Protein Dynamics Control the Progression and Efficiency of the Catalytic Reaction Cycle of the Escherichia coli DNA-Repair Enzyme AlkB*

Background: AlkB is an iron/2-oxoglutarate dioxygenase that repairs alkylated DNA. Results: Enzymological and biophysical methods are used to develop a quantitative mechanistic scheme. Conclusion: A conformational transition controls the order of substrate binding and the coupling of the two sequential sub-reactions catalyzed by AlkB. Significance: These results provide a striking example of the importance of protein dynamics for efficient catalysis of a multistep reaction. A central goal of enzymology is to understand the physicochemical mechanisms that enable proteins to catalyze complex chemical reactions with high efficiency. Recent methodological advances enable the contribution of protein dynamics to enzyme efficiency to be explored more deeply. Here, we utilize enzymological and biophysical studies, including NMR measurements of conformational dynamics, to develop a quantitative mechanistic scheme for the DNA repair enzyme AlkB. Like other iron/2-oxoglutarate-dependent dioxygenases, AlkB employs a two-step mechanism in which oxidation of 2-oxoglutarate generates a highly reactive enzyme-bound oxyferryl intermediate that, in the case of AlkB, slowly hydroxylates an alkylated nucleobase. Our results demonstrate that a microsecond-to-millisecond time scale conformational transition facilitates the proper sequential order of substrate binding to AlkB. Mutations altering the dynamics of this transition allow generation of the oxyferryl intermediate but promote its premature quenching by solvent, which uncouples 2-oxoglutarate turnover from nucleobase oxidation. Therefore, efficient catalysis by AlkB depends upon the dynamics of a specific conformational transition, establishing another paradigm for the control of enzyme function by protein dynamics.

Extensive kinetic and spectroscopic studies have demonstrated that Fe(II)/2OG dioxygenases face a series of mechanistic challenges in achieving efficient catalysis (15)(16)(17)(18)(19)(20)(21)(22). These enzymes catalyze similar multistep reactions in which the Fe(II) cofactor and 2OG co-substrate activate O 2 for hydroxylation of a bioorganic substrate, an alkylated nucleobase in the case of AlkB (12,13). This so-called "primary substrate" occludes the binding sites for Fe(II) and 2OG in available crystal structures (as shown for AlkB in Fig. 1B (23)), consistent with the conclusion from kinetic studies (24 -26) that the enzymes bind the various cofactor and substrates sequentially, with the primary substrate binding only after both Fe(II) and 2OG (Fig. 1A). Subsequent binding of O 2 to the Fe(II) cofactor triggers relatively rapid decarboxylation of 2OG to generate carbon dioxide (CO 2 ), succinate (Suc), and a reactive oxyferryl (Fe(IV)ϭO) intermediate (16,(27)(28)(29) that slowly hydroxylates the nucleobase substrate (21,22). This long lived oxyferryl intermediate must remain tightly sequestered in the active site to avoid reaction with oxidizable compounds in the buffer (solvent-quench-ing), which results in uncoupled 2OG turnover (12,13,30) and release of toxic reactive oxygen species. Several mechanistic features contribute to the very slow occurrence of these adventitious processes in well characterized Fe(II)/2OG dioxygenases (26). First, O 2 does not bind to Fe(II) efficiently until after the binding of the primary substrate (23,31). Second, the release rate of the unreacted primary substrate must be slower than its hydroxylation rate by the oxyferryl intermediate. However, overly slow release of the hydroxylated product would impair catalytic efficiency. Therefore, substrate/product release kinetics must be tuned to optimize the catalytic efficiency of Fe(II)/ 2OG dioxygenases.
Although kinetic and spectroscopic studies have identified these mechanistic challenges, little information is available on the biophysical mechanisms employed by Fe(II)/2OG dioxygenases to surmount them. Significant ligand-dependent protein conformational changes have been observed in several Fe(II)/ 2OG dioxygenases, including AlkB (32), as well as in some homologous halogenases for which crystal structures have been solved in open or closed conformations depending on 2OG binding (33,34). Changes in resonance frequencies and line shapes in 1 H one-dimensional and 1 H-15 N two-dimensional NMR spectra also suggest that the conformational dynamics of AlkB are ligand-dependent (32,35), and these data have been interpreted to reflect increased backbone dynamics that serve to accelerate ligand release from the enzyme upon oxidation of 2OG to Suc. However, given the slow reaction rate of the oxyferryl intermediate generated in parallel with Suc, such accelerated ligand release would be expected to promote quenching of the oxyferryl intermediate (due to premature release of the primary substrate) and adventitious release of reactive oxygen species.
In this study, we explicitly verify that enhanced conformational dynamics in AlkB can increase turnover of the 2OG cosubstrate uncoupled from hydroxylation/repair of the primary substrate. Furthermore, we establish a quantitative mechanistic description of the catalytic cycle of AlkB. A key feature of this mechanism is a specific conformational transition in E. coli AlkB that simultaneously controls both the proper order of substrate binding and the kinetics of sequestration of the enzyme-bound oxyferryl intermediate (Fig. 1, C and D). Sophisticated biophysical methods have been used to demonstrate that protein dynamics play an important role in controlling substrate binding and product release in some enzymes (36 -45). The results reported in this paper demonstrate that protein dynamics can also control the coupling of sequential sub-reactions in a multistep enzyme-bound reaction sequence. Our results provide a striking example of the importance of protein dynamics for efficient catalysis of a complex chemical reaction.

