MHJ_0125 is an M42 glutamyl aminopeptidase that moonlights as a multifunctional adhesin on the surface of Mycoplasma hyopneumoniae

Bacterial aminopeptidases play important roles in pathogenesis by providing a source of amino acids from exogenous proteins, destroying host immunological effector peptides and executing posttranslational modification of bacterial and host proteins. We show that MHJ_0125 from the swine respiratory pathogen Mycoplasma hyopneumoniae represents a new member of the M42 class of bacterial aminopeptidases. Despite lacking a recognizable signal sequence, MHJ_0125 is detectable on the cell surface by fluorescence microscopy and LC-MS/MS of (i) biotinylated surface proteins captured by avidin chromatography and (ii) peptides released by mild trypsin shaving. Furthermore, surface-associated glutamyl aminopeptidase activity was detected by incubation of live M. hyopneumoniae cells with the diagnostic substrate H-Glu-AMC. MHJ_0125 moonlights as a multifunctional adhesin, binding to both heparin and plasminogen. Native proteomics and comparative modelling studies suggest MHJ_0125 forms a dodecameric, homopolymeric structure and provide insight into the positions of key residues that are predicted to interact with heparin and plasminogen. MHJ_0125 is the first aminopeptidase shown to both bind plasminogen and facilitate its activation by tissue plasminogen activator. Plasmin cleaves host extracellular matrix proteins and activates matrix metalloproteases, generating peptide substrates for MHJ_0125 and a source of amino acids for growth of M. hyopneumoniae. This unique interaction represents a new paradigm in microbial pathogenesis.


Summary
Bacterial aminopeptidases play important roles in pathogenesis by providing a source of amino acids from exogenous proteins, destroying host immunological effector peptides and executing posttranslational modification of bacterial and host proteins. We show that MHJ_0125 from the swine respiratory pathogen Mycoplasma hyopneumoniae represents a new member of the M42 class of bacterial aminopeptidases. Despite lacking a recognizable signal sequence, MHJ_0125 is detectable on the cell surface by fluorescence microscopy and LC-MS/MS of (i) biotinylated surface proteins captured by avidin chromatography and (ii) peptides released by mild trypsin shaving. Furthermore, surface-associated glutamyl aminopeptidase activity was detected by incubation of live M. hyopneumoniae cells with the diagnostic substrate H-Glu-AMC. MHJ_0125 moonlights as a multifunctional adhesin, binding to both heparin and plasminogen. Native proteomics and comparative modelling studies suggest MHJ_0125 forms a dodecameric, homopolymeric structure and provide insight into the positions of key residues that are predicted to interact with heparin and plasminogen. MHJ_0125 is the first aminopeptidase shown to both bind plasminogen and facilitate its activation by tissue plasminogen activator. Plasmin cleaves host extracellular matrix proteins and activates matrix metalloproteases, generating peptide substrates for MHJ_0125 and a source of amino acids for growth of M. hyopneumoniae. This unique interaction represents a new paradigm in microbial pathogenesis.

Introduction
Mucosal droplets released from Mycoplasma hyopneumoniaeinfected swine during bouts of coughing are important in the cycle of re-infection of naive animals within the confines of a commercial swine production facility. Upon inhalation, M. hyopneumoniae is confronted by the mucociliary clearance machinery, a key innate immune barrier to infection. Mycoplasma hyopneumoniae must also traverse mucus barriers and initiate colonization by adhering to cilia if infection is to proceed. Ciliostasis, cilial loss and epithelial cell death are pathological features of infection by M. hyopneumoniae [1][2][3]. Infection by M. hyopneumoniae stimulates the expression of proinflammatory cytokines such as interleukin-1b, tumour necrosis factoralpha and interleukin-6 that elicit an acute inflammatory response attracting neutrophils and monocytes to the infected airways [3][4][5]. Although chronic inflammation is a hallmark of M. hyopneumoniae infection, the mechanisms by which M. hyopneumoniae initiates cell damage and influences inflammation are not well understood. Mycoplasma hyopneumoniae is detectable in the spleen, liver and kidneys of artificially challenged pigs and their cohorts [6][7][8][9][10], but further studies are needed to determine if it plays a pathogenic role at these distal tissue sites. Mycoplasma hyopneumoniae must display an elaborate and extensive repertoire of surface antigens to interact with extracellular matrix (ECM) components and diverse cell types.
Mycoplasma hyopneumoniae presents members of the P97 and P102 paralogue families of adhesin proteins on the cell surface, where their primary role is to adhere to respiratory cilia. These multifunctional adhesins undergo endoproteolytic processing such that the N-terminal cleavage product retains the signal peptide, and central and C-terminal fragments are released from the preprotein but remain associated with the cell surface [4,[11][12][13][14][15][16][17][18][19]. The pattern of proteolytic cleavage fragments is often consistent among strains derived from different geographical origins, but some fragments are strain specific or are more prominent because the efficiency of cleavage at a specific site may vary [17][18][19]. Cleavage occurs predominantly at the carboxyl side of phenylalanine residues that reside within the motif S/T-X-F -X-D/E [11] but at least two other cleavage motifs have been described [16,17].
Cleavage fragments generated from P97 and P102 paralogue families are multifunctional. Many bind sulfated glycosaminoglycans and fibronectin [11][12][13][14][15][16][18][19][20], and these interactions are critical for successful colonization of the respiratory tract [15,21] and porcine epithelial cells [4,14,15,19]. Cleavage fragments also bind plasmin(ogen) [4,12,13,18]. Plasminogen is readily detectable in the fluid lining ciliated epithelial surfaces in the porcine respiratory tract and is sequestered onto the surface of M. hyopneumoniae. Mycoplasma hyopneumoniae surface-bound plasminogen is readily converted to the serine protease plasmin by tissue plasminogen activator (tPA) and is capable of degrading fibrinogen [4]. Plasmin levels are consistently elevated in bronchoalveolar lavage (BAL) fluids of pigs infected with M. hyopneumoniae compared with BAL fluid collected from the same animals prior to challenge [4]. The recruitment of plasmin(ogen) to the surface of M. hyopneumoniae is likely to play an important role in its ability to colonize the respiratory tract, traverse ECM/basement membrane and colonize sites distal to the respiratory tract.
In the present study, MHJ_0125 was cloned and expressed as a functionally active polyhistidine fusion (rMHJ_0125) to determine its function. We also investigated the subcellular location and three-dimensional structure of MHJ_0125. The biochemical properties of MHJ_0125 and the potential role(s) in pathogenesis of this moonlighting protease are discussed.

