Altered architecture of substrate binding region defines the unique specificity of UDP-GalNAc 4-epimerases

Abstract

UDP-hexose 4-epimerases play a pivotal role in lipopolysaccharide (LPS) biosynthesis and Leloir pathway. These epimerases are classified into three groups based on whether they recognize nonacetylated UDP-hexoses (Group 1), both N-acetylated and nonacetylated UDP-hexoses (Group 2) or only N-acetylated UDP-hexoses (Group 3). Although the catalysis has been investigated extensively, yet a definitive model rationalizing the substrate specificity of all the three groups on a common platform is largely lacking. In this work, we present the crystal structure of WbgU, a novel UDP-hexose 4-epimerase that belongs to the Group 3. WbgU is involved in biosynthetic pathway of the unusual glycan 2-deoxy-L-altruronic acid that is found in the LPS of the pathogen Pleisomonas shigelloides. A model that defines its substrate specificity is proposed on the basis of the active site architecture. Representatives from all the three groups are then compared to rationalize their substrate specificity. This investigation reveals that the Group 3 active site architecture is markedly different from the “conserved scaffold” of the Group 1 and the Group 2 epimerases and highlights the interactions potentially responsible for the origin of specificity of the Group 3 epimerases toward N-acetylated hexoses. This study provides a platform for further engineering of the UDP-hexose 4-epimerases, leads to a deeper understanding of the LPS biosynthesis and carbohydrate recognition by proteins. It may also have implications in development of novel antibiotics and more economic synthesis of UDP-GalNAc and downstream products such as carbohydrate based vaccines.

Introduction

UDP-hexose 4-epimerases belong to the short chain dehydrogenase/reductase enzyme superfamily.¹ These enzymes typically have distinct N- and C-terminal domains. The N-terminal domain is a modified Rossmann fold that binds one molecule of NAD(H) with the relatively small C-terminal domain being mainly responsible for substrate binding.² These enzymes are important in myriad biological pathways such as lipopolysaccharide (LPS) biosynthesis^3,4 and galactose metabolism.⁵ LPS is implicated in several facets of host-pathogen interactions such as resistance to serum-mediated killing, phagocytosis and killing by cationic peptides^6–10 whereas a malfunctioning UDP-hexose 4-epimerase can lead to epimerase-deficiency galactosemia in humans.^11,12 General details of the epimerization reaction have been thoroughly investigated by Wilson and Hogness,¹³ Bauer et al.,² Thoden et al.,¹⁴ Liu et al.,¹⁵ Wohlers et al.¹¹ among others. Briefly, the reaction proceeds via abstraction of a proton from the 4'-hydroxyl group by a tyrosine acting as an active site base, its pK_a being lowered by a neighboring lysine. A serine residue is thought to facilitate this abstraction by formation of a low barrier hydrogen bond, hence the term SYK catalytic triad.¹⁴ This proton abstraction and subsequent stereospecific hydride transfer to NAD⁺ leads to formation of a keto intermediate. A net rotation of 180° in this keto intermediate allows a nonstereospecific return of hydride at the 4' position resulting in formation of either of the epimers. These details are illustrated in Scheme 1 of the Supporting Information.

Despite the conservation of catalytic residues and the mechanism of catalysis, UDP-hexose 4-epimerases (henceforth referred to as “the epimerases”) can recognize different UDP-hexose substrates in a specific manner. Ishiyama et al. classified the epimerases into three different groups based on their substrate specificity: Group 1 epimerases recognize nonacetylated UDP-hexoses; Group 2 epimerases are promiscuous toward N-acetylated and nonacetylated UDP-hexoses and Group 3 epimerases strongly prefer N-acetylated UDP-hexoses.⁴ In the context of current work, WbgU belongs to the Group 3 epimerases and is critical for synthesis of 2-deoxy-L-altruronic acid (2Ac-AltUA) in Plesiomonas shigelloides.³ 2Ac-AltUA, in turn, is one of the two glycans that form the O-unit of the LPS of this Gram-negative bacterium. Previous work on the substrate specificity of the epimerases has been mainly focused on the Group 1 and the Group 2 epimerases, WbpP being the only representative of the group 3 epimerases that has been investigated for its substrate specificity.^4,16 In this work, we first propose a structure based model to define substrate specificity of the Group 3 epimerases. We then put this model of substrate recognition in the perspective of substrate recognition by the Group 1 and the Group 2 epimerases and finally, we attempt to consolidate certain previously reported investigations regarding UDP-hexose 4-epimerases.