EXPERIMENTAL PROCEDURES
Overview of Experimental Conditions-Enzymological, fluorescence, and circular dichroism experiments were conducted in 1 H 2 O buffer containing 75 mM KCl and either 10 mM Tris-Cl or 20 -50 mM Na-HEPES at pH 7.6 (23,31). NMR experiments were conducted in 99% 2 H 2 O buffer containing 75 mM KCl and either 50 mM NaH 2 PO 4 at pD 5.5 or 50 mM Na-HEPES-d 12 at pD 7.6. Calibration experiments demonstrate that E. coli AlkB exhibits similar turnover rates, similar affinities for the 2OG co-substrate and the alkylated DNA substrate (Table 1), and essentially equivalent 2OG/DNA coupling ratios in the deuterated acidic buffer used for most of the NMR experiments and the neutral pH buffer used for most other experiments (data not shown). Some experiments employed Mn(II) as a redox-inactive analog of Fe(II), whereas others, including all NMR experiments, employed Zn(II) instead because it is diamagnetic and therefore does not interfere with NMR measurements. Measurements of thermal stability in the presence and absence of 2OG demonstrate that these three divalent cations interact with E. coli AlkB with roughly similar affinities but that Zn(II) produces somewhat greater thermal stabilization of the enzyme than Fe(II) or Mn(II) (data not shown). In general, buffers and reagent solutions were passed through uncharged Profinity IMAC resin (Bio-Rad) to remove trace metal ions. Mechanistic characterization focused on experiments at 10°C to minimize interference from protein aggregation, which is enhanced at higher temperature. AlkB catalyzes repair of alkylated nucleobases at a similar rate at either 10 or 37°C (data not shown), consistent with quantum mechanical processes being rate-determining in the reaction cycle (46).
Cloning and Site-directed Mutagenesis-PCR cloning was used to insert a TEV protease recognition sequence (ENLYFQG) after the start codon of AlkB-⌬N11, which has the first 11 N-terminal amino acids of the native E. coli protein removed (23). Following amplification of a construct of the truncated protein containing a C-terminal TEV-cleavable hexahistidine tag, the resulting PCR product was cloned into the NdeI and XhoI restriction sites in plasmid pET28b to yield a protein construct with TEV-cleavable hexahistidine tags at both its N and C termini. Cleavage of this construct with TEV protease yields an AlkB-⌬N11 protein differing from the previously studied construct (23) only in the substitution of a glycine residue for the methionine at the N terminus (i.e. the single residue preceding Gln-12 from the native E. coli protein). Most experiments in this study use this construct starting with Gly. For the small number of experiments in which the original construct was used, an asterisk has been added to the protein name in the figure labels and legends. Mutants were made using the QuikChange kit (Agilent Technologies, Santa Clara, CA).
Protein Expression and Purification-Unlabeled wild-type and mutant proteins were expressed and purified as described previously (23,31). Proteins labeled with [ 13 C ⑀ ]methionine were expressed in E. coli strain B834(DE3), which is auxotrophic for methionine. Cells were grown in M9 medium containing L-[ 13 C ⑀ ]methionine, NH 4 Cl, glucose, and 1ϫ Cellgro MEM vitamins to an A 600 of 0.6 -0.7 at 37°C. Following induction with 0.8 mM IPTG, growth was continued for 18 h at 20°C prior to purification of the protein using the published methods (23,31). For expression and purification of the deuterated protein, every component of the M9 medium was prepared in 99% 2 H 2 O except 1ϫ Cellgro MEM vitamins. To increase labeling efficiency, [ 2 H]glucose was added to the M9 medium at a 1:5 ratio to total glucose. The cells were adapted for growth in 2 H 2 O by first inoculating them from LB into M9 medium with 50% 2 H 2 O and then inoculating the cells grown in this medium into M9 medium with 99% 2 H 2 O. After induction with 0.8 mM IPTG, the protein was expressed for 43 h at 20°C prior to purification using the published methods (23,31). AlkB-⌬N11 was then concentrated to between 500 M and 1 mM and exchanged into the appropriate buffer (99% 2 H 2 O buffer containing 75 mM KCl and either 50 mM NaH 2 PO 4 at pD 5.5 or 50 mM Na-HEPES-d 12 at pD 7.6). Purified protein samples were flash-frozen and stored at Ϫ80°C.
Circular Dichroism (CD) Spectroscopy-Measurements were performed using a J-815 CD spectrometer (Jasco, Easton, MD) equipped with a PFD-425 Peltier temperature controlled cell. Wavelength scans were conducted at 50 nm/min from 205 to 300 nm on samples at 10°C. Thermal denaturation experiments used a 5°C/min temperature ramp and were monitored either at 217 nm from 13 to 90°C (samples in Tris buffer) or at 230 nm from 12 to 80°C (samples in HEPES buffer). Data from these experiments were smoothed using the Savitzky-Golay method with an 11-nm window. Experiments evaluating the effect of Mn(II) and 2OG on the wild-type (WT) enzyme were conducted in 75 mM KCl, 50 M EDTA, 10 mM Tris-HCl, pH 7.6, whereas experiments comparing WT and mutant enzymes were conducted in 75 mM KCl, 2 mM EDTA, 20 mM Na-HEPES, pH 7.6. The enzyme displays similar binding affinities and enzymatic activities in both buffers. Experiments using Zn(II) were conducted in 99% 2 H 2 O buffer containing 75 mM KCl, 50 M EDTA, 50 mM sodium phosphate, pD 5.5. The Tris and phosphate buffers were demetallated using uncharged IMAC resin as described above. The concentrations of the M61L, L99M, and I119M mutants were determined via quantitative amino acid analysis following HCl hydrolysis (W. M. Keck Foundation Laboratory, Yale School of Medicine). The concentrations of the WT and W89Y enzymes were determined based on absorbance at 280 nm in 6 M guanidine hydrochloride, 20 mM sodium phosphate, pH 6.4. The extinction coefficient in this buffer was determined based on the quantitative amino acid analysis results from the M61L mutant.
Intrinsic Tryptophan Fluorescence Quenching Assays-Tryptophan fluorescence was monitored as 2OG, Suc, or DNA substrate (5Ј-CAmCAT-3Ј), or product (5Ј-CACAT-3Ј) was titrated into 0.2 M protein in the presence of 100 M MnCl 2 at 10°C in a PTI QuantaMaster C61 fluorimeter using excitation and emission wavelengths of 297 and 329 nm, respectively, and 5 nm slits. Titration steps were less than 30 s apart to minimize time-dependent protein aggregation at sub-saturating 2OG or Suc concentrations. Unless indicated otherwise, experiments were performed in freshly degassed 75 mM KCl, 50 mM HEPES, pH 7.6. DNA titrations were conducted in buffers containing either 2 mM EDTA or 100 M MnCl 2 and, when present, 100 M 2OG or 2 mM Suc. The dissociation constant (K d ) for each titrant was estimated using the Levenberg-Marquardt algorithm as implemented in the program PRISM (GraphPad, San Diego) to fit the observed data to a quadratic binding Equation 1, where B is the total fluorescence emission intensity observed at total titrant concentration x; B 0 is the background intensity in the absence of the titrant; ⌬B is the change in intensity at saturating concentration of the titrant; and c is the concentration of the protein.
