Murein hydrolases appear to be widespread in the virions of bacteriophages infecting Gram-positive or Gram-negative bacteria. Muralytic activity has been found in virions of the majority of a diverse collection of phages. Where known, the enzyme is either part of a large protein or found associated with other structural components of the virion that limit enzyme activity. In most cases, the lack of genetic and structural characterization of the phage precludes making a definitive identification of the enzymatic protein species. However, three proteins with muralytic activity have been unequivocally identified. T7gp16 is a 144 kDa internal head protein that is ejected into the cell at the initiation of infection; its enzyme activity is required only when the cell wall is more highly cross-linked. P22gp4 is part of the neck of the particle and is essential for infectivity. The activity associated with virions of Bacillus subtilis phage ø29 and its relatives lies in the terminal protein gp3. These studies lead to a general mechanism describing how phage genomes are transported across the bacterial cell wall.
In order to infect, bacteriophages must transport their genome across the bacterial cell envelope (Dijkstra and Keck, 1996a; Letellier et al., 1999; Molineux, 2001). Gram-negative phage genomes must penetrate through an outer membrane composed of lipopolysaccharide, phospholipids and proteins, a periplasmic space housing a thin layer of peptidoglycan and a cytoplasmic membrane made of phospholipids and proteins (Nanninga, 1998). Gram-positive phage genomes must penetrate a thick peptidoglycan layer and a cytoplasmic membrane (Archibald et al., 1993).
The largest single molecule of the bacterial cell is peptidoglycan, also known as murein. It is composed of long glycan strands, consisting of alternating N-acetyl-glucosamine (NAG) and N-acetyl-muramic acid (NAM) residues, linked to each other by peptide cross-bridges between NAM residues (Holtje, 1998; Dmitriev et al., 1999). Gram-positive bacterial murein is thick and highly cross-linked, whereas that from Gram-negative bacteria is predominately a monolayer (Labischinski et al., 1991). In both bacterial cell types, the peptidoglycan network maintains cell integrity and determines shape.
During cell growth, the peptidoglycan layer in Gram-negative bacteria is enlarged by first forming cross-links with new glycan strands and only then breaking old cross-links (Koch and Doyle, 1985). This approach is thought to be necessary to preserve the integrity of the largely single-layered macromolecule. In contrast, new glycan chains of Gram-positive bacteria are cross-linked to the cytoplasmic side of the murein sacculus (Archibald et al., 1993). In both cases, murein hydrolases are necessary not only for the insertion of new strands but also for murein recycling (Holtje, 1998). About 50% of pre-existing peptidoglycan is recycled during every cell generation (Goodell, 1985). Lysozymes, lytic transglycosylases, N-acetylmuramyl-l-alanine amidases and endopeptidases are the major classes of murein hydrolases.
Lysozymes and lytic transglycosylases are similar enzymes; both break the glycosidic bond between NAM and NAG. However, the limit product of lysozyme action is a disaccharide with a reducing end of NAM, whereas that of the lytic transglycosylase is a disaccharide containing an anhydrosugar. N-acetylmuramyl-l-alanine amidases break the lactal linkage between the C2 of the NAM and the primary d-alanine of the cross-bridge. Endopeptidases cleave the peptide bonds within or between cross-bridges (Holtje, 1998). Some class V chitinases have also been shown to possess muralytic activity (Brunner et al., 1998).
The murein layer of both Gram-positive and Gram-negative cells is a barrier to macromolecular transport, and prevents free diffusion of globular proteins larger than ≈ 50 kDa (Demchick and Koch, 1996). Normal cellular functions that necessitate passage of larger molecules or structures through the murein layer may either take advantage of temporary gaps in the murein network during peptidoglycan recycling or, as in the case of flagellar assembly (Nambu et al., 1999), hydrolyse peptidoglycan as part of the assembly process. It has been suggested that the formation of conjugative pili also involves cell wall-degrading enzymes (Bayer et al., 1995; Mushegian et al., 1996). Bioinformatic analyses of both phage and plasmid genomes have provided additional examples of muralytic enzymes associated with phage virions or plasmid DNA transfer (Koonin and Rudd, 1994; Dijkstra and Keck, 1996b; Lehnherr et al., 1998).
The first step in infection is adsorption; this occurs rapidly, even at temperatures lower than those necessary for cellular internalization of the phage genome. These late steps are highly synchronous, suggesting that penetration of the cell envelope by the tail, tail tube or internal proteins ejected from the virion and DNA transport are well-co-ordinated events. It would therefore seem unlikely that phages simply rely on the normal turnover of peptidoglycan to initiate the process of genome transport into the cytoplasm. Indeed, some phage virions are known to contain cell wall-degrading activities.
The baseplate protein gp5 of T4 is required for lysis from without, and a mutant gp5 complements a lysozyme defect for lysis from within (Kao and McClain, 1980). It was suggested that gp5 normally catalyses localized digestion of the cell wall to facilitate tail tube penetration through the cell envelope. T4gp5 is synthesized as a 63 kDa protein (Mosig et al., 1989) that is subsequently processed during baseplate assembly to yield the mature 43 kDa protein gp5*. The protein has been shown to have the same substrate specificity as the gpE lysozyme, it is a N-acetylmuramidase with activity that increases 10-fold upon removal of a C-terminal fragment (Nakagawa et al., 1985; Kanamaru et al., 1999). Although mutations affecting the active site of the gene 5 lysozyme have been identified (Takeda et al., 1998), the essential role of gp5 as a structural baseplate protein has made it difficult to determine whether the cell wall-degrading activity is important during infection. However, mutants that do not undergo the gp5 → gp5* cleavage, and thus have reduced lysozyme activity, have recently been shown to have a cold-sensitive phenotype (S. Kanamaru, personal communication). This is the same phenotype as T7 mutants altered in the predicted catalytic residue of the lytic transglycosylase motif of the virion protein gp16 (Moak and Molineux, 2000).
