The basis for non-canonical ROK family function in the N-acetylmannosamine kinase from the pathogen Staphylococcus aureus

In environments where glucose is limited, some pathogenic bacteria metabolize host-derived sialic acid as a nutrient source. N-Acetylmannosamine kinase (NanK) is the second enzyme of the bacterial sialic acid import and degradation pathway and adds phosphate to N-acetylmannosamine using ATP to prime the molecule for future pathway reactions. Sequence alignments reveal that Gram-positive NanK enzymes belong to the Repressor, ORF, Kinase (ROK) family, but many lack the canonical Zn-binding motif expected for this function, and the sugar-binding EXGH motif is altered to EXGY. As a result, it is unclear how they perform this important reaction. Here, we study the Staphylococcus aureus NanK (SaNanK), which is the first characterization of a Gram-positive NanK. We report the kinetic activity of SaNanK along with the ligand–free, N-acetylmannosamine–bound and substrate analog GlcNAc–bound crystal structures (2.33, 2.20, and 2.20 Å resolution, respectively). These demonstrate, in combination with small-angle X-ray scattering, that SaNanK is a dimer that adopts a closed conformation upon substrate binding. Analysis of the EXGY motif reveals that the tyrosine binds to the N-acetyl group to select for the “boat” conformation of N-acetylmannosamine. Moreover, SaNanK has a stacked arginine pair coordinated by negative residues critical for thermal stability and catalysis. These combined elements serve to constrain the active site and orient the substrate in lieu of Zn binding, representing a significant departure from canonical NanK binding. This characterization provides insight into differences in the ROK family and highlights a novel area for antimicrobial discovery to fight Gram-positive and S. aureus infections.

Pathogenic bacteria import and utilize a wide variety of carbohydrates. In environments where glucose is limited, amino sugars such as sialic acids are important nutrient sources. In general, the bacterial sialic acid catabolic pathway is a validated antimicrobial target (1)(2)(3)(4), because sialic acids are scavenged from mucins that coat the gastrointestinal and respiratory tracts of humans (5) and are imported (6 -8) and used by pathogenic bacteria as a carbon, nitrogen, and energy source when they colonize these niches (1, 9 -11). This is particularly the case in mucosal surfaces, where the amino sugar GlcNAc and the sialic acid, N-acetylneuraminate, are widely incorporated into glycoconjugate mucins and cell surfaces (1). These carbohydrates can be scavenged from host glycoconjugates and metabolized by mucosal pathogens, such as Staphylococcus aureus (2), as a nutrient source (3)(4)(5)(6).
Just as the phosphorylation of glucose in the early phase of glycolysis (sometimes referred to as the "energy investment phase") drives otherwise unfavorable reactions in the payoff phase as glucose is catabolized, in the bacterial sialic acid import and degradation pathway (Fig. 1A), phosphorylation serves to prime the amino sugar for downstream catabolism through glycolysis, or anabolism by synthetic pathways, including cell wall synthesis (10,12). Here, we focus on the second step of the catabolic pathway: the phosphorylation of N-acetylmannosamine catalyzed by N-acetylmannosamine kinase (henceforth abbreviated to NanK).
In bacteria, enzymes from two distantly related protein families carry out ATP-driven sugar phosphorylation (13): the glucokinase family (14) and the Repressor, ORF, Kinase (ROK) 3 family (15). Both families have the RNase H-like protein-fold, a sugar substrate-induced conformational change, ATP-binding DXGXT and GTGXG motifs, and a sugar-binding EXGH motif (13,15). The sugar substrate-binding site is also somewhat con-served, with ligand-bound crystal structures showing five conserved sugar-binding residues (Fig. 1B) (13). The key difference between the families is that ROK proteins have a Zn-binding motif (CXCGXXGCX(E/D)) that buttresses the EXGH motif histidine, placing it in the correct position for substrate binding (Fig. 1C). In contrast, the glucokinase family lack the Zn-binding motif and instead have an additional ␣-helix to structure the region (16).
We conducted a sequence alignment of NanK enzymes from bacteria, which are expected to be part of the ROK family (13). Curiously, we found that the majority of Gram-positive bacterial NanK sequences lack the Zn-binding motif and the ␣-helix used in the glucokinase family (Fig. S1). This is best shown in an alignment of representative NanK sequences (Fig. S2), which reveals that the cysteine residues usually involved in forming Figure 1. A, a schematic of sugar flow through the bacterial sialic acid catabolism pathway. NanK catalyzes the conversion of N-acetylmannosamine and ATP to form N-acetylmanosamine-6-phosphate and ADP. This proposed pathway is based on experimental evidence from S. aureus (4), Bacillus subtilis and Escherichia coli (55), as well as from KEGG genome data (56). For clarity, substrate and enzyme names are abbreviated. Enzymes used were: sialic acid transporter (6 -8), N-acetylneuraminate lyase (57,58), NanK, N-acetylmannosamine-6-phosphate 2-epimerase (59,60), GlcNAc-6-phosphate deacetylase, and glucosamine-6-phosphate deaminase (12). Substrates used were: N-acetylneuraminate, N-acetylmannosamine, N-acetylmannosamine-6-phosphate, GlcNAc-6-phosphate, glucosamine-6-phosphate, and fructose-6-phosphate. B, structures of sugar-bound ROK and non-ROK sugar kinases have a highlyconserved sugar-binding site. Glucokinase family E. coli glucokinase is shown in pink (PDB 1SZ2), ROK S. griseus glucokinase in green (PDB 3VGL), ROK B. subtilis fructokinase in cyan (PDB 3LM9), and ROK human bi-functional UDP-GlcNAc 2-epimerase/NanK in red (PDB 2YHY). C, this image shows the orientation of the EXGH motif (green) and ROK Zn-binding motif (red) relative to the sugar-binding site (PDB 3VGL).

