Biochemical Analysis and Structure Determination of Bacterial Acetyltransferases Responsible for the Biosynthesis of UDP-N,N′-Diacetylbacillosamine*

Background: A connection between glycoproteins containing N,N′-diacetylbacillosamine and pathogenicity has previously been shown in Campylobacter jejuni. Results: Structural and kinetic studies of two bacterial acetyltransferases show the diversity within the binding pockets responsible for UDP-N,N′-diacetylbacillosamine production. Conclusion: Carbohydrate acetyltransferases from O-linked glycosylation pathways exhibit significant divergence from their N-linked counterparts. Significance: Acetyltransferase characterization increases our understanding of the diverse nature of bacterial glycosylation. UDP-N,N′-diacetylbacillosamine (UDP-diNAcBac) is a unique carbohydrate produced by a number of bacterial species and has been implicated in pathogenesis. The terminal step in the formation of this important bacterial sugar is catalyzed by an acetyl-CoA (AcCoA)-dependent acetyltransferase in both N- and O-linked protein glycosylation pathways. This bacterial acetyltransferase is a member of the left-handed β-helix family and forms a homotrimer as the functional unit. Whereas previous endeavors have focused on the Campylobacter jejuni acetyltransferase (PglD) from the N-linked glycosylation pathway, structural characterization of the homologous enzymes in the O-linked glycosylation pathways is lacking. Herein, we present the apo-crystal structures of the acetyltransferase domain (ATD) from the bifunctional enzyme PglB (Neisseria gonorrhoeae) and the full-length acetyltransferase WeeI (Acinetobacter baumannii). Additionally, a PglB-ATD structure was solved in complex with AcCoA. Surprisingly, this structure reveals a contrasting binding mechanism for this substrate when compared with the AcCoA-bound PglD structure. A comparison between these findings and the previously solved PglD crystal structures illustrates a dichotomy among N- and O-linked glycosylation pathway enzymes. Based upon these structures, key residues in the UDP-4-amino and AcCoA binding pockets were mutated to determine their effect on binding and catalysis in PglD, PglB-ATD, and WeeI. Last, a phylogenetic analysis of the aforementioned acetyltransferases was employed to illuminate the diversity among N- and O-linked glycosylation pathway enzymes.

pathogenic strain (ATCC 17978) of A. baumannii was recently characterized (9) and found not to include diNAcBac.
Previous structural characterization of the diNAcBac biosynthetic pathway has focused on the acetyltransferase PglD, an N-linked glycosylation pathway enzyme from C. jejuni (10,11). Additionally, genetic studies have shown that deletion of the pglD gene in C. jejuni results in the loss of the final heptasaccharide and dramatic reduction of colonization in a chick animal model; however, a low level of glycosylation was still detected by lectin blotting and mass spectrometry (12). PglD is a member of the left-handed ␤-helix family and consists of two separate domains. The N-terminal domain contains a ␤-␣-␤-␣-␤-␣ Rossman fold motif to accommodate UDP-4-amino sugar binding. A hexapeptide repeat motif defines the C-terminal domain that is responsible for the left-handed ␤-helix and AcCoA binding. The oligomeric state of PglD consists of a homotrimer that utilizes the left-handed ␤-helix motif of two protomers to form the cleft for AcCoA binding. Structures of other bacterial N-acetyltransferases have recently been reported (13)(14)(15), although they are distant homologues of PglD based upon their divergent sugar substrates. However, the sugar acetyltransferases maintain the same overall protein fold by forming a trimer as the biological unit. In addition, they utilize the same left-handed ␤-helix motif from adjacent protomers to form the AcCoA binding pocket. Structures of mammalian acetyltransferases, such as HAT1 (16), belonging to the GNAT (GCN5-related N-acetyltransferase) superfamily, bear no resemblance to their bacterial counterparts. This is most likely due to the considerable difference between their respective acyl acceptor substrates, histone H4 (HAT1) and UDP-4-amino (PglD). Interestingly, AcCoA has been shown to adopt one of two distinct conformations, either bent or curved, depending upon the specific acetyltransferase in question (17). Similar to citrate synthase (18), AcCoA bound to PglD adopts a compact conformation with a bend at the pyrophosphate moiety.
To further our understanding of acetyltransferases from the different UDP-diNAcBac biosynthetic pathways and to gain insight into the divergent nature of N-and O-linked protein glycosylation in prokaryotes, acetyltransferases from N. gonorrhoeae (PglB-ATD) and A. baumannii (WeeI) were investigated. To this effect, these enzymes were purified and crystallized, and the structures were solved to high resolution. In addition, a co-crystal structure of PglB-ATD bound to AcCoA was determined. In this context, a comparison between these structures and the previously solved C. jejuni acetyltransferase (PglD) crystal structures (10) was explored. Interestingly, the assumption that these bacterial acetyltransferases should closely resemble each other because they catalyze the identical reaction is not founded. Surprisingly, the substrate binding pockets for each of these enzymes vary considerably. Based upon this structural comparison, a series of active site mutations was carried out on all three acetyltransferases, and the enzymes were characterized kinetically for both AcCoA and UDP-4-amino substrates to gain insight into the catalytic mechanism. These studies suggest that although each enzyme catalyzes the acetyltransferase reaction with identical substrates, key residues within the binding pockets lead to a diverse set of catalytic efficiencies. Last, a phylogenetic analysis of acetyltransferases that catalyze the conversion to UDP-di-NAcBac in N-and O-linked glycosylation pathways is examined. The three acetyltransferases presented exhibit a high level of evolutionary diversity despite their ability to generate the identical final UDP-diNAcBac sugar. Unexpectedly, PglB-ATD from the O-linked glycosylation pathway shares a more common ancestral lineage with the PglD (N-linked) when compared with WeeI (O-linked).