Results

Overall structure

WbgU crystallizes in the space group P3₂ with generally well defined electron density, exception being a short region between His287 and Ile293. The electron density for the cofactor NAD(H) as well as the substrate UDP-GlcNAc is also fairly well defined (Fig. 1). The data processing and refinement statistics are listed in Table I. The asymmetric unit of the WbgU crystal structure contains a tetramer. Two of the four subunits of this tetramer are related by a 2-fold symmetry axis that associates them via a 4-helix bundle (Fig. 1, Supporting Information). Since there is a unique axis of symmetry in the asymmetric unit and since WbgU is known already known to exist as a dimer in solution, ³ this is the most likely dimeric interface observed in solution state. The 4-helix bundle formed at the dimeric interface is the commonly observed structural motif in several other epimerases,^2,4 thus further corroborating this position as the physiological dimeric interface. The overall tertiary structure of WbgU is comprised of two domains, N-terminal domain and C-terminal domain (Fig. 2). The N-terminal domain is a Rossmann type fold with strand order 3214567 and contains two “topological switch points,” where the helices connecting the consecutive β strands switch planes. The first topological switch point is formed near carboxy edge of Strand 2 and Strand 3 and results in formation of a crevice containing the adenine moiety of NAD(H). The second topological switch point in the β sheet is at the carboxy edge of Strand 6. This switch results formation of another crevice between carboxy edge of Strand 5 and Strand 6. GlcNAc moiety of the substrate UDP-GlcNAc sits in this crevice. The C-terminal domain contains substrate binding region of the active site and is contributed by Phe194-Tyr238, Ala264-Ala308, Ile323-Gly342. NAD(H) sits on and parallel to the carboxy edge of the β sheet with UDP-GlcNAc approximately perpendicular to it.

Figure 1.

Difference electron density map at the active site of WbgU. F_o-F_c electron density map in the region of UDP-GlcNAc and NAD(H) depicted at an absolute electron density of 0.15 e/ų corresponding to 3.0σ cutoff. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Table I.

Data Collection and Refinement Statistics

Data collection statistics
Source	Beamline 9-2 SSRL
Space group	P32
Unit cell parameters
a, b, c (Å)	78.1, 78.1, 231.9
α, β, γ (deg)	90, 90, 120
Temperature (K)	103
Wavelength (Å)	0.97
Resolution (Å)	44.03–2.50 (2.56–2.50)a
Rmergeb	0.13 (0.9)
Completeness (%)	94.7 (88.8)
〈I/σ(I)〉	13.5 (1.4)
Number of unique reflections	51939 (3563)
Average redundancy	4.3 (4.1)
B factor from Wilson plot (Å2)	44.9
PDB entry	3LU1
Refinement statistics
Resolution (Å)	44.03-2.50
Number of reflections	51878
Rfactorc	20.7
Rfreec	25.8
Number of protein atoms	10684
Number of water atoms	182
Number of ligand atoms	332
RMS deviation from ideal values for bond distances (Å)d	0.01
RMS deviation from ideal values for bond angles (deg)d	1.15
Average B-factors
Main chain (Å2)	36.6
Side chain and waters (Å2)	37.0
All atoms (Å2)	36.8
Ramachandran plote
Favored (%)	95.8
Disallowed (%)	0.3

^aThe numbers in parentheses refer to the highest resolution shell.

^b

^c where R_free refers to the R_factor for 5% of the data that were excluded from the refinement.

^dIdeal values used from Engh and Huber.¹⁷

^eValidation by Molprobity.¹⁸

Figure 2.

