Recombinant PaMnaA epimerase (MnaA.His₆), encompassing amino acids 1–384 with a Leu-Glu linker followed by a C-terminal His₆-tag, was expressed from a pET22b (+) (Novagen, Merck, Vienna, Austria)-based construct in E. coli BL21 (DE3) cells as described previously (Hager et al., 2018). Site directed mutagenesis at predicted catalytic and allosteric sites of PaMnaA, including Q42A, Q69A, E135A and H241A, was performed following the method described by Blacker and colleagues (Blackler et al., 2018) using oligonucleotides listed in Table 3 and pET22b (+) vector for cloning.

Overexpression of wild-type PaMnaA and variants at 37^oC with 200 rpm shaking in selective LB broth (10 g L^-1 tryptone, 5 g L^-1 yeast extract, 10 g L^-1 NaCl) supplemented with 100 μg mL^-1 ampicillin was induced at the mid-exponential growth phase (OD₆₀₀ ∼0.6) by adding isopropyl-β-d-thiogalactopyranoside (IPTG) to a final concentration of 0.6 mM. The cells were incubated for an additional 4 h after IPTG induction and subsequently harvested by centrifugation at 6,500 rpm for 25 min using a Beckmann JA-14 rotor (Beckman Coulter, Brea, CA, United States).

For X-ray crystallography, the cell pellet obtained from a 3-L E. coli BL21 (DE3) PaMnaA overexpression culture was resuspended in 100 mL of lysis buffer (5% (v/v) glycerol, 20 mM imidazole, 1 M NaCl, 0.2 M arginine, 50 mM NaH₂PO₄, pH 7.4), supplemented with a SIGMAFAST™ Protease Inhibitor Tablet (Sigma-Aldrich, St. Louis, MO, United States), 1 mL of 17 mg mL^-1 phenylmethylsulphonyl fluoride (PMSF) in isopropanol, 2 mM tris(2-carboxyethyl) phosphine (TCEP), and 3 mg of DNase I. The resuspended cells were disrupted using a Misonix Sonicator 3,000 (5 min total time, 10 s on, 10 s off, initial power output 6). After sonication, Triton® X-100 was added to a final concentration of 0.1% (v/v). The cell lysate was then centrifuged at 6,500 rpm for 25 min using a Beckmann JA-17 rotor (Beckman Coulter, Brea, CA, United States); the resultant supernatant was removed and centrifuged again at 10,000 rpm for 60 min to further remove cell debris. The final supernatant fraction (cell crude extract) was incubated for 30 min at 4°C with 3 mL of Thermo Scientific™ HisPur™ Ni-NTA resin (Thermo Fisher Scientific, Waltham, MA, United States) pre-equilibrated in lysis buffer. The lysate-Ni-NTA mixture was then loaded onto a column and washed with 50 mL of wash buffer (5% (v/v) glycerol, 25 mM imidazole, 1 M NaCl, 50 mM NaH₂PO₄, 2 mM TCEP, pH 7.4). Recombinant PaMnaA protein was eluted using a gradient of wash buffer (50 mL) and elution buffer (50 mL; 5% (v/v) glycerol, 250 mM imidazole, 1 M NaCl, 50 mM NaH₂PO₄, 2 mM TCEP, pH 7.4), and collected at a flow rate of 1 mL min^-1 in 1.5-mL fractions. Fractions were analyzed for purity by SDS-PAGE gel (12.5% polyacrylamide gel) stained with Coomassie Brilliant Blue G250 (CBB). Those fractions determined to be of the highest purity were pooled.

For enzyme assays, PaMnaA wild-type and mutants were purified from 500-mL batches of E. coli BL21 (DE3) PaMnaA-overexpression cultures according to Hager and colleagues (Hager et al., 2018) immediately before use.

The recombinant, purified PaMnaA sample was dialyzed in 4 L of dialysis buffer (20 mM HEPES, pH 7.5) for 4 h at 4°C with gentle stirring, using 3.5-K MWCO Thermo Scientific™ SnakeSkin™ dialysis tubing (Thermo Fisher Scientific, Waltham, MA, United States). The dialyzed sample was then concentrated to 17 mg mL^-1 in a 3-K MWCO Pall Macrosep® Advance Centrifugal Device (Pall Corporation, Port Washington, NY, United States) by centrifugation at 5,000 rpm in a JA-17 rotor kept at 4°C. Protein concentration was measured at A₂₈₀ using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States).

