Oxidative decarboxylation of pyruvate by 1-deoxy-d-xyulose 5-phosphate synthase, a central metabolic enzyme in bacteria

The underexploited antibacterial target 1-deoxy-d-xyluose 5-phosphate (DXP) synthase catalyzes the thiamin diphosphate (ThDP)-dependent formation of DXP from pyruvate and d-glyceraldehyde 3-phosphate (d-GAP). DXP is an essential intermediate in the biosynthesis of ThDP, pyridoxal phosphate, and isoprenoids in many pathogenic bacteria. DXP synthase catalyzes a distinct mechanism in ThDP decarboxylative enzymology in which the first enzyme-bound pre-decarboxylation intermediate, C2α-lactyl-ThDP (LThDP), is stabilized by DXP synthase in the absence of d-GAP, and d-GAP then induces efficient LThDP decarboxylation. Despite the observed LThDP accumulation and lack of evidence for C2α-carbanion formation in the absence of d-GAP, CO2 is released at appreciable levels under these conditions. Here, seeking to resolve these conflicting observations, we show that DXP synthase catalyzes the oxidative decarboxylation of pyruvate under conditions in which LThDP accumulates. O2-dependent LThDP decarboxylation led to one-electron transfer from the C2α-carbanion/enamine to O2, with intermediate ThDP-enamine radical formation, followed by peracetic acid formation en route to acetate. Thus, LThDP formation and decarboxylation and DXP formation were studied under anaerobic conditions. Our results support a model in which O2-dependent LThDP decarboxylation and peracetic acid formation occur in the absence of d-GAP, decreasing the levels of pyruvate and O2 in solution. The relative pyruvate and O2 concentrations then dictate the extent of LThDP accumulation, and its buildup can be observed when [pyruvate] > [O2]. The finding that O2 acts as a structurally distinct trigger of LThDP decarboxylation supports the hypothesis that a mechanism involving small molecule–dependent LThDP decarboxylation equips DXP synthase for diverse, yet uncharacterized cellular functions.

The increasing spread of antimicrobial resistance demands the development of novel antimicrobial agents. 1-Deoxy-D-xyluose 5-phosphate (DXP) 2 synthase is an underexploited antimicrobial target that catalyzes the formation of DXP from pyruvate and D-glyceraldehyde 3-phosphate (D-GAP, Fig. 1) in many human pathogens (e.g. Mycobacterium and Plasmodium falciparum). DXP is an essential branch point metabolite that serves as a precursor in the biosynthesis of thiamin diphosphate (ThDP), pyridoxal phosphate, and indispensable isoprenoids in bacterial pathogens ( Fig. 1) (1)(2)(3)(4). Given the widespread essentiality of DXP in pathogens and its absence in humans, selective inhibition of DXP synthase could offer a new antimicrobial strategy (5)(6)(7)(8).
The requirement for ternary complex formation during catalysis was first proposed based upon 14 CO 2 trapping studies demonstrating a low rate of CO 2 release in the absence of D-GAP (13). Circular dichroism (CD) and NMR studies subsequently revealed the accumulation of LThDP on DXP synthase in the absence of D-GAP, and provided evidence that D-GAP plays a role to induce efficient LThDP decarboxylation (15). These findings are consistent with a mechanism involving ternary complex formation on DXP synthase; however, these results do not explain the earlier observation that DXP synthase catalyzes multiple turnovers and CO 2 release in the presence of excess pyruvate. DXP synthase-dependent acetolactate formation was demonstrated (18), providing a possible explanation for the observed CO 2 release in the absence of D-GAP. However, the acetolactate synthase activity of DXP synthase was later shown to be exceedingly inefficient (15). LThDP stabilization on DXP synthase, coupled with inefficient acetolactate formation and a lack of evidence for C2␣-carbanion generation in the absence of D-GAP (by detection of acetaldehyde formation from hydroxyethyl-ThDP or 2,6-dichlorophenolindophenol (DCPIP) reduction, Fig. 2) (15) are collectively inconsistent with the extent of CO 2 release observed by Eubanks and Poulter (13). Given the unique requirement for a trigger of LThDP decarboxylation on DXP synthase, resolving this discrepancy is important.
Here, we report studies that reconcile these conflicting observations. Our results indicate that DXP synthase catalyzes oxidative decarboxylation of pyruvate by a mechanism involving O 2 -dependent LThDP decarboxylation and subsequent oxidation of the C2␣-carbanion to produce peracetic acid as an intermediate in acetate formation. Removal of O 2 does not significantly impact the rates of LThDP formation, D-GAP-dependent LThDP decarboxylation, or DXP formation suggesting O 2 does not influence the DXP-forming activity of DXP synthase. However, the stability of LThDP depends upon the presence or absence of O 2 , indicating that mechanistic studies of LThDP formation and decarboxylation in the absence of D-GAP should be conducted under anaerobic conditions. This study provides key insights into our understanding of a DXP synthase mechanism and raises interesting questions about the function of the unique requirement for ternary complex formation on DXP synthase.

