Tuning of pKa values activates substrates in flavin-dependent aromatic hydroxylases

Hydroxylation of substituted phenols by flavin-dependent monooxygenases is the first step of their biotransformation in various microorganisms. The reaction is thought to proceed via electrophilic aromatic substitution, catalyzed by enzymatic deprotonation of substrate, in single-component hydroxylases that use flavin as a cofactor (group A). However, two-component hydroxylases (group D), which use reduced flavin as a co-substrate, are less amenable to spectroscopic investigation. Herein, we employed 19F NMR in conjunction with fluorinated substrate analogs to directly measure pKa values and to monitor protein events in hydroxylase active sites. We found that the single-component monooxygenase 3-hydroxybenzoate 6-hydroxylase (3HB6H) depresses the pKa of the bound substrate analog 4-fluoro-3-hydroxybenzoate (4F3HB) by 1.6 pH units, consistent with previously proposed mechanisms. 19F NMR was applied anaerobically to the two-component monooxygenase 4-hydroxyphenylacetate 3-hydroxylase (HPAH), revealing depression of the pKa of 3-fluoro-4-hydroxyphenylacetate by 2.5 pH units upon binding to the C2 component of HPAH. 19F NMR also revealed a pKa of 8.7 ± 0.05 that we attributed to an active-site residue involved in deprotonating bound substrate, and assigned to His-120 based on studies of protein variants. Thus, in both types of hydroxylases, we confirmed that binding favors the phenolate form of substrate. The 9 and 14 kJ/mol magnitudes of the effects for 3HB6H and HPAH-C2, respectively, are consistent with pKa tuning by one or more H-bonding interactions. Our implementation of 19F NMR in anaerobic samples is applicable to other two-component flavin-dependent hydroxylases and promises to expand our understanding of their catalytic mechanisms.


Hydroxylation of substituted phenols by flavin-dependent monooxygenases is the first step of their biotransformation in various microorganisms. The reaction is thought to proceed via electrophilic aromatic substitution, catalyzed by enzymatic deprotonation of substrate, in single-component hydroxylases that use flavin as a cofactor (group A). However, two-component hydroxylases (group D)
, which use reduced flavin as a co-substrate, are less amenable to spectroscopic investigation. Herein, we employed 19 F NMR in conjunction with fluorinated substrate analogs to directly measure pK a values and to monitor protein events in hydroxylase active sites. We found that the single-component monooxygenase 3-hydroxybenzoate 6-hydroxylase (3HB6H) depresses the pK a of the bound substrate analog 4-fluoro-3-hydroxybenzoate (4F3HB) by 1.6 pH units, consistent with previously proposed mechanisms. 19 F NMR was applied anaerobically to the two-component monooxygenase 4-hydroxyphenylacetate 3-hydroxylase (HPAH), revealing depression of the pK a of 3-fluoro-4-hydroxyphenylacetate by 2.5 pH units upon binding to the C 2 component of HPAH. 19 F NMR also revealed a pK a of 8.7 ؎ 0.05 that we attributed to an active-site residue involved in deprotonating bound substrate, and assigned to His-120 based on studies of protein variants. Thus, in both types of hydroxylases, we confirmed that binding favors the phenolate form of substrate. The 9 and 14 kJ/mol magnitudes of the effects for 3HB6H and HPAH-C 2 , respectively, are consistent with pK a tuning by one or more H-bonding interactions. Our implementation of 19

F NMR in anaerobic samples is applicable to other two-component flavin-dependent hydroxylases and promises to expand our understanding of their catalytic mechanisms.
Hydroxylation of aromatic compounds is important in catabolism and transformation of xenobiotics, for a wide variety of organisms. Such reactions are mostly catalyzed by metal-, pterin-, and flavin-dependent enzymes (1)(2)(3)(4)(5). These enzymes also metabolize drugs and thereby modify their therapeutic effects (4). Enzymatic hydroxylation reactions are important for synthesis of fine chemicals and have been employed in various biocatalytic applications (6 -10). The enzymes are able to activate molecular oxygen (O 2 ) and control the fate of the resulting reactive oxygen intermediates. Flavin-dependent monooxygenases catalyze aromatic hydroxylation by forming a reactive flavin-hydroperoxide adduct that can add a hydroxyl group to aromatic rings (11). Understanding how hydroxylases control the reactivity of activated O 2 is both fundamentally important and potentially valuable for improving the utility of hydroxylases for practical applications.
There are two broad categories of flavin-dependent aromatic hydroxylases: single-component hydroxylases where flavin reduction and oxygenation occur on the same polypeptide chain, and two-component hydroxylases that require separate flavin reductase and substrate oxygenase components (11)(12)(13). Members of the first type utilize flavin as a cofactor, which remains bound throughout the catalytic cycle, and among the second type the oxygenase component binds reduced flavin as a substrate and releases oxidized flavin as a product. Both types of hydroxylases catalyze the reaction of reduced flavin with O 2 to form a reactive C4a-hydroperoxyflavin (C4a-OOH) 4 intermediate that reconciles the triplet nature of O 2 with the singlet substrates and products (14,15) and hydroxylates phenolic substrates (16).
For pyranose 2-oxidase and the oxygenase component (C 2 ) of HPAH, density functional theory and transient kinetics indicate that reduced flavin reacts with O 2 via proton-coupled electron transfer to form a radical pair consisting of flavin semiquinone and ⅐ OOH radicals, before recombination of the radical pair to form the C4a-OOH hydroperoxyflavin adduct (17)(18)(19). Subsequent hydroxylation of phenolic substrates is thought to occur via electrophilic aromatic substitution, based on theoretical calculations (20), and because the rates of hydroxylation are correlated with the favorability of flavin C4a-alkoxide formation, suggesting that this is a leaving group of the hydroxylation step (21,22). However, recent quantum mechanical/molecular mechanical (QM/MM) calculations suggest that a hydroxyl radicalcoupled electron transfer mechanism applies in 3-methylorcinolaldehyde monooxygenase from Talaromyces stipitatus (23).
Flavin-based hydroxylases are thought to activate their substrates for electrophilic attack by deprotonating them (24 -30). Substrate binding to para-hydroxybenzoate hydroxylase appears to lower the substrate pK a by approximately two pH units (27,29,31). Similarly, substrate is deprotonated upon binding to 3-hydroxybenzoate (3HB) 6-hydroxylase (3HB6H), wherein a nearby His residue is proposed to serve as the base (32). NMR studies found that substrate is (partially) deprotonated to the phenolate when bound to reduced phenol hydroxylase (33). Similarly, 2-methyl-3-hydroxypyridine-5-carboxylate oxygenase appears to preferentially bind its substrate as the zwitterionic form in which the phenol group is deprotonated (34). Thus, stabilization of bound phenolates is well-established among single-component hydroxylases, and these therefore provide good systems in which to validate new methods for detecting and quantifying pK a depression applied to substrate upon binding. However, novel approaches are sorely needed to investigate two-component hydroxylases.
Much less is known about two-component hydroxylases because they bind substrates when their flavin is reduced and spectroscopically muted. Nevertheless, a low substrate pK a was found to correlate with rapid hydroxylation by HadA (a dehalogenating monooxygenase), implying that substrate deprotonation is a rate-contributing step (35). Similarly, the phenolic pK a of substrate bound to C 2 may be unusually low (Fig. 1) because the hydroxylation rate constant and product yield are independent of pH from pH 6 to 10 (36). Active-site mutants that preserve positive charge at the position of His-120 retain activity consistent with a mechanism involving the phenolate form of substrate (37). However, quantitative studies of the effect have yet to be performed. There is a need for a method able to discern the protonation state of substrate when bound in the enzyme-active site, while maintaining an inert atmosphere.
Ideally, one would like to know the pK a value of a substrate in the presence of the flavin C4a-hydroperoxide; however, the latter is very short-lived in the presence of substrate. As a proxy, we have studied the FMNH Ϫ -bound state of the C 2 oxygenase, in which we can assume that the enzyme has adopted the conformation in which formation and reaction of the hydroperoxide occurs.
To directly observe the influence of the active-site environment on the pK a values of bound substrates, we used 19 F NMR. 19 F NMR provides high sensitivity comparable with 1 H NMR, but with superior responsiveness to changes in the environment as well as effects of deprotonation of the fluorinated molecule itself (38). Moreover, with 19 F incorporated only in the substrate analogs, the 19 F NMR spectra are not complicated by contributions from the protein or the buffer, so we could replicate conditions under which turnover could occur if O 2 were provided, without regard for the complexity of the 1 H spectra (Fig. S1A). Although individual pK a values of the mono-fluorinated substrates are depressed somewhat by their fluorine substitutions, this should not greatly affect the change in pK a value associated with binding to enzyme. However, it is important to check that use of the analog does not result in mechanistic changes (39,40).
We validated the method by applying it to the fluorinated substrate 4F3HB that binds to the single-component hydroxylase 3HB6H. We then used it to characterize the state of the substrate 3F4HPA when bound to the C 2 component of HPAH. By using internal pH indicators monitored by 1 H NMR, we eliminated use of a pH electrode. This, in turn, enabled the use of a septum-sealed anaerobic NMR tube and observation of the reduced (FMNH Ϫ -bound) state of C 2 , which is the state in which FMN binds to the protein (41). Because the NMR chemical shift indicators are small molecules, they can easily be identified by their sharp lines despite the presence of equal or greater concentrations of protons from the protein (Fig. S1A). This combination of 19 F NMR to observe the bound substrate, with 1 H NMR to measure the pH (Fig. S1B), permits monitoring of pH titrations of bound (and free) ligand to test the hypothesis that the pK a will be depressed upon binding to the enzyme.
The results indicate that the phenolate state of 4F3HB is stabilized by 1.6 pH units upon binding to 3HB6H, consistent with prior work (32). The method was then applied to a two-component hydroxylase, which proved more complicated. The dissociation constant describing 3F4HPA binding to C 2 was found to be strongly pH-dependent consistent with preferential binding of the phenolate state to protonated protein and a pK a of 6.49 for bound 3F4HPA. However, contributions from an additional ionization event attributable to the protein were also detected via 19 F chemical shifts and were tentatively assigned to an active-site histidine via specific amino acid substitutions.
Thus, a one-and a two-component hydroxylase are unified mechanistically in that both deprotonate substrate as part of binding, and we have quantified the extent to which the deprotonated form is stabilized. Although much is known about single-component hydroxylases in this regard, two-component hydroxylases are less well-documented. In this example, the exquisite sensitivity of the 19 F NMR chemical shift enabled us to observe deprotonation of substrate analogs and also to infer the occurrence of additional deprotonation events in the active site. This approach can have great value in additional members of this fascinating and useful enzyme family.