Fluorescence Anisotropy Assays on Labeled DNA-Fluorescence polarization anisotropy was monitored using 494 and 521 nm excitation and emission wavelengths, respectively, with 5 nm slits as AlkB-⌬N11 was titrated onto a 10 nM concentration of a 6-FAM-labeled DNA oligonucleotide at 10°C in freshly degassed 75 mM KCl, 50 mM HEPES, pH 7.6. Mn(II) was used at 20 M, 2OG at 100 M, and EDTA at 2 mM. Dissociation constants were determined by fitting Equation 1 above with x and c representing the total concentrations of AlkB-⌬N11 and 5Ј-CAm-CAT-(6-FAM)-3Ј, respectively. Release rates were measured by adding 10 nM 5Ј-TmAT-FAM-3Ј or 5Ј-CAmCAT-FAM-3Ј to buffer at 10°C containing 20 M MnCl 2 and 2 mM Suc, followed by addition of the protein and then unlabeled 5Ј-TmAT-3Ј or 5Ј-CAmCAT-3Ј to initiate competitive displacement. The release rate (k off ) was determined by fitting these data to the exponential decay function given by Equation 2, where Y is the observed anisotropy at time x; B is the anisotropy upon completion of DNA release; and ⌬B is the difference between that value and the initial value at the start of the experiment.
NMR Spectroscopy-All buffers, protein solutions, and titrant stocks (2OG, Suc, or DNA oligonucleotide) were passed through uncharged Profinity IMAC resin (Bio-Rad) to remove trace metal ions, some of which altered the NMR spectral properties of AlkB-⌬N11 when not removed. Immediately before use, protein samples were thawed, centrifuged, and diluted to a final volume of 550 l.  (47). The 1 H-13 C spectra were recorded using between 16 and 512 scans per t 1 increment and 1024 ϫ 512 points for the t 2 ϫ t 1 dimensions. The spectral widths were 7.0 ϫ 7.8 kHz, 9.6 ϫ 8.9 kHz, and 11.0 ϫ 10.0 kHz at 16.5, 18.8, and 21.1 T, respectively. Relaxation data were collected at 21.1 T using 512 t 1 increments with 16 -32 scans and a t 1 spectral showing the crystal structure of AlkB-⌬N11 with bound Fe(II) (orange), 2OG (blue), and 5Ј-TmAT-3Ј (yellow) (PDB code 2FD8). The Fe(II)/2OG core and NRL are colored cyan and red, respectively, and the residues used as spectroscopic probes are shown in space-filling representation (Trp-69, -89, and -178 in magenta, Met-49, -57, -61, and -92 in darker green, and Leu-99 and Ile-119 in lighter green). C, diagram colored equivalently to B schematizing the two-state conformational transition controlling enzyme activity. Backbone amide hydrogen-deuterium exchange data in the presence and absence of DNA were reported previously (23). D, catalytic mechanism and dissociation/rate constants at 10°C. The two alternative reaction pathways (lower right) produce 2OG turnover coupled to (left branch) or uncoupled from (right branch) DNA repair, depending on the relative rates of substrate hydroxylation versus release. K d values come from Fig. 3, conformational transition rate constants from Fig. 5, and catalytic rate constants from Fig. 13 and Yu and Hunt (31). Dissociation constants were calculated as described under "Experimental Procedures." The upper limit for the rate of 2OG oxidation is the maximum observed rate of 2OG turnover (Table 4), and methylated nucleobase oxidation rates are equal to k cat (31). The equivalence of k cat for trimer and pentamer substrates with the same alkylated base (31), despite trimer being released ϳ50-fold faster (Fig. 13, D and E), suggests that product release is not rate-limiting. The inference that nucleobase oxidation by the oxyferryl intermediate (i.e. quantum chemistry) is rate-limiting is supported by the equivalence of the observed catalytic but not conformational rate constants at 10 and 37°C (Tables 1 and 4).

Protein Dynamics Control Catalytic Efficiency of AlkB
width of 5.6 kHz. The 1 H-13 C single and multiple quantum correlations were measured using sensitivity-enhanced HSQC with gradient selection (48 -50) or SOFAST-HMQC (51) pulse sequences. Fast time scale (picosecond to nanosecond) dynamics were measured using a 1 H-1 H cross-correlated relaxation experiment (52) with a relaxation delay of 2-30 ms. Doubleand zero-quantum relaxation rates, R ZQ and R DQ , were measured using a Hahn-echo experiment (53) with a relaxation delay of 3.91-31.2 ms.
NMR Data Analysis-Data were processed with NMRPipe (54, 55) using exponential line broadening or a cosine bell for apodization in the t 1 dimension. Assignment and extraction of peak heights were performed in Sparky (T. D. Goddard and D. G. Kneller, University of California, San Francisco). NMR spin-relaxation parameters were determined using nonlinear least squares regression with Python scripts (55-57); parameters were fit to peak heights using either a monoexponential decay (Hahn-echo) or Equation 3 (cross-correlated relaxation) (52), Residues Met-49 and Met-57 were assigned identical values for , R ZQ , and R DQ due to resonance degeneracy in the presence of Zn(II). In the case of , the observed trend is not affected, because this value is the smallest of any Met residue over the titration. In the case of R ZQ and R DQ , the value of k 1 only increased by ϳ25% when the values of R MQ and ⌬R MQ determined for Met-61 in the presence of Zn(II) were substituted in the calculation.
The rate constant for transition to the closed state, k 1 , was determined by simultaneously fitting both R MQ and ⌬R MQ for Met-49 to the parameters of the standard two-site exchange Equation 4, In Equation 4, R 2 and R 20 are the transverse relaxation rate constants in the presence and absence of conformational exchange, respectively. The populations of the open and closed states are p a and p b , respectively. The difference in chemical shift between these two states is represented by ⌬, and the conformational exchange rate is denoted by k ex , which is the sum of the opening (k Ϫ1 ) and closing (k 1 ) rates. The closing rate was assumed to remain constant throughout the titration. The value of R 20 was estimated using the respective relaxation rate constant, R MQ or ⌬R MQ , determined for AlkB in the presence of DNA. The values of p a and p b were determined from 13 C chemical shifts.