Product formation on DXP synthase in the absence of D-GAP
To obtain a comprehensive profile of products formed by the action of DXP synthase in the absence of D-GAP, the DXP synthase reaction using uniformly [ 13 C 3 ]pyruvate was monitored over time by NMR. 1D 13 C (Fig. 3) and 2D HCC-COSY experiments (Fig. S1) )) and bicarbonate (␦ 160.36 ppm (s)) as the major products, indicating that CO 2 release occurs under conditions in which accumulation of LThDP is evident by CD analysis (Fig. 3, A-C). Orthogonal biochemical assays were conducted to measure CO 2 release and pyruvate consumption. A 14 CO 2 trapping assay was conducted as previously described (13), and showed DXP synthase and pyruvate-dependent 14 CO 2 release over time, consistent with bicarbonate formation detected by NMR (Fig. 3D). Concomitant with CO 2 release, DXP synthase-dependent pyruvate consumption was observed by the lactate dehydrogenase (LDH) assay (Fig. 3E). Taken together, these results indicate that DXP synthase catalyzes the oxidative decarboxylation of pyruvate in the absence of D-GAP. These findings are consistent with the evidence for CO 2 release reported by Eubanks and Poulter (13), yet puzzling in light of the CD evidence for LThDP accumulation and D-GAP-dependent LThDP decarboxylation on DXP synthase under these conditions (15). Thus, experiments were conducted toward understanding oxidative decarboxylation and determining the trigger for CO 2 release on DXP synthase under these conditions.

O 2 -dependent LThDP decarboxylation on DXP synthase
decomposition to acetate in HEPES, pH 8 ( Fig. S2) (20,28). Elimination of H 2 O 2 from 3 (path B, Fig. 4) would afford acetyl-ThDP 4, which could undergo hydrolysis to release acetate and cofactor. Finally, in path C (Fig. 4), a second electron transfer from 2 to superoxide could occur resulting in the production of H 2 O 2 and 4 en route to acetate. Notably, only LThDP has been detected on DXP synthase by NMR following an acid quench, with no detectable formation of acetyl-ThDP 4 (15,16).
Each proposed pathway to acetate requires O 2 consumption and formation of a radical ThDP intermediate (2, Fig. 4). To confirm that O 2 is utilized as a substrate by DXP synthase, the concentration of O 2 was monitored over time in the presence of varying concentrations of DXP synthase and pyruvate. Indeed, enzyme and pyruvate-dependent O 2 consumption was detected providing further support for DXP synthase-catalyzed oxidative decarboxylation of pyruvate (Fig. 5, Table 1). Ϫ . B and C, 13

O 2 -dependent LThDP decarboxylation on DXP synthase
Electron paramagnetic resonance (EPR) studies of DXP synthase in the presence of unlabeled, 13 C 3 -and 2 H 3 -labeled pyruvate revealed the formation of a radical species at concentrations of ϳ1% of DXP synthase active centers (Fig. 6, solid lines). The spectra generated with [ 2 H 3 ]pyruvate (Fig. 6A) displayed an observable narrowing with respect to the unlabeled sample ( Fig. 6B), as would be expected if the electron spin of the radical species experienced a decreased (albeit unresolved) hyperfine coupling to the deuterons derived from [ 2 H 3 ]pyruvate relative to the protons of the unlabeled pyruvate. Moreover, spectra generated with [ 13 C 3 ]pyruvate produced a spectrum ( Fig. 6C) with observable broadening with respect to the unlabeled sample, which is expected if the radical experienced the additional (unresolved) hyperfine coupling to the 13 C nuclei (nuclear spin 1 ⁄ 2), which is absent in the 12 C (nuclear spin ϭ 0) sample.
The impact of the labeled pyruvate on the EPR spectrum of the radical species clearly implicates the contribution of carbon and hydrogen derived from pyruvate to the structure of the radical species. To determine whether the EPR spectra are consistent with the ThDP-enamine radical species 2 (Fig. 4), simulations were carried out (Fig. 6, dashed lines) using parameters (Table S1) previously determined for similarly structured radical species generated by other ThDP-dependent enzymes (22)(23)(24)(25)(26). All parameters were held constant in the three simulations, with the exception of the appropriate addition or scaling of hyperfine interactions for the labeled nuclei previously determined (22)(23)(24)(25)(26). The fits provide strong evidence that the radical intermediate generated by DXP synthase is consistent with 1-electron oxidation of the enzyme-bound C2␣-carbanion intermediate 1 to form 2 en route to acetate.
To distinguish between paths A and B/C (Fig. 4), we utilized two previously described biochemical assays that specifically monitor peracetic acid (path A) or H 2 O 2 (paths B or C) formation (20,29). DXP synthase-catalyzed peracetic acid formation was probed using a coupled assay with thionitrobenzoate (TNB, max ϭ 412 nm), which is readily oxidized by peracetic acid at a significantly higher rate (ϳ3000-fold higher) compared with its oxidation by H 2 O 2 (20), thus providing a selective detection method to differentiate between peracetic acid and H 2 O 2 production. Control experiments were conducted to confirm selective detection of peracetic acid under the reaction conditions used here (Fig. S3). Observation of DXP synthase-dependent oxidation of TNB (Table 1) suggests that peracetic acid is formed in the oxidative decarboxylation of pyruvate by DXP synthase. Furthermore, the rate of DXP synthase-dependent peracetic acid formation is comparable with the rate of O 2 consumption under the same conditions ( Table 1), suggesting that peracetic acid formation coincides with O 2 consumption, similar to acetolactate synthase and pyruvate decarboxylase (20). In an attempt to ascertain whether acetate formation occurs via either path B or C, we employed a discontinuous assay for selec-    Table S1, which are consistent with the radical structure 2, in Fig. 4. Simulation details are provided in the supporting information.

