Structure-function analyses of alkylhydroperoxidase D from Streptococcus pneumoniae reveal an unusual three-cysteine active site architecture

During aerobic growth, the Gram-positive facultative anaerobe and opportunistic human pathogen Streptococcus pneumoniae generates large amounts of hydrogen peroxide that can accumulate to millimolar concentrations. The mechanism by which this catalase-negative bacterium can withstand endogenous hydrogen peroxide is incompletely understood. The enzyme alkylhydroperoxidase D (AhpD) has been shown to contribute to pneumococcal virulence and oxidative stress responses in vivo. We demonstrate here that SpAhpD exhibits weak thiol-dependent peroxidase activity and, unlike the previously reported Mycobacterium tuberculosis AhpC/D system, SpAhpD does not mediate electron transfer to SpAhpC. A 2.3-Å resolution crystal structure revealed several unusual structural features, including a three-cysteine active site architecture that is buried in a deep pocket, in contrast to the two-cysteine active site found in other AhpD enzymes. All single-cysteine SpAhpD variants remained partially active, and LC-MS/MS analyses revealed that the third cysteine, Cys-163, formed disulfide bonds with either of two cysteines in the canonical Cys-78-X–X-Cys-81 motif. We observed that SpAhpD formed a dimeric quaternary structure both in the crystal and in solution, and that the highly conserved Asn-76 of the AhpD core motif is important for SpAhpD folding. In summary, SpAhpD is a weak peroxidase and does not transfer electrons to AhpC, and therefore does not fit existing models of bacterial AhpD antioxidant defense mechanisms. We propose that it is unlikely that SpAhpD removes peroxides either directly or via AhpC, and that SpAhpD cysteine oxidation may act as a redox switch or mediate electron transfer with other thiol proteins.

Streptococcus pneumoniae is a Gram-positive, catalase-negative, facultative anaerobe associated with a variety of infections in humans (1). During aerobic growth, S. pneumoniae produces millimolar concentrations of hydrogen peroxide (H 2 O 2 ) due to the activities of pyruvate and lactate oxidases, and the absence of significant disposal mechanisms. These levels are sufficient to kill other bacteria and even host cells (2)(3)(4)(5). However, it is unclear how S. pneumoniae protects itself from high levels of endogenous H 2 O 2 . Elucidation of protective mechanisms could lead to the development of novel therapeutics to disrupt this process, thereby making the pathogen more susceptible to oxidative stress.
Several aspects of the S. pneumoniae response to oxidative stress have been described (6), including enzymes such as superoxide dismutase (7), NADH oxidase (8), and thiol peroxidase (9). Additionally, spr0370 from S. pneumoniae strain R6 encodes a putative alkylhydroperoxidase D (SpAhpD). 3 Knocking out spr0370 reveals that AhpD is a pneumococcal virulence determinant involved in the response to oxidative stress (10).
Alkylhydroperoxidase (Ahp) family enzymes are found in both Gram-positive and Gram-negative bacteria and convert peroxides to alcohol and water (6). AhpC/F is the most common bacterial Ahp family peroxidase system (Fig. 1A). AhpC is a highly active two-Cys peroxiredoxin that reacts directly with peroxides, whereas the NADH-dependent flavoprotein AhpF reduces AhpC via a thiol-disulfide exchange reaction, regenerating the enzyme for another cycle (11).
The best-described AhpD is that of Mycobacterium tuberculosis (12) (Fig. 1B). Here, the ahpD gene is immediately downstream of ahpC and replaces the need for ahpF. MtAhpD contains a thioredoxin-like CXXC motif. Initially, it was shown to exhibit weak peroxidase activity in parallel with MtAhpC in vitro by using AhpF from Salmonella typhimurium as the reductase (12). It was later discovered that MtAhpD has a more important function of mediating electron transfer from dihydrolipoamide dehydrogenase (Lpd) and dihydrolipoamide succinyltransferase (SucB) to oxidized MtAhpC via thiol-disulfide exchange reactions (13), linking the M. tuberculosis Lpd-SucB metabolic pathway to antioxidant defense.
Over the last few years, the core active site sequence motif (E(X) 11 CXXC(X) 3 H) from MtAhpD has been used to identify putative AhpD proteins in other bacterial species (10, 14 -16) (Fig. S1). Although the CXXC motif is responsible for redox activity, Glu-118 and His-137 of MtAhpD form a proton shuttle that activates a peroxidatic cysteine by deprotonation. These residues are also conserved in AhpD from S. pneumoniae (SpAhpD) (Fig. 1C).
SpAhpD is encoded in a bicistronic operon, downstream of the gene spr0371, which is predicted to be an integral membrane protein of unknown function (10). Given the structural and functional diversity of AhpD proteins, along with the differences in genome organization, it is difficult to infer how AhpD contributes to the oxidative stress response in S. pneumoniae. As such, we report the first biochemical characterization and crystal structure of SpAhpD to better understand this aspect of the pneumococcal antioxidant defense mechanism. We demonstrate that SpAhpD exhibits only weak peroxidase activity toward both H 2 O 2 and alkyl hydroperoxides, showing rates similar to those of AhpD proteins from other organisms. However, unlike the AhpC/D system in M. tuberculosis, SpAhpD is unable to reduce SpAhpC. The crystal structure of SpAhpD, determined at 2.3-Å resolution, reveals several unique structural features.

S. pneumoniae AhpD exhibits weak DTT-dependent peroxidase activity
The ability of SpAhpD to reduce H 2 O 2 , ter-butyl hydroperoxide (tBuOOH), and cumene hydroperoxide (CuOOH) was tested using the ferrous-oxidation xylenol orange (FOX) assay, with DTT as the reducing agent. Under assay conditions, SpAhpD exhibits net peroxidase activity (rate in presence of SpAhpD minus rate of peroxide with DTT alone) of 0.124 M/min toward H 2 O 2 , 0.127 M/min toward tBuOOH, and 0.575 M/min toward CuOOH ( Fig. 2A). Although these values are greater than the rate of peroxide decomposition in the absence of protein, they correspond to turnover numbers of less than 0.1 s Ϫ1 . Compared with SpAhpC, which shows greater initial velocities by nearly 2 orders of magnitude under the same assay conditions, the peroxidase activity of SpAhpD is weak.
The role of each cysteine residue of SpAhpD was investigated by testing the activity of single mutants against H 2 O 2 under the same assay conditions. All three mutants, C78S, C81S, and C163S, resulted in decreased peroxidase activity (Fig. 2B).  (13) report that because M. tuberculosis lacks AhpF, the Lpd and SucB systems act as the reductase, and AhpD mediates electron transfer from Lpd and SucB to AhpC, rather than being a peroxidase itself. C, local alignment of S. pneumoniae AhpD against the core sequence motifs of all AhpD sequences available from the Protein Data Bank. Multiple sequence alignment was carried out using ClustalOmega (29), and the figure was generated using ESPript 3.0 (30).