EXPERIMENTAL PROCEDURES
Common Materials-All chemicals were purchased from Sigma-Aldrich unless otherwise stated. The UDP-4-amino sugar was biosynthesized as described previously from the C. jejuni enzymes PglF and PglE (19).
Molecular Biology-The acetyltransferase domain (ATD) of the pglB gene from N. gonorrhoeae FA1090 was identified through a Clustal Omega alignment (20) with the C. jejuni acetyltransferase (PglD). The gene encoding this domain was amplified via the polymerase chain reaction (PCR) with the forward primer  NOVEMBER 8, 2013 • VOLUME 288 • NUMBER 45 5Ј-CGCGGATCCATGGCGGGGAATCGCAAACTCG-3Ј and the reverse primer 5Ј-GCAACCCGGCAAAGCCCCTTTAGCT-CGAGCGG-3Ј from the N. gonorrhoeae FA1090 strain (8). The weeI gene was amplified via PCR from the genomic DNA from the A. baumannii AYE strain (ATCC BAA-1710) (21). BamHI and XhoI restriction sites were engineered to facilitate cloning of each construct into a modified pET30b(ϩ) vector (Novagen) containing an N-terminal His 8 tag followed by a tobacco etch virus protease site prior to the BamHI site. Amplifications were accomplished with the PfuTurbo DNA polymerase (Stratagene) as described by the manufacturer. Amplicons were purified and double-digested with BamHI and XhoI restriction enzymes (New England Biolabs). Digested inserts and linearized vectors were fractionated by agarose gel electrophoresis and purified with the Wizard SV Gel and PCR Cleanup Kit (Promega). Ligations were conducted with the T4 DNA ligase kit (Promega) using a 15-min incubation at room temperature. Sequencing by Genewiz (Cambridge, MA) confirmed the presence of all gene products. Site-directed mutagenesis was accomplished utilizing the QuikChange protocol (Stratagene) with pglD-pET24a(ϩ), pglB-ATD-pET24a(ϩ), and weeI-pET24a(ϩ) (from C. jejuni, N. gonorrhoeae, and A. baumannii, respectively) as the template plasmids from previous studies (4,5,19).

Structure and Function of Bacterial Sugar Acetyltransferases
Protein Expression-The modified pET30b(ϩ) plasmid containing each gene was used to transform Escherichia coli BL21(DE3)pLysS RIL competent cells (Stratagene). One liter of LB medium containing 50 g/ml kanamycin and 30 g/ml chloramphenicol was inoculated with 8 ml of an overnight culture of cells. The cells were then allowed to grow at 37°C while shaking until an optical density of ϳ0.8 ( ϭ 600 nm) was reached. The culture was cooled to 16°C and induced with 0.5 mM iso-␤-D-thiogalactosylpyranoside. After incubating for 18 h with shaking at 16°C, the cells were harvested by centrifugation (2600 ϫ g, 30 min) and stored at Ϫ80°C until needed.
Protein Purification-Each protein purification step was carried out at 4°C. For crystallization experiments, the cell pellet (ϳ3 g) was resuspended in 40 ml of 50 mM HEPES, pH 7.4, 100 mM NaCl, 30 mM imidazole (Buffer A) and then lysed by sonication. The lysate was then cleared by centrifugation (145,000 ϫ g, 60 min) and added to 2 ml of nickel-nitrilotriacetic acid resin (Qiagen). The slurry was allowed to tumble for 3 h and then packed into a fritted PolyPrep column (Bio-Rad). The resin was washed with 20 column volumes of Buffer A and then eluted with a buffer containing 50 mM HEPES, pH 7.4, 100 mM NaCl, 300 mM imidazole. Fractions containing the purified protein as analyzed by SDS-PAGE were pooled and dialyzed against 50 mM Tris, pH 8.0, 5 mM EDTA, 5 mM ␤-mercaptoethanol in the presence of 6 M tobacco etch virus protease for 24 h to remove the His 8 tag. Removal of this tag was monitored by Western blot analysis using an anti-His 4 antibody (Qiagen). The reaction was diluted 10-fold in 25 mM HEPES, pH 7.6, and excess tobacco etch virus was then removed with a HiTrap Q HP Sepharose anion exchange column (GE Healthcare) utilizing a linear NaCl gradient. Fractions containing the protein were pooled and dialyzed for 24 h in 50 mM HEPES, pH 8.0, 150 mM NaCl (SEC buffer). After concentrating to a volume of 1.5 ml using a 10,000 Da molecular weight cut-off Amicon Ultra-15 centrifugal filter unit (Millipore), the protein was loaded onto a Superdex 200 16/60 column (GE Healthcare) and subjected to size exclusion chromatography in SEC buffer. Fractions containing the monodispersed protein were pooled, concentrated, and used within 24 h for crystallization experiments. Protein concentrations were calculated based upon the predicted extinction coefficients at ϭ 280 nm.