The lipid-containing phages PM2 (Tsukagoshi et al., 1977), ø6 (Mindich and Lehman, 1979) and PRD1 (Rydman and Bamford, 2000; 2002) are also known to contain virion-associated murein hydrolases. PRD1 virions have been shown to contain two activities, the lysin P15 and P7. Neither protein is essential for infectivity, but P7 plays an accessory role in genome penetration of the infected cell. Cells infected by PRD1 virions lacking P7, but not virions lacking P15, show a delay in phage DNA replication and lysis. Double mutants have not been constructed.
Bacteriophage T7 protein gp16 contains the catalytic motif of the Escherichia coli lytic transglycosylase Slt (Engel et al., 1991). Mutations affecting the catalytic residue of the putative gp16 lytic transglycosylase could be correlated with a defect in establishing genome penetration under conditions in which the cell wall was expected to be more highly cross-linked (Moak and Molineux, 2000). These observations strongly suggested that gp16 contains peptidoglycan hydrolase activity, but assays of enzyme activity were not performed. Here, we use peptidoglycan-containing acrylamide gels (zymograms) to show that the gp16 in the T7 virion does contain muralytic enzyme activity. We also show that several members of the T7 family of phages contain a virion-associated enzyme that degrades peptidoglycan and, furthermore, that comparable activities are widespread in the virions of diverse phages.
The T7 group
In order to demonstrate that T7 virions possess peptidoglycan hydrolytic activity, they were purified by equilibrium CsCl density gradients and disrupted in a denaturing buffer. Proteins were then electrophoretically separated on a zymogram and, after incubation to allow protein renaturation and enzyme action, the peptidoglycan was stained using methylene blue. A parallel denaturing gel lacking peptidoglycan was stained with Coomassie brilliant blue in order to reveal the major structural proteins of the virion. Wild-type T7 virions gave a single band of peptidoglycan hydrolytic activity with an electrophoretic mobility that corresponds to a protein of ≈ 140 kDa, close to that predicted (143.84 kDa) for gp16 (Fig. 1 and Table 1).
Table 1. . Muralytic activity in phage virions.
Approximate size (kDa)
The putative active site residue of gp16 is glutamate 37; by sequence alignment, this residue corresponds to glutamate 478 of E. coli Slt and, by changing the Slt residue to glutamine, hydrolytic activity was abolished (Thunnissen et al., 1994). T7 virions containing gp16E37Q or gp16E37F are defective in initiating genome penetration of cells that were growing at low temperature or, but only in the case of gp16E37F, had been grown at 30°C to high cell density (Moak and Molineux, 2000). T7 virions containing gp16E37Q or gp16E37F were therefore prepared and electrophoresed on the same gel as the wild type. Neither mutant particle contains a protein that hydrolyses peptidoglycan in this assay (Fig. 1). We conclude that T7gp16 does hydrolyse peptidoglycan and that it is important that phage particles contain this activity for efficient infection under certain growth conditions.
Both the coliphage T3 and the yersiniophage øYeO3-12 are close relatives of T7. The predicted sequences of T3 and øYeO3-12 gp16 are, respectively, 66.5% and 67.2% identical to that of T7, and all three proteins are essentially the same size (Pajunen et al., 2001; 2002). The residues known to be important for E. coli Slt activity are totally conserved in all three phage proteins. As expected therefore, virions of both T3 and øYeO3-12 also contain peptidoglycan hydrolytic activity associated with an ≈ 140 kDa protein (Fig. 1 and Table 1).
Two phages that are more distantly related to T7 are the Salmonella phage SP6 (Zinder, 1961) and the coliphage K1-5 (Scholl et al., 2001). The genome sequences of these phages show that most of the structural genes are collinear with, and homologous to, their T7 counterparts (D. Scholl and I. J. Molineux, unpublished observations). Zymogram analysis of the virion proteins of SP6 and K1-5 show that both contain a protein of about 110 kDa that hydrolyses peptidoglycan (Fig. 1 and Table 1). SP6 and K1-5 virions contain, respectively, a 978 and 982 residue protein, both of which exhibit sequence similarity to the active site regions – including the catalytic glutamate – to the Bacillus subtilis phage B103 endolysin and to T4 lysozyme gpE.
P22 virion protein gp4 hydrolyses peptidoglycan
A zymogram of purified P22 virions shows a 15–18 kDa protein with peptidoglycan hydrolytic activity (Fig. 2). The size is consistent with both virion protein gp4 (18.0 kDa) and gp19 lysozyme (16.1 kDa). Lysozyme is not known to be a virion protein and, to help to substantiate the conclusion that gp4 is a peptidoglycan hydrolase, the protein band corresponding to enzyme activity was eluted from the corresponding Coomassie-stained gel. The N-terminal sequence of the protein was determined to be MQIKTKGDLVRAAL, which corresponds to residues 1–14 of the translated gene 4 sequence.