AhpD from S. pneumoniae
Interestingly, both C78S and C81S retained about 50% of the activity, whereas C163S retained 80% of its activity.
We confirmed that the ability of DTT to reduce the disulfide bond in SpAhpD was not rate-limiting, and therefore the reason for the low activity of SpAhpD. The free thiol content of SpAhpD and SpAhpC was measured with Ellman's reagent. DTT rapidly reduced both SpAhpC and SpAhpD (Fig. 2C). In contrast, when H 2 O 2 was added to pre-reduced SpAhpD, the thiol content decreased at a much slower rate than that of SpAhpC (Fig. 2D).

AhpD from S. pneumoniae does not mediate electron transfer to AhpC
To test whether SpAhpD can reduce SpAhpC via a thioldisulfide exchange reaction, as for MtAhpD (13), we used a gel-shift assay based on 4-acetamido-4Ј-maleimidylstilbene-2,2Ј-disulfonic acid (AMS). In this assay, AMS irreversibly alkylates thiols, adding 0.5 kDa for each free thiol, which is observed as a shift on nonreducing SDS-PAGE gels (17). Although oxidized AhpC forms intermolecular disulfide bonds and can easily be distinguished from reduced AhpC, AMS labeling is required for visualizing the redox state of SpAhpD. Pre-reduced SpAhpD in 5-fold excess was mixed with oxidized SpAhpC. No change in the redox state of oxidized SpAhpC was observed over the course of 30 min (Fig.  3A). Reduced thioredoxin from S. pneumoniae (SpTrx) was mixed with either pre-oxidized SpAhpC or SpAhpD for use as positive controls, and almost all of the oxidized SpAhpC was reduced by SpTrx upon mixing (Fig. 3B). Although it was Figure 2. DTT-dependent peroxidase activity of AhpD from S. pneumoniae toward three types of peroxide substrates. A, DTT-dependent peroxidase activity toward three types of peroxide substrates measured using FOX assay. AhpC from S. pneumoniae was used as a positive control. A negative control was performed by omitting protein. Peroxide concentration was measured using FOX reagent, and 300 M DTT was used as the reducing agent to induce turnover of peroxidases. Each reaction contained 10 M protein. Reactions were initiated by adding peroxide to a final concentration of 60 M. Bars represent the initial rate in M/min measured as the slope of the initial linear range of H 2 O 2 concentration versus time. Treatments were carried out in triplicate. The heights of the bars correspond to the median value in each treatment. B, DTT-dependent peroxidase activity of cysteine mutants toward H 2 O 2 . Bars show net activity as a percentage of the activity of WT AhpD. C and D, quantification of free thiol concentrations in AhpD and AhpC from S. pneumoniae using Ellman's assay. B, rates of cysteine disulfide bond reduction by DTT. C, rates of cysteine-free thiol oxidation by H 2 O 2 .

AhpD from S. pneumoniae
difficult to discern the change in intensity of the band corresponding to oxidized SpAhpD, bands corresponding to reduced SpAhpD and oxidized Trx indicated that thiol-disulfide exchange occurs between SpAhpD and SpTrx, albeit slowly (Fig. 3C). These findings confirm that the disulfide bond in both proteins can be reduced by thiol proteins under the assay conditions, and that SpAhpD does not reduce SpAhpC.
We then tested whether SpAhpD mediates electron transfer in the SpAhpCF system. By measuring both the consumption of NADH (at 340 nm) and H 2 O 2 concentration (using the FOX assay), we demonstrate that SpAhpF generates H 2 O 2 in an NADH-dependent manner ( Fig. 3D and Fig. S2). In comparison, the mixture of SpAhpC and SpAhpF has significantly reduced H 2 O 2 production, and the H 2 O 2 concentration decreases rapidly after initial generation (Fig. 3D, red line). We then added SpAhpD in 5-fold excess to test whether it influences activity of the SpAhpCF peroxidase system (Fig.  3D, blue line). However, SpAhpF generated H 2 O 2 at the same rate regardless of whether or not SpAhpD was present, and H 2 O 2 production and subsequent reduction by the SpAhpC/F system is also unaffected by SpAhpD (Fig. 3D). Taken together, these experiments demonstrate that SpAhpD does not mediate electron transfer to SpAhpC.

Crystal structure of AhpD from S. pneumoniae
Given that SpAhpD is functionally different from MtAhpD, we solved the crystal structure of recombinant SpAhpD to a resolution of 2.3 Å (Fig. 4A). The X-ray crystallography data collection and refinement statistics are shown in Table 1 (PDB 6E8L).
In contrast to the MtAhpD trimer and other hexameric AhpD structures in the Protein Data Bank, the SpAhpD structure assembled as a face-to-face dimer (Fig. 4A). The dimeric assembly of SpAhpD was confirmed by subsequent small-angle X-ray scattering and sedimentation velocity centrifugation experiments (discussed below). The structures of the three dimers in an asymmetric unit are almost identical (RMSD ϭ 0.297 and 0.367, respectively) (Fig. S3). Formation of the dimer results in a buried surface area of ϳ4,000 Å 2 , with an average ⌬G int of Ϫ31.4 kcal/mol.
Each SpAhpD monomer consists of nine ␣-helices, each ranging in size from 6 to 20 residues (Fig. 4A). SpAhpD has very low sequence homology (ϳ20% identity), when compared with the functionally well-characterized MtAhpD. However, like all AhpD structures in the Protein Data Bank, both proteins belong to a class of globin-like ␣-proteins and comprise nine ␣-helices.
The putative active site motif E(X) 11 CXXC(X) 3 H is located on helices ␣4 and ␣5 (Fig. 4, A and B), which are connected by a