MHJ_0125 resides on the surface of Mycoplasma hyopneumoniae and is secreted extracellularly
A subset of tryptic peptides detected by LC-MS/MS that were generated by mildly enzymatically shaving the surface of M. hyopneumoniae cells mapped to a protein (MHJ_0125, UniProt number Q4AAK4) annotated as a putative aminopeptidase ( figure 1a). This was a surprising observation because MHJ_ 0125 lacks a signal peptide, has no apparent transmembrane spanning regions and is predicted to reside intracellularly by PSORTb. In control experiments performed in triplicate, proteins identified by mass spectrometric analyses of tryptic digests of proteins released from M. hyopneumoniae incubated under identical conditions used for the shaving, but in the absence of trypsin, identified four proteins. All four proteins belonged to a subset of proteins localized to the cell surface. Tryptic digestion of secreted proteins identified peptides that mapped to MHJ_0125 (figure 1a). On no occasion did we observe proteins in comparable concentration as MHJ_0125 as determined by spectral counting, that we failed to identify in our surface studies. These observations indicated that cell lysis was not responsible for detecting MHJ_0125 extracellularly. Consistent with this, we did not observe appreciable cell lysis with incubations over time from 30 s to 30 min by SDS-PAGE (see the electronic supplementary material, figure S1).
To further address the possibility that MHJ_0125 resides on the surface of M. hyopneumoniae, surface-accessible proteins were labelled with Sulfo-NHS-LC-Biotin, captured by avidin chromatography and separated by two-dimensional gel electrophoresis. Tryptic peptides mapping to MHJ_0125 were generated from a biotinylated protein spot (data not shown) with a mass of approximately 40 kDa. The MHJ_ 0125 protein (356 amino acids) has a predicted molecular mass of 39.2 kDa and a theoretical pI value of 6.21. These surface proteome studies indicate that MHJ_0125 both resides on the cell surface and is shed into the extracellular milieu.
To determine if MHJ_0125 exists solely on the extracellular side of the membrane, M. hyopneumoniae cells were rsob.royalsocietypublishing.org Open Biol 3: 130017 exposed to a range of trypsin concentrations. Cell lysates stained with Flamingo stain (figure 1b) showed the gradual proteolysis of proteins with increasing trypsin concentrations, especially in the high mass regions. Many proteins, however, remained unaffected even at the highest trypsin concentration (500 mg ml 21 ), indicating that the cell membrane remained intact. These observations are consistent with previous experiments [14 -17]. Cell lysates of shaved cells were blotted onto polyvinylidene fluoride (PVDF) membrane and probed with rabbit anti-MHJ_0125 antibodies. MHJ_0125 was detectable in the control lane (no trypsin) and at all trypsin concentrations up to 500 mg ml 21 (figure 1c(ii)). Anti-MHJ_0125 antibodies detected a single protein species with a mass of 39 kDa, confirming it is monospecific for MHJ_0125 (see the electronic supplementary material, figure S2). The same blot was re-probed with P97 N-terminal serum that recognizes the central portion of the P97 adhesin [17]. There was clear evidence that P97 was readily digested at a trypsin concentration of 1 mg ml 21 (figure 1c(i)). These data indicate that MHJ_0125 resides in the cytosol of M. hyopneumoniae, but that a subset of MHJ_0125 molecules are both bound to and secreted from the cell surface.
Fluorescence microscopy was also used to examine the cellular location of MHJ_0125. Mycoplasma hyopneumoniae cells were stained with anti-MHJ_0125 antibodies and detected using goat anti-rabbit antibodies conjugated with Alexa Fluor 488. Widefield conventional fluorescence microscopy suggested that all M. hyopneumoniae cells produced a fluorescent ring surrounding the cell surface (electronic supplementary material, figure S3a). When M. hyopneumoniae cells were pretreated with Triton X-100 to permeabilize the Mycoplasma membranes, the fluorescence pattern instead filled the entire cell body and the fluorescence intensity was brighter compared with images generated with non-permeabilized cells imaged at the same intensity (see the electronic supplementary material, figure S3b). In control experiments, goat anti-rabbit Alexa Fluor antibodies did not contribute to staining of M. hyopneumoniae cells in the absence of anti-MHJ_0125 antibodies (data not shown). To further confirm the subcellular localization of the MHJ_0125 proteins, the cells were examined using super-resolution three-dimensional structured illumination microscopy (3D-SIM). These studies confirmed that MHJ_ 0125 resides both on the surface (figure 1d) and in the cytosol (figure 1e) of all M. hyopneumoniae cells growing in vitro. (b) Cell lysates of M. hyopneumoniae cells exposed to different concentrations of trypsin (0 -500 mg ml 21 at 378C for 15 min). Protein profiles indicate that the cell membrane remains intact under these experimental conditions. (c) Western blot of M. hyopneumoniae lysates from cells exposed to different concentrations of trypsin probed with rabbit anti-P97 N-terminal serum (i) and anti-MHJ_0125 serum (ii). The P97 cilium adhesin (top empty arrow) is reported to be highly sensitive to trypsin [17] and is recognized by anti-P97 N-terminal antibodies. MHJ_0125 was not completely degraded by trypsin because of its presence inside the cell cytosol. (d,e) Structured illumination microscopy images of M. hyopneumoniae cells probed with rabbit anti-MHJ_0125 followed by goat anti-rabbit HRP conjugated to Alexa Fluor 488. Fixed, non-permeabilized cells (d ) show MHJ_0125 only on the cell surface, whereas permeabilized, fixed cells (e) show MHJ_0125 is also present inside the cell. Images were taken on an OMX Deltavision microscope. Scale bar, 0.5 mm. (f ) Cell surface-associated glutamyl aminopeptidase activity was confirmed by culturing live M. hyopneumoniae in the presence of the fluorescent substrate H-Glu-AMC, with or without the aminopeptidase inhibitor bestatin. Data are shown as the relative enzyme activity, expressed as a percentage against no bestatin (none) that is taken to be 100% activity. Data represent mean fluorescence units from three independent assays+s.d.

