Streptococcus pneumoniae bacteriophages (phages) rely on a holin–lysin system to accomplish host lysis. Due to the lack of lysin export signals, it is assumed that holin disruption of the cytoplasmic membrane allows endolysin access to the peptidoglycan. We investigated the lysis mechanism of pneumococcal phage SV1, by using lysogens without holin activity. Upon phage induction in a holin deficient background, phage lysin was gradually targeted to the cell wall, in spite of lacking any obvious signal sequence. Our data indicate that export of the phage lysin requires the presence of choline in the teichoic acids, an unusual characteristic of pneumococci. At the bacterial surface, the exolysin remains bound to choline residues without inducing lysis, but is readily activated by the collapse of the membrane potential. Additionally, the activation of the major autolysin LytA, which also participates in phage-mediated lysis, is equally related to perturbations of the membrane proton motive force. These results indicate that collapse of the membrane potential by holins is sufficient to trigger bacterial lysis. We found that the lysin of phage SV1 reaches the peptidoglycan through a novel holin-independent pathway and propose that the same mechanism could be used by other pneumococcal phages.
The holin–lysin system is the main strategy adopted by bacteriophages to lyse the bacterial hosts at the end of the infective cycle in order to release their progeny (Young, 2005). The phage lysin is characterized by having peptidoglycan degrading activity, whereas holins are proteins with transmembrane domains that cause lesions in the host cytoplasmic membrane (Wang et al., 2000; Young, 2005). In phage λ, enzymatically active endolysin accumulates in the cytoplasm, since it lacks an intrinsic secretory signal sequence, gaining access to the peptidoglycan target when holins disrupt the cytoplasmic membrane at a defined time (Wang et al., 2000; Young, 2005). Although the holin–lysin paradigm was long though universal, it was recently demonstrated that holins are not an absolute requirement for lysin export to the cell envelope (São-José et al., 2000; Xu et al., 2004). The Lys44 lysin from oenococcal temperate phage fOg44 is translocated to the extracytoplasmic environment by the host Sec machinery involving proteolytic removal of its N-terminal signal peptide (São-José et al., 2000). Moreover, the lysins of Escherichia coli phages P1 and 21 have atypical signal sequences (SAR, signal-arrest-release) in the N-terminal domain that mediate their Sec-dependent translocation without cleavage (Xu et al., 2004; 2005; Park et al., 2007). Growing evidence indicates that other phage exolysins, mostly of Gram-negative hosts, possess secretion signals and therefore are likely to be exported also independently of holin function (Young, 2005; Kuty et al., 2010; Briers et al., 2011). However, secretory lysins require the action of phage-encoded holins for activity (Xu et al., 2004; 2005; Park et al., 2006; 2007; Nascimento et al., 2008). For instance, SAR lysins accumulate in the periplasm as enzymatically inactive proteins anchored to the membrane by their N-terminal SAR. Dissipation of the membrane proton motive force (pmf), an event achieved by the holins, releases the SAR domain thus generating the active, soluble form of the enzyme in the periplasm (Xu et al., 2004; 2005; Park et al., 2006; 2007).
Pneumococcal phages studied so far have typical holin–lysin cassettes in their genomes (Díaz et al., 1996; Martín et al., 1998; Obregón et al., 2003a; López and García, 2004). Nevertheless, there are few reported studies on the molecular mechanisms underlying this lysis system in Streptococcus pneumoniae. Functional analysis of the lytic Cp-1 and temperate Ej-1 phages showed that holins form lesions in the cytoplasmic membrane that complement the λ holin in producing phage plaques (Díaz et al., 1996; Martín et al., 1998). Furthermore, all pneumococcal phage lysins lack any known secretory signal sequence (López and García, 2004). Therefore, it was proposed that the lysin is released to its peptidoglycan substrate through the activity of the holin according to the canonical holin–lysin system (Martín et al., 1998; Haro et al., 2003; López and García, 2004).
Recently, we observed that in addition to the phage lysin, holins trigger the major pneumococcal autolysin LytA by compromising membrane integrity (Frias et al., 2009). LytA can remain inactive in the cell wall mainly attached to choline, a structural component of the cell wall teichoic acids (WTA) and cell membrane-linked lipoteichoic acids (LTA) (Höltje and Tomasz, 1975b; Briese and Hakenbeck, 1985; Díaz et al., 1989; Mellroth et al., 2012). Interestingly, the genes encoding pneumococcal phage lysins share as much as 87% identity with lytA, as demonstrated for the Hbl lysin of the HB-3 phage (Romero et al., 1990). Both phage and bacterial lysins also exhibit a bimodular structure with an N-terminal catalytic domain and a C-terminal choline-binding domain (López and García, 2004) and module shuffling between these lysins resulted in chimeric proteins that maintain the enzymatic function (Díaz et al., 1990; López and García, 2004). Moreover, constitutive expression of pneumococcal phage lysins Hbl or Cpl-1 in a strain lacking LytA restored the ability of pneumococci to undergo lysis in stationary phase and after exposure to deoxycholate (DOC), two cellular responses that are dependent on LytA activity (Romero et al., 1993). S. pneumoniae phage lysins could thus share with LytA similar cellular localization and physiological control mechanisms. These findings, taken together with the increasing recognition of holin-independent export mechanisms of phage lysins, led us to investigate the lysis system of pneumococcal bacteriophage SV1, particularly the targeting of the choline binding lysin Svl to the cell wall.
Deletion of holin function suppresses phage-mediated lysis
Although phages are very abundant among pneumococci (Bernheimer, 1979; Ramirez et al., 1999; Romero et al., 2009), little is known about their lysis system. In order to study how the holin–lysin mechanism operates in S. pneumoniae phage SV1, we constructed holin-deficient Δhol lysogens of the pneumococcal strain SVMC28 (Table 1).