O 2 -dependent LThDP decarboxylation on DXP synthase
tive detection of H 2 O 2 (29). Unfortunately, control experiments conducted during this study indicate that H 2 O 2 is unstable under these reaction conditions and, therefore, cannot be accurately detected by this method (Fig. S4). Although DXP synthase-catalyzed H 2 O 2 formation cannot be definitively ruled out, the comparable rates of TNB and O 2 consumption ( Table 1) and lack of acetyl-ThDP detection on DXP synthase to date support path A as the major route to acetate.
The newly revealed understanding that DXP synthase catalyzes the oxidative decarboxylation of pyruvate suggests a model ( Fig. 7) to explain the simultaneous observations of LThDP accumulation and CO 2 release in the absence of D-GAP. In the presence of O 2 , LThDP decarboxylation occurs by an unknown mechanism, to form the C2␣-carbanion, which is then oxidized to form peracetic acid (Fig. 7). This reaction cycle consumes pyruvate and O 2 and depletes a proportion of LThDP-bound DXP synthase; thus, the relative concentrations of O 2 and pyruvate dictate the extent to which LThDP accumulation and/or oxidative decarboxylation is observed. Over time, DXP synthase deoxygenates the solution to a point where LThDP decarboxylation no longer occurs efficiently, permitting stabilization of LThDP over a longer time scale. To provide support for this model, we conducted a series of experiments to determine the effect of O 2 on the activity of DXP synthase.

Characterization of triggers of CO 2 release on DXP synthase
To determine whether O 2 acts as a trigger of CO 2 release, we conducted the CO 2 trapping and pyruvate consumption assays under anaerobic conditions, in the presence or absence of added O 2 (established by purging reaction mixtures in an environment of 90% N 2 , 10% H 2 mixture). Under anaerobic conditions in the absence of D-GAP, CO 2 release and pyruvate consumption are significantly decreased (Fig. 8, A and B), suggesting O 2 is required for efficient release of CO 2 in the absence of D-GAP. The observed low levels of CO 2 release and pyruvate consumption under anaerobic conditions could result from incomplete removal of O 2 or slow acetolactate formation. Addition of O 2 to the anaerobic sample results in significant CO 2 release (Fig. 8A), providing direct evidence for O 2 -induced decarboxylation. Importantly, under anaerobic conditions, addition of a deoxygenated D-GAP solution results in significant CO 2 release, which is consistent with our previous characterization of D-GAP as a trigger of efficient LThDP decarboxylation on DXP synthase (Fig. 8A) (15).

Behavior of LThDP on DXP synthase under aerobic and anaerobic conditions
If the model depicted in Fig. 7 is correct, the stabilization of LThDP on DXP synthase should be observed by CD under anaerobic conditions, and the LThDP species should be unstable in the presence of a trigger of CO 2 release (O 2 or D-GAP). As expected, addition of pyruvate to DXP synthase under anaerobic conditions leads to the emergence of a positive CD signal at 313 nm (Fig. 9A), which has previously been characterized as LThDP on DXP synthase (15). The apparent K d(pyruvate) under anaerobic conditions is 45 Ϯ 2 M ( Table 2). As illustrated in    Fig. 9D), which further supports both D-GAP and O 2 as triggers of LThDP decarboxylation and is consistent with the model shown in Fig. 7. Importantly, this suggests that under aerobic conditions, the reactive nature of the LThDP intermediate may lead to unreliable measurements of K d(pyruvate) , depending upon the extent of O 2 introduction through mixing. Varied aeration will change the amount of pyruvate required for DXP synthase to effectively deoxygenate the solution and permit observation of significant LThDP accumulation. In contrast, pyruvate is not turned over when measurements are conducted in an anaerobic environment, permitting a more reliable measurement of the apparent K d(pyruvate) .

Steady state kinetic characterization of DXP formation under anaerobic conditions
DXP formation is the only known physiological function of DXP synthase, and our results indicate the possibility that O 2 may compete with D-GAP in the LThDP decarboxylation and/or carboligation steps leading to DXP formation. Thus, we investigated the steady kinetics of DXP formation in the absence of O 2 to determine the effect of O 2 on the component steady state kinetic parameters, K m and k cat , of WT Escherichia coli DXP synthase. The results indicate that there is no significant difference in the kinetic parameters measured in the presence or absence of O 2 (Fig. 10, Table 2).

Pre-steady state characterization of LThDP formation and decarboxylation under anaerobic conditions
Steady state CD analysis of LThDP accumulation on DXP synthase in the presence and absence of O 2 provides support for

