Conformational Changes during Nucleotide Selection by Sulfolobus solfataricus DNA Polymerase Dpo4*

The mechanism of nucleotide selection by Y-family DNA polymerases has been the subject of intense study, but significant structural contacts and/or conformational changes that relate to polymerase fidelity have been difficult to identify. Here we report on the conformational dynamics of a model Y-family polymerase Dpo4 from Sulfolobus solfataricus. Hydrogen-deuterium exchange in tandem with mass spectrometry was used to monitor changes in Dpo4 structure as a function of time and the presence or absence of specific substrates and ligands. Analysis of the data revealed previously unrecognized structural changes that accompany steps in the catalytic cycle leading up to phosphoryl transfer. For example, the solvent accessibility of the αB-loop-αC region in the finger domain decreased in the presence of all four dNTP insertion events, but the rate of deuterium exchange, an indicator of conformational flexibility, only decreased during an accurate insertion event. Of particular note is a change in the region surrounding the H-helix of the thumb domain. Upon binding DNA and Mg2+, the H-helix showed a decrease in solvent accessibility and flexibility that was relaxed only upon addition of dCTP, which forms a Watson-Crick base pair with template dG and not during mispairing events. The current study expands upon a previous report from our group that used a fluorescent probe located near the thumb domain to measure the kinetic properties of Dpo4 conformational changes. We now present a model for nucleotide selection by Dpo4 that arises from a synthesis of both structural and kinetic data.

The mechanism of nucleotide selection by Y-family DNA polymerases has been the subject of intense study, but significant structural contacts and/or conformational changes that relate to polymerase fidelity have been difficult to identify. Here we report on the conformational dynamics of a model Y-family polymerase Dpo4 from Sulfolobus solfataricus. Hydrogen-deuterium exchange in tandem with mass spectrometry was used to monitor changes in Dpo4 structure as a function of time and the presence or absence of specific substrates and ligands. Analysis of the data revealed previously unrecognized structural changes that accompany steps in the catalytic cycle leading up to phosphoryl transfer. For example, the solvent accessibility of the ␣B-loop-␣C region in the finger domain decreased in the presence of all four dNTP insertion events, but the rate of deuterium exchange, an indicator of conformational flexibility, only decreased during an accurate insertion event. Of particular note is a change in the region surrounding the H-helix of the thumb domain. Upon binding DNA and Mg 2؉ , the H-helix showed a decrease in solvent accessibility and flexibility that was relaxed only upon addition of dCTP, which forms a Watson-Crick base pair with template dG and not during mispairing events. The current study expands upon a previous report from our group that used a fluorescent probe located near the thumb domain to measure the kinetic properties of Dpo4 conformational changes. We now present a model for nucleotide selection by Dpo4 that arises from a synthesis of both structural and kinetic data.
The mechanisms utilized by DNA polymerases to catalyze replication and/or repair of genomic material provide a fascinating example of how an enzyme can select a single substrate from multiple candidates, all with similar structural and chemical properties. DNA polymerases inside the cells of every living organism select from a pool of (four) dNTP substrates to catalyze phosphoryl transfer and then extend from a nascent primer strand opposite a DNA template. It has been emphasized that the endpoint of each nucleotide selection event cannot be determined solely by the thermodynamics of a given base pair (1,2). Instead, the molecular features intrinsic to DNA polymerases are generally thought to guide the free-energy landscape of phosphoryl transfer toward the selection of "Watson-Crick" pairs (3,4), at least among the four canonical bases. It is now apparent that individual DNA polymerases use variations upon a general mechanism to determine nucleotide selectivity (4 -8). However, the mechanistic differences that define each polymerase class are only now being elucidated in any molecular detail.