Biochemical characterization of functionally active recombinant MHJ_0125
Recombinant MHJ_0125 was purified from E. coli cell lysates by nickel affinity chromatography. A protein of approximately 42 kDa was resolved by SDS-PAGE and its identity was confirmed as rMHJ_0125 by LC-MS/MS analysis (data not shown). rMHJ_0125 efficiently cleaves the acidic amino acid glutamic acid from the N-terminus of synthetic fluorogenic peptide substrates (figure 3a) but not aspartic acid which, unexpectedly, it cleaved relatively poorly. Interestingly, rMHJ_0125 also cleaved H-Ala-AMC and to a lesser extent H-Leu-AMC (figure 3a). Very limited or no hydrolysis was observed against the fluorogenic substrates H-Arg-AMC, H-Pro-AMC, H-Val-AMC and H-Phe-AMC. These data are consistent with the classification of MHJ_0125 as a member of the M42 glutamyl aminopeptidase family. One of the best characterized bacterial M42 glutamyl aminopeptidases is Streptococcus pneumoniae PepA (SpPepA) [25]. A comparison of the substrate specificity of MHJ_0125 with that of SpPepA showed that while both enzymes have a strong preference for glutamic acid, they differ in their ability to cleave aspartic acid residues; SpPepA shows 70 per cent relative activity towards aspartic acid [25] compared with 54 per cent shown by MHJ_0125. Both enzymes were also able to cleave alanine residues to some extent (approx. 20% relative activity by SpPepA [25] and 54% by MHJ_0125). The activity of both enzymes against substrates containing other amino acids was negligible. These differences in substrate specificity may be due to variation within the S1 substrate-binding pocket (see §3.8). rMHJ_0125 activity was reduced to 57.1, 30.1 and 13.3 per cent after incubation with 0.01, 0.1 and 1.0 mM bestatin, respectively. Similarly, rMHJ_0125 activity was reduced to 3.7 per cent following incubation with 10 mM EDTA, demonstrating that metal ions are necessary for enzyme activity (table 1). A study of the effect of various divalent metal ions (final concentration 5 mM) on the activity of the enzyme (against H-Glu-AMC) showed that it was enhanced by Co 2þ (52%). By contrast, Mn 2þ considerably reduced enzyme activity (by 84.5%), whereas Cu 2þ (5 mM) almost abolished enzyme activity (96.5% reduction). Mg 2þ and Zn 2þ did not appreciably affect enzyme activity compared with control assays where no metal ions were added (figure 3b). The purified enzyme exhibited aminopeptidase activity against H-Glu-AMC between pH 6.0 and 8.5 with optimal activity at pH 7.5 (figure 3c). Activity against the fluorogenic substrate H-Glu-AMC was detected with freshly cultured, washed M. hyopneumoniae cells, confirming the presence of surfaceassociated glutamyl aminopeptidase activity. This activity was inhibited in situ by 1 mM bestatin (figure 1f ).

MHJ_0125 binds and activates porcine plasminogen
The presence of MHJ_0125 on the surface of M. hyopneumoniae suggested that it may function as a virulence factor. Our previous observations showing plasmin activity is elevated in the BAL fluid of pigs infected with M. hyopneumoniae [4] and that C-terminal lysines and arginine residues on endoproteolytic cleavage fragments of the P97 and P102 adhesin families bind plasminogen and facilitate its activation to plasmin [12,13,18] led us to investigate whether MHJ_0125 plays a role in plasminogen binding. Consistent with such a role, MHJ_0125 possesses a C-terminal lysine residue. While undertaking a systematic two-dimensional ligand blotting study of M. hyopneumoniae plasminogen-binding proteins, we identified two proteins approximately 40 and 80 kDa in size that bound biotinylated plasminogen in the Triton X-114 aqueous phase fraction (figure 4a). LC-MS/MS of tryptic fragments of these proteins identified MHJ_0125 (Mascot scores of 650 and 120; peptide coverage of 50% and 25%, respectively). We subsequently found that rMHJ_0125 bound to purified porcine plasminogen in a concentration-dependent manner (figure 4b).
In the presence of tPA, there was a clear increase in the rate of plasmin activity (figure 4c) when there was a molar excess (4 : 1 and 8 : 1) of rMHJ_0125 relative to plasminogen; however, rsob.royalsocietypublishing.org Open Biol 3: 130017 rMHJ_0125 did not activate plasminogen in the absence of tPA. These results suggest that binding of MHJ_0125 to plasminogen results in a conformational change leading to enhanced plasmin activity; however, MHJ_0125 does not itself cleave plasminogen to plasmin.