Proteins subjected to mutagenesis were purified using 2 ml of nickel-nitrilotriacetic acid as above. Following elution from the resin, fractions containing the pure protein were dialyzed in a 4-liter volume against 50 mM HEPES, pH 7.4, 100 mM NaCl for 24 h to remove the imidazole. Purity for each protein was assessed by SDS-PAGE to be Ͼ95% (Fig. 2). Full-length constructs were confirmed through Western blot analysis probing with antibody for the T7 and His tags. This solution was concentrated as described above to ϳ10 mg/ml and supplemented with 15% glycerol. Aliquots of each protein were stored at Ϫ80°C until needed.
Crystallization and Data Collection-All crystals were grown as hanging drops by combining 1.5 l of a 10 mg/ml protein solution in SEC buffer with 1.5 l of reservoir solution at 25°C. Each well contained a final volume of 500 l of reservoir solution. For the co-crystallization of PglB-ATD with AcCoA, the substrate was added to the protein so that the final concentration was 10 mM and incubated for 45 min at 25°C. The reservoir solution for apo-PglB-ATD contained 0.1 M sodium acetate, pH 4.6, 0.02 M calcium chloride, and 30% 2-methyl-2,4-pentanediol. The AcCoA-bound PglB-ATD reservoir solution contained 0.1 M BisTris, pH 5.5, 3.0 M NaCl. For apo-WeeI, the well solution contained 0.1 M sodium acetate trihydrate, pH 4.5, 3.0 M NaCl, 0.7% 1-butanol. After the crystals were fully grown (ϳ24 h), they were cryoprotected in reservoir solution containing 20% glycerol. For AcCoA-bound PglB-ATD, this solution was also supplemented with 10 mM substrate. Diffraction data were collected on beamline X25 (National Synchrotron Light Source, Brookhaven National Laboratory, Upton, NY) at 100 K using a Pilatus 6M detector. Data sets were processed using HKL2000 (22), MOSFLM (23), TRUNCATE (24,25), and SCALA (24). Parameters from the data collection are listed in Table 1. Structure Determination and Refinement-Preliminary electron density maps for the PglB-ATD and WeeI structures were generated in PHASER (26), utilizing the previously solved PglD structure (Protein Data Bank code 3BSW) (10) as the molecular replacement search model. Refinement and model building of each structure were accomplished with COOT (27) and PHENIX (28). Water molecules were added using COOT, and the AcCoA ligand was modeled into PglB-ATD after the R free value was Ͻ30%. Refined structures were validated using MolProbity (29). Composite omit maps for the AcCoA-bound PglB-ATD structure were generated with PHENIX. The final refinement statistics are listed in Table 1.
Acetyltransferase Activity Assay-Enzyme mutants were analyzed for activity utilizing a 5,5Ј-dithiobis-(2-nitrobenzoic acid) spectrophotometric assay as described previously (4). Briefly, each assay was carried out at 50 mM HEPES, pH 7.4, 2 mM MgCl 2 , 0.05% BSA, 0.001% Triton X-100, and 1 mM 5,5Јdithio-bis-(2-nitrobenzoic acid). The substrate concentrations of AcCoA and UDP-4-amino were varied separately while holding the other substrate at 2 mM. Reactions were completed in duplicate, and initial rates were measured in the linear portion of the reaction curve over a 5-min time period at 25°C.

RESULTS
Structure of the N. gonorrhoeae Acetyltransferase PglB-ATD-PglB from N. gonorrhoeae is a bifunctional enzyme containing an N-terminal phosphoglycosyltransferase domain and a C-terminal acetyltransferase domain (ATD) that are homologous to the C. jejuni enzymes PglC and PglD, respectively (5). For crystallographic studies, the membrane-bound phosphoglycosyltransferase domain was removed based upon a Clustal Omega alignment with PglD, thus leaving behind the acetyltransferase domain referred to herein as PglB-ATD. The structure of the apo-form of PglB-ATD was solved by molecular replacement utilizing the previously solved acetyltransferase PglD (sequence identity ϭ 34%) (10). Difficulties in crystallization of this protein were addressed by removing the final 10 amino acid residues from the C-terminal tail based upon a sequence alignment with PglD. The removal of these PglB-ATD residues, which are not present in corresponding PglD sequence, results in a comparable C-terminal tail between the two constructs. PglB-ATD was crystallized in the cubic space group P2 1 3 with a single protomer in the asymmetric unit. Previous work has indicated that bacterial acetyltransferases trimerize in solution (10,35). Whereas the structure of PglB-ATD shows a single molecule in the asymmetric unit, the homotrimer can be observed through crystallographic symmetry centered on a 3-fold axis (Fig. 3A). This acetyltransferase contains two distinct domains that are responsible for the catalysis of UDP-4-amino to UDP-di-NAcBac using the AcCoA cosubstrate. The N-terminal section R free was calculated for 5% of the reflections randomly excluded from the refinement. d Ramachandran plot statistics are given as core/allowed/generously allowed and are for all chains.
