Mycobacteriophages encounter a unique problem among phages of Gram-positive bacteria, in that lysis must not only degrade the peptidoglycan layer but also circumvent a mycolic acid-rich outer membrane covalently attached to the arabinogalactan–peptidoglycan complex. Mycobacteriophages accomplish this by producing two lysis enzymes, Lysin A (LysA) that hydrolyses peptidoglycan, and Lysin B (LysB), a novel mycolylarabinogalactan esterase, that cleaves the mycolylarabinogalactan bond to release free mycolic acids. The D29 LysB structure shows an α/β hydrolase organization with a catalytic triad common to cutinases, but which contains an additional four-helix domain implicated in the binding of lipid substrates. Whereas LysA is essential for mycobacterial lysis, a Giles ΔlysB mutant mycobacteriophage is viable, but defective in the normal timing, progression and completion of host cell lysis. We propose that LysB facilitates lysis by compromising the integrity of the mycobacterial outer membrane linkage to the arabinogalactan–peptidoglycan layer.
Upon completion of lytic bacteriophage growth the integrity of the host cell wall must be compromised to release progeny phage particles (Wang et al., 2000; Young et al., 2000). Double-stranded DNA (dsDNA)-tailed phages typically encode an endolysin that degrades the peptidoglycan mesh, together with a means of allowing this enzyme access to its substrate, either via a membrane-localized holin, or through the action of holin-independent SAR (signal-anchor-release) endolysins (Wang et al., 2000; Young et al., 2000; Xu et al., 2005). For phages infecting Gram-positive hosts, not only are these two requirements sufficient for cell lysis (Loessner et al., 1998), but exogenously added endolysin can efficiently kill host cells, with considerable therapeutic potential (Loeffler et al., 2001; Schuch et al., 2002; Fischetti, 2008). In Gram-negative hosts, the outer membrane presents a further barrier to lysis, and phages typically encode an additional set of lysis proteins (Rz/Rz1 proteins or spanin equivalents) that are proposed to complete lysis by fusing the inner and outer membranes (Summer et al., 2007; Berry et al., 2008).
Although the mycobacteria are members of the Gram-positive Actinomycetales, they are unusual in possessing a mycolic acid-rich outer membrane that is covalently attached to the arabinogalactan–peptidoglycan complex (Hoffmann et al., 2008; Zuber et al., 2008). Mycolic acids are α-alkyl, β-hydroxy C60−90 fatty acids with a relatively short (C20−25) saturated α and a longer (C60) meromycolate chain (the β-hydroxy branch) containing double bonds, cyclopropane rings and oxygenated groups, depending on the Mycobacterium species (Hamid et al., 1993; Watanabe et al., 2001). The outer membrane mycolic acids are linked to the cell wall via an ester linkage to the terminal pentaarabinofuranosyl components of arabinogalactan (McNeil et al., 1991; Brennan and Nikaido, 1995; Hoffmann et al., 2008; Niederweis, 2008), which in turn is covalently anchored to peptidoglycan (McNeil et al., 1990; Brennan and Nikaido, 1995). In addition to being part of the mycobacterial outer membrane, mycolic acids are constituents of trehalose dimycolate (TDM, also known as cord factor), a secreted molecule important for mycobacterial pathogenesis (Brennan, 2003). Mycolic acid synthesis is essential for viability and is the target of the major antituberculosis drug isoniazid (Vilcheze and Jacobs, 2007). The mycobacterial outer membrane also plays roles in nutrient acquisition (Niederweis, 2008) and as an immune target (Karakousis et al., 2004). Furthermore, it presents a potential barrier to phage-mediated lysis.
We show here that LysB of mycobacteriophage D29 is a novel mycolylarabinogalactan esterase that cleaves the ester linkage joining the mycolic acid-rich outer membrane to arabinogalactan, releasing free mycolic acids. We propose that LysB acts at a late stage in lysis, severing the connection of the mycobacterial outer membrane to the cell wall and completing lysis of the host.
Organization, conservation and location of mycobacteriophage lysis cassettes
Comparative analysis of mycobacteriophage genomes reveals them to be highly diverse with mosaic architectures (Pedulla et al., 2003; Hatfull et al., 2006). The 6858 predicted protein products can be assorted into 1523 phamilies (Phams) of related sequences and their distributions among the mycobacteriophages analysed (Hatfull et al., 2006) (our unpublished observations). While there are few Phams represented in all 60 completely sequenced genomes, one of these is the lysA phamily (Pham 66), and the LysA proteins are predicted from sequence comparisons to have peptidoglycan hydrolysing activity (Garcia et al., 2002; Hatfull et al., 2006; Marinelli et al., 2008). We have confirmed this for three LysA proteins (Corndog gp69, Bxz1 gp236, Che8 gp32) all of which catalyse peptidoglycan hydrolysis in zymograms (Fig. 1A). The lysB gene is also implicated in lysis, primarily because of its linkage to lysA (Fig. 1B) and the demonstration of lipolytic activity by Ms6 LysB (Gil et al., 2008). LysB homologues are present in 56 of the 60 completely sequenced mycobacteriophage genomes and are located downstream of lysA and separated from it by no more than four intervening genes (Fig. 1B). Some of the intervening genes encode putative holins and exhibit holin-like function (e.g. D29 gene 11, data not shown), whereas others (e.g. Omega gene 51) code for putative homing endonuclease HNH motifs (Fig. 1B), or have homologues elsewhere in other mycobacteriophage genomes (e.g. Troll4 gene 37).
