Crystal structure of the Propionibacterium acnes surface sialidase, a drug target for P. acnes-associated diseases

Abstract

Propionibacterium acnes, though generally considered part of the normal flora of human skin, is an opportunistic pathogen associated with acne vulgaris as well as other diseases, including endocarditis, endophthalmitis and prosthetic joint infections. Its virulence potential is also supported by knowledge gained from its sequenced genome. Indeed, a vaccine targeting a putative cell wall-anchored P. acnes sialidase has been shown to suppress cytotoxicity and pro-inflammatory cytokine release induced by the organism, and is proposed as an alternative treatment for P. acnes-associated diseases. Here, we report the crystal structures of the surface sialidase and its complex with the transition-state mimic Neu5Ac2en. Our structural and kinetic analyses, together with insight from a glycan array screen, which probes subtle specificities of the sialidase for α-2,3-sialosides, provide a basis for the structure-based design of novel small-molecule therapeutics against P. acnes infections.

Introduction

Sialic acids are a family of structurally diverse monosaccharides found at the nonreducing termini of glycoproteins, glycolipids and polysaccharides (Angata and Varki, 2002). They are nine-carbon α-keto sugars typically exhibiting a negative charge (Varki and Schauer, 2009). Due to the cell surface location of sialoglycoconjugates, sialic acids represent an important interface in various biological processes, including host–pathogen interactions. Indeed, many human pathogens, for example, influenza virus, Vibrio cholerae, Clostridium perfringens and Streptococcus pneumoniae, have evolved mechanisms to exploit these features and express sialidase enzymes which degrade sialyl residues from host cell surface to (i) facilitate viral spread, (ii) expose recognition sites for microbial adhesion, (iii) trans-sialylate their own surface with these host molecules for immune evasion or (iv) acquire the released sugars as nutrient (Russell et al. 2006, Moustafa et al. 2004, Rood, 1998, Manco et al. 2006, Vimr et al. 2004, Corfield 1992).

Sialidases classified as EC 3.2.1.18 are exo-α-glycosidases catalyzing the hydrolysis of distal, α-ketosidically linked sialic acids. With the exception of the recently discovered inverting GH156 exo-sialidases (Bule and Chuzel et al. 2019) the vast majority of these are members of CAZy families GH33 and GH34, which effect hydrolysis with retention of the α-anomeric configuration in the product N-acetyl-D-neuraminic acid (Chong, et al. 1992). Depending on their specificity they can hydrolyze α(2,3)-, α(2,6)- or α(2,8)-linked terminal sialic acids and in the case of trans-sialidases they can also transfer these to asialoglycoconjugates (Kim, et al. 2011). As observed in the crystal structures of viral (Varghese, et al. 1983), bacterial (Crennell, et al. 1994, Crennell, et al. 1996, Gaskell, et al. 1995, Newstead, et al. 2008, Park, et al. 2013, Tailford, et al. 2015, Xu, et al. 2008, Xu, et al. 2009, Zaramela, et al. 2019), trypanosomal (Buschiazzo, et al. 2002, Buschiazzo, et al. 2000), fungal (Telford, et al. 2011), leech (Luo, et al. 1998) and human (Chavas, et al. 2005) sialidases available to date, their catalytic domains share a β-propeller fold domain, with six blades of antiparallel β-sheets centered around a pseudo 6-fold axis. In addition, the nonviral sialidases contain two conserved sequence motifs, namely, the RIP/RLP motif (Arg-Ile/Leu-Pro) and the bacterial neuraminidase repeat (BNR repeat) or Asp-box motif (Ser/Thr-X-Asp-[X]-Gly-X-Thr-Trp/Phe) (Quistgaard and Thirup, 2009, Taylor 1996) which adopts a structural role in forming the β-hairpins. Besides the similar fold, the common evolutionary origin and mechanism of action of exo-sialidases are also reflected in several invariant features of its active site. Out of seven strictly conserved residues, an Asp and a Tyr/Glu pair participate directly in substrate catalysis via a transiently linked β-sialosyl tyrosine intermediate (Damager, et al. 2008, Newstead, et al. 2008, Watson, et al. 2003). This conserved tyrosine acts as a nucleophile in exo- and trans-sialidases in contrast to other retaining glycosidases where the nucleophile is the carboxyl group of usually a glutamate or an aspartate (Pierdominici-Sottile, et al. 2011, Watts, et al. 2003, Watts, et al. 2006). A reason for that might be the repelling negative charge common to all sialic acids. Three Arg residues, another Glu residue associated with one of the conserved Arg and a hydrophobic pocket composed of less conserved residues further define the catalytic center. Despite these common features, viral sialidases form tetramers and bind Ca²⁺, whereas bacterial sialidases are monomers without divalent cation binding properties (Crennell, et al. 1993). One other class of sialidases is that of the phage endo-sialidases involved in polysialic acid cleavage. These have quite distinct structures and are in fact inverting enzymes (Morley, et al. 2009).