N-Acetylmannosamine kinase from S. aureus
the Zn-binding motif are absent in sequences of Gram-positive S. aureus and Streptococcus mitis species, and the histidine of the EXGH motif is substituted with a tyrosine (EXGY) (17). This is surprising, as Zn-binding is required for activity in ROK proteins (15,18), is present throughout Gram-negative NanK isozymes (13,19), and the Zn-binding motif differentiates the ROK family from the glucokinase family (15).
Given that many Gram-positive NanK enzymes have been traditionally annotated as ROK enzymes, but are missing the essential Zn-binding motif, and instead have a tyrosine in place of histidine from the sugar-binding EXGH motif, we studied how Gram-positive S. aureus NanK (SaNanK) functions without these canonical features. Here we report the first functional and structural study of a Gram-positive NanK. We discovered that the enzyme is active without Zn-binding, has a unique structured arginine -stack in a position equivalent to the ROK Zn-binding motif, or glucokinase family ␣-helix, and that the tyrosine in the EXGY motif provides specificity for the N-acetylmannosamine substrate.

SaNanK N-acetylmannosamine kinase catalytic function
We first verified whether SaNanK was a N-acetylmannosamine kinase using an established assay for measuring initial rate data (19). The enzyme is active with N-acetylmannosamine and ATP. Initial rate data, where both acetylmannosamine and ATP concentrations are varied, are best fitted with the ternary complex kinetic model ( Fig. 2A), with an R 2 ϭ 0.999 (see "Experimental procedures" for a discussion on how this model was chosen). This gives the following kinetic parameters: K m N-acetylmannosamine ϭ 0.30 mM, K m ATP ϭ 0.73 mM, and k cat ϭ 262.6 s Ϫ1 (V max ϭ 497 mol min Ϫ1 mg Ϫ1 ) ( Table 1). The k cat /K m for N-acetylmannosamine is 870 mM Ϫ1 s Ϫ1 , which is on the same order as the only other bona fide bacterial NanK enzyme characterized (from Escherichia coli) (19), and other ROK kinases (19 -23). Thus, SaNanK is catalytically active and kinetically comparable with other ROK kinases, despite missing conserved and defining active site motifs thought to be important for function.
Three ROK Streptomyces glucokinases have been reported to have a compulsory-ordered mechanism, where glucose binds the active site before ATP (16,24). When the initial rate data are presented as double-reciprocal plots the pattern provides information on whether the ternary complex kinetic mechanism is of the random-or compulsory-ordered type with respect to substrate binding (25). The plots for SaNanK are consistent with a random-ordered ternary complex for catalysis (Fig. S5). These results demonstrate that SaNanK binds substrate differently than Streptomyces ROK glucokinases, which presumably have a structural feature that prevents ATP from binding first.
Glucose kinase activity is thought to be the ancestral function for the ROK family (13). Glucose differs from N-acetylmannosamine in that a equatorial (R)-hydroxyl at the C2 position is substituted with an axial (S)-N-acetyl group. To probe substrate specificity for the C2 position, we compared the catalytic func-  N-Acetylmannosamine kinase from S. aureus tion with N-acetylmannosamine against glucose and GlcNAc, which are C2 substrate analogs. The apparent K m (K m app ) of GlcNAc (260 M) is comparable with N-acetylmannosamine (300 M) (Fig. 2B). However, the k cat (23 s Ϫ1 ) is 11-fold lower. The K m app for glucose (15 mM, Fig. 2C) is 60-fold higher than the N-acetyl substrates, highlighting that the C2 N-acetyl group is important for binding to the enzyme. In ROK glucokinases, the C2 position is bound by the EXGH motif histidine (13), which in SaNanK is changed to tyrosine and represents an uncharacterized change within the ROK protein family. This result demonstrates that the N-acetyl substrate group is important for substrate binding and consequently the residues of the enzyme that coordinate the N-acetyl are important for substrate specificity.
Overall, we demonstrate that SaNanK is a bona fide N-acetylmannosamine kinase with catalytic constants comparable with canonical ROK kinases that have the EXGH and Zn-binding motif. We show that SaNanK catalysis has a random-ordered ternary mechanism, in contrast to Streptomyces glucokinases, which bind glucose first. The difference in kinetic mechanism suggests that SaNanK has novel structural features, which may be related to the altered protein sequence. We also demonstrate that the C2 N-acetyl group of the substrate is important for substrate binding.

Structure of SaNanK
To investigate the structural implications of the missing canonical ROK motifs and the altered kinetic mechanism relative to other glucokinases, we solved the crystal and solution structure of SaNanK in the ligand-free, N-acetylmannosamine-bound, and GlcNAc-bound forms. The ligand-free structure was solved to 2.33 Å resolution. Because we were unable to solve the structure using molecular replacement, selenomethionine-substituted enzyme was crystallized as previously described (17) and initial phases were determined using SHELX (26). The asymmetric unit contains four NanK monomers where chains A, B, and C are complete, but chain D has three regions where electron density is ambiguous and cannot be modeled (Gln-28 -Pro-33, Pro-82-Pro-85, Asn-142-Gln-144). N-Acetylmannosamine-bound and GlcNAc-bound SaNanK structures were each solved at 2.20 Å resolution, using the ligandfree structure for molecular replacement. Electron density was present and distinguishable for all residues, except Met-1. The data collection and refinement statistics are given in Table 2.

The SaNanK monomer
The SaNanK monomer features two domains, an N-terminal ␣/␤ domain (Fig. 3A, red) and a larger ␣ϩ␤ domain (Fig. 3A, pink) that together form the RNase H-like-fold. The domains are orientated such that the monomer resembles an open clam, with the two separated domains being joined at a single hinge region (Fig. 3A, green). The active site is situated between the two "half-shell" domains, where residues from both domains are used to bind substrate and support catalysis. The structure is consistent with canonical ROK kinases, with the secondary structure regions matching E. coli NanK (Fig. S6) and comparison with TM-align (27) provides a structure match score of 84.4 and 86.6% with E. coli NanK (PDB 2AA4) and Streptomyces griseus glucokinase (PDB 3VGK), respectively.
The active sites of ROK kinases are generally well-conserved (13). Mapping the sequence alignment to the structure demonstrates that it is the most highly-conserved area of SaNanK (Fig.  3B). However, from our sequence analysis, there is a major change to the region with the absence of a Zn-binding motif, which in canonical ROK partially encloses the active site from the sugar side (PDB 3VGL). In SaNanK, the region equivalent to the Zn-binding motif is shorter (13 versus 20 residues) and is folded away from the active site area, interacting with residues of the large ␣ϩ␤ domain.