Single-component hydroxylase: pH dependence of 4F3HB alone
Before studying the fluorinated substrate (ligand) when bound to enzyme, we first characterized the free ligand to obtain a reference point. The 19 F chemical shift of 4F3HB dissolved in water was measured at 0.3 pH unit intervals from pH 2.0 to 12.0 and plotted as a function of pH (Fig. 2). Two separate transitions could each be described by the Henderson-Hasselbalch equation (see Equation 3), yielding pK a values of 4.1 Ϯ 0.01 and 8.7 Ϯ 0.03 in excellent agreement with those obtained via titrations monitored by UV-visible spectrophotometry (4.1 Ϯ 0.05 and 8.5 Ϯ 0.01, see Fig. S2, uncertainties quoted are errors of the fits). The identities of the events responsible for the pK a values were determined by 13 C NMR (Fig. 2B).
To evaluate the extent to which the presence of a fluorine affects the pK a values, we also determined the pK a values of authentic substrate, obtaining values of 4.2 Ϯ 0.08 and 10.2 Ϯ 0.08 (Fig. S2) in agreement with literature (pK a, 1 ϭ 4.0 and pK a, 2 ϭ 9.7 (32)). The carboxyl pK a value is thus relatively unaffected; however, the phenolic pK a of the 4F3HB is 1.5 pH units lower than that of 3HB, consistent with electron withdrawal from the ring by the electronegative fluorine substituent (39). Our use of 4F3HB will therefore result in a lower phenolic pK a value relative to authentic substrate, but nonetheless it provides an informative probe of effects of the 3HB6H active site on its bound substrate.
Crucially, these controls demonstrate 19 F NMR's ability to detect deprotonation of each of the compound's ionizable groups and to distinguish between the two events ( Fig. 2A): deprotonation of the phenol group produced a small shift of the 19 F resonance to a less-negative chemical shift (0.32 ppm), and deprotonating the carboxyl group resulted in a large change in chemical shift to more negative values (Ϫ4 ppm). Thus, the 19 F chemical shift displays a nuanced response not just to accumulating negative charge via the inductive effect, but also to interactions through space and ortho effects (42,43), producing an exquisitely responsive probe of local events (38).

pH dependence of the 19 F chemical shift of 4F3HB in the presence of WT 3HB6H
For 4F3HB in the presence of a 2-fold excess of WT 3HB6H at pH 6.5 (Fig. 3A), the 19 F resonance was shifted Ϫ0.20 ppm to a more negative chemical shift relative to its position in the

Activation of bound substrates by deprotonation
absence of enzyme, and it was significantly broadened consistent with a larger correlation time for reorientation as part of a much larger entity (44).
The sample's pH was adjusted from 6.5 to 9.0 in small increments by repeated additions of 0.2 M KOH and from pH 10.0 back to 6.5 by additions of 1 M MES. At each point, the sample pH was determined from the 1 H chemical shifts of the indicator molecules, and the 19 F NMR spectrum of the ligand was collected. At pH values above pH 9.0, an additional sharper signal was observed (inset of Fig. 3A). This was determined to be free 4F3HB by addition of excess 4F3HB to the enzyme solution, resulting in a single stronger sharp signal at an unchanged chemical shift. Appearance of free ligand was accompanied by protein precipitation, indicating that liberation of the ligand was due to protein denaturation rather than elevation of the dissociation constant (K d ). No further measurements were made on such samples.
A plot of the chemical shift of bound 4F3HB versus pH and fitting with the Henderson-Hasselbalch equation (Equation 3) yielded a pK a of 7.1 Ϯ 0.2 (Fig. 3B). We attribute this to ioniza-tion of the phenolic OH because it is associated with a 0.27 ppm change to a less negative chemical shift similar to the higher-pH ionization of free 4F3HB (Fig. 3B, blue curve). Thus, binding to the enzyme depressed the phenolic pK a of 4F3HB by 1.6 pH units relative to that of free 4F3HB (8.7 Ϯ 0.03).