(Asn 199 -Leu 285 ) comprises a binding pocket for the UDP-4amino sugar substrate through a ␤-␣-␤-␣-␤-␣ Rossman fold motif. The C terminus (Pro 286 -Leu 403 ) is composed of a lefthanded ␤-helix motif that, in conjunction with an adjacent PglB-ATD protomer in the trimeric state, forms an extended cleft that is utilized for AcCoA binding. Although the N. gonorrhoeae acetyltransferase catalyzes the same reaction as PglD from C. jejuni and has the same general fold (r.m.s. deviation ϭ 0.79 Å), there are a few notable differences in the structures. The PglB-ATD structure contains a flexible loop (Arg 233 -Thr 246 ) that is not observed in PglD (Fig.  3B). This loop is tucked in between ␣-helices 1 and 2 in the N-terminal sugar-binding domain and makes numerous backbone interactions with the second ␤-sheet (Phe 229 /Asp 231 / Asp 232 ). For instance, the side chain amide nitrogen of Asn 239 has a hydrogen-bonding interaction with the Glu 216 acid moiety on helix ␣1. Similarly, the backbone amide nitrogen of Leu 248 and Leu 249 on helix ␣2 interacts with the hydroxyl and carbonyl moiety from the Thr 246 loop residue, respectively. In the apo-state, PglD contains a cofactor gate, comprising the final 10 C-terminal residues that interacts with the adjacent, active site protomer (10). To accommodate AcCoA binding, this gate undergoes a conformation change such that an interaction is formed with the cognate protomer in a coiled motif. Surprisingly, the apo-structure of PglB-ATD reveals that no such cofactor gate is evident (Fig. 3B). Instead, the apo-state structure exists as the coiled motif resembling the AcCoAbound structure of PglD (3BSY). Additional structures of the apo-form of PglB-ATD were solved under distinct crystallization conditions that further supported the absence of the cofactor gate (data not shown).
Multiple attempts to crystallize PglB-ATD in the presence of UDP-4-amino were unsuccessful. Therefore, a structural alignment of apo-PglB-ATD and UDP-4-amino bound PglD (3BSS) was explored due to the minimal changes within the N-terminal domain upon sugar binding in the PglD structures (r.m.s. deviation ϭ 0.70). PglD residues Asp 35 (Asp 231 in PglB-ATD), Asp 36 (Asp 232 ), and His 125 (His 333 ), which accept hydrogen bonds from the ribosyl 3Ј-hydroxyl group, uridine imide, and pyranose C4-amine, respectively, are strictly conserved between the two structures. Only two notable changes between the structures are observed. The Asn 162 amino acid in PglD, which interacts with the carbonyl oxygen of the pyranose C2-acetyl group, is modified to the homologous Gln 370 residue in the PglB-ATD structure. Interestingly, Ser 13 in the PglD structure, which plays a significant role in the sugar binding pocket by hydrogen bonding to the ␣-phosphate of UDP-4-amino and Nʦ of Lys 38 , is replaced by Gly 208 in PglB-ATD. One cannot rule out the significance of the aforementioned loop in PglB-ATD with respect to sugar binding. Upon UDP-4-amino binding to PglD, an unwinding of helix ␣2 (Met 40 -Thr 45 ) to accommodate sugar binding and allow for optimal interactions is apparent in the crystal structure. The PglB-ATD flexible loop is located adjacent to this helix (Fig. 3B) and, upon sugar binding, could elicit a conformational change in this enzyme to mimic missing interactions within this site. Clearly, a PglB-ATD UDP-4-amino-bound structure would be necessary to confirm this hypothesis.
Structure of the N. gonorrhoeae Acetyltransferase PglB-ATD Bound to AcCoA-The AcCoA-bound PglB-ATD structure was solved by molecular replacement using the apo-PglB-ATD structure. This protein was crystallized in the tetragonal space group P4 3 2 1 2 with three PglB-ATD protomers in the asymmetric unit. Three AcCoA molecules were observed between the clefts formed by adjacent left-handed ␤-helices in a compact conformation with a bend at the pyrophosphate moiety (Fig. 4). AcCoA binds to PglB-ATD in a similar fashion with respect to PglD; however, there are noticeable differences between coenzyme and binding pocket residues. Notably, PglB-ATD utilizes a series of seven residues from both protomers to bind AcCoA. In particular, Ser 350 and the backbone amide nitrogen of Gly 351 form a hydrogen bond to the carbonyl oxygen of the thioester (Fig. 5A). This is in stark contrast to PglD, where the acetyl group is rotated 180 o and forms hydrogen-bonding interactions with His 134 and Asn 118 (Fig. 5B). Although the contacts between protomer B in PglD and AcCoA are mainly hydrophobic, this protomer plays a much larger role in PglB-ATD. Both hydrogen-bonding interactions of the thioester carbonyl originate from this protomer. Likewise, the water hydrogenbonding network binding the pyrophosphate moiety and the 3Ј-phosphate is replaced by Arg 368 and Lys 401 , respectively. Only two water molecules are observed binding to AcCoA in the PglB-ATD structure, whereas seven water molecules are contributing factors in the PglD structure. In fact, there are no conserved residues responsible for binding to AcCoA observed when comparing the two structures. However, backbone interactions between Ala 381 (PglB-ATD) and Gly 173 (PglD) serve a similar purpose by hydrogen bonding to a carbonyl oxygen in the pantetheine moiety and the C6 amine on the adenine ring. Further hydrogen-bonding interactions in the PglB-ATD structure can be observed from Thr 363 (protomer A) and Gln 369 (protomer B) on the amide adjacent to the thioester in the pantetheine moiety.