Assays for Coupling of 2OG Decarboxylation to DNA Demethylation-Wild-type AlkB-⌬N11 at 0.5 M, the W89Y mutant at 1.2 M, or the M61L mutant at 0.7 M were incubated with 30 M TmAT at 37°C for 1, 5, 10, or 30 min in a standard assay buffer containing 150 M 2OG, 75 M Fe(NH 4 ) 2 (SO 4 ) 2 , 500 M ascorbate, 150 M sodium pyruvate, 75 mM KCl, 50 mM Na-HEPES, pH 7.6. Activity on 24 M 5Ј-CAmAAT-3Ј was assayed using the same conditions but with 0.2 M WT or 0.5 M W89Y enzyme at 1-, 5-, 10-, or 20-min time points. After quenching with 8 mM EDTA, the reactions were split to measure 2OG (300 l) or DNA (30 l) turnover. Samples for DNA analysis were diluted 10-fold prior to injection onto a Luna C18 column (Phenomenex Inc., Torrance, CA) run as published previously (23,58). Samples for 2OG analysis were brought to 3% HClO 4 (w/w) and then loaded onto an Aminex HPX-87H column (Bio-Rad) at 45°C running isocratically at 0.5 ml/min in 4 mM H 2 SO 4 on a Beckman System Gold HPLC (Beckman Coulter Inc., Indianapolis, IN). The absorbance of the effluent was monitored at 210 nm. Because the extinction coefficient of . Equivalent trends are observed in the assays conducted in either buffer in these assays at 10°C and also in assays conducted at 25°C (Table 1) the Suc co-product is much lower than that of the 2OG co-substrate, turnover was quantified by monitoring the decrease in 2OG concentration, which was determined by normalizing its peak area to that of 150 M pyruvate, an internal standard that is stable under the assay conditions. Peaks were integrated using Version 7.0 SP1 of Beckman's 32 Karat chromatographic software system.
The catalytic rate constant, k cat , and Michaelis-Menten constant, K m , were determined as described previously (23). Fig. 1D were calculated based on the model for 2OG binding shown in Scheme 1, which yields Equations 5-7 describing the relationships between the indicated microscopic dissociation constants and the observed dissociation constant for 2OG (K d-app ). (The asterisk in the name indicates that a single Met residue precedes residue Gln-12 from E. coli AlkB, whereas a single Gly residue occurs at this location in most enzymes used in this paper; see "Experimental Procedures.") These data demonstrate that the three Trp residues in the WT enzyme are quenched to a roughly comparable extent during the conformational transition coupled to 2OG binding, possibly reflecting the occurrence of fluorescence resonance energy transfer between these residues (which are ϳ8, ϳ13, and ϳ20 Å apart in the closed state of the enzyme, Fig. 1B). C, relative Trp fluorescence of 0.2 M WT (black), M61L (blue), or W89Y (red) enzymes upon titration with 2OG at 10°C in 100 M MnCl 2 , 75 mM KCl, 50 mM Na-HEPES, pH 7.6. The M61L mutation reduces 2OG binding affinity only slightly, as confirmed by NMR spectroscopy (Fig. 9), whereas the W89Y mutation reduces it by ϳ20-fold at 10°C (Table 1), presumably due primarily to destabilization of the closed state of the enzyme (Fig. 5F, inset).

TABLE 1 Kinetic and equilibrium constants of WT and mutant AlkB-⌬N11 enzymes
Dissociation constants were obtained by intrinsic tryptophan fluorescence or fluorescence polarization anisotropy measurements. An asterisk in the protein name indicates that the first residue (preceding Gln-12 in the ⌬N11 construct) is methionine rather than glycine (see "Experimental Procedures"). The entry (NMR) in the Conditions column indicates that the experiment was performed in NMR buffer, i.e. in a demetallated buffer containing 50 mM NaH 2 PO 4 , 75 mM KCl, and 99% 2 H 2 O, pD 5.5.

Ligand
Protein Based on the NMR analyses described in the main text, K c-free Ϸ0. 19 and K c-2OG Ϸ1.8. The same approach was used to calculate the microscopic dissociation constants for Suc. An equivalent approach was used to calculate the microscopic dissociation constant for DNA substrate binding to the closed state of the enzyme assuming negligible binding to the open state. The free energy of binding inferred from this calculation was added to the free energy of the protein's conformational transition in the presence of 2OG to determine the free energy of DNA substrate binding to the open conformation of the enzyme with bound 2OG, and thereby the dissociation constant for this binding reaction which is given in Fig. 1D.

RESULTS
Conformational Change in AlkB Coupled to Co-substrate (2OG) or Co-product (Suc) Binding-We used far-UV circular dichroism spectroscopy (CD) (Fig. 2) to evaluate the conformation and thermal stability of apo and ligand-bound AlkB. Our biophysical studies employed Mn(II) or Zn(II) as well established redox-inactive analogs of Fe(II) and N-terminally trun-cated AlkB constructs that start at residue Gln-12 in the native sequence and retain full enzyme activity (23). The CD spectrum of AlkB does not change significantly upon binding Mn(II) and 2OG ( Fig. 2A). Although these observations indicate that there is not a substantial change in the net secondary-structure content of the protein, some change in local secondary structure remains possible due to complexities in the spectral properties of short ␤-strand structures (59). CD shows that thermal unfolding proceeds via a single transition (Fig. 2, B and D). Under the conditions of these experiments, the midpoint of this transition increases by 3°C in the presence of the ligands, demonstrating that they thermodynamically stabilize the protein.
Calorimetric studies demonstrate substantial enthalpy-entropy compensation (60, 61) upon binding the metal cofactor and also the 2OG co-substrate, 5 suggesting the occurrence of a significant change in solvent interaction or protein dynamics even if there is not a substantial change in net secondary structure. The  OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 NMR studies presented below show that there is a substantial change in protein dynamics and tertiary structure coupled to 2OG binding.