P22 virions contain several internal head proteins that are ejected into the cell at the initiation of infection and/or are necessary for the stability of packaged DNA (Israel, 1977; Strauss and King, 1984). These proteins are not essential for particle formation and, by growing P22 amber mutants in a non-suppressing host, non-infectious virions can be prepared. A zymogram of P22 virions made in the absence of gp4, gp7 or gp26 shows that only those made in the absence of gp4 lack enzyme activity.
In some experiments with P22 virions lacking gp7 or gp26, a reduced amount of peptidoglycan-hydrolysing activity with the same electrophoretic mobility of gp4, relative to wild-type virions, was observed (not shown). Gp26 has been proposed to be the plug that keeps DNA within the particle before infection (Israel, 1977), and both gp4 and gp26 have been described as DNA stability proteins (Strauss and King, 1984). Particles lacking either protein are known to lose packaged DNA spontaneously and, consequently, may also lack additional proteins. The reduction in enzyme activity occasionally observed in particles lacking either gp7 or gp26 may thus be either a consequence of a generally unstable capsid or a requirement that one or both proteins are necessary for correct insertion of gp4 into the phage head. Nevertheless, only P22 virions lacking gp4 were consistently devoid of all peptidoglycan-degrading activity, strongly suggesting that gp4 normally penetrates the infected cell where it hydrolyses peptidoglycan. Virions of gene 4 amber mutants grown in a non-suppressing host do not form plaques on a suppressing host (data not shown). Thus, the presence of gp4 in the infecting particle is essential, but whether peptidoglycan hydrolysis or some other activity is the essential function remains to be determined.
Other phages infecting Gram-negative hosts
The results described above encouraged us to examine other phage virions for the presence of peptidoglycan hydrolase activity. The lipid phages ø6, PRD1 and PM2 are already known to have murein hydrolases associated with their virions (Tsukagoshi et al., 1977; Mindich and Lehman, 1979; Rydman and Bamford, 2000), and the central domain of the P1 orf47 tail tube or baseplate protein has been shown to degrade peptidoglycan in vitro (Lehnherr et al., 1998; H. Lehnherr, personal communication). These phages were therefore not tested.
Purified T5 virions contain a 33 kDa protein that degrades peptidoglycan (Fig. 3 and Table 1), and its electrophoretic mobility is consistent with either of two minor head proteins present at an estimated < 12 and 26 copies per virion (Zweig and Cummings, 1973). Preparations of N4 virions contain an ≈ 75 kDa protein that, however, does not correspond to a major N4 structural protein (Fig. 3; Falco et al., 1980). Attempts to sequence the N-terminal portion of the protein resulted in mixed sequences, suggesting that the activity might result from proteolytic fragments of a larger protein. More careful studies with both T5 and N4 are necessary to identify the enzymatically active species. Two recently described lytic coliphages are C1 and C6; both resemble T1 by morphology but use different receptors: C1 adsorbs to BtuB, whereas C6 uses FhuA (Likhacheva et al., 1996; Samsonov et al., 2002). C1 virions contain a murein hydrolase of ≈ 30 kDa, but no activity was seen with C6.
Purified T4 virions contain two proteins that hydrolyse peptidoglycan (Fig. 3 and Table 1), an ≈ 40 kDa band that probably corresponds to gp5*, the cleaved form of gp5 found in mature virions (Mosig et al., 1989; Kanamaru et al., 1999), and an 18 kDa band that corresponds to the e gene product lysozyme. The presence of gpE in purified phage may seem surprising, but it has been shown that ≈ 0.5 molecules of lysozyme are associated with each particle (Emrich and Streisinger, 1968). However, gpE lysozyme has no role in the initiation of infection, and virions completely devoid of the protein infect cells normally.
A less well-known lytic phage is Xp10, the host of which is Xanthomonas campestris pv. oryzae (Kuo et al., 1967; Chang et al., 1985). Xp10 has lambda-like morphology, and its genome sequence has been determined recently (Yuzenkova et al., 2003). A zymogram of purified Xp10 particles revealed a protein of about 24 kDa that hydrolyses peptidoglycan (Fig. 3). The size of this protein corresponds to the major tail protein of Xp10. However, the predicted amino acid sequence of the tail protein reveals no similarities to lysozymes or lytic transglycosylases. The tail region of the Xp10 genome sequence predicts a 17 kDa Xp10 protein that contains a chitinase motif. The sequence of this protein has weak similarity to λ gpK and the NlpC/P60 family of peptidoglycan binding proteins, which have been shown to hydrolyse peptidoglycan in vitro (Mellroth et al., 2003). If the predicted Xp10 17 kDa protein is responsible for the muralytic activity associated with virions, some post-translational modification of the protein may be necessary to explain its apparent size on zymograms. Clearly, further characterization of the virion protein containing muralytic activity is necessary.
In contrast to most of the phages described above, zymogram analyses using CsCl-purified virions of Mu, P2, λ and HK022, and the single-stranded DNA phages øX174 and G4, failed to reveal a protein that could hydrolyse peptidoglycan. We also tested Ur-λ, the original λ phage isolated from Escherichia coli K-12 (Hendrix and Duda, 1992), in case the side tail fibres, which are missing from common laboratory stocks of λ, contained muralytic activity. For all these phages, a variety of pH values, of both Triton X-100 and MgCl2 concentrations, and of the presence of CaCl2 or EDTA in the protein renaturation step were used, but no conditions were found that revealed peptidoglycan hydrolysis. For Mu and P2, the inability to detect hydrolytic activity may be a failure to renature an active enzyme in the zymogram; purified virions of both phages gave positive spot tests using a lawn of chloroform-killed E. coli as substrate (data not shown). However, λ and HK022, and øX174 and G4 were all negative even in this less rigorous test.