The Y-family DNA polymerases (pols) 3 represent one class of polymerases with distinctive structural and functional features (8,9). Like most other DNA polymerases, the Y-family pols have three domains: the finger for dNTP selection, the palm for catalysis, and the thumb for double strand DNA contact/orientation, which together form a structure that has been likened to a right hand. The Y-family polymerases also possess a unique domain that has been called the little finger or palm-associated domain. Most of the additional domains beyond these four core domains are involved in a complex web of protein-protein interactions and cellular localization events, although there are clearly exceptions to this rule (e.g. the N-clasp of pol and the N-digit of REV1) (10,11). The Sulfolobus solfataricus DNA polymerase Dpo4 has served as the prototypical Y-family polymerase, because it is especially amenable to structural analysis and it shares many features and properties of other Y-family members. Numerous Dpo4 crystal structures have been reported in the literature (12)(13)(14)(15)(16)(17)(18). However, structural alignments of binary and ternary structures have failed to reveal the "open" to "closed" transition observed for some other pols (19). It is clear that, upon binding DNA, the little finger of Dpo4 probably undergoes a dramatic translation/rotation through space (20,21), but few other conformational changes have been observed. The lack of obvious conformational rearrangements has led to the proposal that the Y-family DNA polymerases possess a "pre-formed" active site (22). However, there is substantial kinetic evidence in support of the view that a non-covalent step or steps occur in the Dpo4 reaction cycle prior to nucleophilic attack upon (what is assumed to be) a deprotonated 3Ј-hydroxyl group at the primer terminus, at least during formation of Watson-Crick geometry (23,24). Previous work from our group used tryptophan fluorescence to monitor changes during dNTP insertion and following phosphoryl transfer (25). In the present study we sought to identify structural changes that occur during the Dpo4 reaction cycle by using hydrogen-deuterium exchange in tandem with mass spectrometry (HDX-MS). The method of HDX-MS has been used successfully to probe conformational dynamics of proteins and enzymes (26 -29). The temporal and structural resolution provided by HDX-MS can serve as an effective bridge between kinetic analysis, which provides evidence of changes but cannot identify locations, and the inherently static nature of crystal structures. Our results provide new insight into Y-family polymerase catalysis, which may lead to a better understanding of how these enzymes select nucleotide substrates and, ultimately, how they contribute to mutagenesis.

EXPERIMENTAL PROCEDURES
Materials-Wild-type Dpo4 was expressed in Escherichia coli and purified to electrophoretic homogeneity as described previously (17). Following storage in Tris-HCl (pH 7.4) containing 1 mM DTT, 1 mM EDTA, and 50% glycerol (v/v), the enzyme was dialyzed against 500 ml of 50 mM potassium phosphate buffer (pH 7.4) containing 50 mM NaCl, 1 mM DTT, 1 mM EDTA, and 50% glycerol (v/v) overnight. The concentration of the enzyme following dialysis was estimated by measuring the absorbance at 280 nm (⑀ 280 ϭ 24,100 M Ϫ1 cm Ϫ1 ). The dialyzed protein was filtered through a 0.2-m syringe tip filter and aliquoted into 100-l samples (160 M final concentration), flashfrozen in dry ice/ethanol, and stored at Ϫ80°C. Samples were thawed and used for HDX experiments on the same day. Pepsin was purchased from Worthington Biochemicals (Lakewood, NJ) and dissolved in 10 mM ammonium acetate (pH 5.0) just prior to performing the hydrogen-deuterium exchange reactions. All unlabeled dNTPs were obtained from Amersham Biosciences. All oligonucleotides used in this work were synthesized by Midland Certified Reagent Co. (Midland, TX) and purified by the manufacturer using high-performance liquid chromatography, with analysis by matrix-assisted laser desorption time-of-flight MS. The 13-base primer sequence used in the analyses was 5Ј-GGGGGAAGGATTC-3Ј. A dideoxy-terminated 13-mer primer of identical sequence was used in all experiments that included a dNTP. The template DNA sequence was 5Ј-TCACGGAATCCTTCCCCC-3Ј. All of the MgCl 2 , dNTP, and DNA stock solutions used in the HDX analysis were lyophilized to dryness and resuspended in D 2 O (100%, Acros Organics, Geel, Belgium).
Identification of Dpo4 Peptides-Dpo4 was diluted 10-fold in deionized H 2 O prior to addition of potassium phosphate (pH 2.3, to a final concentration of 0.1 M), and then pepsin (10 molar equivalents) was added. Dpo4 was digested with pepsin for 15 min on ice prior to LC injection. LC-MS/MS analysis was performed on a Waters Acquity UPLC system (Waters, Milford, MA) connected to a Finnigan LTQ mass spectrometer (Thermo Fisher Scientific, Waltham, MA), operating in the ESI positive ion mode. An Acquity UPLC BEH octadecylsilane (C 18 ) column (1.7 m, 1.0 mm ϫ 100 mm) was kept at 30°C with the following LC conditions: buffer A contained 98% H 2 O/2% CH 3 CN/0.01% CF 3 CO 2 H (v/v), and buffer B contained 2% H 2 O/98% CH 3 CN/0.01% CF 3 CO 2 H (v/v). For peptide sequencing the following gradient program was used with a flow rate of 50 l min Ϫ1 : 0 -10 min, linear gradient from 100% A to 95%A/5% B (v/v); 10 -20 min, linear gradient to 85% A/15% B (v/v); 20 -60 min, linear gradient to 47%A/53% B; 60 -64 min, linear gradient to 5%A/95% B; 64 -68 min, hold at 5% A/95% B; 68 -70 min, linear gradient to 100% A. ESI conditions were as follow: source voltage, 4 kV; source current, 100 A; auxiliary gas flow rate setting, 20; sweep gas flow rate setting, 5; sheath gas flow setting, 34; capillary voltage, Ϫ49 V; capillary temperature, 350°C; and tube lens voltage, Ϫ90 V. MS/MS conditions were as follows: normalized collision energy, 35%; activation Q, 0.250; and activation time, 30 ms. Data-dependent spectral acquisition with dynamic exclusion and charge-state monitoring was used to obtain fragmentation patterns for peptides as they eluted from the column.