SaNanK is a dimer, consistent with other NanK enzymes
To determine the relevant oligomeric state in solution, sedimentation velocity analytical ultracentrifugation and small angle X-ray scattering experiments were conducted. Both experiments demonstrate that SaNanK is a dimer (across the concentration range tested).
Sedimentation velocity data were fit with the c(s) and c(M) models using SEDFIT (28), which give a single peak corresponding to 3.95 S and 62 kDa (Fig. 4A); thus, ligand-free SaNanK is a dimer in solution given a monomeric mass of 31.7 kDa calculated from the sequence. A single peak is evident at each concentration, suggesting that the equilibrium constant for dimer dissociation (K D 2-1 ) is very low. Small angle X-ray scattering data are also consistent with a dimer. The theoretical scatter for the ligand-free SaNanK canonical dimer formed in the crystal structure (generated by CRYSOL from the ATSAS package (29)) fitted closely to the scattering data ( 2 ϭ 1.7) (Fig.  4B). We also tested a possible dimer formed by a secondary interface in the asymmetric unit (as defined by PDBe PISA (30), and the full asymmetric unit of the ligand-free structure (two dimers) (Fig. S7), both of which fitted poorly to the scattering data. The dimeric configuration is consistent with other NanK crystal structures (19,31,32) and MALDI-TOF MS analysis of (ROK) glucokinase from Bacillus subtilis (15).
From the crystal structure, the dimeric configuration orients the active sites facing outward on opposite sides of the assembly, whereas the hinge regions are near the center (Fig. 4C). Residues Table 1 Kinetic parameters for wildtype SaNanK The V max , k cat , and K m were calculated in GraphPad Prism. The error is the mean Ϯ S.E. of the fit. N-Acetylmannosamine parameters were produced from global fits by varying ATP.

N-Acetylmannosamine
Glucose a GlcNAc a a Glucose and GlcNAc are apparent parameters, using a saturating ATP concentration.

N-Acetylmannosamine kinase from S. aureus
from the large ␣ϩ␤ domain form the dimerization interface. This configuration is consistent with E. coli, Fusobacterium nucleatum, and Homo sapiens (19,31,32). PDBe PISA analysis indicates that the SaNanK interface has a surface area of 1533.8 Å 2 , featuring five salt bridges and 18 hydrogen bonds. The residues identified by PDBe PISA contribute to three positive and three negatively charged regions that charge complement the opposing monomer ( Fig. 4D).
In summary, we demonstrate that SaNanK is a dimer. The dimeric interface consists of three complementary positive and negative patches on the surface of large ␣ϩ␤ domain.

Substrate-induced changes in conformation
A conformational change upon sugar binding, whereby the lobes of the enzyme close over the substrate, has been demonstrated in both the ROK and glucokinase families (14,16,31). Whether SaNanK undergoes a similar conformational change was determined by comparing the sugar substrate-bound and ligand-free X-ray crystal structures. In the N-acetylmannosamine-bound SaNanK structure the two domains are shifted 7 Å inwards and close around the substrate, like the shells of a clam enclosing a pearl (Fig. 5A). As a dimer, the   Table S1. Active sites residues are highly conserved (maroon) and residues elsewhere are poorly conserved (cyan). Image produced using ConSurf (62) and UCSF Chimera software (63).

N-Acetylmannosamine kinase from S. aureus
movement occurs mainly in the N-terminal ␣/␤ domain, pivoting at the hinge region, as the large ␣ϩ␤ domain is set by the dimeric interface. Overall, the conformational change of SaNanK is consistent with other ROK kinases (14,16,31). The purpose of a conformational change is hypothesized to ensure that ATP hydrolysis occurs only when sugar substrate is present, providing efficient catalysis (31). To investigate, the catalytic residues identified from the sequence alignment ( Fig.  S2) were compared between ligand-free and N-acetylmannosamine-bound structures. In the catalytic mechanism for sugar kinases, ATP is primed for nucleophilic attack at the ␥-phosphate by electron withdrawal through the ␤-phosphate, which then undergoes nucleophilic attack by the sugar (31). For SaNanK, electron withdrawal likely occurs via Asp-7 binding Mg 2ϩ associated with the ATP ␤-phosphate, given its proximity, whereas Thr-11 and Thr-134 hydrogen bond with the other side of the ␤-phosphate group. Alignment of Mg⅐ADP from the human NanK structure demonstrates that when N-acetylmannosamine is bound, Thr-134 is shifted toward the expected ␤-phosphate position (Fig. 5B, pink) and both Asp-7 and Thr-11, which are part of the N-terminal ␣/␤ domain, make a large shift toward the ATP position (Fig. 5B, red). Importantly, this shows that Asp-7, Thr-11, and Thr-134 are in position to provide electron withdrawal in the N-acetylmannosamine-bound structure and not the ligand-free structure. In short, this is consistent with a conformational change that ensures ATP is only hydrolyzed in the presence of substrate (discussed in Ref. 31).
As conformational change is important for catalysis in ROK kinases (discussed above), the GlcNAc and N-acetylmannosamine-bound structures were compared to ration- The fit of the model to the data (Fig. S4) provides a molecular mass of ϳ62 kDa, corresponding to a dimer (monomer mass is 31.7 kDa). B, small-angle X-ray scattering of SaNanK (6 mg ml Ϫ1 ). Ligand-free SaNanK experimental data (red circles) corresponds with theoretical scatter (red line), with a reduced 2 of 1.7 versus 4.9, compared with the N-acetylmannosamine crystal structure scatter (blue line). C, structure of the SaNanK dimer, which is formed through residues of the larger ␣ϩ␤ domain (pink), with the hinge region (green), and large N-terminal ␣/␤ domain (red) free in solution. D, the dimeric interface of SaNanK. The electrostatic surface (generated using PyMOL) shows three opposing negative and positive areas, which pair when the dimer interface forms. Residues involved with dimerization were determined using PDBe PISA and were mapped onto the surface of SaNanK.