Some bacterial sialidases also carry up to two lectin domains which can be located upstream or downstream or in between loops of the β-propeller fold (Buschiazzo, et al. 2002). The trypanosomal trans-sialidase lectin domain and the Micromonospora viridifaciens galactose-binding domain are located at the C-terminal side of the active site (Buschiazzo, et al. 2002, Newstead and Watson, et al. 2005), whereas the leech sialidase lectin domain is located N-terminal to its active site (Luo, et al. 1998). The V. cholerae sialidase possesses two lectin domains flanking the active site, one of which has an additional binding site for sialic acid (Moustafa, et al. 2004).

Present as part of the human skin microflora, the Gram-positive Propionibacterium acnes (recently renamed Cutibacterium acnes) is implicated in the development of inflammation in acne vulgaris and folliculitis. Acne vulgaris is the most common skin disorder worldwide yet current treatment of this chronic disease is inadequate due to increasing resistance to standard antibiotics and the known teratogenicity of isotretinoin (Bhambri, et al. 2009, Eady, et al. 2003). P. acnes also causes postsurgical complications such as endocarditis, endophthalmitis and prosthetic joint infections (Jakab, et al. 1996, Levy, et al. 2008). The P. acnes genome sequence offers insights into the potential virulence strategies of the organism in acne pathogenesis, and provides alternative targets for improved treatments (Bruggemann, et al. 2004). In addition to hemolysins, adhesins and surface antigens involved in phase variation to escape host immune surveillance, the genome reveals a large set of enzymes that can degrade host tissue components, of which three bear sequence similarities to sialidases (Rosen, 2007). In particular, a surface sialidase (accession number: gi|50840637) containing a C-terminal Leu-Pro-X-Thr-Gly cell-wall anchoring motif has been demonstrated to have virulence potential besides serving a pure nutritional role (Nakatsuji, et al. 2008a). The sialidase is a member of glycoside hydrolase family 33 (CAZy family GH-33). There are no lectin or galactose-binding domains in the P. acnes sialidase (PaNA) based on its primary structure.

The recombinant sialidase is immunogenic in mice, and the generation of anti-sialidase antibody results in decreased cytotoxic effects of P. acnes and reduced pro-inflammatory cytokine production in both cell culture, mouse and tissue-chamber models, with the latter serving to mimic P. acnes infections in vivo. Based on its immunomodulatory properties, a vaccine specifically targeting this P. acnes sialidase (PaNA), instead of a previously tested formulation based on the inactivated microorganism, was proposed as an alternative strategy for future acne therapy (Nakatsuji, et al. 2008a, Nakatsuji, et al. 2008b) but it has to be administered in early childhood. A passive immunization based on antibodies against the Christie-Atkins-Munch-Peterson factor of P. acnes has been tested in vitro (Liu, et al. 2011).

As a major virulence factor of P. acnes its sialidase also serves as a potential drug target. Inhibitors for sialidases of other pathogens have been identified, e.g. a mechanism-based inhibitor for the trans-sialidase of Trypanosoma cruzi, the etiologic agent of Chagas disease (Carvalho, et al. 2010). In this report, we describe the crystal structure of PaNA, which is implicated in acne pathogenesis, in its ligand-free form and in complex with the transition-state mimic Neu5Ac2en. The structural information, combined with the kinetic and glycan array data, provides a basis for the design of novel therapeutics for this severe infection of the skin.