Hydrogen-Deuterium Exchange Mass Spectrometry-HDX experiments were performed by diluting 2.5 l of Dpo4 (160 M) into 22.5 l of deuterated solution. For experiments with DNA, MgCl 2 , and/or dNTPs, the ligands were lyophilized to dryness and resuspended in D 2 O. All experiments were performed at 25°C. After the indicated time the exchange was stopped by addition of 25 l of 0.1 M potassium phosphate buffer (pH 2.3), and the reaction tube was placed on ice. Next, 10 molar equivalents of pepsin was added to the quenched solution and allowed to digest for 15 min on ice prior to injection. A zero time control (m 0% ) was performed to assess the amount of deuterium incorporated after acidification and during the pepsin digestion. The zero time control was prepared by adding 2.5 l of Dpo4 (160 M) to 25 l of 0.1 M potassium phosphate (pH 2.3) on ice prior to adding 22.5 l of deuterated solution. The amount of deuterium incorporated during each time course was corrected for deuterium loss during digestion and UPLC separation (i.e. "back-exchange") by using a 6-h time point (m 100% ) as a reference for each experimental condition, which should take into account experimental variability from one run to the next. LC-MS/MS analysis was performed as described above with the following distinctions. For HDX experiments the autosampler was cooled to 5°C and the injector bypassed so that the UPLC column could be packed in ice. Solvents were kept on ice during the analysis. The following gradient program was used with a flow rate of 50 l min Ϫ1 : 0 -3 min, linear gradient from 100% buffer A (see above) to 97%A/3% B (v/v); 3-4.5 min, linear gradient to 80% A/20% B (v/v); 4 -5.5 min, linear gradient to 100% B; 5-5.5 min, hold at 100% B; 5.5-6.5 min, linear gradient to 100% A; 6.5-9.5 min, hold at 100% A. The column was placed in ice for the duration of the experiment. ESI conditions were as described above.
Quantification of Deuterium Incorporation-Data sets were analyzed in Matlab (Mathworks) using a modified version of the DoGEX software developed by our group (30). Peaks defining the isotopic envelope in a given m/z and time range were detected. A weighted average was used to deconvolute the peptide envelope (supplemental Fig. S1). The deconvolution was calculated by summing all the isotopic intensities in the envelope and multiplying them by their respective m/z values. The m/z intensity sum was divided by the sum of all the intensities.
Each deconvoluted m/z value corresponded to a different degree of deuterium incorporation corresponding to a particular time in the reaction for a given peptide. The deuterium content of each peptide was adjusted for gain or loss during the experiment using Equation 1, where D is the deuterium content in a given peptide and m t , m 0% , and m 100% are the weighted average m/z values calculated for a peptide at a given time (m t ), for the exchange-in experiment (m 0% ), and for a maximally deuterated sample (m 100% ), respectively. The total number of amide protons (N) depends upon the peptide sequence (N ϭ [length of peptide Ϫ (N terminus) Ϫ (any prolines)]). The number of deuterons incorporated was plotted as a function of time, and individual data sets were fit using GraphPad Prism (GraphPad Software, La Jolla, CA) to the sum of first-order rate terms (Equation 2).
The structural representations were prepared using Pymol (44).

Peptide Identification and Rapid Exchange Maps-Dpo4
was initially incubated with pepsin and H 2 O to identify the peptides generated during the digestion step. A long chromatographic step was coupled with data-dependent fragmentation of the resulting peptides. The resulting MS/MS spectra were submitted to the Mascot server (Matrix Science) to identify Dpo4 peptides based on their fragmentation patterns. The resulting map gave 96% coverage of the 352 amino acids that comprise the Dpo4 polypeptide (supplemental Table S1). However, during the HDX analysis a shorter chromatographic gradient was used to minimize loss of deuterium prior to MS analysis. The shorter gradient effectively reduced the working coverage to ϳ69% of the Dpo4 sequence (Fig. 1A). The working peptide map consisted of 28 ions and covered all four domains of the polymerase (supplemental Table S2).