AhpD from S. pneumoniae
180°(hairpin) turn motif (14). In SpAhpD, this section contains the redox-active CXXC motif, the highly conserved His-85 and Asn-76 residues, and a semi-conserved Glu-66 residue (Figs. 1, C and B). His-85, Glu-66, and a structural water molecule are thought to form a proton shuttle toward the CXXC motif to deprotonate the free thiols in AhpD from both M. tuberculosis (Fig. 4C) and P. aeruginosa (14,18), and these residues are conserved across most species.
The positioning of these conserved residues in SpAhpD ( Fig.  4B) is similar to that in MtAhpD (Fig. 4C), as is the apparent interaction between His-85 and Glu-66 at a distance of 2.6 Å. The water molecules in the SpAhpD active site are inconsistently positioned among the six protomers in an asymmetric unit and have a high temperature factor (average of 56 Å 2 comparing to global average of 46 Å 2 ), suggesting that the water molecule involved in the SpAhpD proton shuttle may have high mobility, whereas still interacting with His-85 to mediate the deprotonation of Cys-81. Overall, the crystal structure of SpAhpD shows that it has the similar all ␣-helical topology as AhpD from all other organisms and structurally conserved active site residues (E(X) 11 CXXC(X) 3 H), but the biological assembly is a distinct dimer.

S. pneumoniae AhpD forms an unusual three-cysteine active site buried in a deep pocket
Despite the conserved features described above, the active site of SpAhpD has several key differences that could make

AhpD from S. pneumoniae
SpAhpD mechanistically distinct and explain the observed functional difference. The most obvious feature is that the active site consists of three cysteines, rather than the usual two ( Fig. 4B). In addition to the conserved Cys-81 and Cys-78 residues, Cys-163 from helix ␣9 is also present. Although AhpD proteins from P. aeruginosa (PaAhpD; PDB 2O4D) and Ralstonia eutropha (ReAhpD; PDB 2PRR) also contain three cysteines, the third cysteine sits distant from the active site (Fig. S4). Cys-163 in SpAhpD is uniquely positioned in close proximity to both Cys-81 (5.1 Å) and the structural water (3.4 Å) of the proton shuttle, indicating that it may also react with Cys-81 to form a disulfide bond, which is a previously undescribed mechanism in AhpD proteins. It is somewhat further away (8.1 Å) from the second cysteine, Cys-78.
The active site residues of SpAhpD are located in a deep pocket formed between helices ␣1, ␣3, ␣4, and ␣5, with helix ␣9 on the bottom. Cys-81 and Cys-163 are located in the bottom of the pocket, whereas Glu-66 and His-85 are more solvent-accessible and positioned closer to the entrance of the pocket (Fig.  5A). Cys-78 is also solvent-accessible via the opposite face of the monomer. The active sites of most other AhpD proteins are more exposed or in a shallower groove, as seen in MtAhpD (Fig. 5B). In contrast, the pocket on SpAhpD appeared to be fully enclosed and planar in geometry, with a depth of ϳ15 Å. It was only accessible via an opening ϳ9 Å in width. Generating vacuum electrostatic using PyMOL predicted the interior of this pocket to be mildly hydrophobic to slightly positive electrostatic.
In MtAhpD His-132 has been shown to play a functional role by deprotonating the resolving cysteine (equivalent to Cys-78 in SpAhpD) (18). In SpAhpD the equivalent residue is replaced by Phe-80, which is unable to deprotonate the resolving cysteine (Cys-78), and there is no basic residue near Cys-78 in the crystal structure. Therefore, Cys-78 of SpAhpD is likely to be less reactive than the equivalent resolving cysteine in MtAhpD.

The third cysteine residue, Cys-163, forms novel disulfide bonds with both cysteine residues in the CXXC motif
We used LC-MS (LC-MS/MS) to investigate the reactivity of Cys-163 toward the other two cysteine residues. This was performed on chymotryptic digests of SpAhpD treated with various concentrations of H 2 O 2 followed by N-ethylmaleimide (NEM), which specifically alkylates free thiols.
First, we identified the three free cysteine-containing peptides alkylated with NEM: NG-Cys-78 -AF, Cys-81-VAGHTAF, and Cys-163-NY, and confirmed their fragmentation patterns in the reduced sample (Fig. S5). Their respective relative abundance decreased as they were treated with increasing concentrations of H 2 O 2 (Fig. 6, A-C). The relative abundance of the Cys-163 peptide decreased more slowly (Fig. 6C), suggesting that it may be less reactive than the other two cysteines. However, when treated with 0.5 mM H 2 O 2 , all three free cysteine containing peptides had near undetectable abundance, meaning all of them are either involved in disulfide bonds or hyperoxidized.
All three possible combinations of disulfide bonds were detected when SpAhpD was treated with H 2 O 2 (Fig. 6, D-F, and Fig. S5). Interestingly, their relative abundance peaked at 0.1-0.5 mM and decreased at higher H 2 O 2 concentrations (Fig. 6, D-F). This is likely due to the formation of irreversible oxidation products, such as sulfinic (-SOOH) and sulfonic acids (-SO 3 H) (Fig. S6). The relative abundance of the disulfide bond formed within the CXXC motif and between Cys-78 and Cys-163 peaked at 0.1 mM H 2 O 2 treatment, whereas the disulfide bond between Cys-81 and Cys-163 peaked at 0.5 mM, suggesting that the latter may be slightly less favorable. These concentrations of H 2 O 2 are generated by S. pneumoniae grown in culture media. These experiments demonstrate that Cys-163 is redox active and able to form a disulfide bond with either Cys-78 or Cys-81 under physiological conditions.