MHJ_0125 binds heparin
Heparin is used widely to mimic highly sulfated regions present in glycosaminoglycans and effectively blocks the binding of M. hyopneumoniae to porcine cilia [26]. Mycoplasma hyopneumoniae binds heparin on its cell surface in a dose-dependent and saturable manner [19]. MHJ_0125 carries a lysine-rich motif (EKMKKY) similar to the consensus heparin-binding site (XBBXBX) at amino acid positions 5-10 in its N-terminus. LC-MS/MS analysis of heparin-binding proteins identified tryptic peptides that mapped to MHJ_0125 (Mascot score 541, 7 unique peptides identified, 28% coverage; data not shown). We had previously determined that MHJ_0125 migrates as a complex with a nominal mass of 600 kDa during native PAGE (figure 2).
To determine if this aminopeptidase complex bound heparin, proteins that were retained on the heparin column were separated by one-dimensional native PAGE. LC-MS/MS analysis of tryptic fragments (Mascot score 1277, 14 unique peptides, 49% coverage) generated from the 600 kDa protein complex matched to MHJ_0125 (figure 5).

MHJ_0125 is recognized by convalescent swine antibodies
An antibody response against MHJ_0125 was detected in the serum of commercially reared pigs determined to be infected with M. hyopneumoniae by ELISA and western blot (figure 6).

MHJ_0125 is related to other bacterial aminopeptidases
We assessed the phylogenetic relationship between the M. hyopneumoniae sequence and glutamyl aminopeptidases from bacterial and eukaryotic species. Figure 7 shows that MHJ_0125 is a member of a distinct clade (     figure 9. Because size exclusion chromatography indicated that the MHJ_0125 biological unit is approximately 440 kDa, and given that this protein displays homology to 12-subunit proteases (including TET protease), we reasoned that MHJ_0125 forms a dodecameric complex. Therefore, the comparative model generated for MHJ_0125 contains 12 identical subunits with tetrahedral symmetry based on the biological unit of P. horikoshii TET protease (figure 9a). Given that the protein databank (PDB) coordinates for TET protease contain zinc in the metal-binding site, and that    figure S4). Our MHJ_0125 model indicates that the conserved residues identified in the sequence alignment (figure 9b), which are involved in catalysis and the co-ordination of the metal ions in other aminopeptidases, are in the correct steric position to fulfil these functions in MHJ_0125 (figure 9b). This suggests that MHJ_0125 has a similar catalytic mechanism to other aminopeptidases within the superfamily. Interestingly, the substrate-binding pocket of MHJ_0125 differs somewhat from that of the PepA aminopeptidase from S. pneumoniae despite the fact that they both display a substrate preference for glutamic acid. Sp PepA contains an arginine residue at position 257 (Arg 257 ) that confers a positive charge on the S1 pocket of the binding site that is believed to be juxtaposed to the negatively charged carboxylate group of glutamic acid. In MHJ_0125, an aspartic acid residue (Asp 258 ) replaces Arg 257 ; however, the carboxyl group of Asp 258 appears not to protrude far enough into the S1 pocket to confer a negative charge on this region (figure 9c). It is probable that this variation in S1 subsite topography is responsible for the minor differences in substrate preference exhibited by SpPepA and MHJ_0125; however, this awaits confirmation via further structure-function analyses.
Like other TETs, MHJ_0125 is predicted to contain four large pores that are entry points for the substrate and four small pores that act as exit channels for the products of catalysis. In MHJ_0125, the large and small openings are 16 and 8 Å in diameter, respectively. It is notable that the lining of each exit pore is dominated by lysine residues (figure 10a), conferring a positive charge on the opening and allowing the exit of negatively charged glutamic acid residues. A similar conformation of positively charged residues is observed in the exit pores of Sp PepA (figure 10b). By contrast, the exit pores of TET protease from P. horikoshii are dominated by hydrophobic phenylalanine residues (figure 10c), which may be explained by the fact that TET protease has a preference for leucine [27]. The positive surface potential of the MHJ_0125 small opening created by the lysine residues can be seen in figure 10d.
A motif (XBXBBX) similar to the consensus heparinbinding site (XBBXBX) was identified in the sequence of MHJ_0125. The EKMKKY motif is located at the N-terminus of MHJ_0125 and is exposed on the surface of the MHJ_0125 dodecamer ( figure 11a). An XBXBBX motif has previously been implicated in the binding of heparan sulfate [28]. An examination of the surface potential of the MHJ_0125 dodecamer indicates that the area containing the XBXBBX motif is positively charged ( figure 11b); therefore, these regions may play a role in interacting with the negatively charged sulfate groups of heparin.
The C-terminal lysine residues of each MHJ_0125 monomer are highlighted in figure 11c. C-terminal lysine residues have been implicated in the binding of several M. hyopneumoniae proteins to plasminogen [12,18]. These C-terminal lysine residues are exposed to the surface and are located in a small cleft created at the interface of the individual protein subunits.