Although PglD undergoes a conformational change in the C-terminal tail upon AcCoA binding, no evidence of this change is observed in the PglB-ATD structure. In fact, because the apo-state of PglB-ATD is already in the coiled motif, as observed in the AcCoA-PglD structure, no other conformational change is necessary to accommodate AcCoA binding. This lack of change between the AcCoA-bound state and apostate of PglB-ATD is reflected in the minor change in r.m.s. deviation between the two structures (0.30 Å). However care must be taken in interpreting these results because a small change in r.m.s. deviation may be biased because the AcCoAbound structure was solved by molecular replacement with the apo-PglB-ATD structure. There are only two key conformational changes in the active site cleft necessary for AcCoA binding. Most importantly, Arg 368 in the apo-PglB-ATD structure serves to block access to the channel prior to AcCoA binding. Upon binding, Arg 368 rotates out of the cleft to allow AcCoA access to the binding site. This residue is also essential for the binding of AcCoA because it has a total of four hydrogen-bonding interactions with the coenzyme (Fig. 5A). Although no such C-terminal cofactor gate exists in PglB-ATD, Arg 368 may play a similar role to allow for AcCoA binding. Glutamine 369 also plays a role in coenzyme binding by rotating 90º to form part of the pantetheine binding pocket and picks up a favorable hydrogen-bonding interaction with AcCoA. Analogous residues are not apparent in a structural alignment between PglB-ATD and PglD, adding to the dichotomous nature of these two proteins.
Structure of the A. baumannii Acetyltransferase WeeI-The WeeI structure was solved by molecular replacement using the previously solved apo-PglB-ATD structure (sequence identity ϭ 26%). This acetyltransferase crystallized in the hexagonal space group P3 1 21 and contained six protomers in the asymmetric unit forming a dimer of the biological trimer assembly. Optimization of the crystals was a necessity due to the poor diffraction quality of the original conditions. In particular, the addition of 0.7% 1-butanol to the crystallization buffer improved resolution by 0.8 Å (Hampton additive screen). Similar to PglB-ATD and PglD, WeeI is composed of N-terminal (Met 1 -His 90 ) and C-terminal (Leu 91 -Leu 213 ) domains that are each responsible for binding to UDP-4-amino and AcCoA, respectively (Fig. 6A). A cleft is formed between two adjacent protomers from the C-terminal left-handed ␤-helix domain that accommodates AcCoA binding, as observed previously in the PglB-ATD crystal structure. Unfortunately, multiple screening attempts to solve the WeeI structure bound to the UDP-4-amino sugar and AcCoA proved unsuccessful.
Similar to the apo-structure of PglB-ATD, WeeI contains a flexible loop between helices ␣1 and ␣2 in the N-terminal sugar-binding domain. As previously observed with PglB-ATD, this loop has numerous intramolecular interactions with the protein backbone. Interestingly, the conserved residue Asn 46 (Asn 239 in PglB-ATD) exhibits a similar hydrogen-bonding interaction with Asn 121 on an adjacent protomer. Of note, Pro 49 (Pro 242 in PglB-ATD) is also conserved in this region and serves to stabilize this loop through hydrogen bonding of the backbone carbonyl to the conserved Phe 35 (Phe 229 in PglB-ATD) amide nitrogen. The essential sugar binding residues observed in the PglD/UDP-4-amino structure are strictly conserved in WeeI, including Ser 13 (Ser 11 in PglD), which is conspicuously absent in the PglB-ATD structure. Residues Asp 35 , Asp 36 , and His 125 in PglD, which contribute hydrogen-bonding interactions with UDP-4-amino, are conserved in WeeI (Asp 37 , Asp 38 , and His 138 ). The only exception in this binding pocket is the PglD Asn 162 (Gln 370 in PglB-ATD) residue. In WeeI, the pyranose moiety of the UDP-4-amino binding pocket is formed by a seven-amino acid loop (Gln 174 -Pro 180 ) from the adjacent protomer. This loop is not observed in the two other acetyltransferase structures (Fig. 6B) and contains two residues (Gln 174 and Thr 176 ) in the vicinity of hydrogen bonding to the carbonyl oxygen of the pyranose C2-acetyl group. Alanine mutagenesis was performed on these two sites to ascertain their relationship to UDP-4-amino binding (see below). In PglD, a conformational change in His 15 is observed to accommodate sugar substrate binding. In the apo-structure, this residue occludes the UDP-4-amino pocket. However, upon substrate binding, this residue tucks into the pocket and interacts with the ␤-phosphate moiety of the sugar. Whereas this residue is conserved in PglB-ATD (His 210 ), the more bulky, hydrophobic phenylalanine residue is found in WeeI. This small change could have a deleterious binding effect on the UDP-4-amino substrate (see below).