Overall tertiary structure of WbgU/NAD(H)/UDP-GlcNAc complex. N-terminal domain is a modified Rossmann type fold with the cofactor NAD(H) sitting on and parallel to the carboxy edge of the β sheet. The C-terminal domain binds the substrate UDP-GlcNAc. β strands are numbered from the N- to the C- terminal. Dashed grey lines represent the six-residue loop from His287-Ile293 that was not modeled due to poorly defined electron density. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Overall sequence and structure comparison

As expected, the polypeptide sequence of WbgU bears the closest resemblance to WbpP (67%) (Fig. 3). The sequence identity with CGne and HGal is 23% and with GalE it is 25%. The most prominent differences in WbgU polypeptide sequence, when compared to the epimerases from group 1 and group 2 are: (i) insertion of an N-terminal α-helix in WbgU at Tyr1-Pro16,^† this region being completely absent in GalE, CGne and HGal; (ii) insertion of a six residue loop in GalE, CGne and HGal at Arg200-Gln201 position of WbgU; (iii) insertion of an eight residue loop GalE, CGne and HGal at the Asp229-Gly230 position of WbgU and (iv) deletion of four residues in GalE, CGne, and HGal near His292-Glu297 region of WbgU. Although the sequence identity is relatively low across the three groups, the overall similarity in tertiary structures of GalE, HGal, WbpP and WbgU is very high with an overall r.m.s.d. of the multiple structure alignment being 1.08 Å, as computed by MultiProt.¹⁹ Despite this high degree of homology in tertiary structure, local changes in the polypeptide fold are evident throughout and are most pronounced in the C-terminal domain (Fig. 2, Supporting Information).

Figure 3.

Multiple sequence alignment between 5 different UDP-hexose 4-epimerases. Regions deemed important for determination of substrate specificity are highlighted in pink. Regions highlighted in yellow are distinct structural variations that do not have a direct bearing on substrate binding or catalysis. WbgU is UDP-GalNAc 4-epimerase from P. shigelloides; WbpP is UDP-GlcNAc 4-epimerase from P. aeruginosa; CGne is UDP-Glc/GlcNAc 4-epimerase (Gne) from C. jejuni; HGal is UDP-Glc/GlcNAc 4-epimerase from Homo sapiens; GalE is UDP-Glc 4-epimerase from E.coli. The numbering and the secondary structure assignment corresponds the structure of WbgU. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Active site and its comparison across the three groups

The cofactor NAD(H) is bound at the carboxy edge of the β sheet of the N-terminal domain. This region is fairly conserved across all the three groups. The catalytically important Tyr166 and Ser142 are also highly conserved. The substrate binding region is contributed by the residues Ser142-Ser144, Phe194-Asn195, Ala209-Trp214, Tyr225-Arg234, and Glu297-Ala308 [Figs. 4(a, b)]. This region contains potentially important variations (Fig. 5): Ser143 in WbgU is substituted Ala in CGne, HGal and GalE; Ser233 in WbgU is substituted by Ile in CGne and by Val in GalE and HGal; Arg268 in WbgU is substituted by Gly in CGne, HGal and GalE; Arg304 is substituted by Ser in CGne, Ala in HGal and Pro in GalE; His305 in WbgU is substituted by Val in CGne and by Ala in GalE and HGal and Ser306 of WbgU is substituted by a Cys in the group 2 and by a Tyr in the group 1 epimerases. The values estimated for active site volume and surface area of all the 4 experimentally obtained structures, that is, GalE, HGal, WbgU, and WbpP reveal that the Group 2 epimerases have the largest active site volumes and surface areas with the Group 3 having the smallest (Table II).

Figure 4.

Architecture of the substrate binding region of WbgU. A salt bridge between Arg304 and Asp229 on one side and a hydrogen bonding network formed between His305, Ser233, and Arg268 launch the substrate binding region in a conformation that results in the formation of a hydrophobic cluster contributed by Val303, Arg234, and Ser303. This hydrophobic cluster directly interact with the CH₃- group of the GlcNAc moiety. In addition, this conformation results in an interaction between Ala308 and Phe194. This interaction places Asn195 in a hydrogen bond with NH- group of the GlcNAc moiety and the oxy bridge of the diphospho moiety. In addition, a hydrogen bond is formed between Ser143 and NH- group of the GlcNAc moiety and between main chain NH- of Ser144 and CO- group of the GlcNAc moiety. (a) Important interactions with the GlcNAc moiety of the substrate UDP-GlcNAc are highlighted as green (polar interactions) or purple (nonpolar interactions). Tyr166 and Ser142 are hydrogen bonded to 4' hydroxyl group. This region defines the catalytic motif and is highly conserved as is Asn195. The regions defining the substrate binding site are mainly contributed by residues 209-214, 225-234 and 297-308. (b) A schematic representation of the interactions deemed most important in the substrate recognition and catalysis. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 5.