Discussion
Little is known about the strategies used by mycoplasmas to secure essential protein components for life. As an obligate, genome-reduced pathogen, M. hyopneumoniae must derive amino acids from the host and transport them into the cell for the synthesis of essential proteins. Here, we describe the biochemical and structural features of a glutamyl aminopeptidase of the M42 family of aminopeptidases that displays the unusual capability to reside in the cytosol and on the surface of M. hyopneumoniae (figure 1). MHJ_0125 is hydrophilic, and lacks evidence of a signal sequence and transmembrane domains. rMHJ_0125 shows maximal aminopeptidase activity at physiological pH, requires divalent metal ions for activity and is inhibited by bestatin (figure 1). Consistent with MHJ_0125 molecules being located on the cell surface, freshly cultured M. hyopneumoniae cells displayed glutamyl aminopeptidase activity that was inhibited by the exogenous addition of bestatin. While MHJ_0125 molecules are both retained on the external side of the membrane and secreted into the extracellular milieu, the mechanism(s) that facilitate secretion of MHJ_0125 across the cell membrane is unknown.
MHJ_0125 forms a 440 kDa homopolymeric structure (figure 2) and was not detected as a monomer during native PAGE or size exclusion chromatography. Residues involved in metal ion binding (His 64 , Asp 179 , Glu 212 , Asp 234 and His 320 ) are well conserved among M42 aminopeptidases that belong to the TET family of dodecameric tetrahedral complexes [25,29]. Our modelling studies were consistent with this view. Asp 236 , Ser 238 , Leu 255 , Arg 257 , Thr 309 and Gly 311 have been shown to comprise the substrate-binding pocket in the M42 aminopeptidase PepA from S. pneumoniae [25]. While several of these amino acids are conserved in MHJ_0125, a notable omission is Arg 257 , which is purported to play a critical role in making electrostatic interactions with acidic N-terminus substrate residues (Asp/Glu) and Gly 311 . Modelling studies of MHJ_0125 show that a negatively charged aspartic acid residue in position 258 (Asp 258 ) replaces a positively charged Arg 257 ; however, it appears that the carboxyl moiety in Asp 258 does not protrude sufficiently into the S1 pocket to confer negative charge to this region. Consistent with its role as a glutamyl aminopeptidase, rsob.royalsocietypublishing.org Open Biol 3: 130017 the exit pore of MHJ_0125 is enriched in lysine residues that confer a positive charge to the pore opening and are likely to facilitate the expulsion of cleaved glutamyl residues (figure 10). Proteins that reside in more than one cellular location typically are multifunctional and many belong to a remarkable subset of molecules known as moonlighting proteins [30]. Motifs needed to perform alternate functions are likely to be confined to regions of the molecule that do not disrupt the structural constraints required for MHJ_0125 to perform as a dodecameric glutamyl aminopeptidase. We show that in addition to aminopeptidase activity, rMHJ_0125 binds plasminogen and, in the presence of tPA, facilitates conversion of plasminogen to the active serine protease, plasmin. C-terminal lysine residues play an important role in binding plasminogen [12,18,31], and our modelling studies show that the C-terminal lysine of MHJ_0125 is accessible on the surface of the dodecamer. We propose that this surface-accessible lysine is involved in binding plasminogen. Several plasminogen-binding proteins have been described on the surface of M. hyopneumoniae, indicating that the capacity to recruit plasminogen to the cell surface is important during the normal course of infection [4,12,13,18]. Indeed, our in vivo studies are consistent with this view; plasminogen is found in close association with ciliated epithelial borders and plasmin activity is consistently higher in the BAL fluid of most pigs following challenge with M. hyopneumoniae [4].
In addition to binding plasminogen, we repeatedly observed MHJ_0125 to bind heparin. Recovery of a 440 kDa MHJ_0125 complex from heparin-agarose indicates that MHJ_0125 binds glycosaminoglycans in its native dodecameric conformation ( figure 5). We identified a lysine-rich, putative heparin-binding motif (EKMKKY) that conforms to a known heparan sulfate-binding domain (XBXBBX) [28]. The motif was shown in our modelling studies to be accessible on the surface of MHJ_0125. Mycoplasama hyopneumoniae is highly dependent on interactions with glycosaminoglycans during colonization of respiratory tract cilia [21] and porcine epithelial cells [19], and the importance of this interaction is underscored by the degree of redundancy shown by members of the P97 and P102 paralogue families [11,[13][14][15][16]18,19,21]. These adhesins promote close contact of M. hyopneumoniae with ciliated epithelium [1,2,5], where there is an abundance of glycosaminoglycans enriched in highly sulfated, heparin-like domains [32]. The secretion of MHJ_0125 from the surface of M. hyopneumoniae may play a role in colonizing new sites both in the respiratory tract and at distal tissue sites. In addition,  Many pathogenic and environmental bacteria secrete hydrolytic enzymes into the extracellular milieu, where they influence adaption to their respective niche [33,34]. Bacterial extracellular proteases form part of an impressive arsenal to inflict significant tissue damage, destroy innate and adaptive immune effector molecules, modify the protein surface topography of both the organism that secretes them and the host (immune evasion), and provide nutrients for microbial growth [11,18,31,35,36]. Surface-accessible aminopeptidases work in concert with endoproteases to generate amino acids and peptides for nutrition and cell proliferation [36,37]. To our knowledge, this is the first report of an aminopeptidase that moonlights on the cell surface where it displays multifunctional binding properties, while retaining aminopeptidase activity. Recruitment of circulatory plasminogen onto bacterial cell surfaces by surface adhesins and its activation to plasmin underpins pathogenic processes used by many microbial agents [31]. Plasmin degrades tissue barriers by targeting heparan sulfate proteoglycans, laminin, type IV collagen and other key ECM components and is involved in the processing and subsequent activation of latent matrix metalloproteinases [31]. In light of our findings, we propose a model ( figure 12) where cell surface MHJ_0125 participates in the recruitment of plasminogen to the surface of M. hyopneumoniae and assists in its activation to plasmin. The multitude of surface-accessible ECM-binding proteins derived from the P97 and P102 paralogue families recruit plasminogen to the cell surface where it degrades the ECM on epithelial surfaces and in the basal lamina of the respiratory tract and, potentially, at distal tissue sites. These digestion products provide a rich source of substrates for surface-accessible aminopeptidases such as MHJ_0125. This strategy is unlikely to be restricted to M. hyopneumoniae and may be widespread among microbial pathogens. Our study underscores the importance of investigating other surfaceaccessible moonlighting proteases, and presents new opportunities for designing therapeutic compounds to control microbial pathogens (figure 12).  (c) Structure of the substrate-binding (S1) pocket of MHJ_0125 (blue) overlaid with the S1 pocket of PepA glutamyl aminopeptidase from S. pneumoniae. PepA contains an arginine residue (Arg 257 ) that confers a positive charge on one end of the S1 pocket that is believed to position adjacent to the carboxylate group of the glutamic acid substrate. Curiously, this residue is substituted for Asp 258 in MHJ_0125; however, it appears that the side chain does not protrude far enough into the S1 pocket to confer a negative charge at this position. Figure 10. Enlarged view of one of the four small exit pores of (a) MHJ_0125, (b) S. pneumoniae PepA and (c) P. horikoshii TET protease. Consistent with the substrate preference of these glutamyl aminopeptidases which would mainly export negatively charged glutamic acid residues, the opening of the small pore in (a) and (b) is dominated by positively charged residues. In contrast and consistent with predominantly leucyl aminopeptidase activity, the TET protease small pore is dominated by 3 hydrophobic phenylalanine residues. The intensely positively charged nature of the MHJ_0125 small pore is highlighted in a rendering of the dodecamer surface overlaid with its surface potential and rotated to look down the axis of the small pore (d, arrow).