Overview of Changes in HDX as a Function of Primer/Template DNA, MgCl 2 , and Incoming dNTP-For any peptide of length N analyzed by MS, the amide hydrogens in that peptide will exchange with variable rates that are primarily dependent upon hydrogen bond participation. Changes in the measured rates of HDX for a peptide may represent changes in either the stability/identity of hydrogen bonding patterns, which may include spatial rearrangements. Either of these events is indicative of changes in the structure of the protein. A qualitative picture of the number of rapidly exchanging hydrogens (k Ͼ 4 min Ϫ1 ) can be obtained by measuring the amount of HDX at a relatively short time scale (e.g. 15 s). It seems intuitive that the exposure to solvent would be reflected in the rate of rapid HDX, and if this were true then identifying regions that have rapidly exchanging hydrogens would be indicative of solvent exposure. However, there is evidence in the literature that HDX is largely insensitive to the degree of solvent exposure for backbone hydrogens (31). A "rapid exchange map" is therefore a very rough estimate of solvent accessibility at best. Moreover, although a single time point is inextricably linked to the kinetics, it is not necessarily predictive of either the rates obtained when fitting a time course to multiple exponential equations or the final amplitudes of HDX observed in time course experiments. In the case of Dpo4, the rapid exchange map is largely confirmatory in nature because of the large amount of structural information already present in the literature. These socalled "rapid exchange" maps therefore provide an initial overview of deuterium exchange occurring under a given set of experimental conditions (i.e. with or without DNA and/or dNTPs). The rapid exchange map for Dpo4 in isolation indicated that, unsurprisingly, the palm domain exhibited the least amount of exchange in the first 15 s of the reaction (Fig. 1B, and supplemental Table S1). Portions of the finger and thumb domains exhibited the highest amounts of deuterium incorporation for the apoenzyme (Fig. 1B).
The addition of primer/template DNA and MgCl 2 reduced the fraction of rapid exchange by Ͼ5% for 12 peptides ( Fig. 2 and Table 1). One peptide located in the linker region between the thumb and little finger domains (residues 229 -239) exhibited a 12% increase in the amount of rapid exchange observed ( Fig. 2 and Table 1). Overall, the protected regions aligned well with what is observed in the ternary crystal structures of Dpo4 bound to primer/template DNA (Fig. 2). For example, helices H, J, and K (represented by peptides that contain residues 180 -202, 210 -218, and 215-229, respectively) all exhibited a decrease in rapid exchange upon addition of primer/template DNA and MgCl 2 . The loop region located between helices B and C in the finger domain (residues 48 -63) was protected from deuterium exchange upon addition of DNA substrate, as are residues 272-293 in the little finger domain ( Fig. 2 and Table 1).
The addition of dCTP resulted in changes of Ͼ5% of the maximal deuterium incorporation for eight peptides, but these were not the same set of peptides noted following addition of primer/template DNA and MgCl 2 ( Fig. 2 and Table  1). It should be mentioned again that 2Ј,3Ј-dideoxy-terminated primers were used in all experiments that contain dNTP in the reaction mixture. The lack of a 3Ј-hydroxyl moiety prohibited phosphodiester bond formation from occurring in the HDX experiment. Six of the peptides that changed upon addition of dCTP showed a decrease in the rapid phase of exchange, but two peptides in the thumb domain (residues 180 -202 and 215-229) showed an increase in rapid exchange upon addition of dCTP. With the exception of one peptide in the palm domain (residues 151-170 comprising the F-helix), the peptides that exhibited changes in the rapid HDX phase upon addition of dCTP are found in the finger and thumb domain.
Changes in HDX Specific to Nucleotide Selectivity-Separate HDX time courses were performed with either Dpo4 alone, with DNA/Mg 2ϩ , or with DNA/Mg 2ϩ /dNTP (supplemental Fig. S2 and Tables S3-S5). In the sequence context used for the HDX experiments, dCTP would represent a normal Watson-Crick pairing mode for insertion, and binding of dGTP would most likely result in formation of a "Type II" mode of binding (14) where the incoming nucleotide skips the template base at

Dpo4 Conformational Dynamics Using H-D Exchange
position 0 and pairs with the base at position ϩ1 (5Ј to position 0). Inclusion of either dATP or dTTP in the HDX experiment is expected to yield such mispairing events.
The majority of the protein did not exhibit changes in HDX during dCTP binding (supplemental Fig. S2 and Tables S3-S5), but some notable changes were observed in the Dpo4 HDX  (20)). The regions shown in gray indicate portions of Dpo4 that were not included in the peptide map. B, the fraction of deuterium exchanged in 15 s is shown for each peptide for either the enzyme alone (black), Dpo4 with primer/template DNA and Mg 2ϩ (blue), or Dpo4 with primer/template DNA, Mg 2ϩ , and dCTP (red).