Phages infecting Gram-positive bacteria
The cell wall of Gram-positive bacteria is composed of multiple layers of peptidoglycan and is much thicker than that of Gram-negative bacteria. Gram-positive bacteria should therefore provide a more impervious barrier to phage infection. If peptidoglycan hydrolase activity is a common property of phages, those infecting Gram-positive hosts are most likely to contain the enzyme as part of their virion.
Five Lactococcus lactis and two Staphylococcus aureus phages were tested for peptidoglycan hydrolase activity. Purified virions of all seven contain an activity that degrades E. coli peptidoglycan (Fig. 4 and Table 1). The relatively weak activity observed on zymograms may result from the structural differences between E. coli and L. lactis or S. aureus peptidoglycan (Archibald et al., 1993; Gopal and Reilly, 1995). Interestingly, the Lactococcal phage enzymes only exhibited significant enzyme activity when the renaturation step of the zymogram analysis was carried out at pH 6.0. The difference in pH optimum of these peptidoglycan hydrolases, relative to that for the other phage proteins in this study, probably reflects the more acidic environment in which the host bacteria thrive.
These phages have not been analysed extensively, and a genetic analysis is therefore not possible. However, the complete genome sequence is known for L. lactis phages øsk1, ør1t and øc2, and S. aureusø11, and motifs or predicted functions are available. øsk1 Orf5 has the same predicted size as the peptidoglycan hydrolytic activity, and its sequence exhibits homology to d,d-carboxypeptidases. The function of Orf5 is unknown. The murein hydrolase activity of øc2 virions may be associated with Orf115. Orf115 has a host range function (Stuer-Lauridsen et al., 2003) with predicted sequence that has homology to an amidase. Virions of ør1t contain a peptidoglycan hydrolytic activity with the size of the endolysin; no other predicted proteins with appropriate homologies are apparent. The muralytic activity of ø11 particles could be associated with the 72.2 kDa Orf51 as its sequence exhibits homology to a cell wall hydrolase (Iandolo et al., 2002). However, Orf51 also has sequence similarity to the SPP1 head morphogenesis protein gp7. As proteolysis of phage structural proteins, significantly including the T4gp5 virion lysozyme (Mosig et al., 1989; Kanamaru et al., 1999), is common among dairy phages (Lubbers et al., 1995; van Sinderen et al., 1996), more studies are necessary before making a definitive identification of the enzymatically active species.
The two lytic phages of B. subtilis tested both contain muralytic activities in their virions. SPP1 contains a protein of about 38 kDa, and SPO1 a protein of about 44 kDa, which hydrolyse peptidoglycan (Fig. 4). The temperate bacillus phage SPβ YomI protein is thought to be encoded in the late operon (Kunst et al., 1997; Lazarevic et al., 1999). The mass of the protein is predicted to be 252 kDa, and it contains a lytic transglycosylase motif (Lehnherr et al., 1998). Because of this motif, it was suggested that YomI functions in lysis of the infected cell (Lazarevic et al., 1999). A zymogram of purified particles of SPβ shows a> 200 kDa band that is associated with peptidoglycan hydrolytic activity, suggesting that YomI does degrade peptidoglycan. However, the presence of the protein in the mature virion indicates that it more likely functions during the early stages of infection. It should be noted that the size of YomI is far larger than is required for a simple murein hydrolase, suggesting that it has additional functions in phage propagation.
ø29 terminal protein gp3 hydrolyses peptidoglycan
Bacteriophage ø29 is well characterized genetically, functionally and structurally (Tao et al., 1998; Meijer et al., 2001). An initial test for peptidoglycan hydrolytic activity associated with purified wild-type ø29 virions showed a major band of activity with an apparent molecular size of ≈ 30 kDa (data not shown). Assuming that the protein of interest had not been proteolytically or otherwise modified, it was thought most likely to be one of three gene products. Candidates included the head fibre protein gp8.5 and the terminal protein gp3, which both have a mass of 29.5 kDa, and the 28 kDa gp15 endolysin, which degrades peptidoglycan from the interior of the cell at the end of the lytic cycle. Particles lacking the non-essential gp8.5 were prepared, and a zymogram showed that peptidoglycan hydrolase activity was still present (Fig. 5A, lane 1). Although the endolysin is not known to be associated with the phage particle, purified virions of an endolysin-defective ø29 gene 15 mutant were tested. Hydrolysis of peptidoglycan remained associated with a 30 kDa protein (Fig. 5A, lane 2). These data strongly suggest that peptidoglycan hydrolase activity is associated with the DNA terminal protein gp3.