Although lysA and lysB are closely linked, this linkage is not a simple consequence of synteny in the context of the broader genome organizations, because these presumptive lysis cassettes are situated in different chromosomal locations (Fig. 1C). Of the 53 phages with a siphoviral morphotype, the 13 genomes constituting Cluster A [including L5, D29, Bxb1 (Hatfull and Sarkis, 1993; Ford et al., 1998; Mediavilla et al., 2000) and Che12 (Hatfull et al., 2006)] have the lysis cassette situated to the left of the virion structural and assembly genes, whereas all of the other genomes have the lysis cassette positioned to the right of the structural genes (Fig. 1C). The invariable linkage of lysA and lysB regardless of genome location further supports a role for LysB in lysis.
Sequence alignment of the phamily of LysB proteins shows that they are highly diverse, and only three residues are completely conserved (Fig. S1). Although the proteins vary in length [from 254 (D29 gp12) to 451 (PG1 gp50) residues] and there are many gaps throughout the alignment, these proteins do not have modular constructions as seen in the LysA proteins (Hatfull et al., 2006). However, the sequence divergence is sufficiently high that some members of the phamily have little identifiable sequence similarity (< 20% identity) to each other (Figs S1 and S2). In a search for conserved domains, 14 LysB proteins (including D29 gp12) are predicted to contain a cutinase domain (pfam01083), and Wildcat gp52 has an esterase/lipase domain (pfam c109107). Eight proteins have a predicted peptidoglycan-binding domain (pfam01471) at their N-terminus, although this region is not present in D29 LysB (Fig. S1).
The three absolutely conserved residues include the putative active site serine (position 82 in D29 LysB) and a GXP motif located ∼40 residues downstream (residues 117–119 in D29 LysB; Fig. S1). The cutinase- and lipase-like conserved domains suggests that these enzymes are serine esterases, which typically have a Ser-Asp-His catalytic triad. Of three aspartic acids and a histidine residue previously implicated in catalysis through their conservation (Gil et al., 2008), only the aspartic acid corresponding to position 166 in D29 LysB is highly conserved among the larger group of proteins, with the single departure being the LysB protein encoded by a putative Mycobacterium avium prophage that has a glutamic acid residue at that position (Fig. S1). The alignment does not reveal a well-conserved candidate for the histidine component of the catalytic triad (Fig. S1). The GXP motif is not absolutely conserved in all serine esterases and its role in LysB functions is not clear.
D29 LysB has lipolytic activity
To further characterize the structure and function of mycobacteriophage LysB proteins, several Pham73 members were cloned and expressed. Although expression and solubility varies among these, we found that the 254-residue D29 gp12 (LysB) expressed well and was readily purified to near homogeneity and high solubility (> 10 mg ml−1) (Fig. 2A). D29 LysB shares only 40% amino acid sequence identity with the previously characterized Ms6 LysB protein (Gil et al., 2008), and lacks 90 N-terminal residues present in Ms6 LysB (Fig. S1). We also constructed, expressed and purified a mutant version of the protein with a substitution of alanine for the putative catalytic serine at position 82 (S82A).
To test for lipolytic activity of D29 LysB we measured hydrolysis of p-nitrophenyl butyrate (pNPB) to release p-nitrophenol (Gilham and Lehner, 2005) (Fig. 2B). We observed a specific activity of 0.72 U mg−1 (Fig. 2B), somewhat higher than the 0.12 U mg−1 reported for Ms6 LysB protein (Gil et al., 2008), or for any of the seven cutinase-like proteins found in Mycobacterium tuberculosis (West et al., 2009). The S82A mutant is inactive, consistent with this residue being part of the catalytic triad (Fig. 2B). We also tested D29 LysB, as well as the purified lipase from Pseudomonas fluorescens for activity on p-nitrophenyl substrates with different carbon chain lengths. While D29 LysB was more active than the lipase on pNPB, we also observed decreasing activity of D29 LysB with longer substrates (Fig. 2B), as was reported for Ms6 LysB (Gil et al., 2008). D29 LysB thus shares lipolytic activity with Ms6 LysB (Gil et al., 2008), with serine-82 providing critical catalytic functions.
Crystal structure of D29 LysB
The structure of D29 LysB was determined at 2.0 Å resolution in crystals of the P 43 21 2 space group containing a single molecule in the asymmetric unit (Table 1). Structural alignment with known protein structures using Dali showed that D29 LysB is similar to members of the α/β hydrolase family, which includes cutinases, acetyl xylan esterases and lipases (Masaki et al., 2005; Holm et al., 2008). The five closest structural relatives of D29 LysB are all members of the cutinase family (Table 2), although they share no more than 21% amino acid sequence identity with D29 LysB.
Table 1. Crystallography data collection and refinement statistics.
The native D29 LysB data set was collected at APS 23ID and processed with HKL2000.
Refinement was done with Phenix.