Binding of 4F3HB versus pH
A binding site that lowers the ligand pK a favors deprotonation of the ligand. Therefore, one predicts that the fluorinated analog with its intrinsically greater tendency to deprotonation should bind more readily than authentic substrate. Indeed, although 3HB's K d of 0.05-0.16 mM for binding to 3HB6H (45,46) predicts that some 60% of substrate would be proteinbound at the concentrations we used, the absence of any signal from free 4F3HB below pH 9.0 argues that this analog binds more tightly than does 3HB. This is consistent with 4F3HB's lower phenolic pK a and preferential binding of the deprotonated phenolate form.
Having confirmed the expected behavior of a single-component hydroxylase using 19 F NMR, we applied the method to a two-component hydroxylase.  13 C NMR was used to identify the functional group whose ionization state changed with each pK a . The carboxyl carbon near 180 ppm was the most affected by changing pH values from below to above the pK a of 4.2, whereas the phenolic carbon that moves from near 146 ppm to near 156 ppm was the most affected by changing the pH from below to above the pK a of 8.7. Thus, the data assign the pK a of 4.1 to deprotonation of the carboxyl and the pK a of 8.7 to deprotonation of the phenol. The signal near 135 ppm at pH 7.1 belongs to the carbon that subtends the carboxyl and the split signal near 158 ppm at pH 12.8 subtends the fluorine.  19 F signal of 4F3HB bound to 3HB6H at pH 6.5 (orange spectrum) and pH 9.0 (light blue spectrum), and the simulations of the NMR lines are shown in red and blue lines for the pH 6.5 and pH 9.0 spectra, respectively. Only bound ligand was observed from pH 6.0 to 9.0. However, the inset shows the 19   As a model for the native substrate p-hydroxyphenylacetate (4HPA), we used 3F4HPA. This compound is itself a good substrate that reacts with the C 2 component in ways that resemble those of authentic substrate. Pre-steady-state kinetics of the reaction of reduced C 2 and 3F4HPA with a series of oxygen concentrations were examined by stopped-flow spectrophotometry. Absorbance changes at 380 and 450 nm (black and blue kinetic traces, respectively) revealed four phases for the reaction with 3F4HPA, similar to the reaction of C 2 with 4HPA (compare Fig. 4 with Fig. 5 in Ref. 41). The first and second phases were dependent on the oxygen concentration, indicating that they reflect formation of a C4a-hydroperoxy-FMN intermediate.

Activation of bound substrates by deprotonation
The second-order rate constants of the first and second phases were calculated from the slopes of the plots of k obs versus oxygen concentration and found to be 1.16 ϫ 10 6 and 8.6 ϫ 10 4 M Ϫ1 s Ϫ1 , respectively. These are similar to the rate constants measured for the reaction of authentic substrate with C 2 (1.1 ϫ 10 6 and 4.8 ϫ 10 4 M Ϫ1 s Ϫ1 in the absence and presence of 4HPA, respectively (41)). Thus, the presence of the fluorinated analog decreased the rate constant of C4a-hydroperoxy-FMN formation as does the authentic substrate, and it appears that 3F4HPA, like 4HPA, may crowd the active site and decrease O 2 's access to reduced FMN (47). This further suggests that 3F4HPA binds in a position similar to that occupied by native substrate. Finally, at high concentrations, 3F4HPA inhibited dehydration of C4a-hydroxy-FMN, similar to the effect of 4HPA ( Fig. 4C) (36). Thus, our pre-steady-state characterizations affirm that 3F4HPA reproduces the environment and perturbations associated with bound substrate.

pH dependence of free 3F4HPA
As a reference point for our 19 F NMR studies of 3F4HPA bound to C 2 , we measured the 19 F chemical shift of the free compound as a function of pH. Fig. 5A shows two discrete transitions that were each described by the Henderson-Hasselbalch equation (Equation 3) and yielded pK a values of 4.2 Ϯ 0.04 and 9.0 Ϯ 0.02, respectively, in excellent agreement with those obtained via UV-visible spectrophotometry (4.0 Ϯ 0.05 and 8.9 Ϯ 0.01, respectively, see Fig. S3). Similar to the case of 4F3HB, the phenolic pK a of 3F4HPA was 1.6 pH units lower than that of nonfluorinated substrate, although the carboxylic pK a was much less affected (Fig. S3A). As for 4F3HB, deprotonation of the p-hydroxyl group of 3F4HPA resulted in a small shift in the 19 F resonance to less negative chemical shifts, in contrast with the effect of deprotonating the acetic acid group.

pH dependence of the 19 F signal of 3F4HPA in the presence of WT HPAH-C 2
To learn whether the pK a value of 3F4HPA is affected by interactions with the enzyme-active site, we compared the pH dependence of bound 3F4HPA with that of free 3F4HPA. The pH of an NMR sample of 3F4HPA with 2.5-fold excess WT-C 2 and FMNH Ϫ was adjusted from 6.6 to 10.0 as described above, except that inert atmosphere was maintained. The highest pH investigated was pH 10.0 because C 2 denatures above this pH. The 19 F NMR spectrum exhibited a sharp signal from the unbound fraction at Ϫ139.312 ppm (pH 6.6) to Ϫ139.155 ppm (pH 10.0) in addition to a broad signal from the bound fraction at Ϫ137.959 (pH 6.6) to Ϫ139.692 ppm (pH 10.0) (Fig. 6). At pH 6.6 the bound ligand 19 F chemical shift was 1.35 ppm less negative than that of the free ligand, but in the course of the titration to high pH it moved to Ϫ0.54 ppm more negative in 19 F chemical shift relative to free ligand (pH 10.0). Thus, interaction with the protein had a significant effect on the ligand 19 F chemical shift, and the effect had the opposite sign at high pH from at low pH, indicating that the nature of the interaction differs depending on the pH. This is in addition to, and different from, the effect of deprotonating the ligand, and therefore it suggests an additional pH dependence residing in the protein.
The 19 F chemical shifts of bound and free 3F4HPA were plotted against pH (Fig. 6B, black and blue, respectively). The pK a values describing each were obtained from nonlinear leastsquare fits using the Henderson-Hasselbalch equation (Equation 3). Back-titration by addition of HCl gave the same results as those observed by addition of KOH indicating that the spectral changes were not due to increasing ionic strength in the course of the titrations nor an irreversible event such as protein denaturation. For free 3F4HPA, the chemical shift at each pH, the shape of the titration curve, and the pK LH value of 8.9 Ϯ 0.09 were within error of those associated with the deprotonation of the p-hydroxyl group of 3F4HPA in the absence of C 2 (9.0 Ϯ 0.02) (Fig. 5A).
The bound form of 3F4HPA behaved differently. The observed pK a value of 8.7 Ϯ 0.03 was similar to the pK a of phenol deprotonation for free 3F4HPA, but the 19 F chemical shift change associated with the event was opposite in sign and much larger (Ϫ1.78 ppm versus ϩ0.2 ppm for free 3F4HPA, ␦ L Ϫ ␦ LH ). Indeed, the chemical shift change was more similar to the signature of deprotonating the carboxylic acid (Fig. 5A).
We considered the possibilities that the pK a of 8.7 represents the following: (a) deprotonation of the bound ligand's phenol (as in 3HB6H); (b) deprotonation of the bound ligand's carboxyl (based on the chemical shift signature); or (c) response of the ligand's 19 F chemical shift to deprotonation of a protein residue nearby ("X") (48). To arbitrate among the possibilities, we exploited the pH dependence of the apparent dissociation constant K d, obs , as it provides an independent report on deprotonation events affecting the interaction between ligand and protein (Scheme 1). Because each of the pK a values of free ligand have been assigned, the pH dependence of K d, obs can be used to test whether the pK a of 8.7 reflects either pK a of the bound ligand, where a given pK a of free ligand is pK LH , and the corresponding pK a of the bound ligand is pK LH(EH) .
Because binding of 3F4HPA to WT-C 2 increased as the pH increased ( Fig. 6A), we employed notation emphasizing binding of the deprotonated ligand, which in turn will interact more favorably with the positively-charged (protonated) enzyme active-site residue X. However, both protonation states of both Activation of bound substrates by deprotonation groups are accounted for by the algebra employed below in Equations 4 and 5.
The predictions of scenarios a and b are plotted in Fig. S4 for comparison with the actual pH dependence of K d, obs values. Neither are compatible with the data. Therefore, the simple model considering only the protonation state of the ligand is unable to account for the pK a of 8.7, and we must adopt the model that it reflects a protein residue. Thus, we now consider that there are two ionizable residues: the ligand phenol and X. This yields the expanded Scheme 2, which posits that the interaction will be favorable when the ligand is deprotonated and X is protonated (Fig. S5). X's pK a can then be extracted from the pH dependence of the chemical shift of bound ligand (pK (L)EH in Scheme 2), whereas the pK a of the bound ligand's phenol can be extracted from the pH dependence of K d, obs (Equations 4 and 5).
Fits to the pH dependences of the chemical shift of bound ligand ␦ obs (pH) and the K d, obs (pH) indeed yielded two different pK a values. This is not unexpected; the chemical shift of bound 3F4HPA only reflects the states in which ligand is bound, and the K d, obs reflects equilibria with states of free ligand and free enzyme as well (Schemes 1 and 2, and Equations 1 and 5).
A fit to the K d, obs data assuming only one labile proton (e.g. pK a values for ligand, bound and free, Equation 1) and fixing the