The highly conserved Asn-76 residue within the AhpD core motif is required for the folded structure of the active site
Multiple sequence alignment of the AhpD core motif reveals that Asn-76 of SpAhpD is also highly conserved (Fig. 1C). We propose that Asn-76 may serve a structural purpose, stabilizing the folded active site structure in all AhpD proteins. In our structure of SpAhpD, Asn-76 forms a hydrogen bond with the side chain of Asn-164 on helix ␣9 (Fig. 7A). Asn-164 is conserved in all AhpD sequences, except for MtAhpD (Fig. S1), and the hydrogen bonding interaction is structurally conserved (Fig. 7B). In MtAhpD, Asn-128 (structurally homologous to

AhpD from S. pneumoniae
SpAhpD Asn-76) hydrogen bonds with the side chain of Asn-82, and possibly Arg-47 (Fig. 7C). In all known AhpD structures, the asparagine next to the CXXC motif (Asn-76 in SpAhpD) always hydrogen bonds with the side chain of another asparagine on the longest ␣-helix in the same subunit (Asn-164 in SpAhpD). This highly conserved interaction indicates that Asn-76 is important for the function of AhpD, probably by stabilizing the folded structure of the active site.
To test this theory, we constructed two mutants, N76A and N76L, by site-directed mutagenesis and attempted to express and purify them. However, in a small-scale expression trial both asparagine mutants were insoluble (Fig. S7), probably due to the inability to fold properly and consistent with the hypothesis that it stabilizes the structure of the active site.

AhpD from S. pneumoniae forms an unusual dimeric quaternary structure
Our crystal structure demonstrates that SpAhpD forms a dimer (Fig. 4A). This was surprising, because MtAhpD forms a trimer in solution (18), and all other AhpD crystal structures were predicted to form hexamers (trimer of dimers) (19) (Fig. 8). All the hexameric and trimeric AhpD proteins have 3-fold rotational symmetry with a central cavity as the axis of symmetry, whereas SpAhpD lacks both the 3-fold symmetry and a central cavity. Analytical ultracentrifugation and small-angle X-ray scattering experiments confirm that SpAhpD is also a dimer in solution, consistent with the crystal structure.
Sedimentation velocity type analytical ultracentrifugation experiments were conducted across a concentration range of 0.66 -4.0 mg/ml. At all concentrations tested, the enzyme is monodisperse and not involved in a self-association with higher order species (Table S1, Fig. S8). The buoyant molecular mass of the peak is 39.5 kDa, which is very close to the mass of the dimer (39.7 kDa) calculated from the sequence (Fig. 9A).
We also collected small-angle X-ray scattering data for both the reduced and oxidized SpAhpD enzyme to test whether the shape of the protein in the crystal is representative of that in solution and examine whether the oxidation state of the protein affects its structure. The redox states of the protein were confirmed using Ellman's assay and the scattering profiles are shown in Fig. 9 (B and C). The Guinier plots for both data sets are linear (R 2 ϭ 0.999), indicating minimal aggregation and no inter-particle interference.
The scattering profile for both the oxidized and reduced samples are in good agreement with the theoretical scattering profile calculated using the dimer from the crystal structure ( 2 values of 0.486 and 0.694, respectively) ( Fig. 9, B and C, red line). By calculating the pairwise distance distribution (P(r)), both the radius of gyration and the molecular weight are estimated to be very similar for both the oxidized and reduced forms of SpAhpD, so is the maximum distance ( Table 2). The (P(r)) plot reveals a more prominent difference, where oxidized SpAhpD had clearly lower P(r) at a distance of 35-50 Å. (Fig. 9D). Comparison of the scattering profiles also revealed small differences at 0.1-0.2 Å Ϫ1 (Fig. 9E). Taken together, these findings confirm that the dimer observed in the crystal structure represents the in-solution structure of SpAhpD and suggests that the oxidation state of the protein may have a small effect on the global structure.

Discussion
In this study, we reveal the unusual biochemical and structural properties of AhpD from S. pneumoniae. The activity of SpAhpD does not appear to fit into the existing models of bacterial AhpD antioxidant defense mechanisms: although it is  (Table S2).

AhpD from S. pneumoniae
oxidized by hydroperoxides it is a weak peroxidase, and does not mediate electron transfer to AhpC. Results of the DTT-dependent FOX assay and Ellman's assay clearly demonstrate that SpAhpD reacts very slowly with H 2 O 2 . Because of this low activity, we could not induce a Michaelis-Menten-type response in the DTT-dependent kinetic assay. However, the resultant turnover (Ͻ0.1 s Ϫ1 for all three peroxides tested) was 4 orders of magnitudes lower than that of the characterized thiol peroxidase, TpxD, from S. pneumoniae (9). Given the H 2 O 2 con-centrations to which S. pneumoniae is routinely exposed, the direct removal of H 2 O 2 by SpAhpD is unlikely to be important for protection against oxidative damage.
We also tested the hypothesis that SpAhpD mediates electron transfer to SpAhpC, similar to its homologue in M. tuberculosis (13,20). However, the AMS gel-shift assay showed that SpAhpD does not reduce SpAhpC, whereas the NADH-dependent FOX assay confirmed that SpAhpD does not affect the putative AhpC/F system in S. pneumoniae.