Accession code: The atomic co-ordinate and structure factor have been deposited in the PDB with the accession code 3HC7 for D29 LysB.
Table 2. Top 5 results of structural alignment of D29 LysB in Dali database.a
The redundant hits of the same protein have been removed.
Acetyl xylan esterase
Acetyl xylan esterase
Of the 253 determined residues in D29 LysB, 170 residues (Ser2-Tyr161, Arg244-Gln253) are distributed in a doubly wound α/β sandwich, having a central parallel β-sheet flanked by two parallel α-helices on each side (Fig. 3A). This fold is typical of the α/β hydrolase family and the structure of LysB is remarkably similar to that of a Cryptococcus cutinase-like protein (Fig. 3A). Despite their low sequence similarity, the rms differences in c-alpha positions are 2.45Å and a search using the Dali server gave a high Z-score (20.3) when LysB was submitted as the query. The remainder of the protein (Ala162-Asn243) forms an 81-residue linker domain composed of four α-helices connecting the C-terminal α-helix (Arg244-Gln253) back to the side of the central β-sheet. It occupies the general location where the acyl chain is positioned in cocrystals of the Fusarium solani cutinase with a covalently linked n-undecanylphosphonate methyl ester inhibitor (Longhi et al., 1996). Given the high B-factor and poor electron-density of the Arg231-Lys237 loop connecting the helical-linker and the C-terminal helix, it is reasonable to propose that movement of the helical-linker helps to modulate the active site to accommodate bulky and hydrophobic acyl chains. We note that the unusual glycine-rich segment corresponding to a 55-amino-acid segment of PG1 LysB (and related proteins) inserted between D29 LysB residues 199 and 200, of which 19 are glycine residues, is located within this linker between the second and third helices (Fig. 3A; Fig. S1).
The catalytic triad Ser82-Asp166-His240 is located at the edge of the central β-sheet between the α/β sandwich and the linker domain. The position of the D29 LysB catalytic triad is very similar to those in other members of the α/β hydrolase family as can be seen in the superimposition of D29 LysB on the Cryptococcus cutinase-like protein (Fig. 3B). Although Ser-82 is invariant and Asp-166 is very strongly conserved in the alignment of LysB proteins (Fig. S1), the sequence alignment around His-240 varies among the LysB members, reflecting poor conservation of the flanking residues and positioning of the catalytic histidine within a loop region (Fig. S1). In contrast, the GXP motif – of which the glycine and proline residues are invariant – lies immediately adjacent to the catalytic triad. In D29 LysB, the Gly117-Asn118-Pro119 segment defines the end of β4 strand and makes a 180 degree turn just beneath Asp166. Besides the turn-makers Gly117 and Pro119, Asn118 forms hydrogen bonds with Met120, Arg121 and Asp160, which may be energetically favourable to the scaffold.
D29 LysB is a mycolylarabinogalactan esterase
The D29 LysB structure confirms its enzymatic function as a serine esterase, and while it clearly has lipolytic activity, the unusual linker positioned adjacent to the active site – together with the inclusion of lysB genes within lysis cassettes – suggests that its substrate could be a cell wall component containing ester-linked lipids. We reasoned that an excellent candidate substrate is mycolylarabinogalactan–peptidoglycan (mAGP), such that LysB hydrolysis would release free mycolic acids and separate the outer membrane from the arabinogalactan–peptidoglycan layer. To test this we isolated a cell wall fraction from Mycobacterium smegmatis that is highly enriched in mAGP (Besra, 1998) and determined whether it is a substrate for D29 LysB hydrolysis (Fig. 4A and B). We observed both time- and enzyme concentration-dependent release of lipids that migrate similarly to mycolic acids released from the same mAGP preparation by alkaline hydrolysis of mAGP with tetrabutylammonium (TBAH) (Hamid et al., 1993; Besra, 1998; Watanabe et al., 2001) (Fig. 4A and B). To further characterize the lipid products we methyl-esterified them with iodomethane, which yielded methyl esters of α, α′ and epoxy mycolates similar to those from TBAH treatment and specific to M. smegmatis (Fig. 4C). Preliminary analysis of the released lipids by mass spectrometry and nuclear magnetic resonance was consistent with them being free mycolic acids (data not shown). To our knowledge, this is the first mycolylarabinogalactan esterase to be described. As expected, hydrolysis of mAGP by D29 LysB is dependent on the catalytic serine residue at position 82 and the S82A mutant enzyme exhibits no observable activity (Fig. 4D). Moreover, mAGP was not a substrate for a Pseudomonas-derived lipase (Arpigny and Jaeger, 1999; Gupta et al., 2004) as no release of free mycolic acids was observed (Fig. 4D).