Results

Overall architecture of P. acnes sialidase

The crystal structure of the P. acnes sialidase/neuraminidase (PaNA) devoid of the LPXTG cell wall-anchoring motif in its ligand-free form was determined to 2.1 Å resolution. Though no electron density was observed for residues 31–80, the structure clearly adopts the canonical β-propeller fold of a GH33 sialidase, with an N-terminal extension covering part of the solvent-exposed region away from the active site (Figure 1). Superposition with the M. viridifaciens sialidase (MvNA) (PDB code: 1EUR) (Gaskell, et al. 1995), with which it shares 58% overall sequence identity over the structurally resolved region here (Supplementary Figure 1), reveals a root-mean-square deviation (r.m.s.d) of 0.95 Å over 349 Cα atoms. Occurring in topologically identical positions on the face opposite the active site are five Asp-box motifs (Ser/Thr-X-Asp-[X]-Gly-X-Thr-Trp/Phe), a signature of microbial sialidases (Figure 1A, B). Each repeat adopts a β-turn between the third and fourth strands of the four-stranded antiparallel β-sheets comprising the six-bladed β-propeller fold. H-bonding interactions between conserved hydrophilic residues of the Asp-boxes stabilize the turns in solvent-exposed regions, while conserved Trp residues at positions 165, 238, 301, 349 and 410 are sequestered in hydrophobic pockets formed between neighboring β-sheets. It is believed that the Asp-box motif simply serves to maintain the structural fold. Of note is the disulfide bond between Cys404 (located within one of the Asp-boxes) and Cys457, which bridges a surface exposed loop in the penultimate β-sheet and adjacent C-terminal helix. The presence of the surface-exposed disulfide linkage further confers stability to the protein in the external environment (Baldwin, 2007). Analysis of PaNA crystal packing by PISA (www.pdbe.org/pisa) identifies the largest interface to be 466.2 Å², suggesting that PaNA is a monomer. Upon examining the electrostatic surface of the sialidase, a large electronegative patch is apparent on the surface away from the active site (Figure 1C and D), helping to direct the substrate-binding site toward negatively charged glycoconjugates.

Fig. 1.

Overall structure and electrostatic surface map of PaNA. (A) Ribbon representation of the six-bladed β-propeller fold with their respective Asp-box motifs drawn as stick models. Neu5Ac2en, in gray, is bound in the active site away from the Asp boxes. (B) The view in (A) rotated 90°. (C) Surface representations of PaNA colored according to electrostatic potential from −4 kT/e (red) to +4 kT/e (blue), calculated using APBS (Baker, et al. 2001). A prominent electronegative patch is present while the surface surrounding the ligand-binding site is partly electropositive. (D) The view in (C) rotated 180°.

P. acnes sialidase in complex with Neu5Ac2en

Binding of the inhibitor Neu5Ac2en, which mimics the more planar ring of the oxocarbenium ion-like transition state, induces minimal conformational changes in the active site. The apo and inhibitor-bound complex at 1.7 Å superimpose with a r.m.s.d. of 0.15 Å over 379 Cα atoms, suggesting the rigid nature of the active site. Neu5Ac2en is bound in a strained half-chair conformation, stabilized by key interactions as observed in other microbial sialidases (Figure 2) (Gaskell, et al. 1995). The carboxylate group of the inhibitor is anchored by the Arg triad (Arg121, Arg329 and Arg395), where Arg121, part of the RIP motif of bacterial sialidases, also interacts with a conserved Glu residue at position 439. Pointing toward C2 of Neu5Ac2en at 2.7 Å below the ring is a conserved Tyr (Tyr423). The role of this residue as the catalytic nucleophile has been confirmed by trapping of its covalent sialyl-enzyme intermediate (Amaya, et al. 2004, Watts, et al. 2003). Further, Tyr423 is positioned within hydrogen-bonding distance of the general base catalyst Glu313, facilitating the nucleophilic attack. The O4 hydroxyl of Neu5Ac2en makes multiple hydrogen bonding interactions with residues common to nonviral sialidases, namely Asp146, the acid/base catalyst located on a loop above the sugar ring, Asp185 and Arg140. Asp185 also coordinates the amide of the N-acetyl group, with further stabilization of the substituent coming from a hydrophobic pocket formed by Ala147, Val202, Phe209, Leu224 and Phe257.

