium azide]. The protein was greater than 95% pure as determined by SDS PAGE and Coomassie blue staining. The concentration of the protein was calculated on the basis of an extinction coefficient of 21200 M-1cm-1 and a molecular weight of 39 kDa. Roche Protease inhibitor cocktail (1:1000 dilution) was added in final NMR sample to prevent degradation.
Pol β samples for NMR assignments were 2H, 13C, 15N-labeled, whereas samples used for relaxation dispersion experiments were 2H,15N-labeled. The Pol β– DNA binary complex was formed by addition of the pre-formed gapped DNA substrate shown below: 5’GCTGATGCGC GTCGG 3’ 3’CGACTACGCGGCAGCC 5’ to samples of isotopically labeled enzyme. Binding of the gapped DNA was monitored by NMR titration by following resonance shifts upon DNA binding. All enzyme and DNA samples were prepared in buffer containing 50 mM HEPES pH 7.4, 100 mM KCl, 2 mM DTT, 0.02% sodium azide and 7.5% D2O. The Pol β concentration in the NMR samples was kept below 0.6 mM to avoid aggregation. In addition, prior to formation of the Pol β-DNA binary complex, 5 mM MgCl2 (also prepared in the NMR buffer) was added to the NMR sample. During addition of both MgCl2 and DNA, the pH of the NMR sample was monitored after each addition and adjusted if necessary to ensure that all observed changes in the NMR spectra were due to DNA binding. The double-stranded, single-gapped DNA substrate was added in 10 μL increments from a 6 mM stock solution. DNA was added until no further change in NMR resonances was observed. Based on this criteria, saturation occurred at a DNA:Pol β = 1:1.
All NMR experiments were performed at 23 °C on Varian Inova spectrometers operating at 600 or 800 MHz at Yale University or at 900 MHz located in the Department of Chemistry and Biochemistry at the University of Colorado-Boulder. The sample temperature was calibrated with a 100% methanol standard prior to each experiment. Backbone assignments of apo Pol β were determined using data obtained from TROSY versions of the HN(CA)CB, HN(COCA)CB, and HNCA experiments. Backbone assignments of the Pol β binary complex were obtained by comparison to published spectra in combination with triple resonance experiments.21
15N TROSY-CPMG experiments26; 27 were collected with spectral widths of 2500 × 12000 Hz and 256 × 2048 points in the t1 and t2 dimensions. The 1H and 15N carrier frequencies were set to the water resonance and 120 ppm respectively. The recycle delay for all relaxation experiments was 2.5 s. The CPMG experiments were acquired with a constant relaxation time37 of 40 ms with τcp delays of 0, 0.625 (x 2), 0.714, 1, 1.25 (x 2), 1.67, 2, 2.5, 3.33, 5 (x 2), and 10 ms. All NMR relaxation data were processed using NMRPipe38and analyzed using Sparky39 in conjunction with in-house written programs for dispersion fitting. Peak heights were quantified in Sparky using the average of nine points from a 3 × 3 grid centered on the peak maximum. Relaxation rates were determined from peak intensities using in-house written programs. Uncertainties in rates were determined from duplicate measurements and the Jackknife procedure.40 Results are reported only for resonances that are not overlapped in the two-dimensional spectra and that have sufficient signal to noise such that reliable quantitation of peak intensities is possible. CPMG relaxation data were analyzed with
in which Rex = papbΔω2/kex. The equilibrium populations are given by pa(pb), kex is the sum of the forward and reverse rate constants for a two–site conformational equilibrium and Δω is the 15N chemical shift difference between the two conformations, a and b.
RBB acknowledges support from NIH Biophysical training grant (T32GM008283). J.P.L. acknowledges financial support from the National Institutes of Health (R01 GM099990). J.B.S acknowledges financial support from the National Institutes of Health (R01 CA080830).
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.
M.S. and R.B.B. contributed equally to this work
Resonance assignments for apo Pol β were deposited in the BioMagResBank under accession numbers 18267.