Giles lysB facilitates completion of host cell lysis
To explore the role of LysB in mycobacteriophage lytic growth, we first asked whether it is an essential function for plaque formation, using a recently described recombineering strategy to delete lysB (gene 32) from the mycobacteriophage Giles genome (Marinelli et al., 2008). We targeted Giles because it is a demonstrated substrate for recombineering (Marinelli et al., 2008), we know that Giles lysA is essential (Marinelli et al., 2008), and we anticipated needing to complement with a lysB gene that is sufficiently different genetically as to avoid recombination (i.e. D29 gene 12). It should be noted, however, that Giles LysB – cloned and purified in the same manner as D29 LysB – was also able to hydrolyse mAGP (data not shown). A 200 bp substrate containing 100 bp sequences flanking lysB was designed to introduce a 1146 bp internal deletion in Giles lysB, fusing 15 codons at the 5′ and 3′ ends of the gene to minimize effects on expression of adjacent genes as well as avoiding genetic polarity (Fig. 5A). Following co-electroporation of the 200 bp deletion substrate and Giles genomic DNA into an M. smegmatis recombineering strain, plaques were recovered on a lawn of M. smegmatis mc2155 pKMC2 cells expressing D29 LysB. Of 22 primary plaques screened by deletion amplification detection assay (DADA)-PCR (Marinelli et al., 2008), two contained mixtures of wild-type and mutant phages (Fig. 5B).
To test for viability of the Giles ΔlysB mutant, a mixed plaque was picked, resuspended in buffer, and plated to recover ∼600 isolated plaques on both a putative complementing strain (M. smegmatis mc2155 pKMC2) or a non-complementing control strain. Secondary lysates representing all of the recovered particles were harvested and tested by PCR for the presence of the deletion mutant (data not shown). The mutant genotype was present in approximately equivalent proportions in both the complementing and non-complementing strains, suggesting that the mutant is viable. Thirty individual plaques from a secondary plating were then screened by PCR and a homogenous mutant plaque identified (Fig. 5C). The Giles ΔlysB mutant was plaque purified, and the structure of the deletion confirmed by DNA sequencing (data not shown).
The Giles ΔlysB mutant forms plaques at equivalent efficiencies on complementing and non-complementing strains, and titers of wild-type and mutant lysates prepared on a wild-type M. smegmatis host under standard conditions are similar; Giles lysB thus is not essential for lytic growth (Fig. 6A). However, we observed that the mutant forms somewhat smaller plaques on lawns of wild-type M. smegmatis compared with the parental Giles, a phenotype that is exaggerated when higher densities of plating cells are used (conditions that favour fewer bacterial doublings and fewer rounds of phage infection) (Fig. 6A). This difference is directly attributable to the loss of lysB because complementation with a plasmid expressing D29 lysB restores plaque size to near that of the wild-type parent (Fig. 6A). To eliminate the possibility that plaque size restoration results from acquisition of lysB by recombination with the complementing plasmid, several plaques were picked and shown to reproduce the mutant phenotype, and by PCR contain only the mutant genotype (data not shown). Furthermore, at a relatively high cell plating density (2 × 108 cfu/plate) the average number of particles in each plaque is ∼100-fold reduced in ΔlysB mutant relative to wild-type Giles plaques (5 × 105 and 4 × 107 pfu ml−1 respectively), consistent with a lysis defect. Therefore, unlike lysA, lysB is not essential for plaque formation but is required for efficient phage release.
To test whether the small plaque phenotype specifically results from changes in the pattern or timing of lysis, we monitored optical densities (OD) of infected M. smegmatis cultures (Fig. 6B). After infection with wild-type Giles the OD increases for approximately 3 h, after which it steadily declines as the cells lyse, up to a period approximately 5 h after infection. When infected with the ΔlysB mutant, the OD does not begin to decline until 3.5 h and is incomplete even 5.5 h after infection (Fig. 6B). In contrast, cells infected with a ΔlysA mutant – which does not form visible plaques on a non-complementing M. smegmatis strain (Marinelli et al., 2008) – cease to grow after 3–3.5 h, with only a modest reduction in OD thereafter (Fig. 6B).
We also monitored the progression of phage infections by measuring ATP release (Fig. 6C), which we anticipated to reflect either lysis or permeabilization of the cell wall. In a wild-type Giles infection, little or no ATP release is seen until 3 h after infection, followed by a steady increase to 4.5–5 h after infection (Fig. 6C), closely mirroring the changes in OD shown in Fig. 6B. The ΔlysB mutant is delayed in the onset of ATP release by about 30 min, and fails to achieve the wild-type level even 5.5 h after infection, which taken together with the OD changes is consistent with a lysis defect. The phenotype is reminiscent of the conditional lysis defect of phage lambda Rz or Rz1 mutants, where the Rz/Rz1 proteins are implicated in the final lysis step in which the E. coli outer membrane is compromised by fusion with the cytoplasmic membrane (Summer et al., 2007; Berry et al., 2008). Interestingly, the ΔlysA mutant shows no defect in ATP release at all, and may even release more ATP than cells infected with wild-type Giles (Fig. 6C).
Finally, we tested whether the ΔlysB mutant produces the same yield of total phage particles as wild-type Giles in a lytic infection, and determined how the particles distribute between those that are released and those that remain trapped in unlysed cells (Fig. 6D and E). As shown in Fig. 6D, both the ΔlysA and ΔlysB mutants show no major defect in the production of phage particles. However, by 4 h after infection, although > 90% of wild-type particles are present in the culture supernatant, about 45% of the ΔlysB particles remain associated with unlysed cells. In contrast, < 10% of ΔlysA particles are released into the supernatant even 5 h after infection (Fig. 6E). These observations are consistent with a strong defect in cell lysis in the ΔlysA mutant, and a milder defect in the ΔlysB mutant.