The WeeI AcCoA binding pocket exhibits a stronger homology to the PglB-ATD site (56% sequence identity) when compared with PglD (34% sequence identity). Not surprisingly, this can also be observed when comparing the crystal structures. Similar to PglB-ATD, WeeI does not appear to utilize a cofactor gate for AcCoA binding (Fig. 6B). From the apo-structure, the C-terminal tail is in a coiled motif that resembles the AcCoAbound PglD structure. WeeI also contains a residue analogous to Arg 368 (PglB-ATD) that may act as a gate to AcCoA binding. Lysine 173 is positioned in a fashion similar to Arg 368 and obstructs the binding cleft in the apo-state. Although no structure of AcCoA bound to WeeI exists, one can hypothesize that this residue plays an analogous role in coenzyme binding. Key residues that interact with AcCoA in the PglB-ATD crystal structure are mostly conserved in WeeI. PglB-ATD residues Gly 351 (Gly 156 in WeeI), Gln 369 (Gln 174 ), and Thr 363 (Thr 168 ) are strictly conserved. Substitutions at Lys 401 (Arg 211 ) and Ser 350 (Asn 155 ) are complementary in nature, and a similar role can be envisioned at these positions. Of note, the 124 EHE (PglD) and 332 DHD (PglB-ATD) motifs that are critical for catalysis are slightly modified in WeeI ( 137 AHD). The carboxylate moiety of PglD (Glu 126 ), PglB-ATD (Asp 334 ), and WeeI (Asp 139 ) is hydro-gen-bonded to the imidazole ring of histidine, increasing its basicity. This enhancement allows for the Nʦ2 nitrogen of histidine to act as a general base in catalysis by deprotonating the C4 amine on the UDP-4-amino sugar. Although the carboxylate moiety in PglD (Glu 124 ) and PglB-ATD (Asp 332 ) may serve to recycle histidine back to its precatalytic state by abstracting a proton from Nʦ2 following substrate turnover (10), this cannot be the case in WeeI due to the alanine moiety at this position.
Analysis of Acetyltransferase Active Site Mutants-To better understand the contributions of particular residues in binding and catalysis, a series of mutations was created in the active sites of PglD, PglB-ATD, and WeeI based upon their crystal structures. While holding one substrate at saturating levels for PglD and PglB-ATD, the other was varied to determine kinetic parameters through initial velocity measurements. Ellman's reagent (5,5Ј-dithiobis(2-nitrobenzoic acid)) was utilized to monitor AcCoA conversion to CoASH through generation of the TNB 2Ϫ chromophore (⑀ 412 nm ϭ 14,150 M Ϫ1 cm Ϫ1 ). Due to the poor binding of UDP-4-amino to WeeI, the AcCoA kinetic parameters were determined at the K m of the UDP-sugar. Typical Michaelis-Menten kinetics were observed for all concentrations of UDP-4-amino and AcCoA. Initial velocity measurements were averaged between two duplicate experiments. UDP-4-amino and AcCoA kinetic parameters for the acetyltransferase mutants are listed in Tables 2 and 3, respectively.
When comparing UDP-4-amino affinity to bacterial acetyltransferases, WeeI exhibits poor binding (K m Ͼ 10-fold). From a structural alignment standpoint, the phenylalanine at position 13 may contribute to such a poor K m in WeeI. This observation is based upon the absence of changes in UDP-4-amino binding pocket residues with respect to PglD and PglB-ATD. The aforementioned His 15 residue in PglD can be classified as a type of gatekeeper moiety due to its ability to tuck into the pocket to accommodate UDP-4-amino binding and interact directly with this substrate. The histidine is conserved in PglB-ATD, resulting in a similar UDP-4-amino K m ; however, this site is a phenylalanine in WeeI. This change in steric bulk, hydrophobicity, and loss of hydrogen bonding at this key position may result in reduced binding affinity. Therefore, a mutation in PglD (H15F), PglB-ATD (H210F), and WeeI (F13A) was explored. This mutation had a deleterious effect on both catalysis and binding on both PglD and PglB-ATD, whereas the F13A WeeI mutation mainly affected turnover (Table 2). In order to ascertain if these mutations have any effect on the adjacent AcCoA binding pocket, kinetic parameters were explored for this substrate. Surprisingly, the PglD H15F mutation resulted in a 10-fold increase in binding affinity to AcCoA while decreasing k cat by 20-fold ( Table 3). Mutation of H210F (PglB-ATD) and F13A (WeeI) resulted in no change in binding affinity; however, k cat decreased considerably. This particular site in the UDP-4-amino binding pocket contributes significantly to binding and catalysis in these acetyltransferases.