The previously characterized lytic cassette of SV1 (GenBank Accession No. FJ65451) contains two genes, svh1 and svh2, just upstream of the svl gene encoding the phage lysin Svl (Frias et al., 2009). The putative Svh1 and Svh2 proteins (140 and 110 amino acid residues) show high amino acid sequence identity (> 80%) to the two putative holins of the S. pneumoniae temperate phages MM1 (Obregón et al., 2003a), VO1 (Obregón et al., 2003b) and Φ23782 (Croucher et al., 2011). Analysis of the predicted structure of the potential Svh holins using SOSUI (http://www.expasy.org) also reveals three and one potential hydrophobic transmembrane regions for Svh1 and Svh2, respectively (with an N-out, C-in topology), a characteristic of class I and III holins (Wang et al., 2000; Tran et al., 2005). Moreover, as previously reported, attempts to clone svh1 and svh2 in E. coli resulted in loss of viability (Frias et al., 2009). These characteristics of the Svh proteins strongly suggest that they correspond to the holins of phage SV1. Therefore, to create lysogens devoid of holin activity, both svh1 and svh2 genes were deleted.
Since holins trigger the activity of both the pneumococcal phage lysin and the autolysin LytA (Díaz et al., 1996; Frias et al., 2009), we then tested whether the elimination of holin activity actually prevented the activation of both lysins upon induction of the SV1 lytic cycle with mitomycin C (MitC). In the absence of holins, despite Svl and LytA expression, phage induction did not result in host lysis at the expected time or even when lysis was already completed in the holin-carrying strains (Fig. 1A). Except the strain lacking the holins and both phage and bacterial lysins (ΔholΔsvlΔlytA), cultures eventually showed lysis but only after a long MitC exposure. This is possibly related to the characteristic autolysis occurring in late stationary phase of S. pneumoniae (Tomasz et al., 1988) exacerbated by the presence of the phage lysin, which was previously shown to functionally replace LytA (Romero et al., 1993). However, it cannot be totally discarded that an extremely slow activation of the phage lysin or the bacterial autolysin occurs even in the absence of holins, contributing to this late lysis phenotype.
To confirm the abolishment of holin activity, membrane integrity was also examined by flow cytometry after phage induction. In this assay, cells with permeabilized membranes resulting from holin activity (Díaz et al., 1996; Frias et al., 2009) allow the uptake of propidium iodide (PI), fluorescing in the FL3 channel, while cells with non-permeabilized membranes internalize only Syto 9, fluorescing in FL1. Thus, gate R2 represents the permeabilized cells and gate R3 the non-permeabilized cells. In the ΔholΔlytA strain, MitC-treated cells are largely found in gate R3 (Fig. 1B), whether cells are analysed before (40 min), at the onset (80 min) or at the end (120 min) of the lytic process observed in strains possessing functional holins (Fig. 1A). Indeed, at 120 min approximately 96% of cells maintain membrane integrity. Similar results were obtained for strains ΔholΔsvlΔlytA, ΔholΔsvl and Δhol (Fig. S1). In contrast, the control ΔsvlΔlytA strain, with intact holin function, becomes increasingly permeabilized after phage induction and at 120 min bacteria were found mostly in gate R2 (Fig. 1B). Collectively, these results demonstrate the successful elimination of holin activity and further verify that holins are required to activate lysins.
Phage lysin is targeted to the cell wall independently of holin activity
The pneumococcal phage lysins, including Svl, exhibit a high structural similarity with the bacterial autolysin (López and García, 2004). LytA contains a choline-binding domain that attaches the protein to choline residues within the teichoic acids of the S. pneumoniae cell surface (Höltje and Tomasz, 1975b; Briese and Hakenbeck, 1985; Díaz et al., 1989) and it was shown previously that a 2% choline solution could compete for the binding sites leading to the release of LytA (Briese and Hakenbeck, 1985; Weiser et al., 1996). We decided to make use of these properties to investigate the location of the phage lysin. The effect of choline on Svl attachment was then tested in the absence of holin function to avoid possible holin-mediated phage lysin escape.
As shown in Figs 2A and S2A, when lysogens expressing Svl (ΔholΔlytA MitC+, cultures treated with MitC) were washed with 2% choline in PBS, 60 min after phage induction with MitC, Svl was found in the choline wash (Scholine) but it was not released from the cells by washing with PBS only (SPBS) or 2% NaCl solution in PBS (SNaCl), ruling out lysin extraction due to the high ionic strength of the choline solution. Moreover, cytoplasmic elongation factor Ts was not detected in any wash fraction excluding contamination with bacteria or cytoplasmic components due to bacterial lysis. Similar results were obtained with CodY, another known cytoplasmic protein previously used to evaluate cell lysis in S. pneumoniae (Price and Camilli, 2009). As a control, LytA was likewise extracted by the choline solution and found solely in Scholine (Figs 2B and S2B). Finally, pneumolysin (Ply), which was found at the cell surface attached by interactions not involving choline (Price and Camilli, 2009), was almost undetectable in the wash fractions Scholine and SPBS (Fig. S2), confirming the specificity of the choline wash in removing only choline-binding proteins such as Svl and LytA. Overall, these findings demonstrate a cell surface localization of the pneumococcal phage lysin, dependent on interactions with choline, and not restricted to the cytoplasm as predicted from its primary sequence. This surface localization in the absence of holins also indicates that holins are not required for the transport of phage lysin across the S. pneumoniae cytoplasmic membrane.
To further explore the localization of phage lysin during the SV1 lytic cycle, ΔholΔlytA lysogens grown in C+Y were treated with MitC and samples were collected at different time points and assayed in the cell pellet (P) and choline wash fractions (Scholine) for the presence of Svl by Western blot. We found that the protein was present in cells at all times including immediately when MitC is added (0 min) (Fig. 3A), which is possibly due to low-level spontaneous phage induction occurring in the cultures (Fig. S3A) (Bossi et al., 2003). However, the cellular concentration of Svl increased in a gradual fashion until the time of lysis at 80 min (Fig. 3A). The same samples were also analysed for the cytoplasmic accumulation of Ts. Only a small increase in Ts was observed. This was expected since phage induction by MitC arrests cell division but cells tend to elongate and the culture continues to increase in mass (Suzuki et al., 1967; Ramirez et al., 1999). By analysing the ratio of total cellular Svl over Ts for cells grown in C+Y, we determined that at the end of the lytic cycle Svl expression is indeed 6.7-fold higher (Fig. 3D). We were also able to elute with choline increasing amounts of Svl with time (Fig. 3A). Again, PBS washes did not promote Svl removal and Ts did not reveal cell lysis since none was detected in any of the wash samples. Immunoblotting intact bacteria spotted onto nitrocellulose membranes further confirmed that Svl accumulated along the lytic cycle at the surface of ΔholΔlytA MitC+ grown in C+Y, whereas the choline-containing teichoic acids (TA), which bind Svl, the cell wall-associated Ply and the choline-binding proteins PspA and LytC, were similarly displayed at the surface at all times (Fig. 3E). No signal was detected for the intracellular protein Ts, except in the control of disrupted cells at 80 min. Taken together, the results show a continuous holin-independent targeting of the phage lysin to the cell wall during the lytic cycle.