AhpD from S. pneumoniae
SpAhpF alone appears to generate H 2 O 2 , in an NADH-dependent manner, consistent with bifunctional AhpF (NADH oxidase and AhpC reductase) and AhpC/F peroxidase systems in certain other species, including Amphibacillus xylanus, Sporolactobacillus inulinus, and Streptococcus mutans (21,22). Given the high degree of sequence similarity between S. pneumoniae and S. mutans AhpF this dual activity was expected, and may contribute to H 2 O 2 generation in S. pneumoniae, in addition to pyruvate oxidase and lactate oxidase. Although the ahpC (GenBank CMAP01000085.1, locus tag: ERS022390_02365) and ahpF (NZ_CHHM010000 86.1) genes are present in certain strains of S. pneumoniae, their expression levels have not been reported.
The finding that SpAhpC does not functionally interact with SpAhpD was also supported by the observed genome organization. In M. tuberculosis, ahpD is adjacent to ahpC in the same operon, whereas spr0370 (encoding AhpD) from S. pneumoniae is expressed in a bicistronic operon with downstream gene spr0371, which encodes a putative voltage-dependent anion channel family membrane protein of unknown function (10,23). In this respect, SpAhpD is similar to AhpD from C. glutamicum, whose corresponding gene is preceded by a fourcomponent gene cluster encoding an ABC-type nickel/peptide transporter (15).
The weak peroxidase activity of SpAhpD is consistent with several previously-characterized AhpD proteins. Weak activity was initially demonstrated for AhpD from M. tuberculosis (12), with later studies reporting similar findings in P. aeruginosa (14) and Anabaena sp. PCC7120 (16). Interestingly, the primary function of MtAhpD was later shown to be the linking of reducing potential from lipoamide-containing metabolic enzymes to the peroxiredoxin AhpC via thiol-disulfide exchange reactions (20), suggesting the peroxidase activity is an artifact of the redox active motif. We propose that this may also be the case for the AhpD proteins from S. pneumoniae, P. aeruginosa, and Anabaena sp. PCC7120 because they all share the same core sequence motif.
We sought to use structural information to unravel the biological function of SpAhpD. The crystal structure obtained in this study confirmed that SpAhpD contains the structurallyconserved core functional motif (E(X) 11 CXXC(X) 3 H). The structure of the putative active site, with favorable conformation for the formation of disulfide bonds between cysteine residues, is consistent with SpAhpD being a redox active protein. Although SpAhpD is ineffective at reducing AhpC it could reduce other thiol proteins, conferring protection from oxidative stress. This would explain why various AhpD proteins appear to contribute to the oxidative stress response in vivo (10, 14 -16), despite only exhibiting weak peroxidase activity. This hypothesis is supported by studies on C. glutamicum AhpD (15), which is strongly linked to the cellular NAD ϩ /NADH ratio in vivo. In M. tuberculosis, AhpD acts between Lpd-SucB and AhpC, but the reactivities between such bacterial peroxidase-related thiol proteins have been shown to be promiscuous and species-dependent in many cases (11,12,24,25), further complicating the question of the biological activity of AhpD. Alternatively, thiol exchange may not occur with other proteins, but SpAhpD may interact with different bacterial proteins depending on

AhpD from S. pneumoniae
whether it is reduced or oxidized. Mammalian thioredoxin dissociates from ASK1 upon oxidation, leading to activation of this kinase (26). Further investigation of the interacting partners of SpAhpD by pulldown assays or genetic complementation studies are required to understand its biological activity.  A and B of 6E8L). Top, residual plot of the fitness between experimental and predicted scattering curves. D, pairwise distance distribution plot [P(r)] of reduced (black) and oxidized (red) SpAhpD. E, comparison of SAXS scattering profile of reduced SpAhpD with oxidized SpAhpD using the I ox (q)/I red (q) plot. F and G, ab initio models (gray bead models) generated from SAXS scattering profile of reduced (F) and oxidized (G) SpAhpD using DAMMIF superimposed with the SpAhpD dimer crystal structure (green).

AhpD from S. pneumoniae
The combination of several conserved structural features with large differences in the folding and sequence of AhpD proteins from different bacterial species has been attributed to convergent evolution (14). Although SpAhpD shares the common features of the "hairpin" motif (E(X) 11 CXXC(X) 3 H) (14) and an all-␣-helical topology with AhpDs from other organisms, analysis of our crystal structure revealed a number of structural features not previously observed in AhpD family proteins. We also demonstrated that a structurally well-conserved asparagine residue (Asn-76 of SpAhpD) within this core motif is important for the folding of this core structure, and it should be considered part of the AhpD family core sequence motif (E(X) 9 NXCXXC(X) 3

H).
First, the redox-active CXXC motif is located in a deep pocket in SpAhpD, whereas it tends to be more exposed in AhpD proteins from other species. This suggests that the reactive cysteines are less accessible to larger molecules and explains why, in contrast to the M. tuberculosis orthologue, SpAhpD did not react with SpAhpC. This also indicates that the active site of SpAhpD may be more selective toward its substrate. The planar geometry and mildly hydrophobic nature of this pocket is indicative of the type of substrate it may bind. This explains the observation during our kinetic experiments that reactivity of SpAhpD toward CuOOH is five times greater than that of H 2 O 2 and tBuOOH. The peroxidase activity of SpAhpD may be selective toward aromatic alkyl hydroperoxide substrates as a result of the structure of the active site pocket. Importantly, this is also a favorable feature for a putative drug target because it enables rational design of a highly selective competitive inhibitor by exploiting the geometry of the active site pocket.
Second, a third cysteine (Cys-163) from helix ␣9 is also present in the deep pocket, near the CXXC motif. The position of the proton shuttle, consisting of Glu-66, His-85, and the conserved water molecule (14), suggests that it may interact with Cys-163, resulting in deprotonation. Although SpAhpD is not the only AhpD containing three cysteines (Fig. S1), the third cysteines in AhpD from P. aeruginosa and R. eutropha are both distant from the active site. In fact, the third cysteines in both PaAhpD and ReAhpD are symmetrically located on the dimer interface, positioned toward the equivalent cysteine from another chain, suggesting intermolecular disulfide bond formation (Fig. S4). This arrangement is obviously divergent from the intramolecular disulfide bond in SpAhpD.
Testing the activity of cysteine mutants demonstrated that they all had attenuated peroxidase activity, but were still reactive with peroxides. Both single mutants of the canonical CXXC motif retained 50% of the WT activity, whereas C163S retained 80% activity. This is in clear contrast to the C130S and C133S mutants of MtAhpD that were almost completely inactive (27). This suggests that the additional Cys-163 is involved in the activity of SpAhpD, and it offers a degree of redundancy than MtAhpD. This may be important for retaining its function in S. pneumoniae under increased endogenous levels of H 2 O 2 , where oxidative inactivation becomes significant for some peroxidases (28). In such circumstance, Cys-163 may react with cysteine-SOH before H 2 O 2 to prevent hyperoxidation, or act as a replacement when one of the canonical cysteines has been hyperoxidized, thereby retain AhpD activity at high H 2 O 2 concentrations.
Our subsequent LC-MS/MS results demonstrated that Cys-163 is able to form disulfide bonds with either cysteine residues in the CXXC motif. As expected, the disulfide bond between the canonical CXXC motif cysteines is the most common, whereas Cys-163 appears to form disulfides at higher H 2 O 2 concentrations. Although Cys-81 was the most reactive free thiol (Fig. 6B) as expected from its activation from the proton shuttle, Cys-81 was the least susceptible to form sulfonic acid (Fig. S6). This could be explained by the fact that a significant proportion of it is protected by disulfide bonding with Cys-163 after treating with high concentrations (1-3 mM) of H 2 O 2 , as shown in Fig.  6F.
Third, Phe-80 in SpAhpD replaces the equivalent His-132 in MtAhpD, which has been shown to play a functional role of deprotonating the resolving cysteine (equivalent to Cys-78 in SpAhpD) (18). Obviously, this interaction cannot occur in SpAhpD, nor is there an adjacent basic residue that could perform a similar function. As a result, Cys-78 in SpAhpD may be less reactive than the equivalent resolving cysteine in MtAhpD. His-132 is unique at this position to M. tuberculosis. Substitution with tyrosine is more common, followed by phenylalanine (Fig. 1C), but neither substitution is likely to replace the function of His-132 in MtAhpD. Therefore, the resolving cysteine in other organisms may also be less reactive than their equivalent in MtAhpD.
Finally, we demonstrated that SpAhpD forms a dimer in solution using AUC and SAXS analyses. The SAXS profile indicates that the structure of the dimer in the crystal is consistent with that in solution. Interestingly, there is a small difference between the oxidized and reduced solution structures, which may be a result of the alternative disulfide bonding in the active site. This is also supported by our LC-MS/MS results that show the Cys-78 -Cys-163 disulfide bond is more readily formed than Cys-81-Cys-163, despite the fact that the distance between them in the crystal structure (8.1 Å) is clearly greater than that between the latter (5.1 Å). Such distance would require protein motion upon oxidation, which is consistent with the subtle difference observed in our SAXS data. The biological assembly of SpAhpD is a twisted oblate-shaped homodimer. This is in clear contrast to the trimeric MtAhpD enzyme and all other hexameric AhpD structures, which have 3-fold rotational symmetry with a central cavity as the axis of symmetry. Although the function of trimeric and hexameric quaternary structure in other AhpD proteins remain unclear, the unique in-solution structure also indicates that SpAhpD AhpD from S. pneumoniae may differ functionally from the better-characterized MtAhpD protein.
In conclusion, our biochemical and structural studies of SpAhpD reveal a number of unique features that indicate the biological role of this protein is more complex than originally anticipated. The cysteine residues of SpAhpD active site are redox active, but the structure makes them ineffective at influencing H 2 O 2 levels either directly or through provision of reducing equivalents to AhpC enzymes. The structure of the unique active site pocket and its higher reactivity toward CuOOH indicate that it may be selective toward certain planar substrates. The additional Cys-163 in the active site provides SpAhpD with a degree of redundancy, which may make it less prone to oxidative inactivation, a feature beneficial for its function within S. pneumoniae. Investigation of the interaction partners of SpAhpD, and assessment of its redox status in vivo will help to illuminate the role of this protein in S. pneumoniae. Moreover, the difference in activities between MtAhpD and SpAhpD highlights that the functions of AhpD enzymes can be species-dependent and need to be examined individually in each species.