Ethics statement
rsob.royalsocietypublishing.org Open Biol 3: 130017 (AEC project number M08/12) and were in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.  Helens, UK). Bestatin, EDTA, trypsin, tributylphosphine, acrylamide, ExtrAvidin peroxidase and 3,3-diaminobenzidine were obtained from Sigma (Sydney, Australia). Pre-cast gels, buffers and molecular weight markers as well as standard molecular biology reagents were purchased from Life Technologies (Australia) unless stated otherwise.

Mycoplasma culture conditions
Mycoplasma hyopneumoniae (strain J) was grown in modified Friis broth [38]. Cells were harvested by centrifugation at 10 000Âg and pellets were stored at 2208C until use as described previously [39].

High-resolution clear native PAGE and twodimensional high-resolution clear native/SDS-PAGE
Freshly harvested M. hyopneumoniae cells were washed extensively (more than three times) in PBS and pelleted by centrifugation (9000Âg for 10 mins at 48C). The resulting 0.1 g cell pellet was resuspended in 1 ml 1 Â native sample buffer supplemented with 1 per cent n-dodecyl b-D-Maltoside and 1 Â complete protease inhibitor cocktail (Roche) and solubilized for 30 min on ice. Insoluble material was removed by centrifugation at 48C and soluble proteins were desalted using a MicroBiospin column (BioRad). Samples were run on 4-16% NativePAGE gels at 48C using high-resolution clear native electrophoresis with cathode buffer supplemented with sodium deoxycholate only [40]. One-dimensional gels were fixed with 40 per cent methanol/10 per cent acetic acid for 30 min, stained with colloidal Coomassie blue overnight before destaining with 1 per cent acetic acid. Alternatively, lanes for denaturing twodimensional gel electrophoresis were excised from the gel and equilibrated in SDS equilibration solution (2% SDS, 6 M urea, 250 mM Tris-HCl pH 8.5, 0.0025% (w/v) bromophenol blue) for 20 min before the second dimension SDS-PAGE. The resulting gels were fixed and stained as above or stained with Flamingo fluorescent stain (BioRad) for 1-2 h and visualized using a Pharos FX-Plus gel scanner (BioRad).

Detection and isolation of Mycoplasma hyopneumoniae surface-expressed and -secreted proteins
Methods used to biotinylate and recover surface proteins labelled with biotin and generate tryptic peptides of surfaceexposed proteins (surface shaving) and characterize them by LC-MS/MS have been described previously [11,16,18]. Procedures used to detect surface-accessible proteins by graded trypsin hydrolysis and western blotting have also been described previously [17,41]. Secreted proteins were recovered from control shaving experiments. Freshly cultured M. hyopneumoniae cells were harvested by centrifugation, washed twice with PBS and resuspended in PBS pH 7.4 at 378C for 5 min (no added trypsin). Mycoplasma hyopneumoniae cells were removed by centrifugation and proteins released into the PBS during this 5 min incubation at 378C (secreted protein fraction) were retained in the supernatant. Their identities were determined by trypsin digestion followed by LC-MS/MS as described previously [11,18].

Proteomics
The preparation of peptides for mass spectrometry analysis has been described in detail previously [11,42]. Briefly, excised gel spots/bands were reduced and alkylated with 5 mM tributylphosphine and 20 mM acrylamide for 90 min at room temperature. The gel pieces were then digested with 12.5 ng ml 21 sequencing grade trypsin at 378C overnight and the resulting peptides were solubilized with 2 per cent formic acid (v/v) prior to analysis by LC-MS/MS using a