The concentrated PaMnaA sample was filtered through a 0.22-μm centrifugal filter immediately before setting up the crystallization conditions. Crystal screens of recombinant purified PaMnaA (17 mg/mL) were prepared using an Art Robbins Instrument Crystal Gryphon robot and Hampton 96-well Intelli-Plates. The crystallization plates were stored at 18°C. Initial crystals of PaMnaA were obtained from PEGs II condition number 36 (0.2 M CaCl₂, 0.1 M Tris/HCl pH 8.5, 20% (w/v) PEG 4000) 1 week after plating. This condition was further optimized by raising the pH and adjusting the concentrations of CaCl₂ and PEG 4000 (optimized condition: 0.16 M CaCl₂, 0.1 M Tris/HCl pH 8.8, and 15% PEG 4000). Larger crystallization drops were also prepared (1.5 μL mother liquor to 1.5 μL 17 mL^-1 MnaA) using hanging drop vapor diffusion at 18°C to generate larger, higher-quality crystals for use in subsequent X-ray diffraction experiments and data collection.

For MnaA crystal ligand soaking, MnaA crystals grown by hanging drop vapor diffusion in optimized PEGs II condition 36 were soaked in mother liquor solution with addition of either 10 mM or 50 mM UDP-α-d-GlcNAc (Sigma-Aldrich, St. Louis, MO, United States) overnight at 18°C.

A crystal obtained by hanging drop vapor diffusion from the optimization of PEGs II condition 36 was resuspended in a mother liquor solution containing 20% MPD for 10–20 s before being flash-frozen at 100 K for data collection. X-ray diffraction data were collected using a Rigaku MicroMax-007HF X-ray generator coupled to a Rigaku Dectris Pilatus3R 200K-A detector. The data were scaled, averaged, and integrated using HKL 2000 (Otwinowski and Minor, 1997). The data set was solved by molecular replacement with a model of PaMnaA predicted by Phyre2 (Kelley et al., 2015) and PHASER from the CCP4 program suite (Winn et al., 2011; McCoy et al., 2007). All subsequent model building and refinement steps were carried out using Coot (Emsley et al., 2010) and Refmac5 from the CCP4 suite (Winn et al., 2011; Murshudov et al., 2011).

Epimerization catalyzed by PaMnaA was measured by RP-HPLC (U3000, Dionex) run in 0.4 M sodium phosphate buffer, pH 6.4, with UV detection at 254 nm using a Hyperclone 5 µm 120 A 150 × 4.6 mm (Phenomenex) column at a flow rate of 0.6 mL min^-1 (Hager et al., 2018). The reaction set-up utilized a 10 mM sodium phosphate buffer, pH 8.0, as recommended by (Mann et al., 2016), with an assay volume of 50 µL. To inactivate the enzyme, the reactions were heated for 1 min at 100°C and then centrifuged at 10,000 g, 5 min at room temperature (RT). Various reaction conditions were evaluated in the forward direction with 0.5 mM UDP-GlcNAc (Sigma) and 420 ng of PaMnaA, including temperature (4°C, RT, 30°C and 37°C), pH (7.0, 7.5, 8.0, 8.5), and salt additions (MgCl₂, NaCl, MgSO₄, CaCl₂, MnSO₄; 50 and 10 mM, each) with a 5-min incubation time. The reverse reaction was investigated to test UDP-GlcNAc as an allosteric activator (1%, 5% and 10%) using 0.1 mM UDP-ManNAc [chemically synthesized as described in (Hager et al., 2018)] and 420 ng of PaMnaA with a 45-min incubation time.

A reaction time course was recorded to determine initial rates. For the forward reaction (0.1 mM UDP-GlcNAc and 50 ng of PaMnaA), aliquots were taken at 1, 3, 5, 8, 10 and 20 min. For the reverse reaction (0.1 mM UDP-ManNAc and 210 ng of PaMnaA), aliquots were taken at 10, 30, 60, 75, 90, 105 and 120 min.