In order to demonstrate the hydrolytic activity of gp3 more directly, purified phage particles were disrupted with 4 M guanidinium hydrochloride, and the gp3–DNA complex was purified by CsCl density gradient centrifugation (Hermoso and Salas, 1980). A portion of this complex was then treated with micrococcal nuclease to degrade the phage genome. Guanidinium-disrupted phage particles retain peptidoglycan hydrolase activity in a zymogram assay (data not shown), but the CsCl-purified genomic gp3–DNA complex exhibited no activity (Fig. 5A, lanes 3 and 4). However, treatment of the gp3–DNA complex with micrococcal nuclease results in the reappearance of activity with an apparent mass of 30 kDa (Fig. 5A, lanes 5 and 6). Similar data were also obtained using purified gp3–DNA kindly provided by D. Anderson (University of Minnesota): the intact genome–protein complex exhibited no activity, but a 30 kDa band of peptidoglycan-hydrolysing activity was present after nuclease treatment (data not shown).
The observations that disrupted virions of ø29 contain a peptidoglycan hydrolase associated with gp3 and that the CsCl-purified genome–gp3 complex is inactive in a zymogram assay indicate that some molecules of gp3 inside a preparation of virions are not covalently attached to DNA. This ‘extra’ gp3 would be removed during purification of the gp3–DNA complex by isopycnic centrifugation. It was also suggested that the two gp3 molecules per genome may have different molecular weights as a result of proteolytic processing (Anderson and Reilly, 1993). Restriction digestion of the gp3–DNA complexes and purification of the genomic left and right ends also revealed activity only after the fragments were nuclease treated (data not shown), supporting the idea that the muralytic activity of gp3 requires that it not be covalently bound to DNA. No size difference between left- or right-end gp3 was detected.
As a final test that the terminal protein gp3 is a peptidoglycan hydrolase, in addition to its well-established roles in DNA metabolism and DNA packaging, a temperature-sensitive ø29 gene 3 mutant was grown at 32°C. Wild-type phage was grown and purified in parallel. After denaturing electrophoresis of the virion proteins, one zymogram gel was incubated at 30°C and another at 43°C for 16 h in order to allow proteins to renature and exhibit enzymatic activity. After renaturation at 30°C, both phage preparations showed enzyme activity, whereas at 43°C, only wild-type gp3 was active (Fig. 5B). Thus, the Ts3(132) mutation prevents refolding of the mutant gp3 into a conformation that can hydrolyse peptidoglycan.
Bacteriophage ø29 is the prototype member of the ø29 family of Podoviridae that infects Bacillus species; the family has been subdivided into three groups based on serological and other differences (Meijer et al., 2001). ø29 is a member of group 1, whereas phages M2 and GA-1 are representatives of groups 2 and 3 respectively. Both ø29 and M2 grow on B. subtilis 168. GA-1 does not; its preferred host is Bacillus GR1, an organism most closely related by 16S rRNA sequences to Bacillus pumilis. The genetic organization of all three groups of phages is similar, and all code for a terminal protein. Major differences are confined to small genes of unknown function near the physical ends of the M2 and GA-1 genomes (Meijer et al., 2001).
A zymogram of purified M2 virions revealed a band of activity with an approximate mass of 31 kDa, a little larger than the ø29 band (Fig. 6), but a value consistent with the size of gp3 from the related group 2 phage B103 (Pecenkova et al., 1997). Unexpectedly, GA-1 virion proteins failed to give activity. However, when proteins were renatured in the presence of EDTA, rather than in the presence of Mg2+, a GA-1 protein of 30 kDa showed hydrolysis of peptidoglycan. The sequences of ø29, M2 and GA-1 gp3 exhibit weak homology to the class V chitinases that also function as lysozymes (Brunner et al., 1998). The putative catalytic residue E42 is conserved in both ø29 and M2gp3 but not in GA-1gp3, a feature of the latter protein that may reflect its muralytic activity being revealed only in the presence of EDTA.
The lytic transglycosylase motif of T7gp16 was shown to play an important role in the efficient genome penetration of highly cross-linked E. coli peptidoglycan at the initiation of infection (Moak and Molineux, 2000). We have now shown directly that this motif is part of an active enzyme that degrades peptidoglycan. Gp16 is part of the internal core structure of the T7 head and is known to be ejected from the virion before the phage genome (Molineux, 2001). One likely function of the protein is therefore to enlarge a hole in the peptidoglycan mesh-like structure, thereby promoting DNA transport. However, this function is completely dispensable for phage growth under optimal laboratory conditions (Moak and Molineux, 2000). The closely related coliphage T3 and the yersiniophage øYe03-12 contain gp16 homologues with intact lytic transglycosylase motifs, and those proteins also show activity on peptidoglycan zymograms, suggesting that they also have this function in making phage DNA penetration more efficient when infected cells are growing at low temperatures or are at high density. Interestingly, Salmonella phage SP6 and coliphage K1-5 virions also contain a murein hydrolase, but one that is more likely to be a lysozyme rather than the lytic transglycosylase of the close-knit T7 group (Scholl et al., 2003). Furthermore, the genes for the SP6 and K1-5 proteins occupy the equivalent genomic position to T7 gene 15 rather than gene 16, suggesting that acquisition of murein hydrolase activity occurred after divergence of SP6 and K1-5 from the close-knit T7 group.