We have shown here that the mycobacteriophage D29 LysB protein is a novel mycolylarabinogalactan esterase required for completion of lysis of host mycobacterial cells, and a model for the role of the LysA and LysB proteins as well as the reaction catalysed by LysB is presented in Fig. 7. The mycolic acid-rich mycobacterial outer membrane presents an unusual problem for phages of Gram-positive bacteria, which typically complete lysis through simple endolysin-mediated degradation of the peptidoglycan layer. The mycolylarabinogalactan linkage is not common among bacteria, because mycolic acids are found primarily in the Corynebacterineae suborder of the Actinomycetales, which includes Corynebacteria, Gordonia, Nocardia, Rhodoccocci and Mycobacteria. Few phages infecting non-mycobacterial members of the Corynebacterineae have been characterized, although these would be good candidates for also encoding mycolylarabinogalactan esterases. We note that neither phage P2101 of Corynebacterium glutamicum (Chen et al., 2008) nor BFK20 of Brevibacterium flavum (Bukovska et al., 2006) encodes a LysB relative, although unlike the Mycobacteria, mycolic acids are dispensable for growth of C. glutamicum (Portevin et al., 2004) and therefore may not pose a barrier to efficient lysis.
While removal of LysB function in mycobacteriophage Giles results in an apparently mild plaque phenotype that can be rescued by expression of D29 LysB, the problems in lysis are made more apparent by measuring changes in OD of the culture, ATP release and phage particle release. The ATP release by the Giles ΔlysA mutant is in contrast to its lysis defect as reflected in topical density measurements. A simple explanation for this observation is that the peptidoglycan layer – which most probably remains intact in the ΔlysA mutant infection – provides no barrier to ATP release at all, whereas both the cytoplasmic membrane and the mycobacterial outer membrane must be compromised for complete ATP release. Presumably, the cytoplasmic membrane is permeabilized through the action of a Giles holin (which has yet to be identified), and the differences observed between the ΔlysA and ΔlysB mutants are consistent with an attached mycobacterial outer membrane providing a significant barrier to ATP release. While it is difficult to eliminate the possibility of secondary effects of the deletion mutation in the ΔlysB mutant on expression of other phage genes (including a putative holin), the complementation studies (Fig. 6A) are consistent with loss of LysB function as the primary cause of the phenotypes observed.
At this time it is unknown how LysB is localized to its substrate, and no signal peptide motifs have been identified (Gil et al., 2008). It would be interesting to see if LysB activity is dependent on a mycobacteriophage holin, similar to endolysins of other bacteriophage and presumably the LysA proteins of mycobacteriophage. However, few mycobacteriophage holins have been predicted bioinformatically, and even fewer functional studies have been conducted; Ms6 gp4, a predicted holin, was lethal when overexpressed in E. coli (Garcia et al., 2002), and experiments in our lab indicate a potential holin function for D29 gp11 (L. Marinelli and G.F. Hatfull, unpubl. obs.).
The question arises as to whether phages Che12, Rosebush, Qyrzula and Myrna that lack lysB encode alternative enzymes performing analogous functions. These phages do not form a closely related group (Fig. S2), although comparison of each with their closer phage relatives is informative. For example, in Che12 the surrounding genes are syntenic with those in phages L5, D29, Pukovnik and Bxz2 and it appears as though Che12 has simply lost lysB (Fig. S3A). The organizations in Rosebush and Qyrzula depart from their close relatives in the inclusion of a gene immediately downstream of lysA that is lacking in other Cluster B phages, all of which do encode LysB (Fig. S3B). This predicted protein is small (178 aa) and is a distant relative of the D29 putative holin (gp11). In Myrna, neither of the two orfs (244, 245) immediately downstream of lysA are related to any other mycobacteriophage proteins (Fig. S3C). Myrna gp244 does have similarity to the N-terminal segment of Rhodococcus protein (RHA1_ro08121) that contains both peptidase and muramidase motifs in its C-terminus, and Myrna gp244 may play a yet undefined role in lysis. We also note that although Proprionobacterium acnes phage PA6 (Farrar et al., 2007) and Streptomyces phage R4 (our unpublished data) encode proteins with sequence similarity to mycobacteriophage LysA proteins, neither encodes a LysB relative. LysB thus appears to be restricted to those phages infecting hosts with mycolic acid outer membranes.
The release of free mycolic acids from LysB-treated cell walls strongly suggests that mAGP is the substrate for the enzyme. The cell wall preparations may contain additional lipid components, but other predominant mycolic acid-containing constituents – such as TDM (cord factor) and trehalose monomycolate – are not cell wall linked and are not major components of the mAGP preparations (Brennan and Nikaido, 1995; Besra, 1998). Preliminary data suggest that D29 LysB can also hydrolyse TDM (A. Ojha, K. Payne and G.F. Hatfull, unpubl. obs.), but it seems unlikely that this has physiological relevance for lysis because TDM is not covalently attached to mycobacterial cells. We propose that cleavage of the mycobacterial outer membrane from the peptidoglycan–arabinogalactan layer is the primary role of LysB.