Tunicamycin was dissolved in DMSO at a concentration of 1 mM and tested as a potential inhibitor of the reaction at concentrations of 0.5 µM, 1 μM, 10 μM, 50 μM, 100 µM and 200 μM, corresponding to a substrate-to-inhibitor ratio of 1:0.005 (0.5 µM), 1:0.01 (0.1 µM), 1:0.1 (10 µM), 1:0.5 (50 µM), 1:1 (100 µM), and 1:2 (200 µM), respectively (Mann et al., 2016), with a 2-h incubation time and a pre-incubation of tunicamycin with the enzyme for 15 min. To test whether DMSO affects the epimerization reaction, a control reaction was done by adding DMSO without tunicamycin.

The enzyme assays were performed in a total volume of 50 μL sodium phosphate buffer (10 mM, pH 8.0) containing 50 mM MgCl₂. UDP-GlcNAc and UDP-ManNAc were varied at concentrations of 0.1, 0.25, 0.5, 0.75, 1, 1.25, 1.5, 1.75, 2, 3, 4, 5, 7.5 and 10 mM, and at 0.083, 0.125, 0.250, 0.375, 0.5, 0.75, 1, and 2 mM, respectively, with either 50 ng or 210 ng of PaMnaA. For the PaMnaA variants, 50–1,440 ng of enzyme were used. The reactions were incubated for either for 5 min at 30°C (forward reaction) or 45 min at 30°C (reverse reaction). Upon varying the UDP-ManNAc or UDP-GlcNAc concentrations, apparent kinetic results were calculated by fitting the data (average values of triplicates) into the Michaelis Menten-equation using GraphPad Prism (version 9.1.2; GraphPad, San Diego, CA, United States).

To obtain a structural model of the unmodeled residues in chain A and chain B of PaMnaA which correspond to flexible loop regions, SWISS-MODEL (https://swissmodel.expasy.org/) (Waterhouse et al., 2018) was used. For comparability with the other epimerases (SaMnaA and NmSacA), in the subsequent docking studies, additionally a structural homology model of PaMnaA was created using SaMnaA (5ENZ) as a template to obtain a closed conformation. After protein structure modelling, docking studies of PaMnaA chain A in comparison to the crystal structure of SaMnaA (5ENZ), chain A and NmSacA (6VLC), all in a closed conformation, with tunicamycin [ZINC database (Sterling and Irwin, 2015)] as a potential inhibitor were conducted using AutoDockTools 1.5.7. The active/binding site for tunicamycin was predicted by selecting the conserved allosteric residues in PaMnaA (E42, H43, E45, Q69, R209 and H241) as well as in NmSacA (R10, M42, H245, E295), based on the residues in SaMnaA (E39, H40, E66, R206 and H238) (Mann et al., 2016). Furthermore, docking was done over the whole protein of PaMnaA and SaMnaA to identify putative other binding sites which might have escaped detection by docking to the allosteric site. The interactions of the inhibitor with the binding-site residues at the lowest free binding energy was used to create interaction diagrams for tunicamycin docked to PaMnaA, SaMnaA, and NmSacA using MOE (Molecular Operating Environment, v2019). Residues of the inhibitor within a distance of 4.5 Å were considered for interactions.

All figures depicting protein structures and surfaces were created using the UCSF Chimera Extensible Molecular Modeling System, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311 (Pettersen et al., 2004).

For X-ray crystallography, purified PaMnaA (Supplementary Figure S2) from a 3-L overexpression culture of E. coli BL21 (DE3) cells was used at a concentration of 17 mg mL^-1. Initial PaMnaA crystal hits were observed in the PEGs II screen condition number 36 (0.2 M CaCl₂, 0.1 M Tris/HCl pH 8.5, and 20% PEG 4000). This condition was optimized to yield crystals of sufficient quality for use in X-ray diffraction experiments by decreasing the PEG 4000 concentration (Supplementary Figure S3), increasing the crystallization drop size (Supplementary Figure S4) and increasing the pH (Supplementary Figure S5). The final condition, consisting of 0.16 M CaCl₂, 0.1 M Tris/HCl pH 8.8, and 15% PEG 4000, yielded the final PaMnaA crystal used for data collection and MnaA structure determination (detailed below).