Materials and methods
Unless otherwise indicated, chemicals were purchased from AppliChem, Roche, or Sigma-Aldrich.

Cloning, protein expression, and purification
The sequences of the genes encoding SpAhpD (spr0370), SpAhpC (GenBank CMAP01000085.1, locus tag: ERS022390_ 02365), SpAhpF (NZ_CHHM01000086.1), and SpTrx (Gen-Bank CJK72264.1) were obtained from the NCBI GenBank TM database and codon-optimized for expression in Escherichia coli K12 (high) using EMBOSS Backtranseq (32). Linear DNA was then synthesized by Thermo Fisher Scientific and cloned into the expression vector pOPINF (a gift from Ray Owens; Addgene plasmid number 26042), linearized with KpnI/Hin-dIII, using an In-Fusion HD cloning kit (Clontech) as per the manufacturer's instructions (33). Plasmids were then purified from an overnight culture of a single E. coli Stellar transformant using a miniprep plasmid purification kit (iNtRON Biotechnol-ogy) and the DNA sequence was confirmed (Macrogen). Purified plasmid was then transformed into E. coli Tuner pLacI cells (Novagen) for overexpression and purification.
Recombinant protein for subsequent experiments was expressed by culturing E. coli Tuner pLacI transformant cells in LB broth to an A 600 of 0.4 -0.6, then inducing with 1 mM isopropyl ␤-D-1-thiogalactopyranoside for 18 h at 25°C. The cells were then collected by centrifugation and lysed by sonication in HisTrap loading buffer (20 mM Tris, 150 mM NaCl, 30 mM imidazole, pH 8.0). The crude lysate was clarified by centrifuging at 33,000 ϫ g (Thermo Sorvall RC-6-plus centrifuge) for 45 min, followed by filtration through a Minisart NML Syringe Filter (0.2-m pore size).
The clarified lysate was then loaded onto a 5-ml HisTrap High Performance column (GE Healthcare) pre-equilibrated with HisTrap loading buffer using an ÄKTA pure protein purification system (GE Healthcare Life Sciences). Bound protein was eluted in a gradient of high imidazole buffer (20 mM Tris, 150 mM NaCl, 300 mM imidazole, pH 8.0) over a 50-ml retention volume. Fractions containing the protein of interest were identified using SDS-PAGE and pooled for subsequent purification.
The His tag was cleaved from recombinant SpAhpD using Human Rhinovirus 3C Protease (HRV3C; Novagen). The Histagged HRV3C protease and cleaved His tag were removed by HisTrap chromatography (Fig. S9). The pooled protein was then concentrated to a final volume of 2 ml before being purified via size-exclusion chromatography (HiLoad 16/60 Superdex 200; GE LifeSciences) using SEC buffer (20 mM Tris, 150 mM NaCl, 3 mM DTT, pH 8.0). Fractions containing the protein of interest were identified using SDS-PAGE, pooled, frozen in liquid nitrogen, and stored at Ϫ80°C until use.
Electrospray-ionization quadrupole-TOF MS was used to confirm the mass of the purified protein. The sequences of all recombinant proteins used in the present study are shown in Fig. S10.