O 2 -dependent LThDP decarboxylation on DXP synthase
a model in which the stability of the LThDP signal depends on the level of O 2 in solution. This suggests that experiments measuring accumulation or depletion of LThDP may be confounded by O 2 -dependent LThDP decarboxylation and pyruvate consumption under aerobic conditions. Therefore, the pre-steady state rates of LThDP formation and decarboxylation (in the presence of deoxygenated buffer, D-GAP, or O 2 -saturated buffer) were determined to obtain more accurate measurements of these kinetic parameters. To determine the rate of LThDP formation on DXP synthase in the absence of O 2 , DXP synthase in one syringe was rapidly mixed with pyruvate in the other syringe under anaerobic conditions. The results indicate a rate of 1.64 Ϯ 0.21 s Ϫ1 for LThDP formation (k 1 ) on DXP synthase under anaerobic conditions that is comparable with the previously reported rate of LThDP formation under aerobic conditions (1.39 Ϯ 0.05 s Ϫ1 ) (15), showing O 2 does not signifi-cantly impact the kinetics of LThDP formation on DXP synthase (Fig. 11, Table 3). It is interesting to note that under anaerobic conditions, following addition of pyruvate to the preformed DXP synthase-ThDP complex, there is a short lag phase (Fig. 11, inset, region 1) followed by a decrease in the CD signal at 313 nm (Fig. 11, inset, region 2) that immediately precedes the signal increase corresponding to LThDP formation (Fig. 11,  inset, region 3). This behavior of the CD signal was not observed under aerobic conditions (10), and could signify binding and/or conformational change(s) occurring on DXP synthase prior to LThDP formation in the absence of O 2 .
A significant decrease in the CD 313 signal upon the addition of a trigger to pre-formed LThDP is interpreted to coincide with LThDP decarboxylation (15). It follows that the CD signal corresponding to pre-formed LThDP should be minimally affected by the addition of deoxygenated buffer lacking O 2 or D-GAP triggers required for efficient LThDP decarboxylation. Consistent with this, a marginal change in the CD 313 signal is observed upon rapid mixing of deoxygenated buffer in one syringe with pre-formed LThDP in a second syringe under anaerobic conditions (Fig. 12A), in contrast to the significant decrease at 313 nm observed in the presence of D-GAP or O 2 (Fig. 12, B and C). The subtle change in CD signal detected in the absence of added trigger (Fig. 12A) is likely due to incomplete removal of O 2 during the deoxygenation process, leading to decarboxylation of a small fraction of LThDP. Although an accurate background rate of LThDP decarboxylation could not be determined in this case, apparent LThDP re-synthesis (k 1r ) is suggested by the data (apparent k 1r ϭ 1.04 Ϯ 0.11 s Ϫ1 ). In contrast, the rates of LThDP decarboxylation (k 2 ) upon addition of O 2 -saturated buffer or deoxygenated buffer containing D-GAP to preformed LThDP (prepared under anaerobic conditions) were 62.0 Ϯ 8.9 and 50.3 Ϯ 7.8 s Ϫ1 , respectively (Figs. 12, A-C, Table 3), comparable with the rate of LThDP decarboxylation in the presence of D-GAP under aerobic conditions previously reported (k 2 ϭ 42 Ϯ 1 s Ϫ1 ) (15). These results suggest that LThDP decarboxylation occurs efficiently in the presence of either D-GAP or O 2 . Notably, previously observed lag phases in . Data points are an average of the CD signal over time from 7 repetitive shots. Error is calculated from the fit of the averaged data to a double exponential rise to a maximum (see Equation 1 under "Experimental procedures").

O 2 -dependent LThDP decarboxylation on DXP synthase
the CD 313 signal occurring immediately before and after the sharp decrease in this signal corresponding to LThDP decarboxylation (16) are observed upon addition of D-GAP or O 2 -saturated buffer to preformed LThDP generated under anaerobic conditions (Fig. 12, B and C, insets). These lag phases are proposed to signify conformational changes occurring prior to LThDP decarboxylation and immediately after LThDP decarboxylation prior to LThDP re-synthesis (16), a hypothesis that is supported by recent HDX-MS studies (12).

Discussion
Although important aspects of the DXP synthase mechanism are uncovered, gaps in our understanding of catalysis on this unique ThDP-dependent enzyme remain. The present study addresses a discrepancy in mechanistic studies of DXP synthase: the observation of slow CO 2 release in the absence of a known trigger LThDP decarboxylation, D-GAP (13), despite the accumulation of LThDP on the enzyme and lack of evidence for the post-decarboxylation intermediate (15). Our results reveal two roles of O 2 : 1) as a distinct trigger of LThDP decarboxylation, and 2) as an acceptor substrate in the oxidation of the C2␣-carbanion to form peracetic acid en route to acetate. Other ThDP-dependent enzymes are known to catalyze similar reactions (19 -21, 30). However, due to the ability of DXP synthase to stabilize LThDP, and the lack of any indication that C2␣carbanion formation takes place in the absence of D-GAP, this oxygen-consuming reaction was not previously characterized. The results show that steady state and pre-steady state kinetic parameters are unchanged under anaerobic conditions suggesting O 2 does not affect catalytic steps prior to and including DXP formation. The apparent K d(pyruvate) measured under anaerobic conditions is a more accurate estimation as anaerobic conditions prevent confounding reactions that occur in the presence of O 2 (i.e. depletion of pyruvate and LThDP). Although the oxygenase activity of DXP synthase does not appear to affect DXP-forming activity of WT enzyme, this side reaction and discovery of O 2 -dependent LThDP decarboxylation have important implications for the study of the DXP synthase mechanism.
In retrospect, the observation of LThDP buildup on DXP synthase was fortuitous given the efficiency of decarboxylation in the presence of O 2 , and points to the unique mechanistic aspects of this enzyme. The oxidative decarboxylation of pyruvate is a key element of the DXP synthase mechanism that explains the accumulation of LThDP and lack of accumulation of the C2␣-carbanion, despite significant CO 2 release (13, 15). Our inability to detect post-decarboxylation intermediates on DXP synthase in the absence of D-GAP is likely a consequence of electron transfer from the C2␣-carbanion to O 2 preventing efficient reaction of this intermediate with other potential acceptor substrates. The simultaneous observations of LThDP accumulation and CO 2 release can be explained by the model in Fig. 7. Under conditions where O 2 is in excess to pyruvate, occurrence of LThDP decarboxylation at a lower rate than C2␣-carbanion consumption would permit the observation of the LThDP intermediate concomitant with CO 2 release and peracetic acid/acetate formation. The presence of O 2 ensures this cycle of LThDP decarboxylation and peracetic acid formation, consuming both pyruvate and O 2 (Fig. 7). As a result, the relative concentrations of O 2 and pyruvate dictate the extent of buildup and the length of time for which LThDP is stable. Therefore, accumulation of LThDP over a long time period requires conditions where pyruvate is in excess to O 2 , which is the case for previously reported CD and NMR experiments (15). Under these conditions, enzyme-catalyzed O 2 consumption deoxygenates the solution, resulting in a net accumulation of LThDP in the absence of O 2 -dependent decarboxylation,