The only known physiological circumstances in which free mycolic acids are released from mycobacterial cells is during maturation of M. tuberculosis biofilms (Ojha et al., 2008), although the source of biofilm mycolic acids is more likely to be TDM than mAGP (A. Ojha and G.F. Hatfull, unpubl. obs.). While seven cutinase-like proteins (culps) encoded by M. tuberculosis have been expressed and characterized (West et al., 2009), they are not evidently related to LysB and have not yet been tested for mAGP hydrolysis. Culp1 and Culp4 have optimal enzyme activity on short (butyrate) p-nitrophenyl substrates, and only Culp6 has significant activity on longer carbon chain substrates (West et al., 2009). Also, the Culp4 homologue in M. smegmatis has been shown to have phospholipase A activity, which is not found in cutinases (Parker et al., 2007). It remains to be seen whether there are any host-encoded enzymes that share mycolylarabinogalactan esterase activity with LysB.
The D29 LysB structure shows it is a member of the α/β hydrolase family with an α/β sandwich and an active site containing Ser, Asp and His. The five closest related structures are all cutinases, although there is no greater than 21% amino acid sequence identity with any of them. The LysB catalytic mechanism is expected to be similar to that for other serine esterases, although the strong conservation of the GXP motif and its juxtaposition to the active site may play a critical role in its stabilization. However, the D29 LysB also contains an unusual four-helix domain that projects from close to the active site and is lacking in other cutinase-like protein structures. While the specific role of the large linker is not known, its absence in all other cutinase-like proteins and lipases and its proximity to the active site would suggest a role in binding of the D29 LysB substrate. This linker contributes to the generally larger size of LysB proteins (254–451 residues) compared with other cutinases (∼225 residues) (Carvalho et al., 1999; Longhi and Cambillau, 1999).
Acquisition of lysB by mycobacteriophages throughout their evolution likely confers a substantial selective advantage over those without it by providing faster and more complete lysis. While considered more similar to Gram-positive bacteria, the existence of the mycobacterial outer membrane composed of mycolic acids and free lipids (Fig. 7A) presents a second barrier analogous to the outer membrane of Gram-negative bacteria (Hoffmann et al., 2008; Zuber et al., 2008). Phages that infect Gram-negative bacteria have the help of spanin proteins and their Rz/Rz1 counterparts (Summer et al., 2007; Berry et al., 2008; Krupovic et al., 2008) to span the periplasmic space and link the inner and outer membrane, enhancing the completion of lysis. We propose that mycobacteriophage have developed an alternative solution to compromise the mycobacterial outer membrane by severing its linkage to the underlying arabinogalactan–peptidoglycan layer.
Exogenously applied phage-encoded endolysins have been shown to have effective antimicrobial activity against Gram-positive bacterial pathogens including Streptococcus pneumoniae and Bacillus anthracis (Fischetti, 2008). They are, however, ineffective against Gram-negative bacteria because the outer membrane blocks access to the peptidoglycan targets. The mycobacteria are likely to be similarly intractable to exogenously added endolysins because of their mycolic acid-rich outer membrane. While we do not yet know whether mycobacteriophage LysB proteins can reach their mycolylarabinogalactan targets from the outside, this possibility raises the intriguing idea that mycobacterial pathogens such as M. tuberculosis may be rendered susceptible to endolysin treatment through co-treatment with LysA and LysB proteins.
Bacterial strains and growth
Mycobacterium smegmatis mc2155 (Snapper et al., 1990) and the recombineering strain containing pJV53 (van Kessel and Hatfull, 2007) expressing the Che9c genes 60 and 61 under an acetamide-inducible promoter (Parish et al., 1997) have been described previously. All M. smegmatis strains were cultured in Middlebrook 7H9 medium or grown on Middlebrook 7H10 agar supplemented with 10% Albumin Dextrose Complex (ADC), 0.2% succinate, 0.05% Tween-80, 1 mM Ca2Cl, carbenicillin (50 μg ml−1), cyclohexamide (10 μg ml−1) and kanamycin (20 μg ml−1) as required. E. coli was grown in L-broth (LB) supplemented with carbenicillin (50 μg ml−1) and kanamycin (20 μg ml−1) as needed, with E. coli GC5 cells (Stratagene) used for cloning and E. coli BL21(DE3) cells (Stratagene) for protein overexpression.
Plasmids pET21a (Novagen), pLAM12 (van Kessel and Hatfull, 2007), pJV53 (van Kessel and Hatfull, 2007) and pNIT-1 (Pandey et al., 2009) have been previously described. Plasmid pLAM3 was constructed by amplifying the D29 gene 12 (lysB) with primers bearing NdeI and HindIII restriction sites and cloning into pET21a to add a C-terminal His tag. To create the inducible D29B active site mutant (pKC20), the pLAM3 plasmid was subjected to site-directed mutagenesis using the QuikChange XL Site-Directed Mutagenesis Kit (Stratagene) to change Ser82 to alanine. Plasmid pKMC2 was made by amplifying D29 lysB from pLAM3 with primers containing NdeI and NheI restriction sites and cloning this into pLAM12. Plasmid pKMC9 was constructed by amplifying D29 lysB with primers containing NdeI and HindIII restriction sites and cloning this into pNIT-1. Plasmids were purified using QIAprep Spin Miniprep Kit (QIAGEN). Oligonucleotides were supplied by Integrated DNA Technologies and are listed in Table S1.