Dpo4 Conformational Dynamics Using H-D Exchange
kinetics following addition of individual dNTPs. First, the loop region between helix B and C in the finger domain (residues 48 -63) exhibited changes in both the amount and rate of deuterium incorporated when an incoming dNTP was present (Fig.  3, A and B). All four dNTPs caused reductions in the amount of deuterium incorporated, which could indicate decreased solvent accessibility and/or a decrease in the number of hydrogen bonding interactions. A decrease in the HDX rate was only observed when dCTP was included in the experiment, decreasing roughly 2-to 3-fold relative to the rate observed when only DNA and Mg 2ϩ was present ( Fig. 3 and Table 2). The decreased rate of HDX was observed for two ions that mapped peptides 48 -63 and 49 -63 in the finger domain, respectively, and can be interpreted as resulting from increased hydrogen bond stability in this region upon binding of dCTP.
The loop region containing residues 48 -63 is in contact with what has been termed the "roof" of the Dpo4 active site (␤-sheet 3), which includes residues 40 -48. The ion representing residues 43-48 showed the largest reduction in deuterium content for both the dCTP and dGTP binding events (Fig. 3C and Table  2). A modest decrease in the rate of HDX for residues 43-48 was also observed upon addition of all four dNTPs. The decreased amplitude of deuterium incorporation in the presence of dCTP is consistent with the results obtained for residues 48 -63 and may represent more stable interactions between ␤-sheet 3 and ␣B-loop-␣C when the nascent base pair is dCTP:dG (Fig. 3D).
Another ion that exhibited decreased rapid exchange upon addition of dCTP was identified as a peptide containing residues 151-170, which comprise the F-helix in the Dpo4 palm domain. The rate constant that defines the first exponential phase of HDX for the F-helix decreased ϳ3-fold upon addition of dCTP, indicating that conformational flexibility in this region is diminished during formation of the nascent base pair. However, the addition of other dNTPs also elicited a similar change in HDX rate, indicating that the decreased flexibility observed in the F-helix is a general phenomenon associated with formation of a ternary complex but one that is insensitive to nucleotide identity (Fig. 4).
Two peptides in the thumb domain revealed quite unique HDX kinetics during the binding of dNTPs. A peptide containing residues 180 -202, which comprises ␣-helix H in the thumb, displayed increased HDX in the rapid exchange map upon addition of dCTP. Inspection of the HDX kinetic profiles revealed that the H-helix exhibits more stable and/or numerous hydrogen bonding interactions upon binding to DNA⅐Mg 2ϩ and either dATP or dTTP ( Fig. 5A and Table 2). The decrease in the rate of exchange was lost upon addition of either dCTP or (to a lesser extent) dGTP ( Fig. 5A and Table 2). The K-helix, which is in close proximity to the H-helix in the Dpo4 structure, also exhibited an increase in HDX when dCTP was present, relative to was observed with only DNA and Mg 2ϩ (Fig. 5B and Table 2). These two helices in the thumb domain appeared to be sensitive to the identity of the incoming nucleotide, and their importance regarding polymerase fidelity and kinetics is discussed below.

DISCUSSION
The models describing a simple open and closed polymerase structure determining fidelity have been supplanted by more refined descriptions of the molecular events associated with fidelity (32). In the past, most reaction schemes describing polymerase catalysis include a non-covalent step prior to phosphoryl transfer with no distinction made for accurate or errorprone insertion events. It has recently been shown that for bacteriophage T7 Ϫ DNA polymerase (pol T7 Ϫ ) there are different conformational states for correct and incorrect substrates (4). Kinetic and structural analysis by our own group has illus-  (13)). The changes in rapid HDX Ͼ 5% of the possible amount for each peptide are shown in either red (decreased exchange) or green (increased exchange).