Activation of bound substrates by deprotonation
pK a free ligand at pK LH ϭ 9.0 (above) display systematic deviations from the data between pH 7.0 and 10.0 (orange dashed curve in Fig. 7). Nevertheless, it replicates the observed tighter binding at higher pH values, with its convergence on a much lower pK LH(EH) , and a lower pK a for bound than for free ligand (6.7 Ϯ 0.1, versus 9.0 Ϯ 0.02). Upon incorporating a second pK a associated with X (Equation 5), the fit to the data was excellent, replicating not only stronger binding at high pH, but also the biphasic behavior most evident near pH 8 (purple dotted curve in Fig. 7). This lends credibility to our two-residue model (i.e. our assumption of "EH" rather than "E" at pH 7 and below). Again the fit yielded a considerably depressed pK LH(EH) ϭ 6.3 Ϯ 0.1, versus pK LH ϭ 9.0, which was again fixed. Regardless of whether or not a second protonation event is considered, the pK LH(EH) value that emerged was consistent with promotion of substrate deprotonation upon binding to the enzyme. However, the two-residue fit yielded a pK a of pK (L)EH ϭ 8.1 Ϯ 0.2 for the enzyme active-site residue, whereas the pK a obtained by fitting the 19 F chemical shift of bound ligand using Equation 3 was 8.77 Ϯ 0.03 (teal curve in Fig. 7).
Simultaneous fitting to both ␦ obs (pH) and K d, obs (pH) provided good agreement with both (Fig. 7, blue and red curves, respectively), although the fit to K d, obs (pH) is visibly inferior to the fit to ␦ obs (pH). The simultaneous fit yielded a pK a for bound ligand of pK LH(EH) of 6.49 Ϯ 0.09 (Table 1). The fact that similar values were obtained with different complexities of data sets and fits supports their validity. The 2.5 pH unit drop in ligand pK a is consistent with stabilization of phenolate upon binding.
The pK a drop mirrors the 1.6 pH unit drop obtained for 4F3HB (above), although in the case of C 2 it was inferred via the pH dependence of K d, obs rather than that of ␦ obs .
Because the simultaneous fit incorporates greater diversity and quantity of data, we adopt the parameters obtained from it as superior descriptors of the system. We emphasize that the model upon which they are based is not unique, and other solutions might be found that also fit our data. Nevertheless, the model in Scheme 2 has the virtues of being based on behavior also observed in 3HB6H, mechanistic studies of both enzymes, and residues known to be present in their active sites. In particular, the pK EH value of 7.7 that emerges from the model and the data is compatible with X being a His residue.

pH dependence of the 19 F signal of 3F4HPA in the presence of C 2 variants H120N, H120R, and S146A
We tested hypotheses for the identity of residue X by comparing our results on WT protein (above) with experiments on single-site variants. The crystal structure suggests that the conserved Ser-146 and His-120 are close to the ligand and could favor phenol deprotonation as indicated by the pH dependence of K d, obs (Fig. 8B) (49). Analogs of C 2 's His and Ser pair are found in other monooxygenase components, including those of cholesterol monooxygenase from Mycobacterium tuberculosis and chlorophenol 4-monooxygenase from Burkholderia cepacia (50,51). Moreover, these residues have been shown to affect the catalytic parameters and their dependence on pH (37).
To test the possibility that His-120 is responsible for the pHdependent environment sensed by bound 3F4HPA, we collected 19 F spectra of 3F4HPA in the presence of H120N-and H120R-C 2 . These were chosen because the proteins are sufficiently stable for study by NMR, and because the H120R variant retains catalytic activity, albeit impaired (37). 19 F NMR spectra collected as for WT revealed only the free form of 3F4HPA in the presence of the H120N or H120R variants. This is consistent with the K d value for binding of 4HPA to H120N/C4ahydroperoxyflavin of 2.2 mM versus 0.18 mM for the analogous WT-C 2 /C4a-hydroperoxyflavin complex (37). Thus, we confirm that His-120 aids in binding 3F4HPA. However, we could not explore the possibility that His-120 interacts directly with the F substituent of 3F4HPA.
As a control, we also studied S146A-C 2 , which eliminates an interaction with the phenol suggested by crystal structures (Fig. 8B) (49). 19 F NMR spectroscopy revealed signals for both bound and free 3F4HPA, as for the WT-C 2 . The plot of the 19 F chemical shift of bound 3F4HPA versus pH revealed a pK a of 8.70 Ϯ 0.08 associated with a more negative chemical shift upon deprotonation, as for WT-C 2 (Fig. S6). Thus, we can rule out Ser-146 as the source of the pK a that affects bound ligand's chemical shift. Moreover, because Ser-146 interacts with the ligand phenol, this strengthens our conclusion that the pK a of 8.7 is not due to the ligand phenol. This leaves His-120 as the remaining reasonable source of the pK a of 8.7, consistent with all the data in hand.

Single-turnover reaction of WT C 2 with 3F4HPA at various pH values
To learn whether the protein residue with a pK a of 8.7 affects reactivity of the F, we turned to product analyses. The F substituent of 3F4HPA results in the molecule not being symmetric about the 1-4 axis and creates the possibility of forming two different products depending on whether hydroxylation displaces the F at position 3 or hydroxylates at position 5 (see Fig. 8B and Jadan et al. (40)). We hypothesized that the interaction affecting the ligand's 19 F chemical shift could also bias the binding orientation of 3F4HPA and produce a pH-dependent ratio of products.
Product yields on the order of 25% were obtained, possibly because the electron-withdrawing F substituent diminishes 3F4HPA's reactivity as a target for electrophilic attack. Interestingly, the distribution of products obtained did not change with pH, revealing F Ϫ displacement in ϳ53% of the reactions and hydroxylation at position 5 in ϳ47%, over the pH range of 6.0 to 10.0 (Table S1). Thus, 3F4HPA appears able to bind in either orientation (Fig. 8B), and the absence of a change centered at pH 8.7 indicates that interaction between ionizable residue X and the F in the FMNH Ϫ state does not greatly bias the branching ratio between the two reactions that can occur. This agrees with Peelen et al. (33), who concluded that deprotonation of 3F-phenol in Trichosporon cutaneum phenol hydroxylase made the two ortho positions equally electron-rich and therefore equally subject to attack, although there was a 3-fold difference between them when the substrate was protonated (52). The pH-independent product ratio down to pH 6.0 agrees with the pH independence of the hydroxylation rate in indicating that bound substrate's pK a value is significantly lower than 6.0 in the state of the enzyme in which the reaction occurs.
The apparent discrepancy with our measured pK LH(EH) of 6.49 Ϯ 0.09 can be because the reaction occurs in the presence of neutral C4a-hydroperoxyflavin, whereas the NMR studies were in the presence of anionic FNMH Ϫ , the charge of which would tend to raise the pK a values of nearby residues.