N-Acetylmannosamine kinase from S. aureus
alize the lower catalytic rates with GlcNAc. The structure of GlcNAc-bound SaNanK is similar to the N-acetylmannosamine-bound structure but appears to be slightly more open (Fig. S8), with an average RMSD for the four chains of 0.42 Å using 262-265 of the 285 residues (PyMOL). This alters the distances between catalytic residues Thr-11 and Thr-134, which provide electron stabilization to the ␤-phosphate of ATP and provides a possible explanation for lower catalytic rates with GlcNAc.
We then examined whether conformational change explains why the enzyme operates via an ordered or random kinetic mechanism. To do this we compared the ligand-free and sugar-bound structures of SaNanK, which has a random ordered kinetic mechanism, with S. griseus glucokinase (PDB 3VGK and 3VGM), which has a compulsory-ordered kinetic mechanism, where glucose binds first and then ATP binds (16). Both SaNanK and S. griseus glucokinase undergo a similar overall conformational shift, pivoting at the hinge region in response to sugar-binding and the ATP pocket is essentially unchanged (Fig. S9). Therefore, conformational change does not seem to explain why some ROK kinases have a random kinetic mechanism, whereas others have a compulsory-ordered kinetic mechanism.
To test whether the conformational change seen in the crystal structure also occurs in solution, small angle X-ray scattering was used. The scattering profile with and without N-acetyl-mannosamine were individually compared with the theoretical scatter of the crystal structures with and without N-acetylmannosamine using CRYSOL (33). The scattering data from the ligand-free SaNanK sample closely matches the ligand-free SaNanK crystal structure, but not the N-acetylmannosaminebound crystal structure, (reduced 2 of 1.7 versus 4.9) (Fig. 4B). In contrast, the scattering data for SaNanK in the presence of N-acetylmannosamine (5 mM) closely matches the N-acetylmannosamine-bound crystal structure, but not the ligand-free SaNanK crystal structure (reduced 2 of 1.9 versus 9.9) (Fig. 5C). This demonstrates that the crystal structures match the solution structures and the enzyme closes around N-acetylmannosamine. The condensing of the N-acetylmannosamine solution structure is further supported by a smaller radius of gyration calculated from the overall scattering intensity (R g ϭ 26.9 versus 27.5 Å) (Fig. 5D).
To summarize, SaNanK undergoes an N-acetylmannosamine-induced conformational change adopting a closed conformation. This ensures ATP hydrolysis only occurs in the presence of sugar substrate. In support, the catalytic residues (Asp-7, Thr-11, and Thr-134) that withdraw electrons from the ␤-phosphate group are only in an active configuration in the closed conformation. Adherence to the closed and open conformations, with and without ligand, respectively, is demonstrated in solution by small angle X-ray scattering.  (green). B, expected catalytic residues of SaNanK. The N-terminal ␣/␤ domain moves bringing Asp-7 and Thr-11 toward the active site, which in conjunction with Thr-134 prime ATP for nucleophilic attack. The expected position of ATP was indicated by aligning the human NanK structure with N-acetylmannosamine and Mg⅐ADP bound (PDB 2YHY). Asp-108 binds and deprotonates the C6 hydroxyl of N-acetylmannosamine, increasing nucleophilicity of the sugar. C, small-angle X-ray scattering of SaNanK (6 mg ml Ϫ1 ) and N-acetylmannosamine (5 mM). Experimental data (blue circles) corresponds with theoretical scatter of the N-acetylmannosamine-bound crystal structure (blue line), with a reduced 2 of 1.9 versus 9.9, compared with the ligand-free crystal structure scatter (red line). D, radius of gyration calculated from intensity during small angle X-ray scattering. The radius of gyration was calculated to be condensed when N-acetylmannosamine is bound (26.9 versus 27.5 Å), which is consistent with the inward shift represented by the crystal structure.

N-Acetylmannosamine kinase from S. aureus The SaNanK active site uses tyrosine to bind the substrate C2 position
Five sugar substrate-binding residues are reported to be wellconserved in ROK kinases (13) and conservation was investigated in SaNanK. Four of the five residues are conserved, including Asn-107, Asp-108, Glu-157, and Glu-172, whereas Tyr-160 is a nonconservative change and replaces the histidine in the canonical EXGH motif (discussed later). The C1 hydroxyl of N-acetylmannosamine is bound by Glu-172, and Glu-157 primarily binds C3, but also binds the C2 position N-acetyl moiety. Both residues are part of the larger ␣ϩ␤ domain (Fig.  6A, magenta). Asn-107 binds the C4 hydroxyl and Asp-108 binds C6 (Fig. 6A). These residues border the N-terminal ␣/␤ domain and the hinge region, where the conserved substrate binding translates to the conformation change between the ligand-free and N-acetylmannosamine-bound structures. In addition to substrate binding, the conserved Asp-108 that binds the C6 position is also expected to feature in the catalytic mechanism by deprotonating the hydroxyl and increasing nucleophilicity (34). Consistent with the mechanism, Asp-108 is predicted to be deprotonated in the ligand-free structure (pK a ϭ 6.03), but protonated in the N-acetylmannosamine-bound structure (pK a ϭ 9.14), using PDB2PQR (35) at pH 7.
The structural role of Tyr-160 in the EXGY motif of SaNanK is a nonconservative substitution of the Zn-binding histidine in the EXGH motif seen in canonical ROK kinases. Indeed, tyrosine at this position is conserved in other Gram-positive NanK lacking the Zn-binding motif (Fig. S1). The hydroxyl of SaNanK Tyr-160 binds the C2 N-acetyl group of N-acetylmannosamine (Fig. 6A). This is analogous to ROK glucokinase, where the equivalent EXGH motif histidine binds the C2 hydroxyl of glucose (16). However, glucose is much smaller and the EXGH motif histidine also binds Zn. Zn-binding NanK enzymes utilize the EXGH motif histidine to bind the C1 hydroxyl and the larger N-acetyl substrate is bound from residues on the other side of the active site (31). A comparison of substrate binding is presented in Fig. 6B. In contrast to histidine found in the Znbinding structures, SaNanK Tyr-160 tilts away from the active site, increasing the size of the active site, and hydrogen bonds with the N-acetyl group. Interestingly, N-acetylmannosamine binds in a "boat" conformation and the N-acetyl group is in a pseudo-equatorial position (Fig. 6B, Figs. S10 and S11). It is unknown whether these observations are typical for NanK without the Zn-binding motif, although our sequence alignment suggests that tyrosine is consistently present at the position.
Our earlier kinetic analysis revealed that the C2 N-acetyl group is important for binding, with the K m of both N-acetylmannosamine and GlcNAc (0.30 and 0.26 mM, respectively) being much lower than glucose (15 mM), which only differs at , and human NanK (pink). In SaNanK, the EXGH motif histidine is replaced by a tyrosine (Tyr-160), which binds the larger N-acetylated substrate by tilting backward from the active site and having greater length than histidine. In human NanK, the EXGH motif histidine bind the C1 hydroxyl and the axial N-acetyl group is bound by other nonconserved residues. C, GlcNAc bound SaNanK. The N-acetyl carbonyl group is not able to hydrogen bond with Tyr-160. D, Tyr-160 provides substrate specificity toward N-acetylmannosamine, over GlcNAc. When tested with N-acetylmannosamine (blue, left axis), WT SaNanK has more substrate specificity (870 mM Ϫ1 s Ϫ1 ) than Y160F-substituted enzyme (170 mM Ϫ1 s Ϫ1 ), which are absent the Tyr-160 hydroxyl. When tested with GlcNAc (orange, right axis), WT SaNanK has less substrate specificity (90 mM Ϫ1 s Ϫ1 ) than Y160F-substituted enzyme (390 mM Ϫ1 s Ϫ1 ), demonstrating that Tyr-160 provides substrate specificity against GlcNAc.