It is possible that some phage-mediated lytic events, occurring before the normal lysis time, could release intracellular Svl that would become available to bind choline at the surface of intact cells. Moreover, besides LytA, other bacterial choline binding lysins such as LytC could also contribute to this process (García et al., 1999; Eldholm et al., 2009). We therefore addressed whether export of the phage lysin to the cell wall is due to cell lysis. With this in mind, ΔholΔlytA cells were treated with MitC after suspension in C+Y containing 2% choline, which prevents cell wall attachment of phage and pneumococcal lysins and consequently prevents cell lysis (Fig. 3C versus Fig. 1A) (Giudicelli and Tomasz, 1984; García et al., 1987; Mellroth et al., 2012). Samples of equal volume were then collected over time and the cell pellet (P) as well as the culture medium supernatant (Sculture) were tested for Svl presence. Similarly to the observed pattern of Svl extraction by choline wash (Fig. 3A), a clear phage lysin accumulation was observed in the culture medium (Fig. 3B) that still continues beyond the normal lysis timing (Fig. S3B), as confirmed by normalization to the respective cellular Ts. As expected, cytoplasmic Ts was not detected in the culture medium indicating inhibition of cell lysis (Fig. 3B). In line with this, in the presence of 2% choline in the growth medium the total cell-associated phage lysin remained constant during the major part of the lytic cycle (up to 60 min) increasing only by 2.7-fold at 80 min when normalized by Ts (Fig. 3B and D). This corresponds to the amount of Svl present intracellularly. Indeed, bacterial surface analysis by dot blot revealed that the cell wall association of Svl during the lytic cycle was inhibited by 2% choline, in contrast to growth in C+Y, despite TA anchors being present at the surface at all times (Fig. 3E). As controls, the choline-binding proteins PspA and LytC were not detected at the surface, contrary to Ply that displayed a similar distribution to that observed in C+Y growth. In addition, the absence of signal for Ts confirmed that we were assaying intact bacteria (Fig. 3E). These data indicate that accumulation of Svl in the cell wall during the lytic cycle is not due to cell lysis. Instead, the phage lysin is continuously exported through a different mechanism.
Involvement of teichoic acids in phage lysin export
Choline is an essential pneumococcal growth factor that is taken up from the medium and used exclusively in the biosynthesis of WTA and LTA (Tomasz, 1967; Briles and Tomasz, 1973; Fischer et al., 1993). Presumably, TA precursors are loaded with choline intracellularly and then flipped across the cytoplasmic membrane and linked to the cell surface (Zhang et al., 1999; Damjanovic et al., 2007). Since choline anchors the phage lysin to the cell wall, we hypothesized that choline loaded teichoic acids might be involved in its extracytoplasmic targeting. The structural analogue ethanolamine can replace choline in vitro, being incorporated instead of choline into TA (Tomasz, 1968). However, in ethanolamine-grown pneumococci, most choline-binding proteins are no longer able to bind to the cell surface (Briese and Hakenbeck, 1985; Yother and White, 1994; López and García, 2004). Therefore, to explore the potential TA contribution to phage lysin export, we determined the Svl cellular localization in MitC-treated ΔholΔlytA cells grown in a chemically defined medium containing ethanolamine (Cden-et). Culture samples were then collected at 180 min and presence of Svl on the surface was assayed by dot blot as well as in cell pellets (P) and culture medium supernatants (Sculture) by Western blot.
Comparing to MitC-treated cultures of ΔholΔlytA grown in Cden with choline (Cden-cho), showing a similar growth profile until 180 min (Fig. 4A), Svl was not present at the surface of cells grown in Cden-et, although TA could still be detected at similar levels on the surface of bacteria grown in the presence of either choline or ethanolamine (Fig. 4B). As expected, in the presence of ethanolamine, the choline-binding protein PspA was not surface located (Yother and White, 1994) and the ethanolamine-containing TA did not influence the surface localization of Ply, which is not dependent on choline for cell wall association (Price and Camilli, 2009). Surprisingly, with ethanolamine in the culture medium, Svl was not found in the Sculture fraction of ΔholΔlytA MitC+ (Fig. 4C). The phage lysin, absent from the cell surface, was not released into the culture medium during the lytic cycle and was only detected in the cell fraction, in contrast to what happens in the presence of choline. Again, Ts established that cell lysis did not occur under both growth conditions. Confirmation that the phage lysin was indeed kept intracellularly in the presence of ethanolamine was obtained by phage induction in Cden-et of the holin-carrying ΔlytA strain that is able to permeabilize the membrane. This resulted in the release of a considerable amount of Svl to the culture medium (Fig. 4C) indicating that the holin lesions are sufficiently large to allow Svl to escape. In agreement, the cytoplasmic protein Ts was also detected in the Sculture of ΔlytA but not in that of ΔholΔlytA. Substitution of ethanolamine for choline in the medium caused, however, substantial PspA release from the ΔholΔlytA cells into the culture medium. Differently from Svl, the PspA protein contains a signal peptide and is therefore expected to be exported to the cell wall regardless of TA, but accumulates in the medium since it can no longer be retained at the cell surface. Thus, taken together, our data demonstrate that the export to the cell wall of the phage lysin, which does not include recognizable signal sequences for surface targeting, involves the pneumococcal teichoic acids in a process requiring the presence of choline.