Ionization state of ligand bound to 3HB6H and identities of residues affecting it
3HB6H served as a test system for validating our 19 F NMR approach. The 19 F chemical shift of bound 4F3HB changed as expected for deprotonation of the phenol and demonstrating a pK a depressed from 8.7 to 7.1 upon binding to enzyme. This contrasts with the earlier finding that the UV spectrum of 3HB did not appear to change significantly upon binding to enzyme at pH 8.0, arguing against deprotonation (32). However, 19 F NMR provides a signal exclusive to the fluorinated ligand that is less subject to interference from other components of the sample (Fig. S1). Our pK a of 7.1 and 19 F NMR spectra at pH values Figure 6. Effect of pH on the 19 F signal of 3F4HPA free or bound to WT C 2 ⅐FMNH ؊ . A, 19 F spectrum of 3F4HPA at a series of pH values shows both bound (B) and free (F) 3F4HPA in the presence of WT C 2 and FMNH Ϫ . pH values reported are pH 6.6 (red spectrum), pH 7.2 (blue spectrum), pH 7.8 (green spectrum), pH 8.3 (magenta spectrum), pH 8.6 (cyan spectrum), pH 9.0 (black spectrum), pH 9.4 (light blue spectrum), and pH 10.0 (orange spectrum). All spectra are presented using the same vertical scale and were collected at 25°C. Inset B shows the pH dependence of the 19 F shifts of 3F4HPA bound to C 2 (black circles) or free (blue circles). The green circles represent data from back-titration and demonstrate reversibility. The solid line provides the fit to the data of the Henderson-Hasselbalch equation (Equation 3) with pK a ϭ 8.7 Ϯ 0.03 for bound 3F4HPA. For free 3F4HPA, a pK a of 8.9 Ϯ 0.09 was obtained (curve not shown). Inset C plots the observed dissociation constants versus the sample pH documenting tighter binding at higher pH, as predicted for favorable interaction between the phenolate form of substrate and a cationic protein site. All data were collected under inert atmosphere. near 8.0 concur that bound ligand is mostly deprotonated near pH 8.0 (Fig. 3).

Activation of bound substrates by deprotonation
This study also demonstrated that replacement of His-213 with Ser eliminates active-site stabilization of the phenolate form of substrate. This is consistent with the larger K d value reported for 3HB binding of 0.72 mM for H213S, versus the WT value of 0.15 mM (32), and it can explain the lowered (28%) efficiency with which H213S-3HB6H produces product (versus 86% for WT). Proton transfer from substrate to nearby His-213 was proposed to produce a His ϩ ⅐phenolate Ϫ pair in which substrate is activated for electrophilic attack (32). His-213's N⑀2 is 3 Å from the phenolic OH of one of the substrate poses found in the H213S structure (Fig. 8A) (53). Moreover, His-213 N␦1 is 3 Å from N⑀2 of another His, His-363, which is itself polarized by an H-bond to the carbonyl O of Gly-359. Thus, His-213 appears to be supported by a relay of H-bonds that could prime it to abduct a proton from the substrate phenol (32, 53).
The crystal structures of 3HB6H also reveal a Cl Ϫ ion bound against the flavin re face, consistent with inhibition of 3HB6H's reactions with O 2 and NADH by Cl Ϫ (53) and the ability of Cl Ϫ to act as a superoxide analog (54). Thus, we speculate that the Cl Ϫ may identify the pocket in which the hydroperoxide functional group forms. Such a location would enable the distal OH of C4a-hydroperoxide to attack the substrate carbon para to the hydroxyl group when the latter is interacting with His-213.

HPAH-C 2 : use of complementary observables to validate a model and extract pK a values for ligand and protein events
We exploited distinct 19 F signals from bound and free ligand to determine K d, obs as a function of pH. K d, obs reflects states of free enzyme and ligand as well as bound states, so K d, obs reports on ionization events additional to those affecting the ligand 19 F chemical shift. The pH dependence of K d, obs ruled out the ligand's own ionization events as sources of the pK a of 8.7 affecting the ligand chemical shift, so this pK a was assigned to a protein residue X. The resulting enlarged model (Scheme 2) explains the pH dependence of K d, obs substantially in terms of a pK a for bound ligand of pK LH(EH) ϭ 6.49, and it also yields a pK a value for X in the absence of ligand (pK EH ) (Fig. S7). Thus, we learned the pK a values of X when bound and free, as well as the pK a values of ligand when bound and free.
The 2.5 pH unit drop in 3F4HPA's pK a upon binding to C 2 demonstrates that the enzyme stabilizes the ligand's phenolate state, thereby activating it for electrophilic attack. The magni- , yielding a pK a of 8.77 Ϯ 0.03 (teal line). When the chemical shift and K d, obs were fit simultaneously in Matlab, the intermediate pH behavior of K d, obs was not as welldescribed (red solid line), but the parameters obtained from the simultaneous fit are comparable with those obtained from fits to each of chemical shift and K d, obs alone, and the chemical shift's pH dependence was quite well-described (blue dashed line). Obtained parameters are provided in Table 1.

Activation of bound substrates by deprotonation
tude of the pK a shift corresponds to stabilization of the phenolate by 14 kJ/mol relative to the phenol. Similarly, the 1.6 pH unit shift in the pK a of 4F3HB upon binding to 3HB6H represents 9.1 kJ/mol stabilization of the deprotonated form relative to the protonated form.
The energies in question can derive from H-bonds, electrostatic interactions with active-site residues, and/or from effects of the polarity of the active site (34,55). For example, the activesite Asp of ␣-lytic protease from Lysobacter enzymogenes has a pK a below 1.5 representing a shift of almost 3 pH units, attributed to H-bonds with backbone amides and side chains of a His and a Ser (56). Similarly, a salt bridge between two Asp residues in RNase H1 from Escherichia coli elevates the pK a of one Asp to 6.1 and depresses the pK a of the other to 2.6 (57), whereas in the ␤ subunit of F1-ATPase from Bacillus PS-3, the active-site Glu has a pK a of 6.8, representing a 2.5 pH unit elevation (58). Thus, there are ample precedents for the sizes of pK a shifts we have measured (59), based on the same side chains as those found in C 2 's active site (60).

Identities of residues affecting the ionization state of substrate bound to HPAH-C 2
Residue X has a pK a of 7.7 in our model, based on our data, making His-120 the most plausible candidate. Crystal structures indicate that the side chain N⑀ of His-120 is 3.0 Å from the hydroxyl group of 4HPA (Fig. 8B), where it is well-positioned to lower 4HPA's pK a via a N⑀H ϩ /phenolate Ϫ Coulomb interaction. Ser-146 is also in position to donate a hydrogen bond (Fig.  1). The flavin hydroquinone is nearby but is ruled out based on its pK a of 6 or lower, inferred from the pH-independence of the Table 1 Summary of pK a values observed and proposed chemical events associated with them a Carboxyl of free 3F4HPA is shown. b (EH) indicates that the proposed nearby enzyme active-site residue is protonated. c (E) indicates that the proposed nearby enzyme active-site residue is not protonated. d (L) indicates that the deprotonated state of ligand is bound, and imH indicates deprotonation of the imidazolium side chain of His. e Fixing pK a of 9.0 for free 3F4HPA is shown.