N-Acetylmannosamine kinase from S. aureus
the C2 position. To investigate further, the GlcNAc-bound and N-acetylmannosamine-bound structures were compared. Both substrates are bound at the C1, C3, C4, and C6 hydroxyls by Glu-172, Glu-157, Gln-107, and Asp-108, respectively (Fig.  6, A and C). These positions are conserved with S. griseus glucokinase and glucose is expected to bind with SaNanK in a similar manner. At the C2 position, the N-acetyl groups are both in an equatorial position, despite being enantiomers. N-Acetylmannosamine adopts a boat conformation with the carbonyl pointing "down" and the N-acetyl is positioned to bind three residues, Gly-81 to the carbonyl, Glu-157 to both the carbonyl and amine, and Tyr-160 to carbonyl (Fig. 6A). In contrast, GlcNAc adopts a "chair" conformation with carbonyl pointing "up" and the N-acetyl group is positioned to bind two residues only, Gly-81 to the carbonyl and Glu-157 to the amine (Fig. 6C). The comparison reveals that the SaNanK active site is structurally optimized to bind N-acetylmannosamine, over GlcNAc and glucose.
Next, we tested the role of the Tyr-160 hydroxyl in substrate specificity by producing substituted enzyme SaNanK-Y160F and comparing its kinetic profile (Table 3 and Fig. 6D). For N-acetylmannosamine, the specificity constant for the WT (k cat /K m ϭ 870 mM Ϫ1 s Ϫ1 ) is greater than the Y160F-substituted enzyme (k cat /K m ϭ 170 mM Ϫ1 s Ϫ1 ). This is consistent with the participation of the Tyr-160 hydroxyl in substrate binding for WT SaNanK. For GlcNAc, the specificity constant for the WT (k cat /K m ϭ 90 mM Ϫ1 s Ϫ1 ) is lower than the Y160Fsubstituted enzyme (k cat /K m ϭ 390 mM Ϫ1 s Ϫ1 ). This suggests that the Tyr-160 hydroxyl may have a role in perturbing GlcNAc binding. In support of this, Tyr-160 is displaced by 1 Å, compared with the ligand-free and N-acetylmannosaminebound structures, indicating that the residue sterically hinders GlcNAc binding. Together, the results demonstrate that the Tyr-160 hydroxyl is important for specificity toward the pathway substrate, with a 5:1 preference for N-acetylmannosamine with tyrosine, compared with a 2:1 preference for GlcNAc when replaced with phenylalanine. Controls with Y160A-substituted enzyme (Table 3) were similar to Y160F and further support that the tyrosine hydroxyl is important for N-acetylmannosamine specificity.
Without the Zn-bound EXGH motif histidine, SaNanK binds the N-acetyl group of N-acetylmannosamine with tyrosine, which is generally conserved in Gram-positive NanK without the Zn-binding motif. This is different to Zn-binding NanK enzymes, which bind the N-acetyl group using residues on the other side of the active site (31). The tyrosine residue is important for substrate recognition and when replaced with phenylalanine, substrate preference shifts to GlcNAc.

The ROK Zn-binding motif is replaced by a stabilizing arginine stack
SaNanK does not have the highly-conserved ROK Zn-binding motif, however, the significance of this is unknown as the purpose of the motif is not well-understood (13). To identify structural differences, SaNanK was compared with a ROK and a non-ROK sugar kinase. In ROK S. griseus glucokinase, the Zn-binding region encompasses 20 residues, starting from the Zn-binding EXGH motif His-157 to the end of the motif, CXCGXXGCX(E/D) (residues 167-176, PDB 3VGK). The EXGH motif histidine and the Zn-motif Glu/Asp bind to the sugar, with the central residues forming a loop that wraps around Zn (Fig. 1C) enclosing the active site (Fig. 7A, green). For SaNanK ROK without the Zn-motif, the equivalent region between the sugar-binding Tyr-160 and Glu-172 (13 residues) forms a loop that associates with the large ␣ϩ␤ domain and does not occlude the sugar active site (Fig. 7A, red). In non-ROK E. coli glucokinase, the equivalent region is 28 residues long (His-160 -Glu-187) and features highly charged residues as part of the dimerization interface (14). This does not occlude the sugar active site (Fig. 7A, yellow) in a position similar to the SaNanK region. The observations suggest that the absence of Zn-binding corresponds with an open sugar active site and the presence of Zn-binding corresponds with a closed active site.
The absence or presence of the Zn-motif may determine the order of substrate binding. We propose that SaNanK has a random-order kinetic model because the absence of the Zn-motif provides a secondary substrate entry point, whereas in the compulsory-order S. griseus glucokinase there is a narrow active site with a just a single entry and glucose must access the pocket at the bottom before ATP binds. Kinetics assays demonstrate that SaNanK uses a random-order ternary complex for catalysis, whereas canonical ROK S. griseus glucokinase is characterized as a compulsory-ordered complex that requires glucose to bind before ATP (16). Initially, conformational change was investigated as a method of conferring order (discussed above), but comparisons of the enzyme structures failed to identify a differentiating feature in ATP-binding site and it is difficult to attribute the difference in mechanism to conformational change. Compulsory-ordered S. griseus glucokinase has a narrow active site with a single opening near the ATP-binding pocket (Fig.  7A). Glucose binds at the bottom of the active site and ATP binding would obstruct glucose binding, thus providing an explanation of why glucose must bind before ATP. In contrast, SaNanK has a secondary opening near the active site that N-acetylmannosamine can access irrespective of ATP binding providing for a random-ordered kinetic mechanism.