Fig. 2.

Architecture of the PaNA active site. (A) Key interactions established by Neu5Ac2en (green stick model) with residues in the PaNA active site and the hydrophobic pocket (in salmon). H-bonding interactions are shown as blue lines. (B) 2mFo-DFc map calculated in the absence of Neu5Ac2en.

Structural comparison with other bacterial sialidases

Crystal structures of bacterial sialidases are available from the organisms V. cholerae, M. viridifaciens, Salmonella enterica serovar Typhimurium LT2, C. perfringens, S. pneumoniae, Pseudomonas aeruginosa, Parabacteroides distasonis, Bacteroides thetaiotaomicron, B. caccae and Ruminococcus gnavus. Although their active site residues are mostly well conserved some structural differences surrounding the glycerol moiety of Neu5Ac2en exist, and these variable interactions with the glycerol group may contribute to different substrate specificity and ligand binding affinities (Figure 3A). For instance, the extensive hydrogen-bonding network found in the MvNA-Neu5Ac2en complex at 2 Å (PDB code: 1EUS) which is involved in stabilizing the O7 hydroxyl in the MvNA-Neu5Ac2en complex is absent in PaNA (Figure 3B). Another noted difference between the two structures is the substitution of hydrophobic (Ile228 and Phe234 in M. viridifaciens) for hydrophilic residues (Arg282 and Gln287) at equivalent positions and the appearance of an associated water molecule (W1) involved in binding the glycerol group in the P. acnes complex (Figure 3B). The extended side-chain of Arg282 is held in position by Gln287, forming an additional hydrogen bond to the O9 hydroxyl via a bridging water, W1, along with direct interactions with the O8 and O9 hydroxyls made by Asp312. This leads to a tilted orientation of the C9-O9 bond relative to C8 of Neu5Ac2en. A second water molecule, W2, also weakly coordinates O9. The flexibility of the glycerol group is also seen in an alternative conformation where the O8 hydroxyl forms hydrogen-bonding interactions with W2, as well as Arg329.

Fig. 3.

Comparison of the overall and active site structure. (A) Superposition of ligand bound PaNA active site (yellow) with sialidase NanI from C. perfringens (blue), sialidase NanB from S. pneumoniae (green) and StNA from S. typhimurium LT2 (gray). (B) Structural comparison of the PaNA (yellow) and MvNA (blue) active sites. Superposition of the two sialidases highlights differences in residues surrounding the glycerol group of Neu5Ac2en and the water networks involved in stabilizing protein–ligand interactions.

Kinetic analysis of P. acnes sialidase

To characterize the enzymatic profile of PaNA relative to close homologs of the same family, its kinetic parameters were determined using 4-trifluoromethylumbelliferyl-α-D-N-acetylneuraminic acid (CF₃MU-Neu5Ac) (Supplementary Figure 2A). Its affinity for the substrate (K_m = (44 ± 6 μM)) is weaker compared to M. viridifaciens, with a K_m of 7 μM for MU-Neu5Ac, which is similar to that of its fluorinated equivalent (Engstler, et al. 1997, Watson, et al. 2003). The inhibitor Neu5Ac2en inhibits most bacterial sialidases with a K_i value in the 10⁻⁶ M range (Taylor, 1996) though higher values in the low mM range have been observed (Hoyer, et al. 1991). The reported K_i for MvNA is (0.14 ± 0.07) μM (Watson and Newstead, et al. 2005). In comparison, Neu5Ac2en is a much weaker inhibitor of PaNA with an estimated K_i of (120 ± 26) μM (Supplementary Figure 2B).