Protein Dynamics Control Catalytic Efficiency of AlkB
We next used intrinsic tryptophan (Trp) fluorescence emission spectroscopy (Fig. 3, A-D) to monitor local conformational changes induced by ligand binding to AlkB. The truncated protein construct used in our studies has three Trp residues (Fig. 1B). Trp-69 is located in the nucleotide-recognition lid (NRL) and makes a stacking interaction with the alkylated nucleic acid base that seals the 2OG co-substrate into the active site. Trp-89 is located at the hinge between the NRL and the dioxygenase core, whereas Trp-178 lines a solvent-filled cavity behind the catalytic Fe(II) center in the core. Trp fluorescence is strongly quenched upon addition of 2OG in the presence of Mn(II). The Suc co-product has a similar effect, although it binds with reduced affinity and quenches fluorescence to a lesser extent (Fig. 3, A and C). The NMR data pre-sented below suggest that the difference in the quenching level in the co-substrate versus co-product complexes does not reflect a significant conformational difference but rather alterations in the electronic interactions of Trp-69 or Trp-178. Their indole groups reside within 8 Å of the metal center and that of Trp-69 also makes van der Waals contacts to two methionine residues (62), as discussed further below. The differential quenching by 2OG versus Suc is likely to derive either from alterations in the electronic properties of the respective metalchelation complexes or from small perturbations in local stereochemistry; either effect would be expected to change the rate of fluorescence quenching via standard mechanisms, including photo-induced electron transfer and resonance energy transfer (62)(63)(64)(65)(66).
Subsequent addition of a methylated DNA substrate (5Ј-CAmCAT-3Ј) to the AlkB complex with Mn(II) and either 2OG or Suc further quenches fluorescence to the same final level (i.e., for both the Mn(II)/2OG and Mn(II)/Suc complexes) (Fig. 3, B  and D). These observations are consistent with binding of either 2OG or Suc being coupled to a conformational change in AlkB that produces an equivalent structure after binding methylated DNA substrate, as observed crystallographically (14,23,31).
Analyses of ligand-dependent fluorescence changes in Trpto-Tyr mutants of AlkB (Fig. 4) demonstrate that all three Trp residues probably are quenched during the conformational transition, likely due at least in part to altered electronic interactions of Trp-69/Trp-89 with the following three proximal methionine residues (62): Met-49 and Met-57 in the NRL and Met-61 in the adjacent ␤-strand in the dioxygenase core. In the available crystal structures (14,23,31), all of which have a metal cofactor and either 2OG or Suc bound, Trp-89 is observed in the same conformation buried in the interface between the NRL and the dioxygenase core. Strong 2OG-dependent  quenching of Trp-89 in the W69Y/W178Y double mutant (Fig.  4) indicates that this interface must change conformation in the apoenzyme, although without substantial alteration in the net secondary structure content of the enzyme (Fig. 2E). Furthermore, the W89Y single mutant shows ϳ18-fold reduced affinity for 2OG as monitored via the conformational transition coupled to 2OG binding ( Fig. 4 and Table 1), even though Trp-89 is remote from the active site of the enzyme (Fig. 1B). This observation indicates that the W89Y mutation biases the transition toward the apo conformation by ϳ1.6 kcal/mol and provides further evidence for a change in the environment of Trp-89 during the transition.
2OG Binding Substantially Increases Affinity for DNA Substrate-Direct DNA binding assays monitoring fluorescence anisotropy of 5Ј-CAmCAT-(6-carboxyfluorescein)-3Ј in the absence or presence of Mn(II)/2OG or Mn(II)/Suc (Fig. 3, E and F) demonstrate that this conformational transition increases the affinity of AlkB for the methylated DNA substrate by approximately 2 orders of magnitude (32,67) (Table 1). Therefore, this transition controls the proper sequential order of co-substrate/substrate binding, thereby preventing premature binding of the DNA substrate from occluding the binding site for the metal cofactor and 2OG (as observed in all DNAbound structures). The similar effects of 2OG and Suc in enhancing binding of the methylated DNA substrate prevent its premature release, an event that leads to quenching of the oxyferryl intermediate by bulk solvent instead of substrate hydroxylation (Fig. 1D). The demethylated DNA product (5Ј-CACAT-3Ј) binds to the enzyme Ͼ100-fold more weakly than the corresponding substrate, as monitored directly by fluorescence anisotropy of labeled DNA (Fig. 3F) or indirectly by protein Trp fluorescence ( Fig. 3B and Table 1). The greatly increased affinity of the enzyme for alkylated substrate versus the repaired product could be influenced by the net positive charge on the methylated nucleobase (12)(13)(14)58), but it is also likely to be influenced by the methyl substituent on the substrate (12)(13)(14)67). Therefore, AlkB has evolved to selectively bind alkylated nucleic acid substrates following a specific conformational transition coupled to binding of the metal cofactor and 2OG co-substrate.
Co-substrate or Co-product Binding Orders Side Chains at the Interface of the NRL-To characterize this conformational transition further, given the likely alteration in the environments of the three methionine residues close to the interface of the NRL with the dioxygenase core, we used 1 H-13 C NMR spectroscopy of AlkB selectively labeled with [ 13 C ⑀ ]methionine (Figs. 5-10). Fully protonated methionine methyl groups ( 13 C 1 H 3 ) were chosen for observation in NMR experiments over approaches involving backbone amide groups due to significantly increased sensitivity, elimination of solvent exchange artifacts, reduction of spectral crowding, and simplification of resonance assignments. The truncated AlkB construct used in our studies also contains a fourth methionine, Met-92, on the surface of the dioxygenase core (Fig. 1B). Two-dimensional 1 H-13 C NMR spectra were recorded after addition of Zn(II), 2OG, or Suc and methylated nucleotide substrates of varying lengths (5Ј-CAmAAT-3Ј, 5Ј-TmAT-3Ј, or methylated ATP). Zn(II) was used instead of Mn(II) to avoid paramagnetic shifts Negligible perturbation of the 13 C ⑀ resonance of Met-92 is observed upon binding 2OG/DNA to the zinc-bound enzyme (Fig. 5A), consistent with minimal alteration of the dioxyge-nase core structure despite the ligand-dependent conformational changes in the NRL. The 13 C ⑀ resonance of Met-92 is split into two sharp peaks in the apoenzyme prior to collapse into a single sharp peak upon binding the metal cofactor, suggesting that this residue is found in two distinct environments in the apoenzyme that do not exchange on the microsecond time scale of the NMR measurements. This splitting is likely to be caused by local variations in protein conformation or electrostatic environment rather than heterogeneity in global backbone conformation, because a single sharp 13 C ⑀ resonance is observed when one additional methionine residue is inserted at buried positions Leu-99 or Ile-119 in the Fe/2OG dioxygenase core (Fig. 10). These data indicate that the core adopts a well defined conformation and are inconsistent with the suggestion that apo AlkB adopts a molten globule-like conformation. This suggestion is also inconsistent with the results of 4,4Ј-dianilino-1,1Ј-binaphthyl-5,5Ј-disulfonic acid titration experiments and differential FIGURE 9. NMR spectra of AlkB-⌬N11 containing single methionine-to-leucine mutations. The methionine-to-leucine mutations analyzed here were used to assign the resonances in the WT enzyme. Data were acquired using a 1 H-13 C SOFAST HMQC experiment, except for C, L, and T, which were generated using a 1 H-13 C HSQC experiment. Measurements were conducted at 10°C in 75 mM KCl, 50 mM NaH 2 PO 4 , pD 5.5. The unlabeled resonances in L and T are attributable to the presence of trace metals, which have been verified not to influence assignments. All spectra from a given mutant enzyme are shown in the same color.