Table 3 Kinetic parameters for SaNanK substituted at Tyr-160
Without the Tyr-160 hydroxyl group, SaNanK-Y160F and SaNanK have a higher enzyme specificity (k cat /K m ) with GlcNAc than for N-acetylmannosamine, the pathway substrate. The V max , k cat , and K m were calculated in GraphPad Prism. The error is the mean Ϯ S.E. of the fit. Values are apparent parameters, using saturating ATP concentration.

N-Acetylmannosamine kinase from S. aureus
ROK Zn-binding has been demonstrated elsewhere to be necessary for activity (15,18), and Zn coordination is hypothesized to organize three nearby active site residues (16). The organization of equivalent residues (Glu-157, Tyr-160, and Glu-172) were examined in SaNanK to investigate why the enzyme is active without Zn-binding. Glu-157, Tyr-160, and Glu-172 are well-organized in the ligand-free structure with lower B-factors (17.5, 13.8, and 18.8) than the model average (37.7), despite little secondary structure organization in the region. The nearby region, equivalent to the ROK Zn-binding region, was examined for potential stabilizing features. Interestingly, the region features two stacked arginine residues (Arg-164 and Arg-174), ionically coordinated by three negatively charged residues (Glu-167, Asp-219, and Glu-222), with two of the negative residues belonging to a large ␣-helix that anchors the arginine stack (Fig. 7B). Potentially, this five-coordinate feature stabilizes the region and organizes the active site residues, serving the hypothesized role of four-coordinate Zn-binding in canonical ROK.
Re-examination of Gram-positive NanK sequence alignments reveals that the arginine stacking residues are novel within the Staphylococcus genus and Arg-164, which is coordinated by four other residues, is not conserved elsewhere (Fig.  S12). Arg-164 was substituted with alanine and we test the effect of this change on thermal stability to test for structural stabilization, and kinetic assays to test whether the residue is important for active site organization (a hypothesized role of the equivalent Zn-binding region (16)). SaNanK-R164A was not recoverable in the soluble fraction and attempts to improve solubility by altering the expression temperature and length were unsuccessful. This suggests the enzyme is potentially unstable without the positive charge or hydrophobic stacking supplied by arginine. However, SaNanK-R164K was successfully purified. Thermal stability analysis reveals that SaNanK-R164K has an average T m of 35.4°C, which is 4.1°C lower than SaNanK (Fig. 7C), and in conjunction with insolubility of SaNanK-R164A, this indicates that arginine interaction is important for stabilization. Kinetic assays show that the substituted enzyme is catalytically compromised (V max ϭ 6.1 Ϯ 0.3 mol/min/mg, k cat ϭ 3.2 Ϯ 0.2 s Ϫ1 , K m ϭ 0.26 Ϯ 0.01 mM) with approximately a 100-fold decrease in activity compared with WT (k cat ϭ 262.6 s Ϫ1 ). This is consistent with arginine stacking contributing to organize the nearby active site residues. This is a hypothesized role of the ROK Zn-binding region, which is equivalently positioned to the SaNanK arginine stack, and we propose that both coordinating features provide stabilization and organization. In non-Zn-binding ligand-free SaNanK (red) the region is positioned away from the sugar active site. In Zn-binding ROK S. griseus glucokinase (PDB 3VGK) (green) the region partially covers the sugar active site. Glucokinase family E. coli glucokinase (PDB 1Q18) (yellow) has a larger region featuring an ␣-helix and is positioned away from the sugar active site. N-Acetylmannosamine and ATP substrates were aligned from N-acetylmannosamine-bound SaNanK and ROK kinase (PDB 3VGL). B, in the region equivalent to the Zn-binding motif (blue), SaNanK features a -stacking arginine pair, which are coordinated by three negative residues. Two of the negative residues are part of a large ␣-helix (pink) stabilizing the protein loop and positioning it away from the active site. Stable positioning of the region is important as it features three active site residues (Glu-157, Tyr-160, and Glu-172), which bind to N-acetylmannosamine (black). C, differential scanning fluorimetry with SaNanK (solid line) and SaNanK-R164K (dashed line). SaNanK-R164K has lower thermal stability with the average melting temperature of the two domains is 4.1°C lower than WT.

N-Acetylmannosamine kinase from S. aureus
The absence of the Zn-binding motif increases the potential of SaNanK as a drug target. Elsewhere, it has been demonstrated that the bacterial sialic acid catabolic pathway is a viable target to lower infectivity and increase antibiotic efficacy (1,3,4,9,10). Targeting the pathway is only useful if toxic interaction with host sialic pathway enzymes can be avoided. The NanK component of human bifunctional UDP-GlcNAc 2-epimerase/N-acetylmannosamine kinase is a ROK family protein but, unlike SaNanK, it retains the canonical Zn-binding motif (31). The structure of the human enzyme (PDB 2YHY) shows that bound N-acetylmannosamine is enclosed by the Zn-binding motif and the N-acetyl group associates in the axial position. Whereas, our characterization demonstrates that SaNanK is open at the N-acetylmannosamine end of the active site (Fig. 7A). Therefore, we propose that synthesized glucose or mannose C2 (or C1) derivatives can be made to target SaNanK and not the human bifunctional enzyme, where the Zn-binding motif occludes larger derivative.