Figure 12. Pre-steady state decarboxylation of LThDP at 6°C upon addition of anaerobic buffer (A), an anaerobic solution of D-GAP (B), or O 2 saturated buffer (C) to preformed LThDP under anaerobic conditions.
Insets show up to the first 300 ms of the full time course to illustrate the lags in the CD signal at 313 nm before and after LThDP decarboxylation. Data points are an average of the CD signal over time from 7 repetitive shots. Error is calculated from the fit of the averaged data to the appropriate equation (see "Experimental procedures").

O 2 -dependent LThDP decarboxylation on DXP synthase
reconciling the discrepancy in our understanding of DXP synthase mechanism.
Although the requirement of a small-molecule (e.g. D-GAP or O 2 ) to trigger efficient LThDP decarboxylation on DXP synthase is established, the mechanism by which this occurs is not fully elucidated. Several mechanistic studies of DXP synthase have revealed compelling evidence for D-GAP-dependent conformational changes (12,16,17). For example, a recent HDX-MS study showed that a donor substrate analog and D-GAP induce closed and open conformations of DXP synthase, respectively (12). It is conceivable that D-GAP binding induces a conformational change that destabilizes the LThDP intermediate, lowering the barrier for chemical decarboxylation of LThDP (12,15,16). Considering the proposal that 2-oxoacid decarboxylation is reversible on ThDP-dependent enzymes (31)(32)(33), it is also possible that an equilibrium between LThDP and the C2␣-carbanion/CO 2 is established on DXP synthase in a conformation that limits release of CO 2 from the enzyme. Subsequent binding of D-GAP could then drive the equilibrium triggering LThDP depletion and turnover. These mechanistic models cannot be distinguished by the tools used in this study. Although O 2 is structurally and chemically distinct from D-GAP, it appears to trigger depletion of LThDP with a comparable rate and imparts similar behavior to the CD 313 signal of the LThDP species (Fig. 12), suggesting a similar mechanism of LThDP decarboxylation is at play in the presence of these two triggers. Further studies are required to determine whether O 2 can induce a conformational change on DXP synthase, toward understanding the mechanism by which O 2 and D-GAP induce LThDP decarboxylation and release of CO 2 , and to determine whether other natural triggers exist.
Moving forward, it is imperative to consider the O 2 -dependent E-LThDP depletion in mechanistic studies of DXP synthase. First, we have shown that addition of O 2 to LThDP by sample mixing causes depletion of the CD 313 signal, indicating that K d determinations for donor substrates should be performed under anaerobic conditions to obtain the most accurate measurement. Second, it is imperative that studies aimed at identifying other inducers of LThDP decarboxylation or investigating the roles of active site residues in catalysis, are conducted under anaerobic conditions to avoid interfering side reactions or variable contributions of oxygenase activity on variants, both of which could confound results. Other challenges to investigating DXP synthase, such as structural studies, may also be resolved by working under anaerobic conditions.
In this study, we have uncovered a novel activity of DXP synthase, adding to the repertoire of chemistry that this unique enzyme can catalyze. Notably, we illustrate that O 2 is a structurally distinct trigger of LThDP decarboxylation, which has important mechanistic and technical implications for the study of DXP synthase mechanism. These findings add a major piece to the mechanistic puzzle of DXP synthase and have enabled resolution of inconsistencies in the literature. Taken together, these results lay a foundation to gain a deeper understanding of the function of the unique mechanism of DXP synthase in the broader context of microbial metabolism.

General methods
Unless otherwise noted, all reagents were obtained from commercial sources. Aerobic spectrophotometric analyses were performed on a Beckman DU800 UV-visible spectrophotometer (Brea, CA). Anerobic spectrophotometric analyses were performed on a Varian Cary 50 UV-visible spectrophotometer (Palo Alto, CA). Anerobic conditions were established either in a Baker-Ruskinn in vivo (Sanford, ME) or a Coy Laboratory Products vinyl anaerobic chamber (Grass Lake, MI). E. coli WT DXP synthase and E. coli MEP synthase (IspC) were overexpressed and purified as reported previously with minor modifications for the purification of anaerobic DXP synthase (14,34). To obtain anaerobic DXP synthase, the protein was overexpressed and purified as previously described (34), however, the second dialysis was carried out in an anaerobic chamber in 50 mM HEPES, pH 8, 5% glycerol, 100 mM NaCl, 10 mM MgCl 2 , and 1 mM ThDP at 0°C for 4 h. The protein was stored in liquid N 2 until use when it was transferred directly to an anaerobic chamber. Radioactivity measurements were made with a Beckman LS6500 scintillation counter. X-band (9 GHz) EPR measurements were made using a Varian E-112 spectrometer equipped with a TE 102 cavity and interfaced to a PC using custom written software. Steady state CD spectra were recorded on an Aviv 420 CD spectrometer (Lakewood, NJ). Presteady state CD spectra were recorded on an Applied Photophysics Pi*-180 stopped-flow CD spectrometer (Leatherhead, Surrey, UK) with a 10-mm path length cuvette. NMR spectra were obtained on a Bruker Avance III 500 MHz NMR spectrometer (Billerica, MA). Oxygen consumption was measured with the Oxytherm ϩ respiration oxygen monitoring system from Hansatech Instruments Ltd. (Norfolk, United Kingdom).