Dpo4 Conformational Dynamics Using H-D Exchange
trated the role of non-productive or alternative complexes in determining catalytic efficiency during lesion bypass (33,34). The idea that an enzyme can exist in multiple conformational states (i.e. different free-energy minima) seems intuitive but a quantitative description of alternate states is difficult to obtain. In the case of pol T7 Ϫ the difference essentially lies in the fact that the reversal of the non-covalent step is much faster for incorrect substrates than it is for accurate substrates. Of course, this description is primarily kinetic in nature and enzymes are best understood by combining structural and functional descriptions. The objective of the current study was to explore conformational dynamics of a Y-family DNA polymerase using HDX-MS to identify regions of the enzyme that respond to the identity of the incoming dNTP. We succeeded in identifying regions of Dpo4 that confer substrate specificity during the catalytic cycle. The HDX-MS results obtained here provide evidence for two structural changes that are important for Dpo4 fidelity. Upon binding to an incoming dNTP the loop region between ␣-helices B and C appears to respond in a manner that stabilizes the hydrogen bonding patterns by decreasing structural fluctuations in the finger domain (Fig. 3). The flexibility of the loop region inferred from changes in the HDX rate is decreased most substantially during the formation of a Watson-Crick base pair in the pol active site. Given the fact that substantial changes in the finger domain are not observed in the crystal structures of Dpo4, it would seem possible that the decrease in the amount of deuterium observed in the loop region is simply due to solvent exclusion by dNTP binding and not a conformational change per se. In general terms, x-ray crystal structures are thought to represent the most thermodynamically stable conformation of a molecule under a given set of conditions (i.e. an endpoint). HDX methods report on regions of the protein that are undergoing structural changes, primarily those associated with breaking and/or formation of hydrogen bonds in the amide backbone. Our data support the idea that there is structural flexibility in the finger domain that is not reported by the crystal structures. As discussed under "Results," HDX rates for a given peptide are not as sensitive to solvent exposure as might be expected on an intuitive level (31), which would preclude changes due to solvent exclusion upon dNTP binding. Moreover, the contacts observed between the protein and incoming dNTP in the loop region are mediated entirely by the amino acid side chains and not main-chain atoms. HDX analysis only FIGURE 3. The kinetics of HDX in the finger domain is sensitive to the identity of the incoming dNTP. A, the kinetic profile for HDX in residues 48 -63 is plotted for Dpo4 alone (black circles), ϩDNA/Mg 2ϩ (blue squares), ϩDNA/Mg 2ϩ /dCTP (red circles), ϩDNA/Mg 2ϩ /dTTP (purple diamonds), ϩDNA/Mg 2ϩ /dATP (green triangles), and ϩDNA/Mg 2ϩ /dGTP (orange inverted triangles). The results are averages of two data sets. B, the kinetic profile for HDX in residues 49 -63 is plotted for Dpo4 alone (black circles), ϩDNA/Mg 2ϩ (blue squares), ϩDNA/ Mg 2ϩ /dCTP (red circles), ϩDNA/Mg 2ϩ /dTTP (purple diamonds), ϩDNA/Mg 2ϩ /dATP (green triangles), and ϩDNA/Mg 2ϩ /dGTP (orange inverted triangles). C, the kinetic profile for HDX in residues 42-48 is plotted for Dpo4 alone (black circles), ϩDNA/Mg 2ϩ (blue squares), ϩDNA/Mg 2ϩ /dCTP (red circles), ϩDNA/Mg 2ϩ /dTTP (purple diamonds), ϩDNA/Mg 2ϩ /dATP (green triangles), and ϩDNA/Mg 2ϩ /dGTP (orange inverted triangles). The observed rate constants are listed in Table 2. D, the structural orientation of the loop region (residues 48 -63 shown in green) and ␤-sheet 3 (residues 42-48 shown in orange) relative to one another and the incoming dNTP (pdb code 2j6t (13)). Hydrogen bonds between amide hydrogens are shown by dotted lines.

Dpo4 Conformational Dynamics Using H-D Exchange
measures exchange for amide protons. Therefore, the decrease in deuterium content does not result from the establishment of hydrogen bonds between the finger domain and the incoming dNTP and can most readily be accounted for by an increase in the stability of the hydrogen bond pattern that results from decreased flexibility in the loop between ␣-helices B and C. Notably, the region of the active site that contains residues in the palm domain contacting the incoming dNTP (Tyr-10 and Phe-8) does not exhibit reduced HDX during the time course, most likely due to the fact that this part of the protein is buried between ␤-sheets 6 and 7 and ␣-helices D, E, and F and is not undergoing conformational fluctuations. The second important change in Dpo4 HDX kinetics occurs in the thumb domain (Fig. 5). A peptide comprising the H-helix (residues 180 -202) was shown to exhibit reduced HDX during binding of DNA and Mg 2ϩ , as well as during misinsertion events (Fig. 5A). During binding of the correct dNTP (i.e. dCTP) and to a lesser extent when a Type II complex is formed (i.e. dGTP) the HDX rates observed in portions of the thumb domain increase (Fig. 5). A previous crystallographic study of Dpo4 bypass of the major oxidative DNA lesion 7,8-dihydro-8oxo-2Ј-deoxyguanosine (8-oxoG) reported on translocation mechanics by comparing pre-insertion binary, insertion ternary, and post-insertion binary complexes (19). The authors of the translocation study note that the contacts between the thumb domain (helices H, K, and J) and the primer/template DNA remain the same when comparing binary and ternary complexes. The model derived from the crystal structures proposes that the finger, palm, and little finger domains all shift forward by one nucleotide during binding of dCTP. However, the authors superimposed the binary and ternary DNA structures to draw their conclusions regarding domain rearrangements. A more accurate way to compare the structures would involve superimposition of the active site of the binary and ternary complexes (i.e. the palm domain) and searching for structural changes, because the DNA will be required to move relative to the catalytic site when the next dNTP binds. If the palm domains of the pre-insertion binary and insertion ternary complexes are superimposed, then comparison of the crystal structures reveals that the thumb domain changes its position relative to the finger, palm, and little finger domains when an accurate ternary complex is formed and the primer/template DNA shifts back to accommodate the incoming dNTP (Fig.  5C). We should note that the HDX results reported here do not directly measure the direction or magnitude of changes suggested by the superimposition of crystallographic results (Fig.  5C). However, based on the crystallographic results, the conformational change in the thumb domain reported by HDX is most likely a result of the DNA rotating "back" out of the active site to accommodate the nascent base pair (Fig. 5C). The contacts remain fixed, as observed in the crystal structures, but the exposure to solvent and equilibrium between binary/ternary DNA positions shifts toward ternary, which results in increased HDX in the thumb domain. The other domains still translocate forward by one nucleotide, as suggested in the crystallographic study, but it is because the DNA is rotated in the Dpo4 active site by one nucleotide. It is effectively the thumb domain combined with a fixation of the dNTP by the finger domain that serves to promote a catalytically active complex.