Bacteriophage P22 also ejects at least four proteins (gp7, gp16, gp20 and gp26) into the cell at the initiation of infection (Israel, 1977), but their functions are not known precisely. In the virion, these proteins and gp4, which was only discovered later, are thought to be associated with the head–tail connector and/or with DNA (Bazinet and King, 1988; Steinbacher et al., 1997). Gp16 was shown to function in trans from infecting particles, whereas gp7 or gp20 did not (Susskind and Botstein, 1978). Gp4, gp10 and gp26 have been described as head-stability proteins and are thought to close the DNA packaging channel below the head–tail connector to form the neck of the phage (Strauss and King, 1984). The tail-spike protein gp9, which adsorbs to and then degrades the cellular O-antigen, is attached to the base of the neck. Structural data suggest that gp9 has a hinge-like region that could open like a flower, potentially exposing the neck proteins to the cell wall (Steinbacher et al., 1997). This idea becomes more attractive with the demonstration that gp4 hydrolyses peptidoglycan; it now provides the first steps in a conceptual framework describing the process of P22 genome internalization. These steps may be functionally equivalent to the protein-hinging and peptidoglycan-digesting mechanism used by T4 (Kanamaru et al., 2002).
The T7 and SP6 groups, and also P22, recognize LPS as their receptor, and all are members of the Podoviridae. All but one of the Podoviridae that have been tested clearly contain a murein hydrolase in their virions, and this may be a feature common to this morphological group. The possible exception is N4, in which the electrophoretic mobility of enzyme activity did not correspond to that of a structural protein. The activity that we detected may result from either an adventitiously bound protein or a proteolytic fragment of a larger virion protein; both are actually found in T4 particles (Emrich and Streisinger, 1968). It is also possible that N4 virions lack murein-hydrolysing activity. The initial receptor for N4 is the outer membrane protein NfrA (Kiino and Rothman-Denes, 1989; Kiino et al., 1993a), but irreversible adsorption also requires both inner membrane and cytoplasmic proteins (Kim and Yoo, 1989; Kiino et al., 1993a,b). Perhaps these additional proteins obviate a requirement for the virion to harbour cell wall-degrading activity.
A murein hydrolase may also be common among tailless phages that contain double-stranded (ds) nucleic acid genomes. Virions of the dsRNA phage ø6 and the dsDNA phages PM2 and PRD1 have all been shown to contain murein hydrolase activity (Tsukagoshi et al., 1977; Mindich and Lehman, 1979; Rydman and Bamford, 2000). Furthermore, like that of T7gp16, the activity of PRD1 p7 has been directly linked to phage DNA penetration (Grahn et al., 2002; Rydman and Bamford, 2002). However, that activity could not be demonstrated using virions of the single-stranded phages G4 and øX174, which also lack tails.
Among the contractile tailed phages, it has long been known that T4 virions contain a lysozyme activity associated with the baseplate protein gp5* (Kao and McClain, 1980; Mosig et al., 1989; Kanamaru et al., 1999). However, T4 virions also normally contain small amounts of gpE, which has no role in the initiation of infection and is only required for cell lysis; virions lacking gpE show normal infectivity (Emrich and Streisinger, 1968). Thus, the presence of lysozyme activity in purified virions is not by itself a reliable indicator that the initiation of phage infection includes degradation of the cell wall. A related problem, failing to detect murein hydrolytic activity on zymograms but obtaining lysis of a killed bacterial lawn by disrupted virions of Mu and P2, was also encountered during these studies. Although the latter example may simply reflect the failure to renature an active enzyme in zymograms, both problems make it difficult to evaluate the necessity for phage virions to contain cell wall-degrading activity.
All the Siphoviridae infecting Gram-positive hosts tested contain murein hydrolases in their virions, although a lack of genetic and structural studies and the proteolytic processing of structural proteins of some of these phages (Lubbers et al., 1995; van Sinderen et al., 1996) make unequivocal identification of the protein with enzymatic activity problematic. There are other examples of Gram-positive phage genomes with open reading frames (ORFs), outside their lysins, that have predicted murein hydrolase activities. The Streptococcus phages Sfi21 and DT-1 and the Lactobacillus phages LL-H, øadh and øg1e all have putative lytic transglycosylase motifs in likely structural proteins (Crutz-Le Coq et al., 2002).
T5, C6 and HK022 all use the FhuA porin as their receptor (Braun et al., 1973; Dhillon, 1981; Samsonov et al., 2002) but, whereas T5 has a virion-associated murein hydrolase, we could not detect activity with C6 or HK022. Thus, not all phages that use proteinaceous receptors may require a virion-associated murein hydrolase. Phage λ, which adsorbs to the LamB porin, also appears to lack peptidoglycan hydrolytic activity. Porins are known to interact with the peptidoglycan (Gabay and Yasunaka, 1980; Menichi and Buu, 1986), and it is possible that a porin could allow a phage tail to penetrate the cell wall efficiently without a requirement for murein hydrolase activity. Alternatively, a hypothetical virion enzyme may simply have failed to renature in the zymogram, or its activity may require specific structural changes, perhaps comparable to those associated with the hexagon to star transition in the T4 baseplate. It could be significant that λ gpK and HK022gp19 both exhibit weak homology to the NlpC/P60 family of peptidoglycan-binding proteins, which were recently shown to hydrolyse peptidoglycan (Mellroth et al., 2003).