Mammalian glycan array screening of P. acnes sialidase D146A

Using the inactive acid/base mutant of PaNA (D146A), extensive screening with potential mammalian glycan targets was performed to probe the substrate specificity of the enzyme (see http://www.functionalglycomics.org/ primscreen_2531). PaNA binds to α-2,3-, α-2,6- and α-2,8-linked terminal Neu5Ac, although with a slight, 4-fold preference for α-2,3-sialosides, of which many are linked to a galactose residue (Supplementary Figure 3). Its lax preference for the sialyl linkage is consistent with the absence of extended loop insertions above the Arg triad, where steric hindrance and stacking interactions from aromatic residues in this region provide the strict specificity of trans- and intramolecular sialidases toward α-2,3-linked sialic acids (Amaya, et al. 2004, Gut, et al. 2008, Luo, et al. 1998). Among the available sialic acid derivatives in the array, modifications at the C5 position seem to play a critical role in substrate recognition. For instance, replacing the N-acetyl group at C5 of Neu5Ac with an N-glycolyl group (Gc) as in Neu5Gca2-6Galb1-4GlcNAcb-Sp0 results in a 58-fold reduction in binding. Moreover, deaminoneuraminic acid (KDN), with a C5 hydroxyl instead of an N-acetyl group, significantly impairs binding toward PaNA, by up to 203-fold as measured for [KDN-(α2 → 3)-Gal-(β1 → 3)-GalNAc-Sp14]. Acetylation at O9 of Neu5Ac also decreases the affinity of the ligand for PaNA, although to a smaller extent compared to KDN (19-fold reduction for [Neu5Ac(9Ac)-(a2 → 6)-Gal-(b1 → 4)-GlcNAc-Sp8]). Of 406 targets, the top hit from the glycan array screen is Neu5Ac α-ketosidically linked directly to a spacer composed of -OCH₂C₆H₄-p-NHCOCH₂NH, with mean fluorescence slightly higher than for those containing carbohydrate moieties in the penultimate position. Exchanging the spacer to (-CH₂CH₂CH₂NH₂), without an O-glycosidic bridge, reduces the binding by 5-fold.

Discussion

Like many bacterial pathogens evolved to use host sialic acids to their advantage, P. acnes sialidases can catabolize host sialoglycoconjugates for sources of carbon and nitrogen since a sialic acid transporter is encoded in the microbe’s genome (Bruggemann, 2005). In particular, a sialidase containing a cell-wall sorting LPXTG motif has been shown to enhance the adherence of P. acnes to human sebocytes, leading to greater cell death. The fact that a vaccine created using this antigen also attenuates the production of interleukin-8 (IL-8) and macrophage inflammatory protein-2 (MIP-2) further underscores the role of the sialidase in mediating acne pathogenesis and other associated diseases (Nakatsuji, et al. 2008a). In the current study, we present a structural characterization of the surface sialidase from P. acnes, together with the kinetic and glycan array data, providing a framework for the design of novel inhibitors against its enzymatic activity.

The crystal structure of the P. acnes sialidase was solved using the M. viridifaciens homolog as a molecular replacement model. Although the two structures overlap closely in the active site region, they exhibit different kinetic properties, with PaNA having a K_m for CF₃MU-Neu5Ac of (44 ± 6) μM and K_i for Neu5Ac2en around (120 ± 26) μM, as opposed to 7 μM and (0.14 ± 0.07) μM, respectively, for MvNA (Watson and Newstead, et al. 2005).

Subtle structural differences observed in the PaNA-Neu5Ac2en complex, in part, explain the overall weaker binding affinity. Specifically, five precisely oriented water molecules, interconnected through a number of hydrogen-bonding interactions, bridge the O7 hydroxyl of Neu5Ac2en with backbone atoms of residues composing the hydrophobic pocket in MvNA. Such an extended hydrogen-bonding network can contribute considerably to stabilizing protein–ligand interactions. Thus, the absence of such surface water-mediated coordination of O7 in PaNA may lead to reduced binding affinity of Neu5Ac2en. At the distal end of the glycerol group in the PaNA-Neu5Ac2en complex, the microenvironment is more polar relative to that of MvNA, partly because of the substitution of Ile228 and Phe234 for Arg282 and Gln287 (Figure 3A). The more hydrophobic environment in MvNA may help to exclude water from this region of the active site and strengthen sequestered hydrogen bonds formed with the O8 and O9 hydroxyls. The reduced contribution of such shielding effects from hydrophobic residues to ligand binding may also account for the lower affinity of Neu5Ac2en for PaNA.