Conclusions
Gram-positive NanK enzymes are annotated as ROK enzymes, but many are missing the identifying ROK Zn-binding motif and have tyrosine in place of the histidine in the EXGH motif. Here we investigate the effect of these changes on the structure and function of SaNanK. We focus on defining the role of tyrosine that replaces EXGH motif histidine and provide insight into the presence or absence of Zn in ROK kinases.
We report the first functional and structural study of a Grampositive NanK. The crystal and solution structures conform to the RNase H-like-fold present in ROK family proteins (36) and we confirm that the enzyme is catalytically active as a kinase. Therefore, SaNanK has the properties expected from a ROK kinase. Investigation of the protein dynamics reveal the structure condenses with sugar substrate in the X-ray crystal form, as reported for other ROK kinases (16,31). For the first time, we confirm this in solution with small-angle X-ray scattering experiments.
Our sequence alignment reports that Gram-positive NanK without the Zn-binding motif have tyrosine in place of EXGH motif histidine. Structural and functional analysis reveals that tyrosine binds C2 N-acetyl of N-acetylmannosamine by tilting backward from the active site. This is substantially different from Zn-binding NanK, where the EXGH motif histidine binds C1 and is restrained in a forward-facing tilt by secondary binding of Zn. As EXGH motif histidine binds Zn, the use of tyrosine in Gram-positive NanK is certainly linked to the loss of Zn-binding. However, as Gram-positive NanK are highly divergent at a sequence level, it is difficult to know whether change of histidine from the EXGH motif or Zn-binding motif residues occurred first.
The role of the Zn-binding motif is not well-understood, therefore the significance of its absence in Gram-positive NanK is unknown. Our examination considered the absence of Znbinding in SaNanK and its presence in canonical ROK. A role hypothesized for Zn-binding is that it organizes the active site residues for catalysis (16). We find a possibly equivalent coordinating feature in SaNanK: an arginine stack that is novel to Staphylococcus. Loss of this feature in SaNanK lowered thermal stability and significantly lowered catalytic rates, consistent with the hypothesis that it is a feature that provides structural stabilization and active site coordination, as opposed to being provided by that Zn-binding in other ROK kinases. Our structural comparison also revealed that absence of the motif provides an opening at the sugar substrate active site. This is beneficial for the active site access of larger substrates, such as N-acetylmannosamine in SaNanK and potentially, disaccharide cellobiose kinases (23). Whereas Zn-binding in E. coli NanK requires increased hinge flexibility for N-acetylmannosamine association (19). We propose that the ROK Zn-binding motif is an adaptation that excludes large substrates, consistent with ROK repressors and kinases generally functioning with small sugars (13) that may also contribute to compulsoryordered catalysis for pathway regulation (16,24). Finally, given the proximity of the arginine stack and EXGY motif, we propose that this is a unique substructure to target with inhibitory compounds that avoid off-target effects with the human homologue, bifunctional NanK.

Bioinformatics
To provide a root for the phylogenetic tree of NanK protein sequences, sequences for glucokinase, fructokinase, and GlcNAc kinase enzymes that were closely related to the canonical ROK sequence were added to the alignment. To select the root sequences, the ROK Hidden Markov Model (HMM) sequence from the Pfam database (https://pfam.xfam.org/ family/PF00480/hmm) 4 was entered into pBLAST (https:// blast.ncbi.nlm.nih.gov/Blast.cgi), alongside the search terms "glucokinase," "fructokinase," or "GlcNAc kinase." The top three sequence results (from unique genera) were included in the alignment. NanK protein sequences were searched for by inputting the SaNanK sequence, ROK HMM sequence, and E. coli NanK sequence, alongside the search term "N-acetylmannosamine kinase" into the tBLASTn database. The first NanK sequence for each unique genus were added to the alignment. Additionally, sequences that had a different variation of EXGH motif or Zn-binding residues, were included. To verify sequences were annotated correctly, only sequences clustered with Nan genes were included.

Cloning, expression, and purification
SaNanK-R164A, -R164K, and -Y160A were constructed by Genscript based on variation of the SaNanK (accession number YP_498886.1) codon optimized for expression in E. coli. SaNanK-R164A was synthesized de novo and inserted into a pET30⌬SE plasmid between NdeI and HindIII restriction sites. SaNanK-R164K was then created by site-directed mutagenesis.
The protocol for expression and purification of WT protein has previously been reported (17). Briefly, E. coli BL21 (DE3) cells were transformed with pET30⌬SE/NanK, then cultured in Luria-Broth containing kanamycin (30 g ml Ϫ1 ) at 37°C, 180 rpm until an OD 600 of 0.6 was reached. Expression was induced with isopropyl ␤-D-1-thiogalactopyranoside at a final concentration of 1 mM and incubated overnight at 26°C. Purification was conducted at 4°C. Cells were harvested at 6,000 ϫ g for 10 min using a Thermo Sorvall RC-6-Plus centrifuge and resuspended in 20 mM Tris-HCl, pH 8.0. The cells were lysed for 10 min (Hielscher UP200S Ultrasonic Processor at 70%, 0.5 s on/off cycle) and lysate containing soluble protein was extracted from the supernatant after centrifugation at 24,000 ϫ g. For anion exchange chromatography, the sample was applied to a 20-ml Q-Sepharose anion exchange column (GE Healthcare) pre-equilibrated with lysis buffer, then eluted with an increasing concentration gradient of 20 mM Tris-HCl, pH 8.0, 1 M NaCl. For hydrophobic exchange chromatography, ammonium sulfate was added to the eluted protein to a final concentration of 1 M and applied to a 20-ml Phenyl-Sepharose FF column (GE Healthcare) pre-equilibrated with 20 mM Tris-HCl, pH 8.0, 1 M ammonium sulfate and eluted using an increasing concentration gradient of 20 mM Tris-HCl, pH 8.0. Finally, size exclusion chromatography was conducted with a HiLoad 16/60 Superdex 200 column (GE Healthcare) using 20 mM Tris-HCl, pH 8.5. Final purity was estimated using SDS-PAGE gels (Fig.  S13). Exceptions to this procedure were as follows; selenomethionine-substituted protein was produced using PASM-5052 autoinduction media (39) and expression trials of R164A mutants were conducted at 20 and 26°C, for 4, 8, and 16 h.