Pmf dissipation is essential and sufficient for triggering phage-mediated lysis
To understand how the phage lysin activity is regulated at the cell surface, we first examined throughout the lytic cycle the ability of the detergent DOC to induce lysis, since it is known to trigger LytA-mediated lysis (Tomasz et al., 1988) and, similarly to the phage holins, permeabilizes the pneumococcal membrane (Frias et al., 2009). Therefore, DOC was added at several time points after MitC-phage induction to the ΔholΔlytA strain lacking the major autolysin activity but containing a functional phage lysin. This resulted in a gradual increase of lysis with time reaching a substantial percentage from 60 min onwards (Fig. 5A) while virtually no DOC-mediated lysis is observed at any time point in MitC-treated ΔholΔsvlΔlytA cultures (Fig. 5B). This reflects the progressive accumulation of pneumococcal phage lysin during the lytic cycle as determined by immunoblotting analysis (Fig. 3A). DOC induced significant lysis even before it is triggered by holins at 80 min (Fig. 5A and C). These results, together with the observed Svl surface location, suggest that phage lysin accumulation in the cell wall does not by itself cause lysis and confirm that lysis is triggered by membrane perturbations. Thus, as expected, an intact membrane is involved in phage lysin control by preventing premature host lysis.
To further determine what type of membrane perturbation induces phage-mediated lysis, we next tested the effect of collapsing the cytoplasmic membrane electrochemical gradient on lysis induction. For this purpose, we used the membrane pmf-dissipating agent N,N'-dicyclohexylcarbodiimide (DCCD) that inhibits the ATPase thus depleting the proton gradient (Jolliffe et al., 1981). As shown in Fig 6A, addition of DCCD to ΔholΔlytA cultures 60 min after MitC-phage induction triggered complete host lysis. In contrast, DCCD did not induce lysis in MitC-treated cultures of ΔholΔsvlΔlytA, which differ from ΔholΔlytA only in the absence of a functional phage lysin. These data indicate that, in the absence of holins, pmf disruption activates the phage lysin. In support of these results, we observed negligible lysis of MitC-untreated ΔholΔlytA in the presence of DCCD. This minimal lysis is likely due to activation of Svl accumulated through spontaneous phage induction (Fig. S3A). Likewise, DCCD treatment triggered lysis of ΔholΔsvl (expressing LytA) independently of MitC treatment (Fig. 6A). Thus, pmf dissipation also activates the autolysin LytA, known to be holin-activated in phage release (Frias et al., 2009).
Even though DCCD is characterized by pmf-dissipating properties, we also investigated its potential effect on membrane permeabilization. To address this, ΔholΔsvlΔlytA cells were challenged with DCCD at 60 min following MitC addition and immediately collected for flow cytometry analysis (0 min) or after 140 min. As expected, immediately upon exposure to DCCD the cells remained almost exclusively contained in gate R3, similarly to the untreated control (at 0 min, Fig. 6B). In contrast, DOC caused a rapid permeabilization of cells, resulting in two well-defined populations evenly distributed in both gates when a 50% mixture of DOC and DCCD-treated cells was analysed (Fig. 6C). Even after 140 min of DCCD exposure (corresponding to 200 min of MitC treatment, Fig. 6A), when DCCD-induced lysis is nearly complete in the presence of either LytA or Svl, we observed only a small fraction of cells in gate R2 (Fig. 6B). This distribution pattern was also observed at 140 min for the ΔholΔsvlΔlytA control treated with DCCD but not with MitC (Fig. 6C). Similar results were obtained in MitC-untreated cultures of ΔholΔlytA (data not shown).
To evaluate if this small fraction of cells in gate R2 with potentially permeabilized membranes had released proteins from the cytoplasm into the medium, ΔholΔsvlΔlytA cells were suspended in C+Y with 2% choline to prevent possible lytic events and treated with DCCD 60 min after MitC was added to the cultures. As a control, cells were treated with DOC also after 60 min of MitC addition. The presence of the cytoplasmic proteins Ts (43 kDa) and CodY (28 kDa), which have molecular masses bracketing the ∼ 36 kDa of Svl and LytA, was then monitored in the culture medium by immunoblotting. With increasing time from DCCD addition, both CodY and Ts remained almost undetectable up to 140 min, but DOC was found to cause a substantial release of both proteins within only 20 min of addition (Fig. 6D). Together, these findings show that lysis induced by DCCD was not the result of membrane permeabilization allowing Svl to reach the peptidoglycan. Instead, DCCD triggers the phage lysin, as well as the autolysin, already translocated and bound to the cell wall. It is this proportion of lysin surface associated that then initiates the lytic process. Thus, our data indicate that holin-mediated membrane pmf dissipation is required and sufficient for lysins activation.
Some secreted phage lysins that are activated by pmf collapse are associated with pinholins that, as opposed to canonical holins, form membrane lesions not large enough for lysins to pass through (Park et al., 2007). Growth experiments in Cden medium with ethanolamine already indicated that SV1 holin lesions are sufficiently large for its 36 kDa lysin counterpart to cross the membrane since leakage of Svl into the medium only occurred in the presence of holins (Fig. 4C). Indeed, holins also allowed the release of the 43 kDa intracellular Ts (Fig. 4C). We performed similar experiments in C+Y supplemented with 2% choline that confirmed these conclusions (Fig. S4). Our results indicate that the control of lysis by membrane pmf collapse is achieved by large-lesion-forming holins that also allow the passage of a fraction of cytoplasmically trapped phage lysin.