Substrate recognition by the group 1 and the group 2 UDP-hexose 4-epimerases from the perspective of the group 3. His305 in WbgU is substituted by Val in CGne and by Ala in GalE and HGal; Ser233 in WbgU is substituted by Ile in CGne and by Val in GalE and HGal; Arg268 in WbgU is substituted by Gly in CGne, HGal, and GalE (side chains are not shown for clarity). Similarly Arg304 is conserved in WbgU and WbpP whereas being substituted by Ser in CGne, Ala in HGal and Pro in GalE. The loss of the hydrogen bonding network at His305 position and the salt bridge at Arg304 position result in an altered conformation of substrate binding loop in the Group 1 and the Group 2 epimerases. In addition, the insertion of an 8 residue loop at the Asp229 causes formation of a salt bridge with Arg299. The absence of the polar interactions in the Group 1 and the Group 2 epimerases at the 304 and 305 positions in combination with the presence of a salt bridge at the 8 residue insertion between Asp229-Arg299, thus dictates the conformation of the substrate binding region in the group 1 and the group 2 epimerases. The common architecture of the substrate binding region in the group 1 and the group 2 epimerases has one important variation: Ser306 of WbgU (and WbpP) is substituted by a Cys in the Group 2 and by a Tyr in the group 1 hence restricting the access of bulky N-acetyl group to the active site of the group 1 epimerases, which in turn makes them specific towards the nonacetylated substrates. The labels and numbering in black color correspond to WbgU; the labels and numbering in orange color correspond to HGal. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Table II.

Active Site Dimensions of UDP-hexose 4-epimerases

Protein	Classification	PDB id	Vol (Å3)	SA (Å2)
GalE15	Group 1	1NAH	3700	2600
HGal14	Group 2	1HZJ	4000	2800
WbpP4	Group 3	1SB8	2900	2100
WbgU	Group 3	3LU1	2600	1900

Discussion

The catalysis carried out by UDP-hexose 4-epimerases has been extensively investigated using biochemical and structural studies. However, a well-defined rational model ceases to exist for the substrate specificity. This is especially true for the Group 3 epimerases that are highly specific toward N-acetylated hexoses. In this work, we investigated the structure of WbgU, a novel UDP-GalNAc 4-epimerase that belongs to the Group 3. This investigation led us to propose a structure based model that rationalizes the substrate specificity of the Group 3 epimerases. According to this model, the region between Glu297 and Ala308 provides the scaffold that defines the substrate specificity of WbgU. We call this region “297-308 belt” (Fig. 4). Two sets of interactions stabilize the conformation of the 297-308 belt: first, a salt bridge formed between Arg304 and Asp229 and second, a hydrogen bond network between His305, Ser233, and Arg268.

The 297-308 belt, in turn, resorts to a conformation that leads to several interactions potentially critical in determination of the substrate specificity (Fig. 4). First, it leads to formation of a hydrophobic cluster contributed by Arg234, Val303, and Ser306. This hydrophobic cluster interacts with the CH₃- group of the N-acetyl moiety of the UDP-GlcNAc. Second, it causes the relatively conserved Ala308 to interact with Phe194. This interaction places the loop containing Phe194 and Asn195 in a position that causes the Asn195 to hydrogen bond with NH- group of the GlcNAc moiety on one side and the oxy-bridge of the diphospho moiety of UDP-GlcNAc on the other side. Thirdly, the conformation of the 294-308 belt orients the substrate in such a manner that the NH- group of the GlcNAc moiety hydrogen bonds with Ser143 and the CO- group of the GlcNAc moiety also forms a weak hydrogen bond (3.5 Å) with the main chain NH- of Ser144. Thus, the change in conformation of the 297-308 belt in the Group 3 epimerases (relative to the Group 1 and the Group 2 epimerases) is transmitted to other regions of the substrate binding pocket, directly by the interaction between Ala308 and Phe194 and indirectly via the substrate to render the observed specificity. It is also noted that Ser143 in the Group 3 epimerases is substituted by an Ala in the Group 1 and the Group 2 epimerases and could be a critical variation.