The 5′ covalently bound DNA terminal protein gp3 of the bacteriophage ø29 has been implicated in DNA replication and packaging, but we show here that gp3 probably also functions in genome penetration of the cell. The putative muralytic motif of ø29gp3 is found near residues 30–90, clearly distinct from the priming, binding and packaging motifs that are more C terminal. DNA-bound gp3 is not an active murein hydrolase, suggesting that DNA can occlude peptidoglycan from the active site. The idea has been raised that free gp3 is packaged, possibly by interacting with covalently bound gp3 (Anderson and Reilly, 1993). After adsorption to its glycosylated lipoteichoic acid receptor (Young, 1967), the complex of gp3–DNA and free gp3 may move down to the tip of the tail, where the non-covalently bound gp3 could digest peptidoglycan. After digestion is complete, the association of the enzymatically active gp3 with the gp3–DNA complex would result in presentation of the phage genome to the outer face of the cytoplasmic membrane.
Where known, the murein hydrolytic activity associated with tailed phage virions resides with the tail structure or, in the case of the T7 and SP6 groups, with an internal head protein that forms part of the extensible tail (Molineux, 2001). The high-molecular-weight bacteriocins resemble phage tails (Bradley, 1967; Daw and Falkiner, 1996) and may be considered to be defective phages. The barriers imposed by the cell wall to phage infection are therefore likely to be the same for cell killing by this group of bacteriocins. The R-type pyocin of Pseudomonas aeruginosa and Serratia plymithicum J7 serracin P and other bacteriocins are related to the phage P2 tail, and the F-type pyocin is related to λ tail proteins (Nakayama et al., 2000; Jabrane et al., 2002). Although the evidence for peptidoglycan hydrolysis by P2 virion proteins is weak, and for those of λ is only through iterative psi-blasts, the possibility that bacteriocins of the phage tail family possess a muralytic activity should not be discounted. The defective B. subtilis phage PBSX has a predicted lytic transglycosylase motif in the structural protein, XKDO (Lehnherr et al., 1998), which may aid in its killing activity.
Lysozymes may be considered to be bacteriolytic enzymes, but their relatives, the soluble lytic transglycosylases, are essential for peptidoglycan turnover during cell growth. A prominent E. coli lytic transglycosylase is SltY, the crystal structure of which is known (van Asselt et al., 1999). The amino acid motif that describes the active site is not confined to chromosomally encoded genes, but is also found in a variety of phage and conjugal plasmid proteins (Koonin and Rudd, 1994; Bayer et al., 1995; Dijkstra and Keck, 1996a; Mushegian et al., 1996; Lehnherr et al., 1998). Conjugation requires that DNA cross the bacterial cell wall and, in the case of Ti plasmid transfer, DNA also crosses the plant cell wall. The most detailed study of conjugal plasmid proteins has been on the plasmid R1 P19 protein. P19 causes small lesions in the peptidoglycan mesh, leading intracellular material to efflux from the cell (Bayer et al., 2001). It was proposed that the intracellular osmotic pressure resulted in the appearance of vesicles surrounded by a membrane bilayer, but the cell membrane was not disrupted. This suggests that P19 causes only a localized opening in the cell wall. Controlled degradation of the cell wall is of course essential to maintain cell integrity and to allow continued viability of both the conjugal donor and its recipient.
A similar event must occur with a phage-infected cell. The cell must remain metabolically active to produce progeny phage, implying that the infecting virions cannot cause an extensive degradation of the peptidoglycan that would cause cell lysis. The murein hydrolase carried within the virion must either have limited enzymatic activity or the activity must be constrained. Combining a murein hydrolase activity with a major structural component of a phage particle is a logical solution to preventing enzyme diffusion. Lysis from without by members of the T-even family is a case of this constraint being overwhelmed by a high multiplicity of infection, with the consequence that the cell wall loses its structural integrity and allows cell lysis.
The observation that R1 P19 causes the cytoplasmic membrane to bleb through the small hole in the peptidoglycan may also apply to some phage infections. The tail structure of many phages is too short to span the distance from the site of adsorption to the cell cytoplasm. Causing the cytoplasmic membrane to move towards the outer membrane layer and thus closer to the phage head may not only alleviate the distance problem for short-tailed phages but also provides a simple explanation for the failure of the periplasmic endonuclease I to degrade the genome as it passes into the cell.
Bacteriophages and their hosts
Phages øX174 and G4 were grown on E. coli C strain. E. coli K-12 (λ+)/F+, a gift from R. Duda (Hendrix and Duda, 1992), provided Ur-λ; λ was from IJM228 (W3110Thy–[λcI857 S7(Am)]). C1 and C6, gifts from S. P. Sineoky (Samsonov et al., 2002), and other coliphages were propagated on IJ1142 [K-12 ΔlacX74 galK150(Am) trp-49(Am) cysI(Am)] or, for mutant T7 particles, isogenic derivatives containing supE44 or a plasmid carrying a glutamic acid or phenylalanine tRNA amber suppressor (Moak and Molineux, 2000). Salmonella phages were grown on MS1883 [Salmonella typhimurium LT2 leuA14(Am) supE hsdR] or, for mutant P22 particles, a single round of growth on its sup+ parent MS1868. P22 derivatives and their hosts were provided by M. M. Susskind (University of Southern California) and A. R. Poteete (University of Massachusetts); the following mutant alleles were used: 4amH1368, 7amH1375 and 26amH204. Xp10 and X. campestris pv. oryzae were gifts from K. Severinov (Rutgers University). ø29 derivatives and M2 were grown in B. subtilis 168 strains RD2 (Su–) and Su44+ (Su+), provided by D. Anderson (University of Minnesota) and P. Guo (Purdue University). D. Anderson also provided GA-1 together with its host Bacillus G1R and phage SPP1. SPO1 was obtained from C. Stewart. S. aureus RN 4220 and its phages ø11 and ø85 were a gift from O. Schneewind (University of Chicago). T. Klaenhammer (N. Carolina State University) provided L. lactis Nck203 and phages ø949 and ø31, L. lactis MG1363 and phages øc2 and øsk1 and L. lactis R1c and its phage ør1t.