SV1, like all other S. pneumoniae phages, relies on a holin–lysin system to achieve host lysis and release the new phage particles at the end of the lytic cycle (Díaz et al., 1996; Martín et al., 1998; Obregón et al., 2003a; López and García, 2004; Frias et al., 2009). Here we provide evidence that SV1 lysin is targeted to the cell wall without requiring holins, in contrast to the proposed model for pneumococcal phages in which endolysins accumulated in the cytoplasm are released by holin-mediated membrane disruption (Martín et al., 1998; Haro et al., 2003; López and García, 2004). We also show that SV1 lysin uses a novel surface translocation mechanism. Differently from other described phages with secretory lysins, the pneumococcal phage lysin is unusual in its cell wall targeting since it lacks any known signal element for membrane translocation such as SAR sequences (Xu et al., 2004; Park et al., 2007). In agreement with the absence of signal peptides, the phage lysin present in the cell wall apparently does not suffer protein processing since it was indistinguishable by SDS-PAGE from its cytoplasmic form, unlike the observed proteolytic cleavage of Oenococcus oeni phage fOg44 lysin (São-José et al., 2000). In the context of phages, lack of known secretory sequences and a concomitant extracytoplasmic pool was only reported for the mycobacteriophage Ms6 lysin, which is assisted in its export by a chaperone-like protein (Gp1) encoded in the complex lytic operon of Ms6 (Catalão et al., 2010). However, SV1 as well as other pneumococcal phage lytic cassettes do not encode other known functions besides holins and lysins (Díaz et al., 1996; Martín et al., 1998; Obregón et al., 2003a; López and García, 2004; Frias et al., 2009) and analysis of their genomes also did not reveal Gp1-related proteins. Likewise, no such proteins were detected in any of the pneumococcal genomes currently available. We found that Svl export is peculiarly dependent on the host choline-containing teichoic acids that compose the cell surface, whereas in the exolysin-carrying phages studied so far including Ms6, lysin access to the peptidoglycan involves the SecA pathway (São-José et al., 2000; Xu et al., 2004; Catalão et al., 2010).
It was shown that, during TA biosynthesis, choline is incorporated into the teichoic acid precursors that are presumably flipped across the cytoplasmic membrane and then incorporated into the cell wall (Zhang et al., 1999; Damjanovic et al., 2007). Since we found that choline is crucial for intracellular Svl to be translocated across the membrane, we thus speculate that the phage lysin may be co-transported with TA to the cell surface implicating interdependence between Svl export and TA biosynthesis. The exact mechanism of such an unusual protein translocation remains to be elucidated. The high similarity between pneumococcal phage lysins and the autolysin, including the absence of any motifs or signals for protein secretion (Díaz et al., 1989; López and García, 2004), indicate that these could share the same export pathway. Although recently it was suggested that accumulation of LytA on the surface could be attributed to lysis and loading from without (Mellroth et al., 2012), several other lines of evidence are consistent with LytA sharing the same TA-dependent transport mechanism. In accordance with our data for the phage lysin, it was reported that when pneumococci were grown in Cden medium containing ethanolamine LytA was not found in the cell envelope (Díaz et al., 1989). Even though it was recognized that LytA could not be bound by ethanolamine containing TA, no amidase activity was detected in the medium (Díaz et al., 1989), suggesting that LytA was also not accessible to the surface from without. In another independent experiment, LytA was found in the cytoplasm of pneumococci grown in the presence of ethanolamine (Briese and Hakenbeck, 1985). Interestingly, it was shown that LytA is found preferentially associated to the central growth zone where the formation of nascent peptidoglycan occurs (Díaz et al., 1989; Mellroth et al., 2012). However, how the autolysin localizes specifically to this region remains unknown but nascent teichoic acid chains are incorporated into the cell wall at this central cell wall growth zone (Briles and Tomasz, 1970; Tomasz et al., 1971). The location of LytA could thus be explained if its transport was associated with TA biosynthesis as suggested here for phage lysins. Further research is required to clarify this and if this mechanism could also apply to the transport of the few other pneumococcal choline-binding proteins lacking specific targeting signals (Tettelin et al., 2001). Curiously, when expressed in E. coli, LytA was found to be weakly attached to the outer face of the inner membrane, thus not requiring interactions with choline to remain associated with the cellular envelope (Díaz et al., 1989). The reasons underlying such behaviour when expressed in a heterologous system remain unknown and may not reflect the natural behaviour of the protein in S. pneumoniae.
Here we have shown that in S. pneumoniae choline is essential for phage-mediated lysis, not only for anchoring it to the cell surface until holin triggered lysis but also for its continuous translocation to the surface. This suggests a fined tuned adaptation of the phage to its specific host since choline is an unusual component of bacterial surfaces (Tomasz, 1967). Moreover, building up an increasing amount of phage lysin already close to its substrate, rather than targeting it to the cell wall only at the end of the lytic cycle, may ensure a more rapid cell lysis once the lytic activity is triggered.
The data we presented show that pmf disruption achieved by the holin permeabilizing effect on the membrane triggers lysin activation, similarly to what happens in other phages encoding exolysins (Xu et al., 2004; Park et al., 2007; Nascimento et al., 2008; Catalão et al., 2010). The possibility that Svl is continuously degrading the cell wall but is not able to overcome the bacterial cell wall repair pathways until they are disrupted by holin action cannot be completely discarded. However, the observation that in the absence of holins Svl still continues to accumulate at the cell surface beyond the normal timing of lysis without observable lysis suggests that Svl must be inactive while accumulating in the cell wall. Furthermore, since the lytic activity of LytA was shown to be inhibited while surface-associated (Mellroth et al., 2012), it seems unlikely that Svl-mediated lysis is a consequence of unbalanced cell wall degradation over synthesis. Nevertheless, it remains unknown how membrane pmf suppresses Svl activity or how pmf dissipation activates Svl. Since we showed that LytA is also sensitive to the membrane energy status, pmf may influence in the same way the activities of both lysins. It is generally accepted that the interaction with the membrane-bound lipoteichoic acids regulates LytA activity (Höltje and Tomasz, 1975a; Briese and Hakenbeck, 1985) and it was also reported that LTA inhibit the activity of the phage Cp-1 lysin (García et al., 1987). Furthermore, it was suggested that membrane potential imposes a specific conformation on the LTA on the surface of Gram-positive bacteria since, for instance, variations of the ionic strength of the medium result in LTA conformational changes (Neuhaus and Baddiley, 2003). It should also be considered that the cell wall of Bacillus subtilis has a low local pH sustained by the membrane pmf and its abolishment results in cellular lysis (Calamita and Doyle, 2002), raising the possibility that pH changes may specifically influence pneumococcal lysin control. Thus, loss of the membrane pmf promoted by the holins may induce structural and chemical rearrangements in the cell envelope that may abolish the inhibitory function of LTA over the lysins. Alterations of such an association could also potentially result from stalling of growth or the specific arrest of cell wall synthesis, which has been reported to activate LytA (Tomasz and Waks, 1975; Mellroth et al., 2012). We therefore suggest that the TA-dependent transport may also constitute a type of regulatory mechanism of the phage lysin, by ensuring that the potentially lethal activity of the phage lysin once targeted to the cell wall remains inactive since its transport is coupled to its regulatory molecule.