Glycan arrays are useful tools for characterizing the molecular determinants important in carbohydrate–protein interactions (Oyelaran and Gildersleeve, 2009). Although the current screen represents only a limited set of naturally occurring mammalian glycans, the results provide valuable information for defining the binding specificity of PaNA. Significant binding energy is derived from the interaction of a C5 hydrophobic extension of Neu5Ac with the pocket created by Ala147, Val202, Phe209, Leu224 and Phe257 as more polar substituents such as a hydroxyl group or N-glycolyl, which extends the 5-N-acetyl by an additional hydroxyl, can decrease the binding substantially. This is consistent with previous observations that KDN-terminated sialosides are, in general, poor substrates of bacterial sialidases, except KDNase, a sialidase specific for this sialyl variant (Chokhawala, et al. 2007).

In terms of ligand design for PaNA, it may be useful to exploit the hydrophobic pocket near C5 for improved recognition. On the other hand, 9-O-acetylation is better tolerated by PaNA although this type of modification also decreases binding. The reduced affinity may be explained by the loss of direct hydrogen-bonding interactions with Asp312 as a result of acetylation and unfavorable steric or electrostatic constraints imposed by residues surrounding O9. Modifying the glycerol extension of Neu5Ac with substituents capable of forming favorable, direct interactions with nearby residues, for instance, Arg282, Gln287 and Asp312, while displacing the bridging water molecule, W1, may lead to stronger interactions with the PaNA active site. Moreover, the glycan screen also reveals PaNA’s flexibility toward the type of sialyl linkage, and the contribution of the penultimate residue to binding. The fact that α-Neu5Ac directly linked to a bulkier spacer, -OCH₂C₆H₄-p-NHCOCH₂NH, ranks higher than those attached to a linear aliphatic chain, -CH₂CH₂CH₂NH₂, or to carbohydrate residues suggests that some affinity is derived from interactions with more hydrophobic residues in this position though it is not readily apparent which ones are involved. Similar observations have been made with other bacterial sialidases. For example, the structurally similar neuraminic acid C-glycosides with more hydrophobic aglycones have been reported to be effective substrate analog inhibitors (compared to Neu5Ac2en) for C. perfringens sialidase NanI, one of the few well-characterized microbial sialidases in terms of inhibition (Wang, et al. 2000).

Although a high-resolution structure of a sialidase in complex with a neuraminic acid C-glycoside is not yet available, understanding the mode of binding of this class of inhibitors will shed light on the contribution of the hydrophobic aglycone, and provide valuable insight on designing compounds with improved inhibitory properties.

The P. acnes genome contains at least three genes with sequence similarity to known bacterial sialidases, namely, sialidases A and B (accession codes: AAT82441 and AAT82440, respectively) and the cell wall-anchored PaNA (AAT83304) (Bruggemann, et al. 2004). Biological functions for each of them have yet to be fully described. Sialidase A precursor has a laminin G-like domain, which is absent in PaNA, N-terminal to the catalytic unit. The additional domain may serve to localize the enzyme to the appropriate substrates and increase its catalytic efficiency as suggested for other modular sialidases containing carbohydrate-binding modules, while mediating adhesion by interacting with cell surface and extracellular matrix components of host tissues (Hohenester, et al. 1999, Thobhani, et al. 2003). Sialidase B differs from PaNA in that it exhibits sequence similarity to intramolecular trans-sialidase with strict specificity toward α-2,3-linked sialic acids and generates 2,7-anhydro-Neu5Ac, which is proposed to be a stable intermediate that can be transferred to other glycoconjugates (Gut, et al. 2008). In light of the different specificities, each may provide a selective advantage in host environments with different distributions of the various sialyl linkages. The S. pneumonia NanA neuraminidase and the P. aeruginosa sialidase-like enzyme PA2794 have been shown to participate in biofilm production, which facilitates the development of respiratory infections (Parker, et al. 2009, Soong, et al. 2006). Interestingly, P. acnes biofilm has long been implicated in the pathogenesis of acne vulgaris as well as medical implant infections, suggesting additional virulent functions for P. acnes sialidases (Coenye, et al. 2008). Further experiments are required to establish each of their roles in P. acnes-associated diseases and validate the sialidases as suitable drug targets.