All data collection and refinement statistics for the unliganded structure of PaMnaA are shown in Table 2. The crystal structure was solved at a final resolution of 2.20 Å in space group P12₁1 with two molecules in the asymmetric unit (AU). A search model generated by Phyre2 (Kelley et al., 2015) was used for molecular replacement in PHASER, part of the CCP4 program suite (Winn et al., 2011). The solution was finalized through iterative refinement and model improvements using REFMAC5 (Winn et al., 2011) and the Coot macromolecular model building software (Emsley et al., 2010). The final structure yielded an average B-factor of 36.1 Ų, with R_work and R_free values of 21.0% and 26.2%, respectively.

As expected, the overall structure of PaMnaA forms a GT-B fold, with both the N-terminal and C-terminal domains adopting distinct Rossmann folds connected by a short linker region (residues 173–200). The space between these two folds forms an interdomain cleft containing the active site (Figure 3). The N-terminal domain (consisting of residues 2–172) shows good electron density for most residues, except for positions 42–45 and 64–69 in molecule A, and those at positions 43–46 and 66–68 in molecule B, which showed no corresponding electron density and, thus, were not modeled. The N-terminal domain forms an extended β-sheet structure with 3-2-1-4-5-6-7 connectivity between the seven β-strands (β1-β7), which are linked by six α-helices (α1-α6) (Figure 3). The unmodeled residues in molecules A and B correspond to flexible loop regions between the β2 and α2, and β3 and α3 secondary structural elements. The linker region connecting the N- and C-terminal domains (residues 173–200) consists of two α-helices (α7 and α8). Residues in this region mostly show good electron density, apart from residues 186–192 in molecule A, and residues 186–190 in molecule B, which show only partial main chain and side-chain electron density. The C-terminal domain (residues 201–347) exhibits good electron density for all residues in both molecules A and B. Similar to the N-terminal domain, the C-terminal domain forms an extended β-sheet structure (with 3-2-1-4-5-6 connectivity between the β-strands) which comprises six β-strands (β8-β13) linked by five α-helices (α9-α13). Additionally, there are three α-helices (α14-α16) after the final β-strand of the C-terminal domain, with the last helix associated with the N-terminal domain of the PaMnaA crystal structure (Figure 3).

Co-crystal screening of PaMnaA with UDP-GlcNAc substrate did not produce crystal hits; therefore, MnaA crystals obtained from optimized PEGs II condition number 36 (0.16 M CaCl₂, 0.1 M Tris/HCl, pH 8.8, 15% PEG 4000) were soaked in a drop of mother liquor supplemented with UDP-α-d-GlcNAc in an attempt to generate a donor-bound structure. Overnight soaks in 10 mM UDP-α-d-GlcNAc produced no change to the resultant crystal structures, with no bound ligand observed (data not shown). Overnight soaking in 50 mM UDP-α-d-GlcNAc caused all crystals to dissolve.

Purified, C-terminal His₆-tagged PaMnaA was obtained at a concentration of 0.42 mg mL^-1 in 20 mM sodium phosphate buffer, pH 7.5. For enzymatic assays, PaMnaA was used directly after recombinant production.

Using our established HPLC assay (Hager et al., 2018), we evaluated the optimal reaction conditions for measuring the initial rate of PaMnaA catalysis, including temperature, pH, salt supplement, and incubation time and temperature. No clear temperature optimum was identified within the range tested, but a significantly reduced turnover was observed at 4°C (not shown). Therefore, 30°C was chosen as the assay temperature. The reaction was mostly unaffected by pH changes within a range of 7.0–8.5 (Supplementary Figure S6), so a pH of 8.0, as reported in the literature, was used. Interestingly, all salts tested, i.e., MgCl₂, MgSO₄, and NaCl, resulted in a significantly better performance of the enzyme compared to the reaction without addition of metal ions (Supplementary Figure S7). The addition of 50 mM MgCl₂ resulted in the best yield and was thus used for the enzyme assays. The optimal incubation time was individually determined for wild-type PaMnaA and the enzyme mutants. For the wild-type enzyme, the initial rate incubation time was determined to be 5 min for the forward reaction and 45 min for the reverse reaction. For the variants, the initial rate incubation time to reach the end of the linear phase of the reaction was determined to be between 5 and 30 min.