The holin lesions could also release some activator factor from the membrane. However, preliminary experiments do not seem to support this hypothesis. Indeed, lysis was not induced when ΔholΔsvl or ΔholΔlytA cultures treated with MitC for 80 min, and thus expressing LytA or Svl, were supplied with the supernatant of ΔsvlΔlytA cells (carrying functional holins) that were treated with MitC for 140 min (at which time, lysis is completed in the parental strain culture) (data not shown). However, further experiments are required to fully clarify these issues.
As reported for other S. pneumoniae phages (Martín et al., 1998; Haro et al., 2003), we found that SV1 holins form lesions large enough to allow lysin passage. This is not surprising since pmf-activated exolysins not associated with pinholins (that form channels only sufficiently large for ions to pass through), have already been described (Park et al., 2007). For instance, the holin of phage P1, which is paired with the secreted SAR lysin Lyz, is in fact a canonical holin (Park et al., 2007). Curiously, the Gp4 holin of the mycobacteriophage Ms6 has in its primary sequence features of pinholins, although it is able to promote the release of the cytoplasmic λ lysin but not of the respective Ms6 lysin (Catalão et al., 2011). Inspection of the amino acid sequence of both SV1 holins did not reveal an N-terminal transmembrane SAR domain (rich in residues that are weakly hydrophobic) characteristic of the pinholin of phage 21 (Park et al., 2006). Since lysis promoted by Svl in the presence of holins was more effective than that triggered by the pmf-dissipating agent DCCD (Figs 1A and 6A), we propose that holin lesions also directly facilitate the release of phage lysin from the cytoplasm, which reaches a peak of intracellular concentration at the normal lysis time of 80 min (Fig. 3B). Likewise, holin-activated LytA, which contributes significantly to phage progeny release (Frias et al., 2009), may also escape the cytoplasmic compartment through holin lesions since, besides being cell wall localized it was also found intracellularly in a considerable proportion (Fig. S2B, Pcholine) (Briese and Hakenbeck, 1985; Mellroth et al., 2012). The released lysins could thus accelerate cell wall degradation at the end of the lytic cycle when the cells are no longer protected from lysin activities due to holin-induced pmf collapse.
In conclusion, we have firmly established the exolysin's nature of a S. pneumoniae phage lysin. The remarkably high structural and functional similarities between Svl and all other known pneumococcal phage lysins, including the presence of choline-binding domains and lack of signal sequences (López and García, 2004), suggest that the same surface translocation mechanism can be used by the majority of pneumococcal phages and is not unique to SV1.
Pneumococcal strains and growth conditions
Streptococcus pneumoniae strains used in this study (Table 1) are derivatives of strain SVMC28 (serotype 23F, sequence type 36), which is a clinical isolate lysogenic for phage SV1 obtained from the Rockefeller University collection (A. Tomasz). Pneumococcal strains were grown at 37°C in tryptic soy agar (TSA) (Oxoid, Hampshire, England) supplemented with 5% sterile sheep blood in a 5% CO2 atmosphere, in the casein-based semisynthetic medium C+Y without aeration (Lacks and Hotchkiss, 1960) or in a chemically defined medium (Cden) (Mosser and Tomasz, 1970) with choline at 5 μg ml−1 (Cden-cho) or ethanolamine at 40 μg ml−1 (Cden-et). Et-containing pneumococci were serially grown in Cden-et for at least 10 consecutive times. Where necessary, 200 μg ml−1 kanamycin (Kn), 100 μg ml−1 streptomycin (Sm), 2 μg ml−1 erythromycin (Ery) and/or 4 μg ml−1 chloramphenicol (Cm) (Sigma, Steinheim, Germany) were added.
General DNA techniques
Routine DNA manipulations were performed according to standard methods (Sambrook et al., 1989). PCR primers are listed in Table S1. Phage DNA was isolated similarly to already described procedures (Su et al., 1998). PCR reactions for the purpose of constructing mutant strains were carried out with High fidelity PCR enzyme Mix kit (MBI Fermentas, Vilnius, Lithuania). PCR products were purified using the High Pure PCR product purification system (Roche, Mannheim, Germany). All oligonucleotides were obtained from Invitrogen (Paisley, Scotland). Nucleotide sequences were analysed using VECTOR NTI Deluxe (Invitrogen, Barcelona, Spain) software.
Construction of mutant strains
Prior to the construction of mutant strains, the S. pneumoniae parental strain SVMC28 was tested for growth in the presence of Kn and Sm by seeding in TSA supplemented with the antibiotics at the appropriate concentration (200 μg ml−1 and 100 μg ml−1 respectively). The Kn susceptibility and Sm resistance background in SVMC28 allowed the usage of Janus (kan-rpsL+) cassette (Sung et al., 2001), in a two-step transformation procedure for holin elimination.
First, strains harbouring a substitution of the hol genes (svh1 and svh2) by the kan-rpsL+ cassette were constructed following the procedure previously described (Sung et al., 2001). A PCR fragment containing the region immediately upstream of svh1 was amplified from SV1 phage DNA with primers AF_H and AR_H_C, and the kan-rpsL+ cassette was amplified from a PCR fragment containing the cassette (kindly provided by D. Morrison) with the Kan5 and DAM351 primers (Table S1). Following purification, the two fragments were mixed and connected to generate fragment A by PCR with primers AF_H and DAM351. A fragment downstream of svh2 gene (containing just 11 bp of svh2) was then amplified from SV1 DNA with the BF_H_C/BR_H primer pair (Table S1), purified and mixed with fragment A for assembly into a unique product through PCR amplification with primers AF_H and BR_H. Strains SVMC28Δhol::kan-rpsL+ and SVMC28Δhol::kan-rpsL+ΔlytA were obtained following transformation of strains SVMC28 and SVMC28ΔlytA respectively, with the final purified PCR product by selection for Knr transformants (Table 1).