Materials and methods

Molecular cloning

The gene corresponding to the cell wall-anchored sialidase at locus PPA1560 in P. acnes KPA171202 (accession code: AAT83304) was PCR-amplified from genomic DNA of a clinically relevant strain (ATCC® 6919). The forward and reverse primers used were: 5′- CCGCAGCATATGTCAGGCAGGGCTCCGGCCCCAGTT-3′ and 5’-CCGCAGGAGCTCTCAAGTTGCAGTGGGCGGGAGAG, respectively. The fragment containing residues 31–481 was cloned into the NdeI and SacI sites of a pET-28(a) expression vector (Novagen).

An acid/base catalyst mutant of PaNA was designed for use in a glycan array screening to evaluate the enzyme’s binding affinity for various ligands and prevent substrate hydrolysis. The D146A mutation was introduced into the wild-type construct using the QuickChange site-directed mutagenesis kit (Stratagene), with the forward (5’-GGGAGCGCCGGCGCTGCGCCTAACCCCAATTC-3′) and reverse primers (5’-GAATTGGGGTTAGGCGCAGCGCCGGCGCTC CC-3′). The presence of the mutation was confirmed by DNA sequencing.

Protein expression and purification

Escherichia coli BL21 (DE3) transformed with the N-terminally His₆-tagged PaNA construct was grown at 37 °C in LB media until mid-exponential phase and induced with 0.5 mM IPTG at 20 °C. Cells were harvested after overnight incubation. The resulting pellet was disrupted using a pressurized homogenizer (Avestin) in lysis buffer (20 mM HEPES pH 6.8, 500 mM NaCl). The clarified lysate containing the target protein was purified sequentially by cobalt-chelating Sepharose, MonoS cation exchange and Superdex-75 HR 10/30 columns (all GE Healthcare) to >95 % purity, as judged by SDS-PAGE. The His6 tag was cleaved by incubating with thrombin overnight after cobalt affinity chromatography. The D146A mutant was purified following the same protocol.

Crystallization

Crystals of PaNA were grown by the microbatch method at 18 °C. The drops consisted of a 1:1 ratio of purified PaNA (6 mg^.mL⁻¹) and the reservoir solution (20 % (w/v) polyethylene glycol monomethylether 2000, 100 mM HEPES, pH 7.5). Crystals for ligand soaking were reproduced by seeding into drops with similar chemical compositions. Crystals grown under the condition 18 % (w/v) polyethylene glycol monomethylether 2000, 100 mM HEPES, pH 6.8 were soaked with 5 mM Neu5Ac2en for 3 h.

Data collection and structure determination

The crystals were cryoprotected by briefly soaking in the mother liquor containing additional 10 % (v/v) glycerol and then flash-frozen in liquid nitrogen. X-ray diffraction data were collected under the cryostream at 100 K, using an in-house Cu-K_α rotating anode X-ray generator coupled to a Mar345 detector.

The data were processed using the HKL 2000 program package (Otwinowski and Minor 1997). A molecular replacement solution was found by Phaser using the M. viridifaciens sialidase structure encompassing residues 55–405 as a search model (PDB code: 1EUR) (McCoy, et al. 2007). After rigid body refinement of the starting poly-alanine template, manual model building was initiated with Coot (Emsley and Cowtan 2004). Refinement based on simulated annealing was carried out using PHENIX (Adams, et al. 2002). Iterative cycles of manual model rebuilding and refinement were performed using Coot and REFMAC5 (Murshudov, et al. 1997), respectively, excluding 5 % of the reflections for the calculation of R_free. TLS (Translation/Liberation/Screw) parameters were included in further rounds of refinement (Winn, et al. 2003). Ideal coordinates for the Neu5Ac2en model and its refinement restraints were obtained from the HIC-Up server (Kleywegt 2007). Data collection parameters and refinement statistics are listed in Table I. All structure images were produced with UCSF Chimera (Pettersen, et al. 2004). Electrostatic surface calculations were carried out using APBS (Baker, et al. 2001).

Table I.