The evidence supporting an important role for rotation of the primer/template DNA during ternary complex formation is not confined to Dpo4. A similar conclusion was derived from the crystal structure of yeast pol bound to cisplatin-containing DNA (35). In the pol structure, two distinct primer/template DNA binding orientations were observed in the asymmetric unit. These modes were termed pre-elongation and elongation complexes. The general conclusion from comparing pre-elongation and elongation modes was that a rotation of the DNA was associated with the induced-fit mechanism of action. An overlay of the pre-elongation and elongation complexes does suggest that the thumb and little finger domains shift their orientation when the DNA rotates, which again highlights the importance of translocating the DNA into an active conformation during nucleotide selection by Y-family DNA polymerases.
Further evidence supporting an active role for the thumb domain during dNTP selection can be found in recent work from our group that used tryptophan fluorescence to measure the kinetics of conformational changes during Dpo4 The results are averages of two data sets. The HDX rate constants are listed in Table 2. B, Dpo4 is shown in schematic form with the domains colorcoded as in Fig. 1. The F-helix (orange) in the palm domain is sensitive to dNTP binding in the absence of direct contacts but does not appear to distinguish one dNTP from another.

Dpo4 Conformational Dynamics Using H-D Exchange
catalysis (25). The wild-type Dpo4 sequence does not contain any endogenous tryptophan residues. The Trp mutant used to analyze conformational kinetics was Dpo4-T239W (25). Importantly, Thr-239 is located in the linker region adjacent to the thumb domain. The primer/template DNA sequence used in the fluorescence study is identical to the one used here. Changes in fluorescence were only observed when the incoming dNTP formed either a "Type I" Watson-Crick base pair or a "Type II" Ϫ1 frameshift deletion and not when dNTPs forming mispairing events (i.e. dATP or dTTP) were included in the reaction. In reactions with dideoxyterminated primers the fluorescence change was a monophasic increase. The maximal extrapolated rate-constant defining the fluorescent increase was ϳ35 s Ϫ1 (25). When phosphodiester bond formation could occur (i.e. 3Ј-hydroxyl-containing primer) a second, slower phase of decreasing fluorescence (k obs ϭ ϳ1.5 s Ϫ1 ) was observed after the fast increase.
The slow decay appears to be first-order in nature because the rate was insensitive to dCTP concentration. The slow fluorescent decay was attributed to a "relaxation" event following phosphodiester bond formation. By measuring the kinetics of pyrophosphate release the relaxation event was determined to occur following pyrophosphate release. Such a model is consistent with the HDX results reported here. In the absence of phosphodiester bond formation the relaxation event in the thumb domain (Fig. 5) did not occur for either the accurate ternary complex (dCTP) or a Type II complex (dGTP), consistent with the increased rate of HDX observed for this region of Dpo4.