A comparison of WbgU with the Group 1 (GalE), the Group 2 (CGne and HGal) and the Group 3 (WbpP) epimerases further reinforced our model. It was found that all the aforesaid residues proposed to be important in determination of substrate specificity of the Group 3 epimerases are conserved in the Group 3 (WbgU and WbpP) epimerases whereas being variant among the Group 1 (GalE) and the Group 2 (CGne, HGal) epimerases (Figs. 3 and 5). The hydrogen bond network corresponding to the His305 position of WbgU is absent in the group 1 and the group 2 epimerases and so is the salt bridge between Arg304 and Asp229. These variations combined with the insertion of an 8 residue loop in the group 1 and the group 2 epimerases at the position corresponding to Asp229-Val230 of WbgU, result in formation of a salt bridge between the Asp229 and Arg299 (numbering follows HGal) in the Group 1 and the Group 2 epimerases. This salt bridge and several van der Waals interactions render a conformation to this region that is very similar in the Group 1 and the Group 2 epimerases while being markedly different from the Group 3. This is also indicated by the fact that the 297-308 belt in the Group 3 is rotated by ∼10° from the corresponding region of the Group 1 and the Group 2 epimerases, the measurement being done from the C-terminal α-helix (Fig. 2, Supporting Information). The surface topology of the active site is very similar in both the Group 1 and the Group 2 epimerases while being significantly different from the Group 3 epimerases (Fig. 3, Supporting Information). The volume of active site is also much lesser in the Group 3 epimerases when compared with the Group 1 and the Group 2 epimerases (Table II). Thus, it follows that the substrate binding region of the Group 1 and the Group 2 epimerases are built on a very similar main chain scaffold, which is markedly different from the corresponding region in the Group 3 epimerases.

Finally, we attempted to consolidate some results previously reported in the literature, using our model of substrate recognition. First, if the substrate binding scaffold of the Group 1 epimerases and the Group 2 epimerases is so similar then what causes their substrate specificities to be different, that is, the Group 1 epimerases being specific toward nonacetylated UDP-hexoses, whereas the Group 2 epimerases having similar specificity toward nonacetylated and N-acetylated UDP-hexoses? The key seems to be the variation at Ser306 position. Ser306 in WbgU is substituted by a Cys in HGal and by a Tyr in GalE. Since the main chain scaffold of the Group 1 and the Group 2 epimerases is very similar, the bulkier side chain of this Tyr in the substrate binding region of the Group 1 epimerases can easily result in a greater specificity toward less bulkier substrates, that is, nonacetylated hexoses. The importance of this variation has been reported previously by a Tyr299Cys mutation engineered in the Group 1 epimerase GalE that switched its specificity from the Group 1 to Group 2,²⁰ by a Ser306Tyr mutation engineered into the Group 2 epimerase from E. coli O86:B7 that switched its specificity from the Group 2 to the Group 1²¹ and the corresponding Cys307Tyr mutation in HGal that switched its activity from the Group 2 to the Group 1.^20,22 Second, can a single point mutation of Ser306Tyr also switch the substrate specificity from the Group 3 to the Group 1? Based on our model of substrate recognition, a single point mutation of Ser306Tyr can not be engineered in the Group 3 epimerases since it would place the Tyr residue in steric clashes in all of the allowed rotamer conformations. This immediately explains the loss in activity of this mutant reported for the Group 3 epimerase WbpP.⁴