scanning calorimetry experiments that fail to support the presence of a significant population of molten-globule structure in the apoenzyme. 5 Substrate and Product Molecules Modulate the Same Discrete Two-state Conformational Transition-Several observations indicate that substrate/product molecules modulate the equilibrium constant for a fast transition on the NMR chemical-shift time scale between two discrete AlkB conformations. One of these is a more dynamic conformation that is dominant in the apo state, while the other is a more ordered conformation that is dominant in the 2OG/Suc/DNA-bound states (Fig. 1, C  and D). We designate these conformations as "open" and "closed," respectively, consistent with longstanding usage of these terms in the enzymology literature as well as with the NMR data reported in this paper and previously reported structural results on AlkB (23,31). The 1 H and 13 C chemical shift perturbations of Met-49, which are larger than those of the other methionines, even though Met-49 makes no direct and potentially confounding contacts with substrates, are linearly correlated upon additions of 2OG or Suc and nucleic acid substrates (Fig. 5B). Similar behavior is observed for Met-57, although within a smaller range of chemical shifts. This observation is consistent with all ligands modulating the same twostate transition. Another observation supporting this inference is that 1 H and 13 C chemical shifts are uniformly downfield for complexes with longer versus shorter nucleic acid substrates, consistent with longer substrates stabilizing the closed conformation somewhat more than shorter substrates. In contrast, the resonance for the more distal Met-92 is minimally perturbed by ligand addition, whereas Met-61 exhibits a complex pattern of shift perturbations likely due to its close proximity to the ligands (Fig. 1B).
Importantly, partial saturation with 2OG, Suc, or nucleic acid substrate yields resolved resonances for unbound versus bound AlkB molecules (Fig. 6). This observation indicates that ligand off-rates (k off ) are slow on the NMR time scale and that FIGURE 10. NMR spectra of the L99M and I119M mutants verify that the conformation of the Fe(II)/2OG dioxygenase core is not significantly influenced by the conformational transition in the NRL. Two-dimensional spectra of WT (A-C), L99M (D-F), or I119M (G-I) enzymes selectively labeled with [ 13 C ⑀ ]methionine were collected using a 1 H-13 C HSQC experiment from samples at 10°C in 75 mM KCl, 50 mM NaH 2 PO 4 , pD 5.5. Unlabeled resonances in the Zn(II)/2OG and Zn(II)/2OG/5Ј-CAmAAT-3Ј complexes are caused by the presence of trace metals, which have been verified not to affect resonance assignments. A, D, and G, spectra of the apoenzymes shown in gray. B, E, and H, spectra of the enzymes in the presence of 1.5ϫ Zn(II) and 10ϫ 2OG shown in green. C, F, and I. spectra of the enzymes in the presence of 1.5ϫ Zn(II), 10ϫ 2OG, and 1.5ϫ 5Ј-CAmAAT-3Ј shown in blue.
the observed chemical shift perturbations reflect the relative populations of the open and closed conformations of AlkB in the bound state, rather than ligand-binding kinetics. The intermediate chemical shifts of the 2OG and Suc complexes indicate that the enzyme exchanges between these two conformations rapidly on the NMR time scale in the presence of either ligand. Furthermore, the similarity of the chemical shifts when saturated with either compound indicates that 2OG and Suc stabilize the closed state of AlkB to a similar extent, which conflicts with an earlier claim (32,35) but is consistent with the need to sequester the oxyferryl intermediate after Suc generation.
Equivalent results are observed when [ 13 C ⑀ ]methionine NMR measurements are performed in the presence of partially versus fully saturating concentrations of 2OG or methylated DNA substrate either in our standard NMR buffer containing 50 mM NaH 2 PO 4 at pD 5.5 or in an otherwise identical buffer containing 50 mM Na-HEPES-d 12 at pD 7.6 (gray versus red contours in Fig. 6, A, B, E, and F). Therefore, the ligand-dependent conformational dynamics of AlkB are similar in both buffers, consistent with the comparable ligand-binding affinities ( Table 1) and enzymological parameters (data not shown) of the enzyme in the two environments.
The picosecond-to-nanosecond time scale dynamics of [ 13 C ⑀ ]Met residues were quantified via the 1 H-1 H cross-correlated relaxation interference rate constant, ϰ S 2 ϫ C , with S 2 being the generalized order parameter of the methyl symmetry axis and C being the protein rotational correlation time (52). Values of are low for the Met residues in and near the NRL of AlkB in the apo/open state, confirming that these side chains are disordered, but their values increase upon ligand addition ( Fig. 5C and Table 2). Values of for Met-49 are linearly correlated with 13 C chemical shifts in the Zn(II), 2OG, and 5Ј-CA-mAAT-3Ј complexes (R 2 ϭ 0.996, Fig. 7), consistent with reduction in its mobility in the closed conformation. With methylated DNA substrate bound, the more interior residues Met-57 and Met-61 have similar values to Met-92, which is relatively ordered in all states.