NMR detection of product formation by DXP synthase in the absence of D-GAP
To monitor reaction progress by NMR, DXP synthase (30 M) was mixed with [ 13 C 3 ]pyruvate (500 M) in 50 mM HEPES-D 18 , pH 8, 100 mM NaCl, 1 mM ThDP, 2 mM MgCl 2 , 1% glycerol, and 10% D 2 O to initiate product formation. 1D 13 C spectra (500 MHz, 256 scans, 0.7 s/scan) were recorded over time at 25°C. A control lacking enzyme was prepared in the same buffer and incubated at 25°C for 4 h after which a spectrum was recorded. Spectra of [ 13 C 2 ]acetate and [ 13 C]bicarbonate standards were obtained in the NMR buffer and with the same acquisition parameters.

CO 2 trapping under aerobic conditions in the absence of D-GAP
To measure 14 CO 2 release from DXP synthase, the same general protocol from Eubanks and Poulter (13) was conducted with some minor modifications. Briefly, reactions (250 l) contained 30 M DXP synthase and [1-14 C]pyruvate diluted with cold pyruvate (0.18 Ci/mol of total pyruvate, 500 M pyruvate total) in reaction buffer (50 mM HEPES, pH 8, 100 mM NaCl, 1 mM ThDP, 2 mM MgCl 2 , and 1% glycerol). Solutions were preincubated in the absence of DXP synthase at 0°C for 5 min in the outer well of a trapping vial (glass vial, 4.5 cm height O 2 -dependent LThDP decarboxylation on DXP synthase and 2.5 cm diameter, with 24/40 ground glass neck and a center well trap, 2 cm height and 1.5 cm diameter). The center well trap contained 500 l of 1 M NaOH, and a 24/40 septum stopper was used to seal the trapping vial. Reactions were then initiated upon addition of DXP synthase to the outer well via syringe and incubated at 0°C. At 0.5, 5, and 60 min, 250 l of 12.5% TCA in 1 M HCl was added to the outer well of the trapping vial via syringe to quench the reaction and drive hydrated 14 CO 2 out of solution (note each time point was a separate reaction). A negative control in which enzyme storage buffer (50 mM Tris, pH 8, 10% glycerol, 100 mM NaCl, 10 mM MgCl 2 , 1 mM ThDP) was added in place of enzyme was also prepared and "quenched" at 60 min. The quenched solutions were incubated at 25°C for 2 h to trap the 14 CO 2 released by DXP synthase. The entire contents of the acid quench solution and the NaOH trap solution were added to 10 ml of Hionic-Fluor scintillation fluid (PerkinElmer Life Sciences), mixed by vortexing, and immediately counted (1-min counting time per sample). Disintegrations per minute were determined from the measured CPMs using the counting efficiency determined from a standard curve of [1-14 C]pyruvate in the acidic medium or [ 14 C]bicarbonate in the basic medium. All reactions were carried out in triplicate. Because this quench and trapping protocol varied slightly from that of Eubanks and Poulter (13), a trapping control was conducted to confirm that Ͼ95% of the 14 CO 2 that is released from the enzyme is collected in the trap (Fig. S5).

Lactate dehydrogenase detection of pyruvate consumption under aerobic conditions
For LDH detection of pyruvate consumption, reactions contained 30 M DXP synthase and 500 M pyruvate in reaction buffer (described above). These solutions were preincubated for 5 min at 0°C in the absence of DXP synthase. Pyruvate consumption was initiated upon addition of DXP synthase or storage buffer for the no enzyme control. At 0.5, 5, and 60 min, a 60-l aliquot of the reaction solution was boiled at 100°C for 5 min to quench the reaction. Precipitated protein was pelleted by centrifugation. The pyruvate remaining in the quenched solution was then determined using LDH. Reactions (400 l) containing 40 l of the supernatant from the quenched reaction solution, 200 M NADH, 32.5 mM KH 2 PO 4 , pH 7.4, and 0.75 units/ml of yeast LDH were initiated upon addition of the pyruvate-containing solution and the absorbance at 340 nm (⑀ ϭ 6220 M Ϫ1 cm Ϫ1 ) was monitored at 37°C for 15 min. The amount of pyruvate in each sample was calculated using the change in the initial and final absorbance. Corrections were made for nonenzymatic oxidation of NADH over the course of the LDH reaction. DXP synthase reactions were conducted in triplicate.

Detection of oxygen consumption
To determine whether DXP synthase consumes O 2 in an enzyme-and substrate-dependent manner, DXP synthase concentration was varied (0 -30 M) in the presence of 500 M pyruvate, and pyruvate concentration was varied (0 -500 M) in the presence of 5 M DXP synthase in reaction buffer (described above). Solutions were prepared in the absence of pyruvate, added to the electrode chamber (calibrated fresh each day at 25°C), and equilibrated at 25°C for at least 4 min until the O 2 signal stabilized. O 2 consumption was initiated upon addition of pyruvate (25 l regardless of the final [pyruvate]) by syringe, and the concentration of O 2 was monitored over time at 25°C. O 2 consumption was monitored under each condition in triplicate.