To conclude, a general basis of the substrate specificity is apparent among the Group 1 and the Group 2 epimerases while the altered architecture of active site of the Group 3 epimerases causes the corresponding substrate specificity to be defined in a significantly different manner. This investigation provides a platform for structure-guided engineering of UDP-hexose 4-epimerases, which, in turn can lead to more economic production of UDP-GalNAc and downstream products such as carbohydrate based vaccines. It also enriches our comprehension of the LPS biosynthesis. A UDP-hexose 4-epimerase from Trypanosoma brucei, a protozoan causing a variety of tropical diseases, has been reported to be essential for survival of its host.^23,24 Thus, a deeper understanding of substrate recognition by UDP-hexose 4-epimerases has potential implications in structure-based development of novel antibiotics. In addition, these results may also provide useful insights regarding other 4-epimerases. For example, a GlcNAc-P-P-Und/GalNAc-P-P-Und epimerase was recently reported to be critical for LPS biosynthesis in E.coli O157.²⁵ Based on our studies and previous literature available regarding UDP-hexose 4-epimerases, it is very likely that the C-terminal domain of the GlcNAc-P-P-Und/GalNAc-P-P-Und epimerase sits close to the inner side of the inner membrane allowing GlcNAc-P-P-/GalNAc-P-P-moiety of the GlcNAc-P-P-Und/GalNAc-P-P-Und to wedge itself in the substrate binding region, whereas the N-terminal domain containing the cofactor sits further away from inner membrane.

Materials and Methods

Cloning, expression, and purification

WbgU was cloned, expressed, and purified with some modifications to the protocol described previously.³ Briefly, wbgU was inserted between XhoI and BamHI restriction sites of pET-15b plasmid (Novagen) with an N-terminal (His)₆ fusion tag. The construct pET/wbgU was transformed into BL21(DE3) (Novagen) for protein expression. A small scale culture of these cells grown overnight in LB medium was transferred to large-scale LB medium in a ratio of 1:100. The large scale culture was allowed to grow to an OD₅₉₅ of 0.8 at 30°C with constant shaking. Following this, the culture was induced with 0.15 mM IPTG and allowed to grow for 5 more hours. The cells were then harvested and stored at −0°C till further use. The pellet thus obtained, was washed and dissolved in buffer A (50 mM Tris-HCl pH 8.9, 500 mM NaCl and 5 mM imidazole), and sonicated over ice after addition of PMSF to a final concentration of 2 mM. The lysate was centrifuged for 40 min at 31,000g and the supernatant was loaded onto a 5 mL nickel-sepharose high performance affinity column (GE healthcare) previously equilibrated with buffer A. The column was washed with 20 column volumes of buffer B (50 mM Tris-HCl pH 8.9, 500 mM NaCl, and 50 mM imidazole) and eluted in buffer C (50 mM Tris-HCl pH 8.9, 500 mM NaCl and 500 mM imidazole). The buffer was exchanged to buffer D (200 mM glycine pH 10.0, 40 mM NaCl) and the protein concentrated to 25 mg/mL using Amicon Ultra centrifugal filter unit (Millipore). UDP-GlcNAc was added to a final concentration of 15 mM and the protein sample was allowed to equilibrate overnight at 4°C.

Crystallization, data collection and data processing

The WbgU/UDP-GlcNAc sample prepared above was used to search for condition(s) that could lead to crystallization using home made screens. The first crystals were obtained in 35% PEG 400, 200 mM ammonium sulfate using a hanging drop vapor diffusion set up. The drop was prepared by mixing 2 μL protein sample with 2 μL crystallization reagent and was allowed to equilibrate with the well solution containing 1000 μL crystallization reagent. The protein failed to crystallize in absence of UDP-GlcNAc. The best diffracting crystals could be obtained after optimization using crystallization screens (Hampton Research, CA) in a microbatch under oil set up,^26,27 where 1 μL of WbgU/UDP-GlcNAc sample was set up with 1 μL of crystallization reagent (250 mM ammonium sulfate, 100 mM Bis-Tris pH 6.5, 25% w/v PEG 3,350) overlaid with 200 μL of mineral oil. These crystals grew to maximum dimensions of 300 μm × 300 μm × 100 μm within 3 weeks. After equilibration in a harvest buffer (250 mM ammonium sulfate, 0.1M BIS-TRIS pH 6.5, 30% w/v PEG 3,350) for 40 min, the crystals were transferred to cryoprotection buffer (250 mM ammonium sulfate, 0.1M BIS-TRIS pH 6.5, 30% w/v PEG 3,350, 5% glycerol), flash frozen in nylon cryoloops (Hampton Research) and stored in liquid nitrogen. A 2.5 Å resolution data set was collected at the Stanford Synchrotron Radiation Lightsource (SSRL). The data were processed with MOSFLM²⁸ in the space group P3 (a = 78.1 Å, b = 78.1 Å, c = 231.9 Å). The data processing statistics are listed in Table I.