NMR Measurements of Exchange Rates between Open and Closed States-Exchange kinetics between open and closed
conformations were quantified using a Hahn-echo experiment (53) to measure the zero-and double-quantum relaxation rate constants, R ZQ and R DQ , which are reported as R MQ ϭ (R DQ ؉ R ZQ )/2 and ⌬R MQ ϭ (R DQ Ϫ R ZQ )/2 ( Fig. 5D and Table 3). Assuming that the 5Ј-CAmAAT-3Ј complex is fully closed, the ratio R MQ /⌬R MQ ϭ 1.18 for Met-49 in the 2OG complex (i.e. based on chemical exchange rate constants only) is similar to the ratio predicted from the 1 H and 13 C chemical shift changes between the Zn(II) and 5Ј-CAmAAT-3Ј complexes, which is 1.11. This correspondence provides further support for the two-state transition model. The Met-49 chemical shift and relaxation data were fit to a two-site exchange model ( Fig. 5E and Equation 4) to determine the chemical shift for the open state (16.6 ppm) and the rate constant for the transition to the closed state (k 1 ϭ 12,700 Ϯ 200 s Ϫ1 ), assuming that k 1 is the same for all complexes, as previously observed for other enzymes (42,68). This model gives the fractions of the closed state as 0.16, 0.65, and ϳ1 in the Zn(II), Zn(II)/2OG, and Zn(II)/ 2OG/5Ј-CAmAAT-3Ј complexes, respectively. Alternatively, broadening in the Zn(II) state may reflect chemical exchange within the open-state ensemble, rather than transitions to the closed state. This model gives k 1 ϭ 8,600 Ϯ 300 s Ϫ1 and the fractions of the closed state as ϳ0, 0.59, and ϳ1, respectively. As expected under the two-state transition model (Fig. 1D), both calculations show a substantial increase in the closed-state population in the presence of 2OG, and this increase accounts energetically for at least half of the observed increase in the affinity for the methylated DNA substrate in the presence of 2OG ( Table 1).
The two-state model is further supported by the linear correlations observed for the Met-49 and Met-57 13 C ⑀ chemical shifts in wild-type AlkB compared with three of the four single Met-to-Leu mutants (Figs. 5F and 9). The slopes of the correlation plots for these two residues are near unity in the M49L, M57L, and M92L mutant proteins. In contrast, in the M61L

TABLE 3 Transverse relaxation rate constants for AlkB-⌬N11/ligand complexes
The zero-and double-quantum relaxation rate constants (R ZQ and R DQ , respectively) were measured using a Hahn-echo experiment (53). As in Table 2 (Fig. 4). Therefore, the W89Y mutation opposes shifting of the conformational ensemble toward the closed state. Accelerated Release of DNA Substrate Causes Uncoupled Turnover of 2OG-Perturbation of the thermodynamics of the conformational transition by W89Y provided an opportunity to evaluate the transition's influence on catalytic efficiency. To determine the degree of coupling of the two stages of the enzymatic reaction, i.e. 2OG decarboxylation and subsequent DNA FIGURE 11. Representative HPLC assays of 2OG and DNA turnover. Decarboxylation of 2OG co-substrate (top) and demethylation of DNA substrates (bottom) in standard reaction buffer at 37°C were quantified in parallel using Aminex HPX-87H and Phenomenex Luna C18 columns, respectively. Because the extinction coefficient of the Suc co-product at 210 nm is too low for reliable quantification, decarboxylation of 2OG was measured based on its depletion compared with pyruvate, an internal control compound that is not chemically modified during incubation or analysis. The chromatograms labeled "-control" come from mock reactions incubated in the absence of enzyme. At every time point, greater 2OG decarboxylation and less DNA demethylation was observed for the W89Y mutant relative to the WT enzyme. Therefore, the W89Y mutation substantially increases the oxidation of 2OG uncoupled from demethylation of the shorter trimer DNA substrate, which is released from the enzyme more rapidly than the longer pentamer DNA substrate (Fig. 13, D and E). Reactions initially containing 150 M 2OG and 30 M 5Ј-TmAT-3Ј were incubated for 5 (pink), 10 (blue), or 30 (green) minutes. The dashed line shows data from a mock (control) reaction without enzyme.

TABLE 4 Catalytic parameters of WT and mutant AlkB-⌬N11 enzymes
The k cat and K m values were determined by Michaelis-Menten analysis. The k off values were measured by monitoring fluorescence anisotropy during competitive DNA-displacement experiments (e.g. as shown in Fig. 13, A-E). Michaelis-Menten analyses and DNA substrate release measurements were performed in the standard assay buffer at 10°C. The 2OG/DNA coupling ratio (i.e. the ratio of 2OG co-substrate turnover to DNA substrate turnover) was measured at the end of a 20-min incubation period for pentamer DNA substrate or a 30-min incubation period for trimer DNA substrate at 37°C.  Fig. 12, A-C). Fluorescence anisotropy assays of the release rate of methylated DNA substrate from the Mn(II)/ Suc complex demonstrate that trimer substrate is released from the co-product complex ϳ60-fold more rapidly than pentamer substrate (Fig. 13, D and E, and Table 4), and this accelerated release accounts for the premature quenching of the oxyferryl intermediate and uncoupled 2OG turnover in ϳ80% of oxidation reactions that occur when trimer substrate is bound. These observations indicate that the lifetime of the oxyferryl intermediate is Ն11 s (k cat for 5Ј-CAmCAT-3Ј (31)) and that the  enzyme releases longer DNA substrates more slowly than the forward reaction rate with methylated nucleobase (Figs. 1D and 13, D-F). Notably, the Mn(II)/Suc complex does not release the pentamer substrate more rapidly than the Mn(II)/2OG complex (Fig. 14), which is contrary to previous claims (32,35) but is consistent with the need to sequester the oxyferryl intermediate in the active site after Suc generation. Protein Dynamics Control the Catalytic Efficiency of AlkB-The observation that the release rate of the DNA substrate is a critical determinant of catalytic efficiency provides a sensitive starting point for analysis of the influence of protein mutations on this process. The M61L mutation, which changes a residue directly contacting the DNA substrate, substantially accelerates DNA substrate release ( Fig. 13C and Table 4) while only slightly perturbing the thermodynamics of the functional conformational transition characterized above (Figs. 5F and 9). This point mutation substantially increases turnover of 2OG uncoupled from DNA demethylation (Fig. 12, A-C), confirming that premature release of methylated substrate causes uncoupling. Strikingly, a similar increase in DNA release rate (Fig. 13, A and B, and Table 4) and uncoupled 2OG turnover (Fig. 12, A-C) is caused by the W89Y mutation, which is remote from the active site ( Fig. 1B) but strongly biases the enzyme conformation toward the open conformation ( Figs. 4 and 8). In contrast, the mutation of methionine residue Met-92, which is located in a region of the protein that is not significantly affected during the ligand-dependent conformational transition (Fig. 5A), alters neither the rate of DNA release nor the efficiency of coupling between 2OG turnover and DNA repair ( Fig. 15 and Table 4). These results demonstrate that, in addition to controlling the proper order of co-substrate/substrate binding (Fig. 3), the conformational transition between open/closed states of the NRL determines the DNA release rate, which critically modulates the catalytic efficiency of AlkB. If this transition were too slow, the rate of DNA repair would be limited due to slow product release. Our results demonstrate that, if this transition is too fast, catalytic efficiency is reduced because of increased uncoupled turnover of the 2OG co-substrate. Therefore, the kinetics of the conformational transition gating DNA substrate binding have evolved to control sequestration of the reactive oxyferryl intermediate and the net efficiency of the catalytic reaction cycle of AlkB.