Site-directed mutagenesis
To construct mutant proteins of SpAhpD, pOPINF construct encoding WT SpAhpD was mutated using the In-Fusion HD cloning kit (Clontech) following the manufacturer's instructions (TaKaRa) (34). The In-Fusion cloning products after mutagenesis were transformed into E. coli Stellar-competent cells. Plasmids were then purified from an overnight culture of a single colony using a miniprep plasmid purification kit (iNtRON Biotechnology) and the DNA sequence was confirmed (Macrogen). Plasmid containing correct mutation was then transformed into E. coli Tuner pLacI cells (Novagen) for over-expression and purification, following the same procedures as the WT SpAhpD.

Preparation of protein redox state
Unless otherwise indicated, all proteins used in the following activity assays were prepared using the following methods to ensure consistent redox state at the onset of the activity assay. To prepare reduced protein, freshly-prepared DTT solution was mixed with protein at a final concentration of 3 mM and incubated for 1 h at 4°C. To prepare oxidized protein, H 2 O 2 was mixed with protein at a final concentration of 1 mM and incubated at 4°C overnight. DTT and H 2 O 2 were removed prior to use by desalting the protein into TBS solution (TBS, 20 mM Tris, 150 mM NaCl, pH 8.0, at 4°C) using a HiTrap Desalting column with Sephadex G-25 resin (GE Healthcare Life Sciences) following manufacturer's instructions.

DTT-dependent FOX assay
To test the ability of SpAhpD to reduce H 2 O 2 and alkyl hydroperoxides, peroxide concentrations in the reaction mixtures were measured using the FOX assay, as previously described (14). Briefly, 10 l of reaction mix were mixed with 190 l of FOX reagent (250 M ammonium ferrous sulfate, 125 M xylenol orange, 100 mM sorbitol, and 25 mM sulfuric acid) in a clear-bottom 96-well-microplate and incubated for 20 min at room temperature. The absorbance at 560 nm was monitored using a microplate reader. Peroxide concentration was calculated using a standard curve for each type of peroxide (H 2 O 2 , tBuOOH, and CuOOH).
To test the peroxidase activity of SpAhpD, 10 M SpAhpD was mixed with 300 M DTT in TBS at room temperature. The reaction was initiated by adding peroxide to a final concentration of 60 M. The concentration of peroxide was determined at various time points up to 2 h, as described above.

Ellman's assay
Ellman's reagent (5,5Ј-dithio-bis-(2-nitrobenzoic acid) reacts readily with free thiol to produce chromophoric 2-nitro-5-thiobenzoate, which absorbs strongly at 412 nm, enabling colorimetric determination of free thiol concentrations. The concentration of free thiol in either SpAhpD or SpAhpC was determined using Ellman's reagent as per the manufacturer's instructions (Thermo Fisher Scientific). Briefly, Ellman's reagent (0.08 mg/ml final concentration) was first dissolved in buffer (0.1 M sodium phosphate, 1 mM EDTA, pH 8.0). A 200-l volume of the reagent solution was then mixed with 20 l of sample in a flat-bottom microplate and incubated at room temperature for at least 15 min. The absorbance was then measured at 412 nm. Thiol concentration was determined using a standard curve generated using cysteine standards.
SpAhpD and SpAhpC were used at a final concentration of ϳ80 M. To test their ability to react with DTT, pre-oxidized protein was mixed with DTT at a final concentration of 240 M. At various time points, aliquots of reaction mixtures were removed and immediately desalted into TBS using Micro Bio-Spin 6 columns (Bio-Rad) as per the manufacturer's instructions. The free thiol concentration was determined using Ellman's assay immediately after desalting. The protein concentration was determined using Bio-Rad Protein Assay Dye Reagent and calculated using a standard curve generated using BSA. The number of free thiols per molecule was calculated by dividing free thiol concentration in micromolar by protein concentration in micromolar. To test the reactivity of SpAhpD and SpAhpC with H 2 O 2 , pre-reduced proteins were reacted with H 2 O 2 at a final concentration of 500 M. Free thiol and protein concentrations were determined at various time points using the same method.

Gel-shift assay using 4-acetamido-4maleimidylstilbene-2,2disulfonic acid (AMS)
A gel-shift assay using AMS was used to determine whether thiol-disulfide exchange occurs between SpAhpD and SpAhpC. Labeling of free thiol with AMS was carried out as described previously (35). Briefly, 20 mM AMS was first dissolved in buffer (50 mM sodium phosphate, pH 7.0). Derivatization was performed by mixing equal volumes of protein sample (Յ1 mg/ml) and AMS solution, followed by incubation at room temperature for at least 10 min. Three standards were prepared for each protein: reduced, oxidized, and native. The reduced and oxidized standards were prepared by derivatizing either reduced or oxidized protein with AMS, whereas the native standard was prepared by derivatizing reduced protein with NEM. The reaction was initiated by mixing 20 M oxidized SpAhpC with 80 M reduced SpAhpD. At various time points, aliquots were removed and mixed with an equal volume of AMS solution. Samples were then diluted 5-fold and analyzed by nonreducing SDS-PAGE on 4 -12% BisTris gels. Positive controls were included for both SpAhpC and SpAhpD to confirm that the thiol-disulfide exchange reaction did occur under the assay conditions. The control reactions consisted of 20 M oxidized SpAhpC or SpAhpD mixed with 80 M reduced SpTrx and were analyzed as described above.

NADH-dependent kinetic assay
To test the reductase activity of SpAhpF toward SpAhpD or SpAhpC, 2 M AhpF was mixed with 2 M SpAhpC and/or 10 M SpAhpD. The reaction was initiated by adding NADH to a final concentration of 300 M. The H 2 O 2 concentration was determined by FOX assay. The rate of NADH reduction was monitored by measuring the absorbance at 340 nm.

Crystallization
Purified AhpD in SEC buffer was concentrated to 19 mg/ml. Crystals were grown in sitting drops at 8°C by mixing 400 nl of protein solution with 400 nl of reservoir solution in a TTP LabTech Mosquito Crystal Unit. The optimal reservoir contained 0.1 M Bistris propane, 0.2 M MgCl 2 , 25% (w/v) PEG 3350, and 6% (v/v) 1,2-propanediol, pH 5.5. Optimum crystal growth was initiated using seed stock prepared from an initial crystallization hit using Seed Beads (Hampton Research), as per the manufacturer's instructions. The crystals appeared as large, transparent, triangular, or quadrilateral plates.