Expression and purification
Escherichia coli BL21(DE3) carrying plasmids pLAM3 or pKC20 (or vector pET21a) were grown to OD600 between 0.4 and 0.6 in LB containing carbenicillin (50 μg ml−1) and induced with 1 mM IPTG for 4 h. Cells were resuspended in TWEB (50 mM Tris-HCl pH 8.0, 300 mM NaCl), sonicated, centrifuged and the clarified lysates passed over TALON Co2+ resin (Clonetech). The resin was washed sequentially with TWEB containing 10 mM and 20 mM imidazole, and bound protein eluted with five volumes of 120 mM imidazole. Eluted fractions were concentrated using Vivaspin concentration columns (molecular weight cut-off 10 kDa; Sartorius) followed by dialysis against storage buffer (50 mM Tris pH 8.0, 50 mM NaCl, 50% glycerol) and stored at −20°C.
D29 LysB was purified for crystallization similarly. E. coli BL21(DE3)pLAM3 cells were induced for 18 h at 25°C, pelleted, resuspended in Buffer A (50 mM Tris, 500 mM NaCl, 5 mM imidazole, pH 8) and disrupted in a French pressure cell three times at 10 000 PSI for three passages. The cell lysate was cleared by centrifugation at 15 000 r.p.m. for 30 min and applied to a 5 ml HisTrap FF column (GE Healthcare, Piscataway, NJ), and eluted with a linear gradient to 100% Buffer B (50 mM Tris, 500 mM NaCl, 500 mM imidazole, pH 8). Fractions were pooled and dialysed against buffer containing 20 mM Tris, 50 mM NaCl, pH 8. D29 LysB was concentrated to at least 10 mg ml−1.
Zymograms were performed as described previously (Piuri and Hatfull, 2006) by incorporation of 0.2% lyophilized Micrococcus luteus cells as a source of peptidoglycan into the gel matrix. Zymograms were developed by renaturation overnight at 37°C in 25 mM Tris (pH 7.5), 1% Triton X-100 and 0.1 mM ZnSO4, stained with 0.5% methylene blue with 0.01% KOH before destaining with water.
Lipolytic enzyme assays
Enzymatic assays for lipolytic activity were adapted from those described previously (Gilham and Lehner, 2005). One millilitre of p-nitrophenyl substrates (50 μM) (Sigma) was incubated with 1 μg of D29 LysB, D29 LysB S82A, lipase (Pseudomonas fluorescens, Sigma), or 5 μl of a mock purified sample (derived from pET21a containing cells) in buffer (20 mM Tris-HCl pH 8.0, 100 mM NaCl, 0.1% Triton X-100) at room temperature for 30 min. Release of p-nitrophenol was determined by measuring absorbance at 420 nm (A420).
Assays for hydrolysis of M. smegmatis mAGP
Mycolylarabinogalactan–peptidoglycan was isolated from M. smegmatis as described previously (Besra, 1998). M. smegmatis cells were grown, collected and washed three times in phosphate buffered saline (PBS; pH 7.4), and resuspended in PBS + 2% Triton X-100 (PBSX). Cells were disrupted by extensive sonication, centrifuged to collect the insoluble cell wall fraction, resuspended in PBSX and agitated overnight at 4°C. After centrifugation the pellet was resuspended in PBS + 2% SDS and incubated at 100°C for 60 min; this was done three times. After three rounds of extraction, the pellet was washed once each with H2O, 80% acetone in H2O, and acetone. Following evaporation of the acetone, the mAGP-enriched cell wall material (mAGP) was resuspended in PBS + 0.1% Triton X-100 (final concentration 10 mg ml−1) and frozen aliquots stored at −80°C.
mAGP material was chemically hydrolysed by addition of an equal volume of 15% TBAH to 1 mg mAGP resuspended in PBS + 0.1% Triton X-100 and incubated at 100°C overnight. Enzymatic assays were performed by incubation of 1 mg mAGP in 100 μl with varying concentrations of protein at 37°C. TBAH- and enzyme-treated samples were prepared for analysis by thin-layer chromatography (TLC) by addition of an equal volume of dichloromethane and incubation for 15 min at room temperature. The lipid-rich lower dichloromethane layer was removed, extracted once with 0.25 M HCl and once with water, and lipids collected by evaporation. Samples were resuspended in dichloromethane, spotted onto silica-aluminium TLC plates, and separated by chromatography in chloroform/methanol (97:3); lipids were identified by spraying with 5% molybdophosphoric acid (in ethanol) and charring for 15 min at 110°C.
Methyl-esterification of lipids was performed by resuspension of enzyme reaction products in 15% TBAH, addition of an equal volume of dicholoromethane and 1/10 volume iodomethane. Reactions were incubated with shaking at room temperature for 15 min and the lower dichloromethane layer recovered and extracted with HCl and water as described above. Lipids were separated by TLC in 95:5 petroleum ether/ethyl ether and recognized by charring as above.