Other than λ, which was grown by induction of a lysogen, Gram-negative phages were grown lytically at 30°C or 37°C in rich medium (1% tryptone, 0.5% yeast extract and 0.5% NaCl) using appropriate host strains. B. subtilis phage strains were grown at 37°C on B. subtilis 168 derivatives, GA-1 on Bacillus G1R, using A416 media (2% tryptone, 1% yeast extract, 1% NaCl, 0.5% glucose, 0.5 mM MnCl2, 2.5 mM MgSO4), supplemented with 10 mM CaCl2 for SPP1. SPβ was induced from B. subtilis Su44+ using 15 µM mitomycin C (Sigma). The lactococcal phages ø949, ø31, øc2, øsk1 and ør1t were grown at 30°C on various L. lactis strains in M17 glucose media (Difco) containing 10 mM CaCl2. The staphylococcal phages ø11 and ø85 were absorbed to S. aureus in M17 glucose media at room temperature for 30 min and then propagated at 37°C.
After removal of bacterial debris from lysates, phages were concentrated by centrifugation at 75 000 g for 3 h or by precipitation with polyethylene glycol. Phages were allowed to resuspend at 4°C overnight in buffer and were then purified at least once by equilibrium density gradient centrifugation using CsCl dissolved in 10 mM Tris-HCl, pH 7.6, 10 mM MgCl2.
ø29 DNA–gp3 complex
The ø29 DNA–gp3 complex was isolated according to the procedure of Hermoso and Salas (1980). A suspension of ø29 particles was made in 4 M guanidinium chloride (pH 8.0) and incubated at 0°C for 1 h. Disrupted particles were then subjected to equilibrium CsCl density gradient centrifugation at 35 000 r.p.m. for 48 h at 4°C. The DNA band was collected and dialysed at 4°C against 50 mM Tris-HCl, pH 7.8, 10 mM MgCl2, 100 mM NaCl.
Preparation of peptidoglycan
Escherichia coli K-12 strain IJ1142 was grown to an A 600 = 1; cells were harvested, washed in 100 mM sodium phosphate, pH 6.8, and then boiled in 4% SDS for 45 min. Insoluble material was collected by centrifugation, washed three times with water and then resuspended in 10 mM MgCl 2 . After incubation with DNase, RNase and proteinase K, the insoluble peptidoglycan was washed extensively with water and stored at −20°C. Before use, the peptidoglycan suspension was sonicated at 220 W for 3 min on ice before addition to the other components of the zymogram.
A purified phage suspension, in the CsCl solution used for purification, was mixed with methanol and chloroform (1:1:0.75 by volume); after vigorous agitation, the layers were separated by centrifugation. The upper layer was discarded, taking care not to disturb the precipitated protein at the interface, and an equal volume of methanol was added to make a single liquid phase. Protein was collected by centrifugation at 14 000 r.p.m. in a microfuge for 6 min. The protein pellet was dried, resuspended in 1% SDS, 6% sucrose, 100 mM dithiothreitol (DTT), 10 mM Tris, pH 6.8, 0.0625% bromophenol blue and boiled for 3 min before loading onto 12% polyacrylamide gels. Gels were cast according to Laemmli (1970) except that only 0.01% SDS was used; in addition, zymogram gels contained 0.1% peptidoglycan (Hoyle and Beveridge, 1983; Bernadsky et al., 1994). After electrophoresis, zymograms were washed for 30 min with water and then soaked for 1–3 days at room temperature, normally in 150 mM sodium phosphate, pH 7.0, buffer containing 0.1% Triton X-100 and 10 mM MgCl2. Different renaturation conditions, including a pH range from pH 6.0 to 8.0, and various concentrations of both Triton X-100 (0.01–1%) and MgCl2 (1 mM to 50 mM) or EDTA (10 mM), were also tested to optimize peptidoglycan hydrolase activity. Zymograms were stained for 3 h with 0.1% methylene blue in 0.001% KOH and destained with water. Peptidoglycan hydrolase activity is detected as a clear zone in a dark blue background of stained peptidoglycan. Gels not containing peptidoglycan were stained with Coomassie brilliant blue.
P22gp4 was purified by SDS–PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane; a partial amino acid sequence of the protein was determined in a core facility.
Chloroform vapour-treated cell test
Spot tests were conducted according to the procedures of Streisinger et al. (1961) and Mosig et al. (1989). Log-phase cells of E. coli were plated in soft agar overlays on LB plates, incubated for 4–6 h at 37°C and then exposed to chloroform vapour overnight. Aliquots of 10–50 µl (≈ 1 × 1010 phage) of each sample in TM buffer (pH 7.8) were spotted on the agar surface and incubated at room temperature. Hen egg white lysozyme (50 ng) was used as a control.
We are grateful to the many people who generously provided phages and bacterial strains. D. Anderson is especially thanked for his interest in these studies. We thank Hansjorg Lehnherr and Shuji Kanamaru for providing information before publication, and Gail Christie for giving many helpful suggestions. This work was supported by grant GM32095 from the National Institutes of Health.