Activation of bound substrates by deprotonation
rate of reaction with O 2 (Fig. 8B) (36,61). Another His in the active site, His-396, was determined to aid formation of the C4a-hydroperoxy-FMN (19), suggesting that His-396's pK a is above 10 due to the pH independence of the reaction rate between pH 6.3 and 9.9 (61). Thus, the viable candidates are Ser-146 and His-120. Because the event and its pK a of 8.7 were retained when Ser-146 was replaced by Ala, but ligand binding was lost when His-120 was replaced, we identify X as His-120.
Only C 2 variants with a positive charge at position 120 produced product (37), consistent with electrostatic stabilization of a phenolate substrate transition state. Finally, our pK a values of pK EH ϭ 7.7 and pK (L)EH ϭ 8.7 are comparable with pK a values measured for other His residues. For example, the active site His in ␣-chymotrypsin from bovine pancreas has a pK a of 7.5 that rises to ϳ10.3-12.1 upon binding of substrate analogs (62).
From an energetic standpoint, a protein/ligand interaction that stabilizes the anionic state of the ligand should conversely stabilize the cationic form of the binding site. Indeed, His-120's pK a is elevated when ligand is bound, from pK EH ϭ 7.7 to pK (LH)EH ϭ 8.7. The magnitude of this change is only 1.0 pH unit, indicating that additional residues(s) contribute to the 2.5 pH unit depression of the ligand's pK a . Based on its position (49) and conservation (50,51), Ser-146 is an obvious candidate, and it has been credited with positioning the substrate (37).
For S146A-C 2 , the pH dependence of the 19 F chemical shift of bound ligand as well as its K d, obs yielded mostly similar parameters to those obtained for WT protein (compare Fig. S6 with Fig. 7). However, pK EH of His-120 is 0.4 pH units lower in S146A-C 2 than in WT, compatible with a modest hydrogen bond between His-120 and Ser-146, and the pK LH(EH) of bound ligand is also lowered to 4.8 versus 6.49 for WT. Thus, it appears that Ser-146 raises the pK a values of both His-120 and bound substrate. The latter effect may be inconsequential because pK LH(EH) is quite low in WT and S146A, but elevation of His-120's pK EH to 7.7 can improve enzyme proficiency by extending the pH range over which the active site contains a cation. The effects of Ser-146 are nevertheless small compared with the effect of the interaction between His-120 and ligand, as pK (L)EH is unchanged at 8.7 in S146A-C 2 .
Here, too, our results concur with the literature, as replacement of Ser-146 by Ala barely changed the K d (to 0.32 mM at pH 8.0 versus the WT K d of 0.35 mM (41)), and S146A-C 2 can perform hydroxylation reasonably well (37), affirming a correlation between activity (36) and stabilization of the phenolate form of substrate demonstrated here.

Mechanistic implications of the residue ionization states that emerge
Our data support a model in which ligand binding to C 2 causes His-120 to gain a proton at pH values between 7.7 and 8.7, while the ligand loses one. In effect, a proton is transferred from ligand to His-120 converting the two neutral functionalities to a counterion pair that stabilizes the anionic substrate (Fig. 9). However, the hydroxylation rate is pH-independent over a larger pH range, from 6.2 to 9.9 (36). This was thought to require that the pK LH(EH) of bound substrate be below 6.0 and the pK (L)EH of His-120 with substrate bound be above 10. The current analysis finds that a His-120 pK (L)EH higher than 8.7 is not necessary, because the ligand deprotonates without assistance from His-120 at higher pH values, due to its pK LH(E) of 7.5 (Fig. 9). Thus, the active site has the effect of stabilizing Ͼ25% of substrate in its deprotonated state from pH 6.0 up, whereas free substrate is Ͼ25% deprotonated only above pH 8.5.
At pH values near 6, we expect that the extent of substrate deprotonation would be greater in the state of the enzyme that actually undergoes reaction. We were constrained to study FMNH Ϫ -containing enzyme, although it is neutral C4ahydroperoxy-flavin that participates in the rate-limiting step. The absence of anionic flavin in that case would tend to lower the pK a value of a nearby ligand, explaining why our value for  by stars (B). The 3HB6H structure was obtained by co-crystallization of the H213S variant of 3HB6H with 3HB (PDB code 4BK1) (53). UCSF's Chimera program (75) was used to model the WT His residue in place of Ser-213. The rotamer of His-213 that best overlays the Ser residue and approaches bound substrate without production of clashes was selected, and the energy of the resulting structure was minimized within Chimera. B, active site of WT-C 2 co-crystallized with 4HPA is shown (PDB code 2JBT (49)). To show the locations that could be occupied by F atoms when 3F4HPA is bound, we designated both the 3 and 5 hydrogens of the 4HPA ligand as F atoms and identified them with stars to underscore that they are not experimentally determined F positions. Although the substrate analog we used has F in only one position, we do not know which of two possible orientations it might prefer when bound or whether both orientations are populated. Thus, the two "Fs" in the figure represent two possible positions of F, neglecting additional possible effects of enzyme dynamics. One of these positions is consistently Ϸ4 Å from the N⑀ of His-120 in our model active sites.

Activation of bound substrates by deprotonation
pK LH(EH) is higher than the pK a Ͻ6.0 determined from kinetics (36), despite our use of a fluorinated substrate. 19 F-containing substrate analogs in combination with the excellent responsiveness of 19 F's NMR chemical shift have afforded unique insight into the identities and energies of proton dissociation equilibria that contribute to the catalytic mechanisms of aromatic hydroxylases. There are previous examples of 19 F NMR of fluorinated ligands bound to enzymes (63); however, ours stands out in having used both 19 F and 1 H spectra to determine the ligand protonation state and pH values in sealed samples that preserve the enzyme's redox state. Moreover, we combined the NMR with kinetic measurements to demonstrate that our 19 F probe is in fact a substrate. Our approach has wide applicability to diamagnetic aromatic monooxygenases, flavin-based or not, and indeed to any enzymes for which fluorinated substrate analogs can be found and shown to mimic the behavior of authentic substrate. Because of 19 F's high gyromagnetic ratio, the NMR spectra are comparable in strength to 1 H spectra, so sub-millimolar concentrations of protein can be studied (38,64,65). However, because 19 F is naturally absent from proteins, the spectra are uncomplicated by protein resonances or signals from most common buffers (Fig. S1). Because the 19 F chemical shift is highly responsive to the environment (66) and there are very few distinct F atoms in the samples, the signals tend not to overlap. Thus, simple one-dimensional (1D) NMR spectra suffice, and the experiments can be applied to proteins too large for study by 2D and 3D NMR. Moreover, the large chemical shift change associated with ligand binding means that faster on-and off-rates nonetheless remain in the slow-exchange limit providing separate bound and free signals.

Broader impact of the methodology
In the case of C 2 , an unexpected chemical shift signature and a large effect on the dissociation constant could both be explained on the basis of deprotonation of ligand in concert with a protein residue whose protonated form favors substrate binding. Thus, 19 F NMR afforded information on a protein residue as well as the fluorinated ligand itself. It would be very interesting to apply this method to the reduced state of parahydroxybenzoate hydroxylase where a pK a of 7.1-7.4 has been assigned to bound substrate when the enzyme is oxidized based on UV spectra (27, 29 -31) but to His-72 when the enzyme is reduced, despite a comparative dearth of data (67).

Conclusions
Our 19 F NMR studies demonstrate that the pK a of substrate is depressed in both a single-component and a two-component flavoprotein hydroxylase, supporting deprotonation of the phenol group as a general mechanism of substrate activation for electrophilic attack among these enzymes. Several one-component oxygenases were known to bind substrate in phenolate form (30,31,33,34), but we now place this on quantitative footing, demonstrating 9.1 and 14 kJ/mol stabilization of substrate phenolate consistent with electrostatic or hydrogen bonding mechanisms.
The case of C 2 demonstrates the capacity of 19 F NMR to distinguish ligand deprotonation from other events, based on companion 13 C NMR experiments that identify specific deprotonation events (Figs. 2 and 5). Thus, our data and model indicate that the protonation state of a nearby amino acid is coupled to that of the ligand. For both enzymes, comparisons with protein variants supported an active site His as substantially responsible for stabilizing bound ligand as the phenolate. In C 2 , our experimentally-validated model supports formation of a His ϩ ⅐phenolate Ϫ pair.