Bacteriophage recombineering of electroporated DNA
A lysB deletion mutant of phage Giles was constructed as described previously (Marinelli et al., 2008). The targeting substrate was generated using a 100 bp oligonucleotide with 50 bp of homology upstream and downstream of the deletion site, which was then expanded to a 200 bp dsDNA by PCR. The deletion was designed to remove 1146 bp between Giles genome co-ordinates 28,384 and 29,529. Phage Giles DNA (350 ng) was co-electroporated with 200 ng of the 200 bp substrate into induced electrocompetent M. smegmatis mc2155 pJV53 cells, recovered at 37°C for 2 h and plated on top agar lawns with M. smegmatis mc2155 pKMC2 cells in the presence of 0.2% acetamide; approximately 100 plaques were recovered. Individual plaques were picked into 100 μl phage buffer (10 mM Tris-HCl, pH 7.5; 10 mM MgSO4, 68.5 mM NaCl; 1 mM CaCl2) and analysed by DADA-PCR (Marinelli et al., 2008) using 1 μl sample and primers GilesB-DiagR3 and GilesB-DADAPCR3 (Table S1). Primary plaques containing both wild-type and mutant alleles were diluted and plated with 300 μl of M. smegmatis cells containing pLAM12 or pKMC2, either with or without acetamide inducer. To test for viability of the mutant, lysates from plates containing approximately 600 plaques were harvested and tested by PCR with primers GilesB-DiagR3 and GilesB-DiagF. Individual plaques from the secondary plating were picked and tested using the same two primers. A mutant derivative was plaque purified and the sequence confirmed by DNA sequencing.
To test for complementation of the Giles ΔlysB small plaque phenotype, mc2155 cells containing pKMC9 or pNIT-1 vector control were grown to OD600 0.6 and 10 μl of dilutions of either wild-type Giles or GilesΔlysB phage containing ∼102 pfu was added to 500 μl of cells, adsorbed for 30 min at room temperature, and plated as top agar lawns in 0.35% MBTA with 1 mM CaCl2 on 7H10 plates with 0–0.4% ε-caprolactam.
Mycobacterium smegmatis cells grown in 7H9 supplemented with ADC, carbenicillin, cyclohexamide and calcium were grown to OD600 0.3–1.0. For the ATP release assay, cells were diluted to OD600 0.03, and infected with phage particles at a moi of 10. After adsorption for 30 min, cultures were shaken at 37°C, and 100 μl samples removed at different times. ATP release was measured by addition of 100 μl of ENLITEN rLuciferase/Luciferin reagent (Promega), and luminescence recorded for a 10 s interval in a Monolight 2010 luminometer. For the OD assay, cells were diluted to an OD600 of 0.25 and infected with phage particles at a moi of 10. After adsorption for 30 min, cultures were shaken at 37°C and 1 ml samples removed at different times. OD was measured at 600 nm using a Beckman Coulter DU 530 Spectrophotometer.
To determine the number of phage released into the supernatant or retained in unlysed cells, M. smegmatis cells were grown as above and diluted to an OD600 of 0.25. These were infected at an moi of 0.1 with adsorption for 30 min followed by incubation with shaking at 37°C. One millilitre of samples was removed at different times and separated by centrifugation into supernatant and pellet fractions. The pellet was resuspended in 1 ml phage buffer and sonicated. Both pellet and supernatant were serially diluted and 5 μl samples were spotted onto top agar lawns containing M. smegmatis in 0.35% MBTA with 1 mM CaCl2 on 7H10 plates.
Crystallization and structure determination
Purified D29 LysB at 10 mg ml−1 concentration was used to screen a large number of commercially available crystallization conditions by vapour diffusion method. After optimization upon the hit condition in the initial screen, diffraction quality crystals were obtained with 2 mg ml−1 protein in the condition of 50 mM Tris pH 8, 20%PEG1000, 100 mM Ca(OAc)2. The best crystals grew in 4 days and were flash frozen in mother liquor with 25% glycerol then stored in liquid nitrogen. Data sets were collected with a Rigaku MM007HF generator with a Bruker Smart 6000 CCD detector, and with synchrotron beamlines at Advanced Photon Source (APS 23-ID), Argonne National Laboratory, Chicago. The data were processed with HKL2000 and the anisotropic scaling was further performed on the diffraction anisotropy server (Strong et al., 2006).
The initial structure was solved by SIRAS combining the native crystal with an iodine derivative obtained by quick soaking of the crystal in the cryo with 0.5 M NaI for 2 min. The model was refined against the native data set iteratively until the R/Rfree reached 0.205/0.252 at 2.0 Å. The Iodine sites were found with Shelxd/e program. The phasing and refinement of the protein structure were carried out with the Phenix package. The model was built in Coot. 20.3 PyMol was used for structure analysis and rendering.
The atomic co-ordinate and structure factor have been deposited in the PDB with accession code 3HC7 for D29 LysB.
We especially thank Laurent Kremer, Xavier Trivelli and Yann Guerardel for their assistance in analysis of the mycolic acid products of LysB hydrolysis of mAGP and for providing the data shown in Fig. 4C. We are grateful to Drs Laura Marinelli and Ting Huang for providing materials and insights into mycobacteriophage lysis systems, and to Dr Anil Ojha for comments on the manuscript. This work was supported in part by NIH Grant AI064494 to G.F.H., and by funds from the Robert A. Welch foundation to James Sacchettini.