The inability to crystallize UDP-4-amino with either PglB-ATD or WeeI prompted a series of alanine mutations to determine specific sites within each binding pocket that contribute to binding and catalysis. Based upon an alignment with the UDP-4-amino PglD structure, PglB-ATD (Q369A and Q370A) and WeeI (Q174A and T176A) mutants were created. In both cases, these changes are within the vicinity of the PglD residue Asn 162 , which interacts with the carbonyl oxygen of the pyranose C2-acetyl group. Although neither PglB-ATD mutation resulted in a change in UDP-4-amino binding, Q369A had a significant effect on turnover (13-fold decrease). Likewise in the WeeI mutations, only T174A resulted in a considerable (270fold) loss in k cat . Due to the proximity of these mutations to the AcCoA binding site, kinetic parameters were also established for this substrate. Mirroring the UDP-4-amino results, a decrease in k cat was only observed for the PglB-ATD Q369A mutant (6-fold) and WeeI Q174A (260-fold). It is apparent that the WeeI Q174A mutant plays an extremely important role in catalysis of this reaction.
There is still an ongoing discussion over the acetyltransferase catalytic mechanism and in particular the protonation state of UDP-4-amino substrate (10,11,15). Although this study does not address this question specifically, the role that Glu 124 (PglD) plays in catalysis was explored. This position has been implicated in returning the catalytic histidine (H125) back to its preturnover state by transferring the proton on the imidazole moiety to the thiolate on CoAS Ϫ (10). Most homologous acetyltransferases incorporate either a glutamate or aspartate at this position; however, this site is occupied by an alanine in WeeI. Interestingly, the catalytic efficiency of WeeI is comparable with that of PglD and PglB-ATD. To better understand the catalysis and binding at this site, mutant variants of PglD (E124A) and PglB-ATD (D332A) were prepared. In both cases, k cat was reduced for both substrates (Tables 2 and 3); however, the loss was more significant in PglB-ATD (20-fold). Mutation to alanine in PglD and PglB-ATD then has a detrimental effect on catalysis, yet the wild-type WeeI is still a competent enzyme with alanine at this position. Therefore, recycling of the active site must be accomplished in another manner. The closest amino acid site that can act as a general base is Lys 15 ; however, that residue is over 6 Å away from the catalytic histidine. The most straightforward solution would rely on the thiolate from CoAS Ϫ (following acetylation of the UDP-4-amino sugar) to act as a base to directly remove the proton from the catalytic histidine. This would regenerate the active site to its precatalytic state and explain the absence of a general base adjacent to His 138 in WeeI. In lieu of these results, the glutamate/aspartate moiety in PglD (Glu 124 ) and PglB-ATD (Asp 332 ) appears to be a non-absolute requirement for catalysis, and its essentiality may have previously been overstated.
Phylogenetic Analysis of Bacterial Acetyltransferases-Although the three acetyltransferases presented here carry out the same reaction and display the same general protein fold, homology within the substrate binding sites is quite divergent. To further our understanding on the evolutionary aspect of acetyl-  transferases within the diNAcBac pathway, a phylogenetic analysis was carried out. Bacteria containing the diNAcBac pathway were first identified by having a Ͼ35% homology to known oligosaccharyltransferases from C. jejuni, N. gonorrhoeae, and A. baumannii. Comparative assessment of these enzymes allowed for classification of PglD as an N-linked glycosylation system and PglB-ATD/ WeeI as O-linked systems. Acetyltransferases were further classified similarly to the oligosaccharyltransferase analysis above and sequentially aligned with the software program MUSCLE. Interestingly, the neighbor-joining dendrogram (Fig. 7) is broken up into multiple clades and exhibits evolutionary diversity, which is observed within the acetyltransferase binding pockets. This is somewhat surprising because the acetyltransferases from C. jejuni, N. gonorrhoeae, and A. baumannii carry out the identical reaction to produce the diNAcBac sugar. Similar results were observed previously using dehydratase and aminotransferase homologs from Campylobacter and Neisseria (36). Of note, homologous A. baumannii acetyltransferases are evolutionarily more distant with respect to C. jejuni and N. gonorrhoeae.

Divergence of Bacterial Acetyltransferases in N-and O-Linked Protein Glycosylation
Pathways-Glycosylation is a ubiquitous post-translational modification and is known for modulating cellular processes, such as protein folding, stability, and cell signaling (37)(38). Significantly, bacteria also utilize protein gly- cosylation for purposes of mediating colonization, adhesion, and invasion of eukaryotic cells (1,12,39). In fact, recent work on the ATCC 17978 strain of A. baumannii has demonstrated a link between pathogenicity and protein glycosylation (9,40). To better understand the module responsible for the biosynthesis of UDP-diNAcBac, research has focused on the specific enzymes that lead to the synthesis of this unusual sugar. Bacterial glycosylation can be classified as N-linked (asparaginelinked glycan) and O-linked (serine/threonine-linked glycan). Both modifications, in the context of UDP-diNAcBac biosynthesis, have been studied extensively in C. jejuni (N-linked) (7,10,19) and to a lesser extent in N. gonorrhoeae and A. baumannii (O-linked) (4,5,8). Although the enzymes responsible for the biosynthesis of this unique, nucleotide sugar are present in these bacteria, they are evolutionarily divergent with regard to their acetyltransferases (Fig. 7). Homologs of these enzymes from their respective organisms are separated into multiple clades within the dendrogram. There are two explanations to account for this observation. First, these enzymes could have covergently evolved by acquiring the biosynthetic enzymes necessary for the production of UDP-diNAcBac. Second, these enzymes could have evolved from a common ancestor and diverged over an extended period of time. This is the simpler explanation and could account for the varying degrees of identity observed within the AcCoA and UDP-4-amino binding pockets (4). For instance, the C. jejuni PglD UDP-4-amino binding pocket shares a higher sequence identity with PglB-ATD from N. gonorrhoeae. Conversely, WeeI from A. baumannii shares a higher homology with its O-linked counterpart, PglB-ATD, in the AcCoA binding pocket. In either case, it is interesting that the acetyltransferases from two O-linked pathogens (N. gonorrhoeae and A. baumannii) are evolutionarily more divergent with respect to the N-linked C. jejuni enzyme. It is currently unknown whether A. baumannii acquired this enzyme from an N-or O-linked pathway. Although the true significance of UDP-diNAcBac is presently unclear, it is important to recognize its ubiquitous nature in pathogenic bacteria. Why specific bacteria acquired the UDP-diNAcBac biosynthetic pathway remains a mystery. Additionally, questions surrounding the motility of the UDP-diNAcBac module between bacteria in lieu of the entire glycosylation pathway remain unanswered. Further work is warranted to address these questions in the context of bacterial fitness and pathogenicity.