Enzymes
The genes for wildtype (WT) and variants of C 2 from Acinetobacter baumanii and 3HB6H from Rhodococcus jostii RHA1 were constructed and expressed, and proteins were purified as described previously (46,68,69). Immediately before NMR data collection, enzymes were exchanged into 100 mM potassium chloride, pH 6.0, by passage through a Sephadex TM G-25 gelfiltration chromatographic column equilibrated with 100 mM potassium chloride, pH 6.0. The compounds used in the experiment (3F4HPA, 4F3HB, maleic acid, acetic acid, 4-MI, and 2,4-DMI) were dissolved in 100 mM potassium chloride to final concentrations of 10 mM, and the resulting solutions were adjusted to pH ϳ6.0. The concentrations of the following compounds were determined using the known extinction coefficients: FMN, ⑀ 446 ϭ 12.2 mM Ϫ1 cm Ϫ1 ; FAD, ⑀ 450 ϭ 11.3 mM Ϫ1 Figure 9. Extent of deprotonation of 3F4HPA free or bound to WT C 2 ⅐FMNH ؊ as a function of pH. The extent to which the two different states containing deprotonated 3F4HPA are populated (dotted lines), along with the total extent to which 3F4HPA is deprotonated (solid red line), based on our model and the pK a values that emerged from our fits to the data. At high pH values above 8.7, the active-site His-120 and bound 3F4HPA are both predominantly deprotonated (dotted green). At lower pH values where free 3F4HPA would normally be protonated (dashed black), the protonated state of His-120 stabilizes deprotonated 3F4HPA (dotted blue). Thus, our model indicates that enzyme-bound substrate is deprotonated to a much lower pH than free substrate (red curve versus dashed black) and that the enlarged pH span for the substrate phenolate is due to cationic His-120 (dotted blue). Our data thereby predict a substantially pH-independent catalytic rate over the pH range of 6.5 to 10.

NMR samples
NMR samples of WT and variants of C 2 were prepared under inert atmosphere in an anaerobic glove box (M-BRAUN UNIlab). The reduced holoenzyme solution (C 2 /FMNH Ϫ ) was prepared by equilibration of 0.5 mM C 2 and 0.5 mM oxidized FMN with O 2 -free atmosphere followed by reduction with a stoichiometric amount of sodium dithionite. Anaerobic solution of 3F4HPA was then added to yield a final concentration of 0.2 mM followed by addition of an aliquot of stock solution of NMR pH indicators (acetic acid, maleic acid, 4-MI, and 2,4-DMI) to yield final concentrations of 1 mM, and sodium trimethylsilylpropanesulfonate (DSS) to yield a final concentration of 0.5 mM (internal chemical shift standard for 1 H). This solution was augmented with 25% (v/v) 2 H 2 O for field locking, and then 700 l were transferred to a 5-mm NMR tube with a septum-lined screw cap (Wilmad catalog no. 535-TR-7). A 1-mm capillary tube was filled with 0.1 mM potassium fluoride in 100 mM potassium chloride, pH 6.0, to serve as a chemical shift reference for 19 F, and flame-sealed. This was placed in the NMR tube, which was then sealed before removal from the glove box.
For 3HB6H experiments, the NMR samples (aerobic condition) contained 0.8 mM 3HB6H enzyme (in the oxidized state, WT, or variants), 0.4 mM 4F3HB substrate analog, 1 mM indicators, 0.5 mM DSS, 0.1 mM KF (in a sealed capillary), and 25% 2 H 2 O in 100 mM potassium chloride, pH 6.0. The provision of excess enzyme favored complete binding of the substrate analog. 19 F NMR spectra were acquired on a Varian INOVA 600 MHz NMR spectrometer using a 600 MHz 1 H-19 F/ 15 N-31 P 5-mm PFG switchable probe with Z gradients, at 25°C (for WT and variants of C 2 ) or 4°C (for WT 3HB6H to accommodate the enzyme's moderate stability). The sample was equilibrated at each temperature for at least 30 min, and the probe was tuned for detection of 19 F. After inserting the sample into the magnet bore, N 2 gas was used to maintain an inert atmosphere, control the temperature, maintain probe cooling, and spin NMR tubes at 20 Hz. 19 F spectra were obtained as 2.323-s acquisitions following 90°excitation pulses (25 s at 25°C and 21.1 s at 4°C) separated by 3.9-s relaxation delays. A total of 6400 transients were averaged per 28,000 Hz-wide 1D spectrum (total time Ϸ12 h). Data were processed using 20 Hz Lorentz line-broadening prior to Fourier transformation. Chemical shift values were referenced to potassium fluoride in a capillary at Ϫ121 ppm. 19 F spectra of 4F3HB bound to 3HB6H were analyzed by Gaussian deconvolution using the software provided by the spectrometer manufacturer (VnmrJ 3.2) to yield the positions, widths, and integrated areas of individual resonances (see Fig.  3A). 1 H 1D NMR spectra were acquired on a Varian INOVA 400 MHz spectrometer using the "Wet1D" pre-sequence with 300 Hz-wide on-resonance eburp1 waveforms to selectively excite the signal of bulk water in preparation for pulse-field gradient suppression, followed by a 90 o excitation pulse to excite the full spectrum and a 1-s acquisition time, with a 1-s relaxation delay time between scans. The pH indicators' chemical shifts were tabulated relative to internal DSS at 0 ppm in 10,000 Hz-wide spectra produced by averaging 64 scans. Data were zero-filled once, weighted with a Gaussian or squared sine bell apodization function, and Fourier was transformed. 13 C NMR spectra were obtained using a Varian INOVA 400 MHz spectrometer (100 MHz for 13 C) at 25°C. The sample solutions for 13 C NMR measurement were 0.1 M 3F4HPA, 10% 2 H 2 O, with CDCl 3 in a sealed capillary serving as an internal standard for 13 C, 77.23 ppm. The 13 C NMR spectra were obtained using a 45°excitation pulse, a 12-s delay between scans, and proton decoupling throughout.

Calculation of the apparent dissociation constant, K d, obs
For the case of 3F4HPA binding to C 2 , we calculated K d, obs at each pH based on the integrated peak areas of the bound and free forms of 3F4HPA. In the absence of chemical exchange, the 6.2-s interval we employed between 90°pulses would allow recovery of Ͼ95 and Ͼ93% of bound and free ligand magnetization, respectively, based on the T 1 values of 2.25 Ϯ 0.1 s for free 3F4HPA and T 1 Յ2 s for C 2 (calculated based on measurements for two other proteins and scaling for c with the assumption that bound ligand tumbles with the protein (70)). The almost-complete recovery and similar degrees of saturation indicate that the two signals will similarly report on the concentration of the species they represent. Moreover, taking into account chemical exchange employing the on-rate of 10 s Ϫ1 reported (41) or adjusted to the concentrations employed here (30 s Ϫ1 ), simulations as per Ref. 71 show that the degree of saturation of the two ligands differs by less than 0.1% with the 6.2 s recycle time we used.
The peak areas were used with knowledge that the total concentration of ligand was 0.2 mM to calculate bound and free ligand concentrations. Then, the total enzyme concentration of 0.5 mM and the concentration of bound ligand (ϭligand⅐enzyme complex, [LigEz]) was used to calculate the concentration of free enzyme, and thereby the dissociation constant in effect (Equation 2): In Equation 2, the protonation states of the ligand and enzyme are not specified (this feature is incorporated below).

pH titrations
The internal NMR pH indicators with pK a values spanning the range of 4.8 to 8.7 were chosen based on the simplicity of their spectra and noninteraction with the enzymes and their substrate binding. The pK a value of each of the indicators used was determined under our conditions (in 100 mM potassium chloride with 10% 2 H 2 O and 0.5 mM DSS at 25°C), and the NMR response was documented by withdrawing NMR samples at 0.3 pH unit intervals in a conventional pH electrode-monitored titration extending from pH 1 to 12. 1 H NMR spectra were acquired on a Varian INOVA 400 MHz as described above. The pK a of each indicator was determined from plots of Activation of bound substrates by deprotonation the indicators' chemical shifts versus the pH values provided by the electrode (Table 2). Thereafter, the known chemical shifts versus pH were used to infer internal pH from the set of indicator chemical shifts without opening the sample to air or inserting a pH probe, because the indicators' chemical shift(s) vary with the pH near their pK a s (Fig. S1).
For experiments on C 2 , an anaerobic solution of potassium hydroxide (KOH) was titrated into 700 l of anaerobic NMR samples as described above. Each addition of KOH was made in an anaerobic glove box so the sample was never opened to air. Each 2 l addition of 0.2 M KOH produced an ϳ0.3 pH unit change. After each KOH addition, a 1 H NMR spectrum was acquired at 400 MHz to determine the new pH, and a 19 F NMR spectrum was acquired at 596 MHz (600 MHz for 1 H) to determine the 3F4HPA 19 F chemical shift. KCl (100 mM) was also present in the enzyme solution to minimize the change in overall ionic strength, and the reversibility of the titration was confirmed by addition of 0.1 M HCl to return the pH to its original range. Fewer points were taken from this back-titration because the enzyme is less amenable to addition of HCl. The 19 F chemical shifts versus pH from the back-titration were compared with those obtained from the upward titration and confirmed that titrations were reversible (Figs. 3 and 6).
For 3HB6H experiments, titrations from pH 6.5 to 10.0 were accomplished by additions of aerobic 0.2 M KOH, and the return to low pH was accomplished via additions of aerobic 1.0 M MES, see above. All titrations were performed at least twice.