Data collection and refinement statistics^a

	PaNA	PaNA:Neu5Ac2en
Data collection
PDB accession code	7LBU	7LBV
Space group	P1	P1
Unit cell parameter
a, b, c (Å)	43.97, 46.85, 49.60	43.98, 46.65, 49.66
α, β, γ (°)	116.51, 98.94, 91.69	116.93, 99.10, 91.73
Resolution (Å)	50.00-2.10 (2.18-2.10)	50.00-1.70 (1.76-1.70)
R merge (%)	5.4 (23.5)	4.0 (24.8)
Completeness (%)	91.8 (77.4)	93.2 (88.2)
I/σ(I)	18.26 (5.51)	20.17 (4.43)
Multiplicity	3.9 (3.6)	3.9 (3.5)
Overall B factor from Wilson plot (Å2)	29.05	20.25
Refinement
Resolution limits (Å)	43.60-2.11	43.44-1.70
No. of reflections	17,676	33,635
R work/Rfree (%)	17.26/23.28	15.17/18.17
B factor (Å2)
Protein	39.34	28.29
Ligand	-	26.40
Water	38.54	34.80
r.m.s. deviations
Bond lengths (Å)	0.017	0.021
Bond angles (°)	1.913	2.047
Ramachandran plot statistics (%)b
Favored	97.09	96.83
Allowed	2.65	2.61
Disallowed	0.26	0.56

^aValues in parentheses belong to the highest resolution shell.

^bStatistics calculated by Molprobity (http://molprobity.biochem.duke.edu).

Kinetic characterization

Kinetic parameters for PaNA were determined using the fluorogenic substrate 4-trifluoromethylumbelliferyl-α-d-N-acetylneuraminic acid (CF₃MU-Neu5Ac) in the assay buffer (50 mM sodium acetate pH 5.4 plus 0.1 mg^.mL⁻¹ BSA) at 37 °C. Briefly, 700 μL reactions containing the enzyme (36 ng^.mL⁻¹) with varying concentrations of CF₃MU-Neu5Ac (2-200 μM) were allowed to react at 37 °C. At 1 min intervals, a 100 μL sample was quenched with 900 μL glycine 0.2 M, pH 10. Fluorescence of these samples, with excitation and emission wavelengths of 357 and 499 nm, respectively, was subsequently read on a Synergy H1 plate reader [BioTek].

Inhibition by Neu5ac2en was measured with a single fixed concentration of substrate (50 μM) and enzyme (10 ng^.mL⁻¹) but varied inhibitor concentration (0.06–0.45 mM). Reaction progress was monitored using a stopped assay as above.

Data analysis was carried out in Graphpad Prism 9 (Graphpad software, USA). The kinetic parameters were determined by fitting all data to the Michaelis–Menten equation using nonlinear regression analysis, and inhibition was estimated by assuming a competitive inhibition model and extrapolating a linear fit in a Dixon plot to its intercept with 1/V_max. Uncertainties represent standard errors of nonlinear regression for kinetic parameters and propagation of errors in both the linear fit and the Vmax determination for inhibition.

Glycan array screening

The glycan array binding analysis was performed at the Consortium for Functional Glycomics (http://www.functionalglycomics.org/static/index.shtml). The PaNA acid/base catalyst mutant D146A, fluorescently labeled with the Alexa Fluor® 488 dye (Molecular Probes), was screened with version 3.2 of the printed mammalian glycan microarray containing 406 ligands in replicates of 6. The highest and lowest fluorescence readings were excluded from statistical analysis of the data in an attempt to eliminate false hits.

Funding

This work was supported by the Canadian Institutes of Health research.

Abbreviations

PaNA, Propionibacterium acnes neuraminidase (sialidase); MvNA, Micromonospora viridifaciens neuraminidase (sialidase); KDN, deaminoneuraminic acid; StNA, Salmonella typhimurium neuraminidase (sialidase)

Conflict of interest

The authors declare that there is no conflict of interest.

Data avilability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request. Atomic coordinates and associated structure factors have been deposited with the PDB with accession codes 7LBU (without substrate) and 7LBV (with Neu5Ac2en). The glycan array data are available at http://www.functionalglycomics.org/ with accession id primscreen_2531.