Electron paramagnetic resonance detection of ThDP-enamine radical species
DXP synthase (32 mg/ml, 508 M) in 0.4 ml of 50 mM HEPES, pH 8.0, containing 0.10 M NaCl, 1.0 mM ThDP, 10 mM MgCl 2 , and 10% glycerol was mixed with unlabeled pyruvate (10 mM), [ 2 H 3 ]pyruvate (10 mM), or [ 13 C 3 ]pyruvate (10 mM) at room temperature. Mixtures were immediately transferred into EPR tubes and flash-frozen in liquid nitrogen after a 30 -40-s incubation at room temperature. During EPR measurements, the sample temperature was held at 77 K using an immersion finger Dewar. Spectrometer parameters used to acquire the spectra are as follows: modulation amplitude, 5 G; microwave power, 10 W; microwave frequency, 9.09 GHz; scan time, 2 min; time constant, 0.5 s; number of scans, 9. The field was calibrated using a standard sample of manganese doped in MgO (35). The concentration of active sites occupied by the ThDP radical species was calculated by comparing the double integral of the radical EPR spectrum with both Cu(II) and nitroxide standard sample spectra (23).

Colorimetric detection of DXP synthase-dependent peracetic acid formation
To determine whether DXP synthase catalyzes the formation of peracetic acid as an intermediate in acetate formation, we utilized the previously described TNB assay (20). Samples were prepared in the reaction buffer described above with 75 M TNB (synthesized and characterized according to the protocol in the supporting information) varying DXP synthase (0 -30 M) in the presence of 500 M pyruvate or varying pyruvate (0 -500 M) in the presence of 5 M DXP synthase. After a 5-min preincubation at 25°C, peracetic acid formation was initiated upon addition of DXP synthase and the oxidation of TNB was monitored spectrophotometrically ( max ϭ 412 nm, ⑀ ϭ 13,600 M Ϫ1 cm Ϫ1 ) at 25°C over time. All reactions were carried out in triplicate.

CO 2 trapping under anaerobic conditions in the presence and absence of triggers
For the detection of 14 CO 2 under anaerobic conditions the general method used for the aerobic 14 CO 2 trapping assay was performed with some modifications. Before addition of the [1-14 C]pyruvate, all solutions and the trapping vials (prepared with 500 l of 1 M NaOH in the center well) were deoxygenated in an anaerobic chamber for 15 min at 30°C and then sealed with septa before removing from the chamber. [1-14 C]Pyruvate was added to the reaction solution that was distributed to the outside wells of the trapping vials via syringe. One trapping vial was purged with 100% O 2 for 10 min at 0°C before addition of the reaction solution. All trapping vials were then incubated at 0°C for 5 min. 14 CO 2 release was initiated upon addition of DXP synthase, and the mixtures were incubated at 0°C for 1 h.

O 2 -dependent LThDP decarboxylation on DXP synthase
The reactions were then quenched, trapped, and counted as described for aerobic conditions.

LDH detection of pyruvate consumption under anaerobic conditions
The same general method for LDH detection of pyruvate under aerobic conditions was repeated for a single 1-h time point under anaerobic conditions. Samples were prepared in the absence of DXP synthase, and both the reaction solutions and DXP synthase were purged in septa-capped Eppendorf tubes in the anaerobic chamber for 15 min at 30°C. All solutions were capped before removing from the chamber, and incubated at 0°C for 5 min. Pyruvate consumption was initiated upon addition of DXP synthase by syringe. After 1 h at 0°C, the samples were transferred to 100°C to quench the reactions. Precipitated DXP synthase was removed by centrifugation, and the pyruvate remaining in the supernatant was determined using the LDH assay as described above. Reactions were carried out in triplicate.

Steady state CD analysis of DXP synthase under anaerobic conditions
DXP synthase (30 M) in 50 mM HEPES, pH 8, 100 mM NaCl, 1 mM MgCl 2 , 1 mM ␤-mercaptoethanol, 0.2 mM ThDP (3 ml total) was deoxygenated in an anaerobic chamber for 30 min at 25°C. The sample was transferred to a septa-capped cuvette (Starna Cells, 1-Q-10-ST-S), which was closed before removal from the chamber. A DXP synthase only scan was obtained from 450 to 280 nm at 4°C with a 1-nm step and 1.6-s averaging time. Pyruvate (7-401 M) was titrated into the sample via syringe, and the CD spectrum after each addition of pyruvate was recorded as described above. The titration was carried out in triplicate. D-GAP (1 mM) or O 2 (via aeration of the sample by pipette) were added after the final pyruvate addition, and a spectrum obtained by the scan method described above. The normalized CD signal at 313 nm, indicative of LThDP on DXP synthase (15), was plotted against the concentration of pyruvate to obtain an apparent K d(pyruvate) and the plot was fit to a one-site specific binding curve in GraphPad Prism Version 6. All K d determinations were performed in triplicate.