Next, the kan-rpsL+ cassette was deleted as follows. The PCR fragment containing the region immediately upstream of svh1 was amplified with primers AF_H and AR_H_C2 and the fragment downstream of svh2 (with 11 bp of svh2) was amplified with primers BF_H and BR_H. Fragments assembly into a unique product was achieved with the AF_H and BR_H primer pair (Table S1). Strains SVMC28Δhol and SVMC28ΔholΔlytA (without the entire cassette) were obtained following transformation of strains SVMC28Δhol::kan-rpsL+ and SVMC28Δhol::kan-rpsL+ΔlytA respectively, with the final purified PCR product generating Kns Smr transformants selected with Sm (Table 1). All strains harbouring a deletion in lytA, were also selected for chloramphenicol resistance. Besides drug selection, mutants resulting from each transformation were further confirmed by PCR with primers AF_H and BR_H. Subsequent DNA sequencing confirmed holins deletion and the integrity of phage lysin svl gene.
A similar strategy was used to construct the mutant strain SVMC28ΔholΔsvl (Table 1). The PCR fragment containing the region immediately upstream of svh1 and the kan-rpsL+ cassette were amplified and connected as before (fragment A). A fragment downstream of svl gene (containing 185 bp of svl) was amplified from SV1 DNA with the 2BFSVL_C/2BR_SVL primer pair (Table S1), and assembled with fragment A with primers AF_H and 2BR_SVL. Transformation of SVMC28 with the final PCR product and selection for Knr transformants resulted in strain SVMC28ΔholΔsvl::kan-rpsL+. The kan-rpsL+ cassette was then deleted as follows. The region immediately upstream of svh1 and the region downstream of svl (with 185 bp of svl) were amplified with primers pairs AF_H/AR_H_C3 and 2BFSVL/2BRSVL respectively (Table S1). Fragments assembly into a unique product was generated with AF_H and 2BRSVL (Table S1). Strain SVMC28ΔholΔsvl (without the entire cassette) was obtained following transformation of strain SVMC28ΔholΔsvl::kan-rpsL+ with the final purified PCR product by selection for Smr transformants (Table 1). Additionally to drug selection, mutants resulting from each transformation were further confirmed by PCR and subsequent DNA sequencing.
The mutant strain SVMC28ΔholΔsvlΔlytA (Table 1) was constructed using the strategy previously described to delete the lytA gene (Frias et al., 2009) in the SVMC28ΔholΔsvl background followed by selection for Sm and Cm resistance. Gene deletion and flanking regions were verified by sequencing.
Transformation of pneumococci was carried out as described (Frias et al., 2009). Lysis upon DOC addition and after culture growth until late stationary phase confirmed the presence of LytA activity in SVMC28Δhol and SVMC28ΔholΔsvl. In strains SVMC28ΔholΔlytA and SVMC28ΔholΔsvlΔlytA, a non-lytic phenotype in both situations confirmed the absence of LytA activity.
Growth and lysis assays
To monitor growth and lysis, pneumococci were grown overnight at 37°C in C+Y or in Cden containing choline or ethanolamine supplemented with the appropriate antibiotics, and diluted 1:100 in fresh medium (without antibiotics). Cultures were kept at 37°C for the rest of the incubation period and the OD600 was measured. In experiments of induction of the phage lytic cycle, MitC (Sigma, Steinheim, Germany) was added when cultures reached OD600 0.2–0.25 to a final concentration of 0.1 μg ml−1 (Otsuji et al., 1959). For DOC activation of phage lysin Svl, MitC-induced cultures grown in C+Y were treated with DOC [0.04% (w/v)] (Sigma, Steinheim, Germany) at different time points. In membrane pmf-dissipating assays, DCCD (Sigma, Steinheim, Germany), an ATPase inhibitor (Jolliffe et al., 1981), was added to cultures grown in C+Y after 60 min of MitC addition at a final concentration of 100 μM. Whenever cultures were not MitC-induced, DCCD was added 60 min after the cultures reached OD600 0.2–0.25, when MitC is added in treated cultures. To prevent lysis, in other experiments bacteria were grown in C+Y and at OD600 0.2–0.25 the cultures were sedimented, the supernatants were discarded and the cells were suspended in the same volume of C+Y supplemented with 2% choline and then challenged with MitC.
Flow cytometry assays
Cultures grown in C+Y were treated with MitC and/or DCCD as described above and cells were collected after exposure at defined intervals and then diluted in sterile-filtered 0.85% NaCl to a concentration of ∼ 1 × 106 cells ml−1. Cell viability was assessed by using the Live/Dead BacLight bacterial viability kit (Invitrogen, Carlbad, USA) as previously described (Frias et al., 2009). As a control for cell death, cultures were treated with DOC. Samples were analysed on a Partec CyFlow space flow cytometer (Partec GmbH, Münster, Germany) with 488 nm excitation from a blue solid-state laser at 50 mW. Green fluorescence (Syto 9), indicating the population of cells without permeabilized cytoplasmic membranes, was detected in the FL1 channel and red fluorescence (PI), indicating the population of cells with permeabilized cytoplasmic membranes, was detected in the FL3 channel. Optical filters were set up such that FL1 measured at 520 nm and FL3 measured above 610 nm. The sample analysis rate was kept below 1000 events s−1. Twelve thousand events were collected for each sample taken. Data were collected and analysed by using FloMax software (Partec GmbH, Münster, Germany) and gates representing the permeabilized cells (gate R2) and non-permeabilized cells (gate R3) were constructed as previously described (Frias et al., 2009).