Site-directed mutagenesis
Since Mycoplasma species use the UGA codon to encode tryptophan rather than signalling the end of translation, the expression of Mycoplasma proteins in E. coli results in the production of truncated recombinants. Therefore, site-directed mutagenesis (by fusion PCR) was used to remove premature stop codons from the MHJ_0125 gene (Genbank accession AAZ44217.1). Primers extending in opposite directions were designed across the target mutagenesis site (MutF125: 5 0 -GA ATTTGGGGAAATGCGGTAATAG-3 0 and MutR125: 5 0 -TTC CCCAAATTCCGCCAATTG-3 0 ) with 12 base overlap (underlined) and including an A to G transition to convert a premature stop codon (TGA) into tryptophan (TGG). The two-step fusion PCR was done initially by pairing the mutagenesis primers with primers designed to anneal at the two boundaries of the target gene (F125: 5 0 -CACCATGTCAATAT TAGAAAAAATGAAAA-3 0 , R125: 5 0 -CTTTAATATCTGATA AACTAATTAGACCATAAG-3 0 ), designed following instructions in the Champion pET directional TOPO Expression kit. The two separate PCR products were then fused in a sevencycle PCR reaction, followed by extension of the fused products with boundary primers to generate the insert for the pET expression system. All PCR reactions were performed using the Platinum Pfx DNA polymerase proof-reading enzyme. The mutagenized MHJ_0125 template DNA was cloned into the pET100/D-TOPO expression vector (Life Technologies), and following DNA sequencing (Macquarie University Sequencing Facility), the construct was transformed into BL21 (DE3) E. coli for expression.

Expression and purification of recombinant MHJ_0125 in Escherichia coli
Soluble His-tagged rMHJ_0125 protein was purified by affinity chromatography using Profinity IMAC Ni-charged resin (BioRad) according to the manufacturer's instructions. Briefly, the cleared lysate was loaded onto a 1.2 ml (50% slurry) IMAC column and washed with six column volumes of wash buffer (300 mM KCl, 50 mM KH 2 PO 4 , 20 mM imidazole). The rMHJ_0125 protein was eluted from the column in the same buffer containing 250 mM imidazole, dialysed into PBS (pH 7.3) and stored in aliquots at 2208C until use.

Enzymatic analysis
The aminopeptidase activity of rMHJ_0125 was determined by measuring the release of the fluorogenic leaving group, AMC, from a range of fluorogenic peptide substrates representative of the various amino acid groupings as previously described [43]. Reactions were carried out in 96-well microtitre plates (200 ml total volume, 1 h, 378C) using a rsob.royalsocietypublishing.org Open Biol 3: 130017 spectrofluorometer (Bio-Tek KC4) with excitation at 360 nm and emission at 460 nm. Generally, purified recombinant enzyme (30 nM), or cultured M. hyopneumoniae cells (1 Â 10 5 cells/well), were incubated in 50 mM Tris-HCl ( pH 7.5) containing 5 mM CoCl 2 for 20 min before the addition of 50 mM substrate. Assays were also performed in a range of 50 mM buffers: sodium acetate ( pH 4.5-5.5), sodium phosphate (pH 5.5-8.0) and sodium borate ( pH 8.0-9.0), and in the presence of a range of 5 mM metal chlorides (CoCl 2 , CuCl 2 , MnCl 2 , MgCl 2 and ZnCl 2 ). The susceptibility of MHJ_0125 to the aminopeptidase inhibitor bestatin and to the metal ion chelator EDTA was determined by performing assays against H-Glu-AMC following pre-incubation (10 min at 378C) with each over the range 0.01-1 mM.

Porcine plasminogen purification
Plasminogen was isolated from porcine plasma as described previously [12,13]. The isolated protein was confirmed as plasminogen following LC-MS/MS analysis. Biotinylation of plasminogen for ligand blot analysis was performed using Sulfo-NHS-LC-Biotin (sulfosuccinimidobiotin) according to the manufacturer's instructions. To remove excess biotin the biotinylated plasminogen was dialysed against PBS at 48C.

Porcine plasminogen activation assay
The approach used to test whether rMHJ_0125 was able to enhance activation of plasminogen is described in detail elsewhere [4]. Porcine plasminogen (50 mg) was pre-incubated for 1 h at 378C with 1 : 1, 1 : 4 and 1 : 8 molar ratios of MHJ_0125 in microtitre plate wells (Greiener Bio One, Frickenhausen, Germany) for these assays.

Western and ligand blot analysis
Blots of lysates from M. hyopneumoniae cells exposed to different concentrations of trypsin in the range 0-500 mg ml 21 were separately probed with rabbit anti-P97 N-terminal serum (1 : 100) as described previously [15,17], and rabbit anti-MHJ_0125 serum (1 : 100) followed by goat anti-rabbit HRP conjugate (1 : 1000). In control experiments, blots probed only with goat anti-rabbit conjugate (1 : 1000) did not produce any detectable bands. The binding of rMHJ_0125 to porcine plasminogen was investigated using ligand blotting. Serially diluted rMHJ_0125 samples were run on 4-12% Criterion Bis-Tris gels and transferred to PVDF membranes at 300 mA for 30 min. Blots were blocked with 5 per cent non-fat dry milk in PBS Tween 20 (0.1% v/v). The membranes were then incubated with a 1 : 250 dilution of biotinylated porcine plasminogen for 1 h at room temperature followed by ExtraAvidin peroxidase (1 : 3000) for 1 h. To determine the immunogenicity of MHJ_0125, blots were probed with a pool of convalescent pig sera (1 : 50) collected during M. hyopneumoniae infection and detected with peroxidase-conjugated anti-pig IgG (diluted 1 : 5000). Both ligand and Western blots were visualized by the addition of 3,3-diaminobenzidine peroxidase substrate.

Generation of rabbit sera against MHJ_0125
Monospecific, polyclonal antibodies against rMHJ_0125 were generated in New Zealand White rabbits using a protocol described previously [41]. All experiments involving animals were conducted with animal ethics approval.