Mutagenesis of the UDP-4-amino Binding Pocket Reveals Kinetic Diversity-The structures of PglB-ATD and WeeI add to the growing number of acetyltransferases that are associated with UDP-diNAcBac biosynthesis. Importantly, these structures represent the first O-linked glycosylation pathway enzymes that result in the production of this bacterial sugar. Although the overall architecture of these proteins is similar with respect to PglD, there are notable differences that contribute to their contrasting kinetic parameters. In particular, WeeI binds to UDP-4-amino with a significantly lower affinity (10fold) in comparison with PglD and PglB-ATD (Table 2). From a structural alignment standpoint, WeeI contains one key residue (Phe 13 ) that may be responsible for this dramatic K m shift. In PglD, this position (His 15 ) undergoes a conformational change to accommodate UDP-4-amino binding and interacts with sugar ␤-phosphate. Site-directed mutagenesis of this position (H15F) resulted in a 10-fold loss in affinity for UDP-4amino binding with PglD. However, the same mutation in PglB-ATD (H210F) produced a more modest loss in binding (3-fold). This position is extremely important for acetyltransferase activity because there is a 100-fold decrease in catalytic efficiency (k cat /K m ) when mutating this residue to a phenylalanine in both PglD and PglB-ATD. Despite the poor binding affinity of UDP-4-amino, WeeI must contain a compensatory effect because this enzyme retains a similar efficiency with respect to PglD and PglB-ATD.
WeeI contains an additional loop (Gln 174 -Pro 180 ) that forms the UDP-4-amino binding pocket near the pyranose moiety. Residue Gln 174 seems to be critical for catalysis because an alanine mutation results in a 270-fold loss in k cat while maintaining its affinity for UDP-4-amino. When aligned to the PglD UDP-4-amino structure, this position is analogous to Asn 162 that interacts with the carbonyl oxygen of the pyranose C2-acetyl group. In the WeeI apo-structure, Gln 174 is within 5 Å of the catalytic base (His 138 ) and 3.6 Å of the AcCoA thioester when aligned to the PglB-ATD AcCoA structure. Likewise, the Q174A mutation has a detrimental effect on AcCoA catalysis with a 260-fold loss in turnover. Clearly, this residue plays a key role in the overall function of WeeI.
Dichotomy among N-and O-Linked Acetyltransferase AcCoA Binding Pockets-A general theme of binding and catalysis among homologous proteins is conservation of key amino acids that result in the comparable activity between enzymes. Although the AcCoA binding parameters of PglD and PglB-ATD are extremely similar (Table 3), the ways in which the enzymes bind the coenzyme are distinct (Fig. 5). AcCoA is mainly held into the binding pocket of PglD by hydrophobic interactions and a network of water molecules. In fact, only two side chains (Asn 118 and His 134 ) contribute to the overall affinity of this substrate in PglD. Surprisingly, these analogous residues in PglB-ATD play no role in binding to AcCoA. Instead, the coenzyme is held in the binding site by a series of hydrogenbonding interactions from a total of seven residues. Interactions between the phosphate moieties in PglB-ATD have replaced the water molecules in PglD with Arg 368 and Lys 401 .
In addition to this major change in binding site functionality, PglB-ATD does not appear to utilize a C-terminal cofactor gate for AcCoA binding. Upon AcCoA binding in the PglD structure, the C-terminal tail undergoes a conformational change to accommodate the coenzyme in the form of a coiled motif. This coiled motif is already apparent in the apo-structure of PglB-ATD (Fig. 3); however, the removal of the final 10 amino acids from the C-terminal tail for crystallization purposes could have elicited this result. However, this is unlikely because this deletion does not remove the residues responsible for this conformational change, as observed in PglD. Furthermore, the apo-WeeI structure is of the full-length protein and also does not exhibit a conformational change in this cofactor gate (Fig. 6). However, one cannot rule out the possibility of a crystallographic artifact when discussing these types of small conformational changes between enzymes. Although both O-linked glycosylation acetyltransferases do not contain this cofactor gate, comparisons between the apo-and AcCoA-bound structures of