DISCUSSION
Like other Fe(II)/2OG dioxygenases, AlkB performs a high energy chemical transformation requiring coordination of the binding of four cofactor/substrate molecules followed by sequestration of a highly reactive enzyme-bound intermediate (2). Previous research suggested that the dynamics of AlkB contribute to several stages in its reaction cycle, including binding of structurally diverse substrates (23), coupling between successive chemical steps in the reaction cycle (23), and facilitating release of the repaired DNA product (32). In this study, by combining fluorescence spectroscopy and NMR spectroscopy with enzymological assays, we demonstrate that a specific protein conformational transition orchestrates the complex multistep catalytic reaction cycle of AlkB (Fig. 1D), controlling both the proper sequential order of cofactor/substrate binding as well as the kinetic sequestration of the reactive oxyferryl intermediate.
This transition changes AlkB from an open conformation that has yet to be observed crystallographically to a closed conformation resembling that in substrate-and product-bound crystal structures (14,23,31). Our spectroscopic data demonstrate that the open conformation has similar net secondary structure content as measured by CD (Fig. 2, A and C) but an altered environment of the Trp and Met residues proximal to the active site and at the interface between the Fe(II)/2OG core and the NRL (Figs. 3, A-D, and 4). Previous amide-exchange measurements showed that the backbone of the NRL has enhanced dynamics in the open conformation (23), and the NMR data presented here show that its methionine side chains are disordered in this conformation (Fig. 5). Therefore, the key conformational transition controlling the catalytic reaction cycle of AlkB involves movement of the NRL away from the active site into a more open position where it participates in FIGURE 14. Oxidation of 2OG to Suc does not enhance the release rate of methylated DNA substrate. The release rate of 5Ј-CAmCAT-FAM-3Ј from WT AlkB in complex with Suc/Mn(II) (A) or 2OG/Mn(II) (B) was measured at 10°C in the standard buffer containing 20 M MnCl 2 and either 2 mM Suc (A) or 100 M 2OG (B). These competitive binding experiments were carried out as described for Fig. 13, A-D. Based on several replicates of each experiment, the DNA release rate is 3.4 Ϯ 0.2 min Ϫ1 from the Mn(II)/2OG complex and 1.80 Ϯ 0.02 min Ϫ1 from the Mn(II)/Suc complex. FIGURE 15. AlkB-⌬N11 mutants with unaffected 2OG/DNA coupling efficiency have DNA substrate release rates similar to the wild-type enzyme. A, coupling of 2OG oxidation to 5Ј-TmAT-3Ј demethylation by WT (black), M92L (red), or I119M (blue) enzymes was quantified as shown in Fig. 11 in reactions conducted at 37°C in the standard assay buffer. Although some uncoupled decarboxylation of the 2OG co-substrate occurs in the absence of alkylated DNA substrate, its addition substantially increases the rate of this reaction (12,13). B, fluorescence anisotropy assays were used to measure the release rate of 5Ј-CAmCAT-FAM-3Ј from WT (left) or M92L (right) enzyme at 10°C in standard buffer containing 20 M Mn(II) and 2 mM Suc. Competitive binding experiments were conducted as described for Fig. 13. fewer constraining interactions with the Fe(II)/2OG core (as schematized in Fig. 1C).
In the metal-bound enzyme, this open state is favored over the closed state by a factor of ϳ5:1, as elucidated by the NMR data presented here. Binding of 2OG or Suc changes this ratio to ϳ0.5:1, which importantly increases the affinity for DNA substrate. Although some variation in the exact ratio is possible with co-substrate versus co-product in the buffers at acidic or neutral pH, our 1 H-13 C chemical shift data (Figs. [5][6][7][8] suggest that any such variations are relatively small. Using these data, the estimated proportion of the closed state in the Zn(II)/2OG complex is 0.64 Ϯ 0.05 at pD 5.5 and 0.59 Ϯ 0.06 at pD 7.6, and the estimated proportion of the closed state in the Zn(II)/Suc complex is 0.61 Ϯ 0.05 at pD 5.5.
Comparing the apoenzyme (i.e. with neither metal cofactor nor 2OG co-substrate bound) with the metal/2OG complex, the affinity for the DNA substrate is increased by ϳ100-fold by Fe(II)/2OG binding, as demonstrated by our fluorescence binding experiments. Thus, the transition toward the closed state controls the proper sequential order of substrate binding. We also demonstrate that DNA binding greatly stabilizes the closed state and that mutations that bias the transition toward the open state substantially accelerate DNA substrate release and enhance uncoupled turnover of 2OG. Therefore, the kinetics of the conformational transition gating DNA binding directly control sequestration of the reactive oxyferryl intermediate and thereby the net efficiency of the catalytic reaction cycle. In wild-type AlkB, the opening kinetics are tuned to mediate release of the DNA substrate/product slightly more slowly than hydrogen abstraction from the primary substrate, to produce efficient repair of this substrate without promoting uncoupled turnover of 2OG and adventitious oxidative side reactions. Dynamic conformational transitions of this kind are likely to control the catalytic efficiency of other Fe(II)/2OG dioxygenases as well as many other enzyme families performing complex biomolecular transformations.