Using UDP-GlcNAc as a substrate, the apparent k _cat value of PaMnaA was determined to be 33.44 s^-1, with a K _M value of 3.91 mM (Figure 4A; Table 5). When UDP-ManNAc was used as the substrate, the apparent k _cat value was 6.02 s^-1, with a K _M value of 2.41 mM (Figure 4B; Table 5). PaMnaA exhibited an approximately three times higher catalytic efficiency towards UDP-GlcNAc (k _cat/K _M = 8.56 mM s^-1) compared to UDP-ManNAc (k _cat/K _M = 2.49 mM s^-1). However, PaMnaA showed an approximately two times higher K _M value for UDP-GlcNAc (3.91 mM) than for UDP-ManNAc (2.41 mM), indicating a lower affinity for UDP-GlcNAc. This lower affinity for UDP-GlcNAc is also seen when compared to other enzymes such as RffE of E. coli (Samuel and Tanner, 2004) or SaMnaA (Mann et al., 2016). Notably, a higher affinity for UDP-ManNAc than for UDP-GlcNAc was reported also for SaMnaA (Mann et al., 2016), but not for NmSacA (Hurlburt et al., 2020) (Table 5). In contrast to PaMnaA, SaMnaA shows a K _M value for UDP-ManNAc that is four times lower than that for UDP-GlcNAc (Mann et al., 2016), whereas for PaMnaA, this difference is only twofold. The determined K _M of PaMnaA for UDP-GlcNAc is almost doubled compared to that of BaMnaA (2.2 mM) and comparable to that of NmSacA (3.6 mM) (Zhang et al., 2016). Furthermore, the k _cat value of 33.44 s^-1 (forward reaction) and 6.02 s^-1 (reverse reaction) of PaMnaA is comparable to that calculated for NmSacA for UDP-GlcNAc (Zhang et al., 2016), whereas BaMnaA shows a lower k _cat of around 7.0 s^-1, comparable to 6.01 s^-1 determined for the reverse reaction of PaMnaA.

Overall, PaMnaA exhibited a lower affinity to UDP-GlcNAc in the forward reaction but an increased turnover and catalytic efficiency; consequently, the increased turnover and catalytic efficiency compensates for the lower affinity and leads to the previously determined 1:9-ratio in the UDP-ManNAc epimerization reaction (Hager et al., 2018).

Two of the PaMnaA variant constructs, Q42A and H241A, which targeted the predicted allosteric site (Figure 1), revealed a significant reduction of the k _cat value (0.05 s^-1 and 0.53 s^-1, respectively) and a 2.5-fold lower K _M value (1.61 mM) for Q42A and a similar K _M value for H241A (3.27 mM) (Supplementary Figures S8, S9). Both mutations showed a pronounced influence on the k _cat contrary to the corresponding BaMnaA variants, where the influence was not as pronounced (Velloso et al., 2008). Another point mutation within the allosteric site of PaMnaA, Q69A, showed a 2.3-fold decrease in affinity to UDP-GlcNAc (K _M of 8.92 mM) and a similar k _cat value of 37.92 s⁻¹) compared to the one of the wild-type enzyme (Supplementary Figure S10). Thus, intervention at the allosteric site had only a small effect on the affinity of the enzyme for UDP-GlcNAc. Conversely, the variants of the homologous residues in BaMnaA targeting the allosteric site showed a more pronounced effect; the K _M values showed a 5- to 8-fold increase, resulting in significantly poorer affinity to the substrate (Velloso et al., 2008) (Table 5). PaMnaA variant E135A was a direct intervention on a putative catalytic residue and led to a drastic reduction in activity and no kinetic data could be obtained.

Next, it was investigated whether UDP-GlcNAc serves as an allosteric activator of the PaMnaA-catalyzed epimerization reaction (reverse reaction), as reported for some UDP-GlcNAc 2-epimerases of other bacteria. We found that, while the reverse reaction did not require addition of UDP-GlcNAc to the assay mixture, the reaction rate was doubled after 5 min when 5% UDP-GlcNAc was included in the reaction mixture. Notably, provision of UDP-GlcNAc at 10% only led to a slight increase of the reaction rate with a bad cost-to-benefit ratio (Figure 5). Thus, PaMnaA shows a similar behavior regarding allosteric activation as the NmSacA enzyme, where UDP-GlcNAc is not necessary for UDP-ManNAc epimerization but enhances the reaction rate, if provided (Hurlburt et al., 2020). Modelling the activity curve using the Hill-equation revealed a Hill-coefficient of 1.31 which is lower as compared to the one of E. coli epimerase (1.81) (Samuel and Tanner, 2004) and therefore supports a lower degree of allosteric control of PaMnaA.