Preparation of pneumococcal proteins for Western blot analysis
Cells were grown in C+Y at an OD600 of 0.2–0.25 and then induced with MitC or left untreated. Samples (7 ml) were taken at the indicated time points after MitC treatment. In the case of untreated cultures, the samples were collected at the same time points after the culture reached OD600 0.2–0.25. Cells were harvested by centrifugation (3200 g, 4°C, 10 min), washed once with 0.5 volumes of PBS 1× (PBS 10× pH 7.2, Gibco, Invitrogen, Paisley, Scotland) and suspended in 200 μl of Tris 50 mM pH 7.5 (this cell pellet fraction was designated P). For choline wash, PBS washed cells were gently suspended in 200 μl of 2% choline chloride (w/v) (Sigma, Steinheim, Germany) prepared in PBS 1× and incubated 30 min at 4°C without agitation. Bacteria were collected by centrifugation (3200 g, 4°C, 15 min) and the supernatant was filtered through a 0.2 μm low-binding-protein membrane (DISMIC-03CP, Toyo Roshikaisha, Japan) to ensure the removal of all bacteria. This choline wash fraction was designated Scholine. As control, cells were incubated in the same conditions with PBS 1× or NaCl 2% (w/v) prepared in PBS 1× and the wash fractions were designated SPBS and SNaCl respectively. The pellet was then washed once with 0.5 volumes of PBS 1× and suspended in 200 μl of Tris 50 mM pH 7.5. Since cells were previously washed with choline, this cell pellet fraction was called Pcholine.
To prepare culture medium fractions (Sculture), bacteria suspended in C+Y with 2% choline or grown in Cden with choline or ethanolamine were treated with MitC at OD600 0.2–0.25 and harvested (7 ml of sample) at the indicated time points. After centrifugation, supernatants were collected, filtered through 0.22 μm membrane filters and concentrated 35-fold (final volume of 200 μl) by ultrafiltration on Amicon Ultra-15 centrifugal filter units (cut-off 10 kDa, Merck Millipore, Cork, Ireland). As before, the cell pellet fraction (P) was washed once with PBS 1× and suspended in 200 μl of Tris 50 mM pH 7.5. To test the membrane permeabilization effect of DCCD over time, cultures grown in C+Y with 2% choline were also challenged with DCCD 60 min after MitC addition and the samples (7 ml) collected were processed as before. Treatment with DOC for 20 min in the same conditions was used as a control.
For SDS-PAGE electrophoresis, 5–15 μl of P fractions and 45 μl of S fractions were boiled for 5 min with loading buffer containing 10% β-mercaptoethanol at a final concentration of 1×. Because growth in the presence of ethanolamine leads to formation of long chains (Tomasz, 1968), to compensate for differences in protein concentration between samples from growing cultures in Cden with choline or ethanolamine, viable counts were determined and the samples were adjusted at 3 × 109 cfu ml−1 for P fractions and at 1 × 109 cfu ml−1 to obtain S fractions prior to loading onto the gel. Samples were electrophoresed in 5% acrylamide stacking and 12% separating gels, proteins were electrotransferred to 0.20 μm nitrocellulose membranes (Whatman GmbH, Dassel, Germany) and subjected to immunodetection. Protein molecular mass marker Precision Plus Protein Standard was used (Bio-Rad, California, USA).
Dot blots and immunodetection
For dot blot analysis, bacteria harvested at different time points after MitC treatment were washed twice with cold PBS 1× and suspended in 200 μl of PBS 1×. As before, for cells grown in Cden with either choline or ethanolamine the bacterial suspensions were further adjusted at 1 × 109 cfu ml−1. In all experiments, 2.5 μl of bacterial suspension was spotted on nitrocellulose membranes and cell surface-exposed proteins were detected using specific antibodies.
Phage lysin detection was performed using an antibody against pneumococcal autolysin LytA that recognizes Svl due to the high similarity to LytA (89.3% identity at the amino acid level). Polyclonal antibodies (pAb) anti-LytA and anti-LytC were kindly provided by P. García (García et al., 1982; 1999). Monoclonal antibody (mAb) 144,H-3, which recognizes streptococcal elongation factor Ts, and pAb anti-PspA were provided by J. Kolberg (Kolberg et al., 1997) and S. Hammerschmidt (Kolberg et al., 2006) respectively. PAb antibody to Bacillus subtilis CodY was a gift of A. Sonenshein (Ratnayake-Lecamwasam et al., 2001), mAb against pneumococcal pneumolysin (Ply) and pAb anti-TA were purchased from Statens Serum Institute (Copenhagen, Denmark). Anti-LytA (rabbit), anti-PspA (rabbit) and anti-CodY (rabbit) were used at a dilution of 1:10000. Anti-Ply (mouse), anti-TA (rabbit), anti-LytC (rabbit) and 144,H-3 anti-Ts (mouse) antibodies were used at a dilution of 1:1000, 1:2000, 1:3000 and 1:5000 respectively. Secondary antibodies anti-rabbit or anti-mouse IgG conjugated to horseradish peroxidase (Santa Cruz Biotechnology, California, USA) were diluted 1:2500 or 1:5000. After incubation, blocking and washing procedures, blots were developed using the Pierce ECL Western blotting substrate (Thermo Fisher Scientific, Rockford, USA), according to the manufacturer's instructions, and exposed to Amersham Hyperfilm ECL (GE Healthcare, Buckinghamshire, UK).
We thank Dr Thomas Hänscheid for helpful assistance during the flow cytometry assays, Inês Domingues for aid with the Western blot procedure and Dr Donald Morrison for providing kan-rpsL+ cassette. We are also grateful to Dr Pedro García (Centro de Investigaciones Biológicas, Madrid), Dr Jan Kolberg (Norwegian Institute of Public Health, Oslo), Dr Abraham Sonenshein (Tufts University, Boston) and Dr Sven Hammerschmidt (Interfaculty Institute for Genetics and Functional Genomics, Germany) for the generous gift of LytA, LytC, Ts, CodY and PspA antibodies. We thank Dr Carlos São-José for helpful discussions.
M.J.F. was supported by Grants SFRH/BD/38543/2007 and SFRH/BPD/79621/2011 from the Fundação para a Ciência e a Tecnologia, Portugal. This work was partially supported by Fundação para a Ciência e Tecnologia, Portugal (PIC/IC/83065/2007), the European Union (CAREPNEUMO – Combating antibiotic resistance pneumococci by novel strategies based on in vivo and in vitro host–pathogen interactions, FP7-HEALTH-2007-223111) and Fundação Calouste Gulbenkian. Dr José Melo-Cristino has received research grants administered through his university and received honoraria for consulting and serving on the speakers bureaus of Pfizer, Bial, GlaxoSmithKline and Novartis. Dr Mário Ramirez has received honoraria for consulting and serving on speakers bureau of Pfizer. The other authors declare no conflict of interest. No company or financing body had any interference in the decision to publish.