Fluorescence microscopy
Glass coverslips (13 mm, no. 1.5 thickness; Gerhard Menzel GmbH, Braunschweig, Germany) were coated in 0.01% poly-L-lysine for 10 min, dried at 558C for 1 h and placed in a 24well microtitre plate. One ml of M. hyopneumoniae strain J culture was centrifuged (10 000Âg for 10 min) and washed three times with 1 ml sterile PBS. Washed cells (200 ml) were added to wells containing glass coverslips and allowed to adhere at 378C for 30 min. Wells were washed once with PBS and fixed in 4 per cent paraformaldehyde at 48C for 30 min.
To permeabilize cells, 0.1 per cent Triton X-100 in PBS was added for 5 min and washed three times with PBS. To quench excess aldehydes, 100 mM glycine in PBS was added for 5 min and washed three times with PBS. Non-specific binding sites were blocked in 2 per cent bovine serum albumin (BSA) in PBS overnight at 48C and washed three times in PBS. Rabbit MHJ_0125 antisera and control rabbit antisera were each diluted 1 : 100 with PBS containing 1 per cent BSA, incubated for 1 h at room temperature and washed three times in PBS. A 1 : 1000 dilution of goat anti-rabbit antibodies conjugated to Alexa Fluor 488 (Life Technologies) was prepared in PBS containing 1 per cent BSA and incubated for 1 h at room temperature and washed three times in PBS. Coverslips were mounted onto microscope slides in 2 ml of VECTASHIELD, sealed using nail varnish and imaged on an Olympus BX51 Upright Epi Fluorescence microscope. Images were captured by an Olympus DP70 Digital Microscope Camera using Olympus DP Controller software with manual intensity exposure settings at 9-10. These same slides were imaged using super-resolution 3D-SIM using a DeltaVision OMX Imaging System (Applied Precision Inc., Issaquah, WA, USA) as previously described [44].

Size exclusion chromatography
Gel filtration chromatography was carried out by loading 10 ml of purified rMHJ_0125 protein (10 mg) onto a Superdex 200 column (3.2 mm ID Â 300 mm) equilibrated with Tris -HCl (20 mM, pH 8.0). Chromatography was performed with a flow rate of 50 ml min 21 using a Waters Alliance 2690 HPLC system, controlled by MASSLYNX v. 4.0. Eluted proteins were detected with a Model 990 PDA UV detector by monitoring absorbance at 220 and 280 nm.

Heparin column chromatography
Mycoplasma hyopneumoniae cells were lysed in 10 mM phosphate buffer by sonication on ice for 3 Â 15 s bursts at 50 per cent power and insoluble material was pelleted by centrifugation. A 1 ml HiTrap heparin column was equilibrated with 10 mM phosphate buffer and the sample was loaded at 0.5 ml min 21 over 10 min, then washed with 10 mM phosphate buffer for a further 10 min to remove non-specific binders before running an increasing linear gradient to 1 M NaCl over 25 min then to 2 M NaCl over 10 min and maintained for 5 min to elute any remaining protein from the column before returning to 10 mM phosphate buffer over 5 min and re-equilibrating. The chromatographic profile (A 280 ) indicated all non-heparin-binding proteins were collected in the first 15 min. Heparin-binding proteins started to elute at approximately 35 min at a NaCl concentration of about 300 mM. Elutions were collected over 20 min and pooled for analysis. SDS-PAGE was also used to monitor the elution profile of proteins from the heparin column. Pooled elutions were analysed by native PAGE to determine the presence of complexes. Heparin-binding proteins, and eluted complexes, were identified by LC-MS/MS using standard procedures [15].

Bioinformatics and phylogenetic analysis
MHJ_0125 orthologues in other Mycoplasma strains and bacterial species, as well as related sequences from metazoans, were identified following BLASTp analyses of the UniProt Knowledgebase (http://web.expasy.org/blast/) using the MHJ_0125 primary sequence as search query. Selected sequences were aligned using CLUSTAL W and ESPRIPT [45,46]. Phylogenetic trees were created using 33 Mycoplasma and bacterial amino acid sequences and six metazoan sequences. The protein sequences were initially aligned using CLUSTAL W [45] and the trees were created using the bootstrapped (1000 trials) neighbour-joining method of MEGA v. 4.0 [47], using the Kimura two-parameter model with uniform rates for all sites. The UniProt accession numbers of the sequences used for alignment and phylogenetic analyses can be found in electronic supplementary material, figure S5.

Modelling of MHJ_0125
The three-dimensional structure of MHJ_0125 was reconstructed using comparative modelling. BLAST analysis was used to identify a number of related structures within the PDB, which were considered as potential templates, and the program MODELLER [48] was used to select the most suitable template for construction of the model. The align2d function in MODELLER was used to align MHJ_0125 to TET protease from P. horikoshii (PDB ID: 1Y0R) according to secondary structure. For comparative modelling, the P. horikoshii TET protease dodecamer was assembled using the BioMT figures provided in the PDB file. MODELLER was then used to generate the 3-dimensional structure of the MHJ_0125 using the TET protease dodecamer as a template. Pyrococcus horikoshii TET protease residues V115-K127 were not determined during crystallography owing to the fact that these residues are buried, therefore the corresponding residues (T116-K128) were excluded from the MHJ_0125 dodecameric model. The quality of the model was checked with ProSA [49,50]. The final structure of MHJ_0125 was rendered using the PYMOL Molecular Graphics System, v. 1.5.0.4, Schrödinger, LLC. To determine the electrostatic potential of the MHJ_0125 surface, Poisson Boltzmann electrostatics calculations were performed using PDB2PQR [51,52] and were rendered in PYMOL using the ADAPTIVE POISSON Boltzmann software plugin [53].

Acknowledgements
We are grateful for the assistance of G. J Eamens for the preparation of anti-MHJ_0125 serum in rabbits. This work was supported by a UTS start up grant to S.P.D. J.L.T. and B.B.R. are recipients of PhD scholarships from UTS. C.B.W. was supported by an Australian NHMRC Senior Research Fellowship (571905).