Enzyme kinetics
Initial rate data were recorded using a pyruvate kinase/lactate dehydrogenase-coupled assay (19). NADH oxidation was monitored using a CARY 100 Bio UV-visible spectrophotometer (Agilent Technologies) based on an extinction coefficient of 6220 M Ϫ1 cm Ϫ1 at 340 nm (40). The reaction mixture incubated for 15 min at 25°C consisted of a pyruvate kinase/lactate dehydrogenase mixture (10 l, purchased from Sigma SKU-P0294), purified WT SaNanK, 0.017 mg ml Ϫ1 , 536 nM (10 l), 50 mM phosphoenolpyruvate (20 l), 8 mM NADH (20 l), ATP varied concentration (20 l), and kinetics buffer, 22 mM Tris-HCl, pH 8.5, 20 mM KCl, 5 mM MgCl 2 (900 l), listed values are initial concentrations. The reaction was initiated with sugar substrate (20 l, at varied concentration), which was also incubated at 25°C. To ensure that the N-acetylmannosamine kinase concentration was directly proportional with the rate, it was checked that the coupling enzymes were in excess and that there was no background reaction. The enzyme appeared to be less stable when incubated at 30°C or above, which was consistent with a differential scanning fluorimetry experiment demonstrating that the melting temperature of SaNanK is 39.5°C (Fig. 7C). Moreover, the enzyme was deactivated when in Tris-HCl buffer at pH 8 or below. The data were processed using GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA). The model that best fitted the data (ternary complex (Equation 1) or substituted complex models (Equation 2)) was determined by evaluating the AICc values (41) (ternary complex ϭ 129.8, substituted complex ϭ 203.4). Single substrate (apparent) kinetic parameters were determined using the Michaelis-Menten model.
Initial rate equations for ternary and substituted complexes (42)

Analytical ultracentrifugation
Analytical ultracentrifugation was performed using an XL-I analytical ultracentrifuge (Beckman Coulter) at 50,000 rpm at 20°C. Sedimentation velocity experiments were conducted with N-acetylmannosamine kinase at 0.3, 0.6, and 0.9 mg ml Ϫ1 (9.5, 18.9, and 28.4 M), in 20 mM Tris-HCl, pH 8.5, 100 mM KCl. Sample (380 l) and reference buffer (400 l) were loaded into 12-mm double-sector quartz cells with standard Epon 2-channel centerpieces and mounted in an An-50 Ti 8-hole rotor (Beckman). Radial absorbance was measured at 280 nm with a step size of 0.003 cm over 80 scans. SEDNTERP software was used to calculate values for solvent density and solvent viscosity and to estimate partial specific volume, based on the amino acid sequence (43). SEDFIT was then used to fit the data with continuous size distribution (c(s)) and continuous mass distribution (c(M)) models (28). Residuals were examined to review that noise was normally distributed and data points were checked for fit with curves (Fig. S4).

Small angle X-ray scattering
Small angle X-ray scattering was undertaken at the Australian Synchrotron on the SAXS/WAXS beamline equipped with a PILATUS 1M detector (170 ϫ 170 mm, effective pixel size 172 ϫ 172 m), as described by Griffin et al. (44). The X-ray wavelength provided was 1.0332 Å and the sample detector distance was set at 1600 mm, which produced a q range of 0.006 -0.4 Å Ϫ1 . SaNanK (6 mg ml Ϫ1 , 189 M, 50 l) was injected onto an inline Superdex 200 5/150 GL (GE Healthcare) using a co-flow setup (61). The column was equilibrated with 20 mM Tris-HCl, pH 8.0, 100 mM NaCl and the radical scavengers, 5% (v/v) glycerol and 0.1% (w/v) sodium azide, using a flow rate of 0.45 ml min Ϫ1 . Additionally, 5 mM N-acetylmannosamine was added to the equilibration buffer above for the sugar-bound sample. Over the course of the elution, scattering data were collected in 1-s exposures using a 1.0-mm glass capillary at 12°C.
To process the scattering data, ScatterBrain IDL (version 2.822, Australian Synchrotron Facility) was employed for initial data reduction, radial integration of individual frames, and the N-Acetylmannosamine kinase from S. aureus conversion of 1D data files. Data were buffer-subtracted using CHROMIXS (45). The ATSAS software suite (version 2.8.0) was used for all further analysis (29), including calculation of the radius of gyration and pairwise distribution functions P(r). Theoretical scattering of the SaNanK crystal structures were calculated using CRYSOL (33).
The Australian Synchrotron MX1 and MX2 beamlines were used for data collection. Data collected for the selenomethionine-substituted enzyme was processed using iMosflm (46) and AIMLESS (47). For enzyme with N-acetylmannosamine and GlcNAc, data were processed with XDS and AIMLESS (47). X-ray diffraction parameters are reported in Table 2.
Experimental phasing was conducted using anomalous scattering of SaNanK from selenium-derivatized methionine (6 sites per monomer). The diffraction data were phased using single anomalous diffraction processed by SHELXC/D/E, including automated model building (26). Electron density of the final resolved phases are demonstrated in Fig. S3. Phases of N-acetylmannosamine-and GlcNAc-bound SaNanK were calculated using the ligand-free SaNanK structure as a molecular replacement model with MOLREP (48). Models were iteratively refined and built, using REFMAC5 and COOT (49,50). Finalization and model validation was conducted using phenix.refine and MOLPROBITY (51,52).
Attempts to solve the phases of ligand-free nonselenomethionine-substituted SaNanK using molecular replacement were unsuccessful. Search models for SaNanK were selected based on the highest sequence identity: 2GUP, 3VOV, 3VGK, and 4HTL (identity ϭ 28.0 -24.6%). Additionally, two NanK structures were also trialled, PDB 2YHY and 2AA4 (identity ϭ 21.9 -21.4%). The models were split into monomeric units and processed with CHAINSAW, to generate the search model, and molecular replacement was attempted using MOLREP and PHASER (48,53). We also tried partial models and lower packing restraints (54), but these were also unsuccessful.