X-ray diffraction data collection, phasing, and refinement
X-ray diffraction data were collected at the Australian Synchrotron on the micro-crystallography (MX2) beamline equipped with an EIGAR 1M detector. The dataset was reduced using AIMLESS (36) via CCP4i2 (37). The space group was P2 1 2 1 2 1 , with unit cell dimensions of a ϭ 65.296 Å, b ϭ 84.233 Å, and c ϭ 183.846 Å. Although Matthew's coefficient (38) suggests that both five and six molecules in an asymmetric unit are equally probable (0.42 and 0.40, respectively), there were six molecules present. The solvent content was 42.12%, and the V M was 2.12 Å 3 Da Ϫ1 .
The phases were solved by molecular replacement using PHASER. AhpD from S. mutans (PDB 3LVY) was used as the AhpD from S. pneumoniae search model, and it was truncated using CHAINSAW (39) prior to molecular replacement. REFMAC5 (40) was initially used for rigid body refinement of the molecular replacement model, and was subsequently used in alternating cycles with Coot (CCP4 program suite) (41), which was used to improve the structure model by restrained refinement.
Recombinant SpAhpD consists of residues 2-182 from the WT AhpD sequence, plus additional N-terminal Gly-Pro residues from the expression vector. Each chain of the refined structural model contained residues 2-183, 3-183, or 4 -183 of the recombinant SpAhpD sequence. The first one to three residues of the N-terminal residues, which form a random coil, have high temperature factors and low electron density on the 2F o Ϫ F c map, and were therefore not included in the structural model.

Liquid chromatography-MS (LC-MS/MS)
Initially, the SpAhpD for LC-MS/MS was pre-reduced with 3 mM DTT. The DTT was removed using a HiTrap Desalting column. The desalted protein solution was divided into five equal aliquots mixed with an equal volume of diluted H 2 O 2 at various concentrations to make final concentrations of 0 -3 mM. The SpAhpD samples were left to incubate with H 2 O 2 at 4°C for 18 h, before adding NEM to a final concentration of 50 mM to alkylate and stabilize the remaining free thiols.
Chymotrypsin was added to all samples at a 20:1 substrate: chymotrypsin weight ratio and incubated at 25°C overnight. Digested samples were analyzed using a Thermo Scientific Velos Pro ion trap mass spectrometer coupled to a Dionex Ulti-Mate 3000 HPLC system with a 50-l injection loop (Thermo Scientific, Waltham, MA). Samples were stored on the autosampler tray at 5°C. A Jupiter 4-m Proteo 90Å column (150 ϫ 2 mm, Phenomenex, Torrance, CA) was used for chromatographic separation using 100% water (0.1% formic acid) as Solvent A and 100% acetonitrile (0.1% formic acid) as Solvent B. The column temperature was set to 40°C. The column was equilibrated with 95% Solvent A and 5% Solvent B for 5 min and then a linear gradient was run for 45 min to 5% Solvent A and 95% Solvent B to achieve separation. The column was then flushed with 95% Solvent A and 5% Solvent B for 5 min and re-equilibrated at initial conditions for 5 min. A flow rate of 0.2 ml/min was used and ϳ30 g of digested protein was injected for each sample. Nitrogen was used as sheath gas. The temperature of the heated capillary was 275°C. Data were analyzed using Thermo Xcalibur Qual Browser 2.2 SP1.48 (Thermo Fisher Scientific Inc., Waltham, MA).
The m/z values of peptides of interest were predicted and collision-induced dissociation-MS/MS spectra in positive-ion mode were acquired for each of them. The collision energy was set at 40. Peptide fragments were manually assigned based on Roepstorff-Fohlman nomenclature (43). A representative fragmentation pattern for each peptide species is shown in Fig. S5. Each peptide species was quantified by post-acquisition filtering the MS/MS spectra obtained for a chosen abundant and characteristic fragment ion (listed in Table S2), and then mea-suring the area under the curve of the resulting peak (peak algorithm: Genesis, peak smoothing: Gaussian 7 points).

Small-angle X-ray scattering
Small-angle X-ray scattering was performed on the SAXS/ WAXS beamline at the Australian Synchrotron. The protein sample (10 mg/ml) was loaded onto an inline Superdex 200 5/150 GL size-exclusion column (GE Healthcare), pre-equilibrated with running buffer (20 mM Tris, 150 mM sodium chloride, 0.1% sodium azide, 5% glycerol, pH 8.0) to remove any aggregate prior to data collection. The fractionated sample was pumped through a capillary where it was exposed to the X-ray beam. The X-ray wavelength was 1.0332 Å. A 1-M Pilatus detector was positioned 1600 mm from the sample. A total of 500 detector images were collected using a 1-s exposure time and a flow rate of 0.45 ml/min. SAXS data were analyzed using the ATSAS program suite AU (44). Data points from the peak of the SEC chromatogram were selected using CHROMIXS (45). Two-dimensional intensity plots were radially-averaged, normalized to sample transmission, and background-subtracted using the SCAT-TERBRAIN software package (Australian Synchrotron). PRIMUS QT was used to generate the Guinier plots and pairwise distance distribution P(r) plots, as well as for calculating the Porod volumes and molecular weight estimates. The theoretical scattering curves based on the crystallographic structure were generated and compared with experimental scattering curves using CRYSOL (46). Scattering profiles of the reduced and oxidized samples were compared using ScÅtter (47).

Analytical ultracentrifugation (AUC)
AUC was performed using a Beckman Coulter XL-I analytical ultracentrifuge. The proteins were prepared in buffer containing 20 mM Tris, 150 mM NaCl, 1 mM Tris(2-carboxyethyl) phosphine, pH 7.5. The samples were loaded into Epon two-channel centerpieces with quartz windows in a four-hole An-60 Ti rotor. Sedimentation velocity experiments were performed at 55,000 rpm at 20°C using protein at various concentrations. Sedimentation profiles were measured at 280 nm using absorbance mode with a step size of 0.003 cm over 200 scans.
All AUC data were analyzed using SEDFIT (48). Buffer density and viscosity, and partial specific volume of the protein, were calculated using SEDNTERP (49). The sedimentation data were fitted to a continuous size-distribution model.