Steady state kinetic characterization of DXP formation under aerobic and anaerobic conditions
DXP formation was monitored using the previously described DXP synthase/IspC-coupled assay with some modifications (14). Pyruvate concentration was varied (6.25-1000 M) while holding D-GAP constant at 500 M, or D-GAP concentration was varied (3.125-1000 M) while holding pyruvate constant at 1 mM with 100 nM DXP synthase, 2 M IspC, and 200 M NADPH in 100 mM HEPES, pH 8, 5 mM NaCl, 2.5 mM tris(2carboxyethyl)phosphine, 2 mM MgCl 2 , and 1 mM ThDP. When the D-GAP concentration was varied, samples were prepared without pyruvate and initiated with pyruvate. When pyruvate concentration was varied, samples were prepared without pyruvate or D-GAP, and enzymatic reactions were initiated with both substrates. For the anaerobic characterization, solutions were prepared from anaerobic stocks in an anaerobic chamber at 26°C (either prepared from solids in the chamber with deoxygenated water or solutions made anaerobic by incubation in the chamber for 15 min before sample preparation). Reaction solutions in septa-topped cuvettes (Starna cells, 9B-Q-10-GL14-S) and initiation solutions in septa-topped vials were sealed before removal from the chamber. Reactions where initiated upon addition of substrate via syringe as described above, and the solutions were mixed by tilting the cuvettes. Samples for the aerobic characterization were treated similarly but with constant exposure to room air. For both the aerobic and anaerobic characterizations, the depletion of NADPH ( max ϭ 340 nm, ⑀ ϭ 6220 M Ϫ1 cm Ϫ1 ) was monitored over time at 25°C following initiation of DXP formation. The rates of NADPH depletion were used to determine the initial rates of DXP formation. Plots of initial rate of DXP formation against [substrate] were fit to the Michaelis-Menten equation in GraphPad Prism version 6. Reactions were performed in duplicate.

Pre-steady state CD characterization of LThDP formation under anaerobic conditions
For all pre-steady state experiments, the CD signal of LThDP at 313 nm was monitored at 6°C over time. All samples were prepared with anaerobic stocks (prepared from solids with deoxygenated water in an anaerobic chamber) in a portable glove box (Belart) under constant flow of 100% N 2 . Trace oxygen inside the glove box was removed with Anaero-Packs (Mitsubushi) and the O 2 level of the chamber was monitored with Anaero-indicator tablets (Mitsubushi). Anerobic conditions were established and maintained in the stopped-flow CD using the anaerobic accessory and anaerobic operation protocol provided by Applied Photophysics (36). To determine the rate of LThDP formation, DXP synthase (60 M) in buffer A (50 mM HEPES, pH 8, 1 mM ThDP, 100 mM NaCl, 2 mM MgCl 2 ) was rapidly mixed with an equal volume of pyruvate (2 mM) in buffer A. The CD signal at 313 was monitored for 10 s at 6°C and the data from 7 repetitive shots were averaged. The data from 0.01 to 10 s were fit to a double exponential association model (Equation 1) using SigmaPlot version 10, y͑t͒ ϭ y 0 ϩ a ϫ ͑1 Ϫ exp͑Ϫk 1 t͒͒ ϩ b ϫ ͑1 Ϫ exp͑ϪkЈ 1 t͒͒ (Eq. 1) where y 0 is the CD 313 at t ϭ 0, k 1 and k 1 Ј represent the rate constants of the faster and slower phases of LThDP formation, respectively, and a and b describe the maximum CD 313 for the faster and slower phases, respectively. Error represents the error of the curve fit.

Pre-steady state CD characterization of decarboxylation rates in the presence and absence of triggers under anaerobic conditions
The effect of O 2 on the rates of LThDP decarboxylation and resynthesis in the presence and absence of D-GAP were determined by conducting the experiments under anaerobic conditions. For the rate of LThDP decarboxylation in the absence of triggers, LThDP was performed in one syringe (60 M DXP synthase and 1 mM pyruvate in buffer A described above) and O 2 -dependent LThDP decarboxylation on DXP synthase mixed with an equal volume of buffer A in a second syringe under anaerobic conditions. The CD signal at 313 nm was monitored at 6°C for 5 s and the data from 7 repetitive shots were averaged. Determinations of the rates of LThDP decarboxylation in the presence of D-GAP and O 2 were conducted similarly except the second syringe contained either 1 mM D-GAP or oxygenated buffer A (prepared by bubbling 100% O 2 through 1 ml of buffer A for 1 h at 25°C), respectively. The data for LThDP decarboxylation were fit to a single exponential decay model (Equation 2) using SigmaPlot version 10, y͑t͒ ϭ y 0 ϩ a ϫ exp͑Ϫk 2 t͒ (Eq. 2) where y 0 is the CD 313 at t ϭ 0, a describes the minimum CD 313 , and k 2 represents the rate constant for LThDP decarboxylation. For the determination of LThDP decarboxylation rates, data points between 0.023 and 0.3 s were used in the (ϩ) O 2 analysis and data points between 0.023 and 0.165 s were used in the (ϩ) D-GAP analysis. Rates of LThDP resynthesis were determined using either the double exponential association model (Equation 1) for the (ϩ) D-GAP experiment or a single association model (Equation 3) for the (ϩ) O 2 experiment using SigmaPlot version 10, y͑t͒ ϭ y 0 ϩ a ϫ ͑1 Ϫ exp͑Ϫk 1r t͒͒ (Eq. 3) where y 0 is the CD 313 at t ϭ 0, a describes the maximum CD 313 , and k 1r represents the rate constant for LThDP resynthesis. Data points between 0.358 and 5 s were used for the analysis of LThDP resynthesis in the presence of O 2 and data points between 0.2 and 5 s were used in the analysis of LThDP resynthesis in the presence of D-GAP. Error represents the error of the curve fit.