Analysis of the pH dependence
The dissociation constants and chemical shifts measured at a series of pH values were modeled as functions of pH, with fitting to different models performed using Kaleida-Graph (Synergy Software Version 4.5.3) and MatLab (Mathworks R2019a).
The pH dependence of the observed 19 F chemical shift ␦ obs was modeled in terms of the Henderson-Hasselbalch equation (Equation 3), in which ␦ LH and ␦ L are the 19 F chemical shifts of the acid and base forms of the ligand (LH and L) and are the asymptotes obtained from the fit. Equation 3 was adapted to analyze absorbance as a function of pH. Observed dissociation constants, K d, obs , are understood to encompass multiple microscopic binding possibilities differing with respect to the protonation states of participating species. To shed light on these, the pH dependence of K d, obs was analyzed in terms of two different models.
1) In the first model, binding of ligand depends only on the protonation state of the ligand itself, and it reflects pK a values of one proton in the enzyme-bound as well as the free state of the ligand pK LH(EH) and pK LH , respectively (Scheme 1, EH denotes enzyme in its protonated state). In our enzymes, the tightest binding is attained for deprotonated ligand (L), so K d, obs (pH) is described relative to binding of L to EH: L ϩ EH N L⅐EH.
2) In the second model, we additionally considered that the protonation state of an enzyme residue might also change in the pH range of interest (Scheme 2, the two-proton model, and Equations 4 and 5). The model incorporates interaction between the protonation states of the ligand (LH or L) and the enzyme (EH or E), where the species responsible for a pK a is indicated in the subscript with the state of the other species providing context in parentheses. Thus pK LH(EH) is the pK a of ligand bound to the protonated state of the enzyme, whereas pK (L)EH indicates the pK a of the enzyme residue when deprotonated ligand is bound. Favorable interaction between a protonated (cationic) enzyme residue and deprotonated (anionic) ligand will tend to raise the enzyme's pK a and lower that of the ligand. Thus, we expect that the ligand's pK a will decrease more upon binding to protonated enzyme than to deprotonated enzyme: pK LH(EH) Ͻ pK LH(E) and that bound ligand will display a pK a of pK LH(EH) when the enzyme is protonated, i.e. when pH Ͻ pK (L)EH , ␦ L, bound (pH) ϭ ␦ LE ϩ ␦ LHE 10 (pKLH(E) Ϫ pH) ϩ ␦ LEH 10 (pK(L)EH Ϫ pH) ϩ ␦ LHEH 10 (pKLH(EH) ϩ pK(L)EH Ϫ 2pH) 1 ϩ 10 (pKLH(E) Ϫ pH) ϩ 10 (pK(L)EH Ϫ pH) ϩ 10 (pKLH(EH) ϩ pK(L)EH Ϫ 2pH) (Eq. 4) and ␦ LHE ϭ ␦ LE Ϫ 0.15 and ␦ LEH ϭ ␦ LHEH ϩ 0.15, where 0.15 is the chemical shift change produced by ligand protonation, obtained from fits to data for free ligand.  (72), except where noted. d The higher of maleic acid's two pK a values is named pK a 2 by Perrin and Dempsey (72). e Data were determined in 100% 2 H 2 O with 100 mM potassium chloride at 25°C (73). L⅐EH (1 ϩ 10 (pKLH Ϫ pH) )(1 ϩ 10 (pH Ϫ pKEH) ) (1 ϩ 10 (pKLH(EH) Ϫ pH) ϩ 10 (pH Ϫ pK(L)EH) (1 ϩ 10 (pKLH(E) Ϫ pH) )) (Eq. 5) In the slow-exchange binding regime that applies to our systems, chemical shifts of bound ligand describe only the environment sensed by the bound population of ligand and reflect the protonation state of enzyme residues that interact with ligand as well as the protonation state of ligand itself (portion of Scheme 2 boxed in green). The K d, obs describes the equilibrium between all bound and all free ligand, and thus additionally reports on the pK a of free ligand (pK LH , known from titrations of free ligand) and free enzyme (pK EH ), in the portion of Scheme 2 boxed in red.

Details of fits
The pH dependence of the apparent dissociation constant K d, obs was fit with Equations 1 (model 1) and 5 (model 2), and the chemical shift of bound ligand was fit with Equations 3 or 4, respectively, to assess the merits of the different models and to extract parameters able to simultaneously describe both pH dependences. Equations 4 and 5 together were used to fit a merged data set of ␦ obs and K d, obs (Scheme 2).
Global fits to ␦ obs and K d, obs were implemented in MatLab, wherein scaled chemical shifts (CS sc ϭ (␦ obs ϩ 137.7)⅐5) were employed so that the magnitude of the total change would be 8.9, comparable with the total change of 9 for the K d, obs . In fitting model 2 to the ␦ obs of bound ligand versus pH, we made the simplifying assumption that the chemical shift would respond similarly to the ligand's own protonation state when bound as when free, regardless of the protonation state of the protein residue. This allowed us to treat the difference ␦ L(EH) Ϫ ␦ LH(EH) as equal to ␦ L(E) Ϫ ␦ LH(E) and ␦ L Ϫ ␦ LH , of which the latter is known from titrations of free ligand. This decreased the number of unknown chemical shifts from 4 to 2 in the global fits. When pK a values obtained from these simplified fits were fixed and all four chemical shifts were optimized independently, the results did not change significantly.

Rapid kinetics experiments
Rapid kinetics measurements were performed with a Hi-Tech Scientific Model SF-61DX stopped-flow spectrophotometer in single-mixing and double-mixing modes. The optical path-length of the observation cell was 1 cm. The reactions were conducted at 4°C. The stopped-flow instrument was made anaerobic by flushing with an oxygen-scrubbing solution consisting of 400 M glucose, 1 mg/ml glucose oxidase (15.5 unit/ml), and 4.8 g/ml catalase in 50 mM sodium phosphate buffer, pH 7.0. The oxygenscrubbing solution was allowed to stand in the flow system overnight, and the system was thoroughly rinsed with anaerobic buffer before experiments were performed. To study the reaction of C 2 with oxygen, the reduced enzyme in the presence of substrate was prepared by rendering a solution of C 2 (100 M) plus FMN (32 M) anaerobic by equilibration in an anaerobic glove box, followed by stoichiometric reduction with sodium dithionite. The resulting solution was placed in a tonometer and transferred to the stoppedflow instrument. The reduced enzyme was mixed with buffers containing various oxygen concentrations (0.13, 0.31, 0.61, and 1.03 mM) and 2 mM 3F4HPA (41). All quoted concentrations are those obtained after mixing. Apparent rate constants (k obs ) were calculated from exponential fits to the kinetic traces, performed using KineAsyst3 software (Hi-Tech Scientific, Salisbury, UK) or Program A (written at the University of Michigan by Rong Chang, Jung-yen Chiu, Joel Dinverno, and David P. Ballou). Rate constants were obtained by fitting plots of k obs versus concentrations of oxygen with a Marquardt-Levenberg nonlinear fit algorithm that is included in the KaleidaGraph software.