Structure determination

A sequence alignment was generated between WbgU and WbpP using CLUSTALW.²⁹ A search model was prepared from crystal structure of WbpP⁴ by mutating the mismatched residues to alanines. Molecular replacement using Phaser³⁰ was used to find a solution to the phase problem. A unique solution could be obtained in the enatiomorphic space group P3₂ with four monomers in the asymmetric unit that corresponds to a Matthews coefficient (V_M) of 2.62³¹ and solvent content of 53%. This solution had a final log-likelihood gain of 6014.3 and an R_factor of 41.2%. 7 cycles of maximum-likelihood based refinement reduced R_factor/R_free to 29.3%/33.5%, where R_free was calculated from 5% of the reflections. All the refinement was conducted using REFMAC.³² In the first round of manual model building using COOT,³³ variations were made in this refined model according to the amino acid sequence of WbgU. At the same time, modifications in main chain and side chains were made according to the 2mF_o-DF_c electron density map. Iterative cycles of model building followed by real and reciprocal space refinement reduced R_factor/R_free to 24.9%/29.8%. At this point well-defined electron density was observed in the 2mF_o-DF_c map for NAD(H) and UDP-GlcNAc in all the subunits, which were then modeled. Subsequent alternating cycles of model building along with real and reciprocal space refinement eventually improved electron density for some regions of the polypeptide and revealed the electron density for some glycine molecules, Na⁺ and SO₄²⁻ ions. Lastly, solvent molecules were added and a final round of refinement was done in a similar fashion as above. The final model consists of four molecules each of WbgU, UDP-GlcNAc, NAD(H) and glycine, three SO₄²⁻ ions, two Na⁺ ions and a total of 182 solvent molecules. The electron density was found to be well defined for most part, notable exception being a loop region extending from His287-Ile293. The average B-factor of the C-terminal domain is higher (47.5 Ų) than the average B-factor N-terminal domain (32.5 Ų). MolProbity¹⁸ was used to validate the quality of the final model. Details of the data collection and final refinement statistics are listed in Table I.

Sequence and structure comparisons

Five epimerases were used to draw out the sequence comparison with at least one representative from each of the groups: (i) UDP-Gal 4-epimerase from Escherichia coli (GalE^‡) (Group 1)²; (ii) UDP-Glc/GlcNAc 4-epimerase from Campylobacter jejuni (CGne) (Group 2)³⁴; (iii) UDP-Glc/GlcNAc 4-epimerase from humans (HGal) (Group 2)¹¹; (iv) UDP-GlcNAc 4-epimerase from Pseudomonas aeruginosa (WbpP) (Group 3)³⁵ and (v) UDP-GlcNAc 4-epimerase from Plesiomonas shigelloides O17 (WbgU) (Group 3)³. The multiple sequence alignment was performed using T-Coffee.³⁶ Among these five protein sequences, crystal structures have been reported for GalE,³⁷ HGal,³⁸ WbpP,⁴ and WbgU (this work). The crystal structure of CGne is not available yet. Its polypeptide sequence was submitted to I-TASSER server³⁹ for structure prediction without providing any reference structural template. I-TASSER server could predict a model with a high C-score of 1.62. Together these five structures were then used to prepare a structure alignment using UCSF Chimera.⁴⁰ The axial rotation of C-termial domains of the Group 3 epimerases relative to the Group 1 and the Group 2 epimerases was determined by computing axes passing through the geometric centroid of atoms corresponding to the main chain N atoms of Arg338, Val210, and Val303 of WbgU using UCSF Chimera.⁴⁰ Finally, active site volumes were calculated for the 4 experimentally determined structures (GalE, HGal, WbpP, and WbpP) using CASTp⁴¹ with precision approximated as per the discussion by Novotny et al.⁴² All the molecular graphics were prepared using UCSF Chimera^40,43 and POV-Ray.⁴⁴

Glossary

Abbreviations:

GalE

UDP-Galactose 4-epimerase from Escherichia coli

HGal

UDP-Galactose 4-epimerase from Homo sapiens

LPS

lipopolysaccharide

Und

undecaprenol.