Based on the study of SaMnaA, demonstrating a dose-dependent impact of tunicamycin on the enzyme reaction (Mann et al., 2016), the impact of tunicamycin on the PaMnaA-catalyzed reaction(s) was investigated. Specifically, tunicamycin was tested at concentrations of 0.5 µM, 1 μM, 10µM, 50µM, 100 μM and 200 µM in forward and reverse reaction. However, no inhibition of PaMnaA by tunicamycin was seen in either direction of the reaction, as shown in Figure 6A (forward reaction) and Figure 6B (reverse reaction), respectively, regardless of the presence of tunicamycin at low or high concentration.

Based on the observed lack of inhibition of PaMnaA by tunicamycin (Figure 6A, B) and the absence of a requirement for allosteric activation (Figure 5), we sought to gain insight into the molecular mechanism of tunicamycin resistance of PaMnaA through molecular docking of the antibiotic to the enzyme’s allosteric site. A complete structural homology model of molecule A was created by SWISS-MODEL using the 5ENZ structure (SaMnaA) as a template to enforce a closed conformation of PaMnaA, for comparability with literature data, and additionally accounting for the missing residues within the crystal structure of PaMnaA. Subsequently, tunicamycin docking was performed into the predicted allosteric site of PaMnaA (Figure 7A) and compared to a similar study on SaMnaA (5ENZ) chain A (Figure 7B), which had previously shown to be inhibited by the antibiotic. The docking studies were conducted using AutodockTools. For comparison, the antibiotic was also docked into a model of NmSacA, since it shows a similar behavior towards the allosteric activator (Hurlburt et al., 2020) as PaMnaA, although there is no data in the literature on the impact of tunicamycin on NmSacA (Supplementary Figure S11). The position of tunicamycin in PaMnaA, SaMnaA and NmSacA differs regarding the spatial orientation of the antibiotic’s side chains. Notably, the opening angle defined as the angle between the Cα atoms of the residues 131, 216, 352 in PaMnaA crystal structure was found to be 45°, while that in the PaMnaA_closed model is 31° which agrees with 31° and 32° in the crystal structure of SaMnaA and NmSacA, respectively (de Azevedo and Nascimento, 2019).

Ligand interaction diagrams did not reveal prominent interactions between that part within tunicamycin that is analogous to UDP-GlcNAc and any of the indicated allosteric site residues in PaMnaA (Figure 8A). This contrasts with the results obtained for SaMnaA (Figure 8B), where two of the indicated allosteric site residues, i.e., Q39, corresponding to Q42 in PaMnaA (Q41 in NmSacA) as well as R206, corresponding to R209 in PaMnaA (R213 in NmSacA), showed interactions with the inhibitor. Both of these residues localize within the allosteric site. In comparison, the docking of tunicamycin to PaMnaA (Figure 8A), revealed the residues L129, S130, R209, E210, E303 and D322 as potential interaction partners with the inhibitor. Out of these residues, only R209 is present within the predicted allosteric or substrate binding site of the enzyme. However, no interaction of this amino acid with the UDP-GlcNAc-analogous portion of the inhibitor were revealed in the ligand interaction diagram, explaining that the activity of PaMnaA remained unaffected by both low and high tunicamycin concentrations. Additionally, docking over the whole protein of PaMnaA (Supplementary Figure S12A) and SaMnaA (Supplementary Figure S12B) was performed. The docking of tunicamycin over SaMnaA resulted in similar predicted interaction partners as those revealed by docking to the allosteric site, whereas in PaMnaA, the antibiotic was docked far away from the predicted allosteric or substrate binding site of the enzyme, underlaying the failure of inhibition by the antibiotic. In the case of NmSacA (Supplementary Figure S11B), residues R10 and E295 present in the substrate binding site were revealed to be exposed to tunicamycin, as well as M42 and H245, of which neither is located close to the allosteric or substrate binding site.