There are dynamic interactions between Dpo4, DNA, dNTPs, and metal ions in addition to the structural rearrangements noted above. As mentioned previously, upon binding to DNA the little finger domain apparently undergoes a dramatic conformational change that includes a large rotation and translocation through space (20). The binding of dNTP and metal ions is thought to be rapid (k forward ϳ10 8 M Ϫ1 s Ϫ1 ), with the rate of dNTP dissociation providing an initial level of nucleotide discrimination that is manifest in the apparent equilibrium binding constant (K D ,dNTP). Based on data reported here and elsewhere (19,20,23,25), dNTP⅐metal ion binding is either accompanied or immediately preceded by thumb-induced translocation of the DNA (i.e. rotation of the primer terminus out of the active site). It seems reasonable to assume that Dpo4 fluctuates between complexes that are not able to accommodate the incoming dNTP (pre-insertion) and those that can (insertion). Stabilization of the translocated state (i.e. insertion complex) is followed by phosphodiester bond formation, and it is most likely defined by how quickly it can be reversed, which is consistent with our HDX results and the model for nucleotide selection proposed by the Johnson group (4). For Dpo4, stabilization of the nascent base pair (i.e. the translocated state) is defined at a molecular level by electrostatic contacts with the phosphosugar moiety (e.g. Thr-45 and Arg-51) and by increased stability in the hydrogen-bonding pattern between FIGURE 5. Changes in HDX kinetics for the thumb domain are sensitive to the identity of the incoming dNTP. A, the kinetic profile for HDX in residues 180 -202 is plotted for Dpo4 alone (black circles), ϩDNA/Mg 2ϩ (blue squares), ϩDNA/Mg 2ϩ /dCTP (red circles), ϩDNA/Mg 2ϩ /dTTP (purple diamonds), ϩDNA/Mg 2ϩ /dATP (green triangles), and ϩDNA/Mg 2ϩ /dGTP (orange inverted triangles). The results are averages of two data sets. B, the kinetic profile for HDX in residues 215-229 is plotted for Dpo4 alone (black circles), ϩDNA/Mg 2ϩ (blue squares), ϩDNA/Mg 2ϩ /dCTP (red circles), ϩDNA/Mg 2ϩ /dTTP (purple diamonds), ϩDNA/Mg 2ϩ /dATP (green triangles), and ϩDNA/Mg 2ϩ /dGTP (orange inverted triangles). The results are averages of two data sets. The HDX rate-constants for A and B are listed in Table 2. C, a superimposition of the palm domain for pre-insertion binary (gray, pdb ID code 2asj (19)) and insertion ternary (green, pdb ID code 2asd (19)) is shown for the region of the thumb domain that exhibits HDX kinetics sensitive to the identity of the incoming dNTP. the loop region (residues 48 -63), ␣B, ␣C, and ␤3 in the finger domain (Fig. 3D).
Dpo4 is classified in the DinB subfamily of the Y-family DNA polymerases (9). It is of interest to consider the implications of change observed in the finger domain (residues 48 -63) of Dpo4 to the human DinB homologue pol . There is substantial evidence that the Y-family DNA polymerase is important during bypass of bulky minor groove adducts, especially products arising from exposure to exogenous polycyclic aromatic hydrocarbons (e.g. benzo[a]pyrene) and/or endogenous products of steroid biosynthesis (36 -40). pol is structurally similar to Dpo4 in overall domain topology with one major exception. pol possesses an N-terminal extension termed the N-clasp, which forms two ␣-helices that were shown to encircle the DNA in a crystal structure of the pol ternary complex (10). The first helix in the N-clasp sits directly atop the finger domain of pol , and the second helix traverses the DNA at a 45°angle before connecting with and packing against the thumb domain. The N-clasp is essential for pol function, because deletion of the first 68 residues results in a dramatic loss of activity, including a substantial decrease in the binding affinity for DNA (10). Given the fact that the N-clasp contacts both domains that appear to be central to structural rearrangements during Dpo4 nucleotide selection, it will be of great interest to investigate whether pol exhibits similar dynamics. A recent study from our group investigated pol -catalyzed bypass of 8-oxoG, 4 and the results obtained there are consistent with more restrained conformational dynamics for pol . Specifically, we found that the error-prone nature of pol bypass is determined by the stabilization of syn-oriented 8-oxoG by residues in the finger domain of pol and a lack of electrostatic interactions needed to stabilize the anti conformation of 8-oxoG. The 8-oxoG adduct is thermodynamically more stable in the syn orientation (41), and its mutagenicity stems from the ability of many polymerases to form a Hoogsteen base pair between dATP and syn-8-oxoG (42,43). Because the N-clasp is in contact with the finger domain of pol , it seems reasonable to assume that any conformational dynamics will be restrained by this interaction, but further study is required to address such a question.
The results obtained in the present study are important for an understanding of Dpo4 catalysis, because they effectively bridge results obtained using x-ray crystallography (19) and kinetic approaches (23,25). The ability of the loop region in the finger domain to efficiently stabilize the incoming dNTP acts in concert with the thumb-mediated equilibrium between binary and ternary DNA orientations to promote nucleotide selection in the absence of steric constraints and/or base specific contacts, which are utilized by more high fidelity pols.