Predicting protein function from structure--the roles of short-chain dehydrogenase/reductase enzymes in Bordetella O-antigen biosynthesis

Abstract

The pathogenic bacteria Bordetella parapertussis and Bordetella bronchiseptica express a lipopolysaccharide O antigen containing a polymer of 2,3-diacetamido-2,3-dideoxy-l-galacturonic acid. The O-antigen cluster contains three neighbouring genes that encode proteins belonging to the short-chain dehydrogenase/reductase (SDR) family, wbmF, wbmG and wbmH, and we aimed to elucidate their individual functions. Mutation and complementation implicate each gene in O-antigen expression but, as their putative sugar nucleotide substrates are not currently available, biochemical characterisation of WbmF, WbmG and WbmH is impractical at the present time. SDR family members catalyse a wide range of chemical reactions including oxidation, reduction and epimerisation. Because they typically share low sequence conservation, however, catalytic function cannot be predicted from sequence analysis alone. In this context, structural characterisation of the native proteins, co-crystals and small-molecule soaks enables differentiation of the functions of WbmF, WbmG and WbmH. These proteins exhibit typical SDR architecture and coordinate NAD. In the substrate-binding domain, all three enzymes bind uridyl nucleotides. WbmG contains a typical SDR catalytic TYK triad, which is required for oxidoreductase function, but the active site is devoid of additional acid-base functionality. Similarly, WbmH possesses a TYK triad, but an otherwise feature-poor active site. Consequently, 3,5-epimerase function can probably be ruled out for these enzymes. The WbmF active site contains conserved 3,5-epimerase features, namely, a positionally conserved cysteine (Cys133) and basic side chain (His90 or Asn213), but lacks the serine/threonine component of the SDR triad and therefore may not act as an oxidoreductase. The data suggest a pathway for synthesis of the O-antigen precursor UDP-2,3-diacetamido-2,3-dideoxy-l-galacturonic acid and illustrate the usefulness of structural data in predicting protein function.

Fig. 1

(a) Silver stain analysis and (b) Western immunoblot of duplicate SDS-PAGE gels showing the LPS profiles of wild-type B. bronchiseptica (CN7635E) and CN7635E-derived mutants in wbmF (CNF0a), wbmG (CNG1a) and wbmH (CNH1d) and mutants carrying the complementation vectors for wbmF (pCompF), wbmG (pCompG) and wbmH (pCompH) or the empty vector (pCompEmpty). The positions of the B. bronchiseptica A-band (A) and B-band (B) species are indicated as well as the position of LPS that contains O antigen (O). The primary monoclonal antibody used in (b) that recognises O-band LPS was D13B11.

Fig. 2

Conserved motifs in extended SDR enzymes. Sections from Wbm SDR protein sequences are aligned with the characteristic motifs of the extended SDR subfamily. Except for the alignments with Wbm sequences, the information in this table is from Ref. . Lowercase letters represent conserved biophysical properties of amino acid side chains: x, any residue; h, hydrophobic; c, charged; p, polar; a, aromatic. Conservation of motif features in Wbm proteins is indicated by shading.

Fig. 3

Nucleotide binding in WbmF, WbmG and WbmH. Proteins are oriented with the Rossmann domains in the lower part of each structure. Cartoons are used to represent protein backbone; bound nucleotides are shown as ball-and-stick and key protein side chains as sticks. The boxed panels show details of UMP-binding pockets in (a) His₆-WbmF, UDP co-crystal, (b) His₆-WbmG crystal soaked with UDP-glucose and (c) His₆-WbmH. In (a) and (b), UMP is shown as sticks, and spheres represent atoms within 3.5 Å of the bound nucleotide. Probable hydrogen-bonding interactions are shown as dashed lines. In (c) spheres represent ordered water molecules in this binding pocket. Carbon atoms are coloured rainbow for protein, white for bound nucleotides; oxygen atoms are coloured red; nitrogen is blue and phosphorus orange.

Fig. 4

The UDP-binding pockets are shown for (a) His₆-WbmF, soaked with UDP and (b) the His₆-WbmG, UDP co-crystal. Spheres represent atoms within 3.5 Å of the bound UDP; NAD indicates the nicotinamide ring of the NAD cofactor. (c) dTDP-glucose bound in the active site of a D128N, E129Q mutant of dTDP-glucose 4,6-dehydratase DesIV from S. venezuelae (PDB ID 1R6D). The substrate-binding pockets are all shown from the same angle to enable comparison of the relative positions of the UDP diphosphates in (a) and (b) with the phosphates in dTDP-glucose in (c).

Fig. 5

Pi-stacking interactions in binding of uracil or thymine. Amino acids that directly interact with the substrate nucleobase are shown for B. bronchiseptica WbmF and WbmG, E. coli GalE, S. venezuelae DesIV and P. aeruginosa WbpP with carbon atoms coloured yellow. UMP or dTDP from each structure is shown with carbon atoms in white. Amino acid side chains that have pi-stacking interactions with the base are labelled, and hydrogen-bonding interactions with the peptide backbone are indicated as dashed lines. Images for GalE, DesIV and WbpP were prepared using PDB files 1LRL,1R6D and 1SB8, respectively.

Fig. 6

Comparison of active-site residues in WbmF, WbmG and WbmH compared with WbpP and GMER. (a) The SDR catalytic triad is conserved in WbmG and WbmH but not in WbmF where the residue normally found as serine or threonine superimposes onto the position of Ala131. (b) GMER is shown as an example of an SDR that catalyses 3,5-epimerisation of its substrate. The GMER epimerase catalytic residues Cys109 and His179 superimpose onto hydrophobic residues in WbmG and WbmH. WbmF has side chains in these positions (Cys133 and Asn213) that may be capable of acid–base chemistry. *Cys133 is not resolved in the WbmF crystal; its position here is inferred from the neighbouring residue (Gly132), which is visible in the electron density.

Fig. 7

A model of the proposed substrate in the binding site of WbmF. (a) Overview of modelled interaction of WbmF with the 4-keto derivative of UDP-d-ManNAc3NAcA. Protein is shown as cartoon, with key (labelled) side chains, NAD and sugar nucleotide in sticks. The elements of the UDP-sugar found in the experimental structure have paler colours to highlight the modelled sections. (b) and (c) Close-up of the modelled sugar ring (shown as spheres) demonstrates that there are no clashes between the proposed ligand and the protein surface. The two views show opposite sides of the sugar: (c) shows the surface with the imaged slab cut at the level of the first sugar atom, with NAD surface also removed for clarity. Carbon atoms are yellow for protein, white for the NAD and cyan for 4-keto UDP-d-ManNAc3NAcA; oxygen is red, nitrogen is blue, phosphorus is orange and sulfur is purple.

Fig. 8

Proposed pathway for the synthesis of UDP-l-GalNAc3NAcA catalysed by WbmF, WbmG and WbmH.