Summary
Temperate phages are pervasive in bacterial genomes, existing as vertically-inherited islands called prophages. Prophages are vulnerable to the predation of their host bacterium by exogenous phages. Here we identify BstA, a novel family of prophage-encoded phage defence proteins found in diverse Gram-negative bacteria. BstA drives potent suppression of phage epidemics through abortive infection. The bstA-encoding prophage itself is not inhibited by BstA during lytic replication due to a self-immunity mechanism driven by the anti-BstA (aba) element, a short stretch of DNA within the bstA locus. Phage-targeting by distinct BstA proteins from Salmonella, Klebsiella and Escherichia prophages is functionally interchangeable, but each possesses a cognate aba element. The specificity of the aba element ensures that immunity is exclusive to the replicating prophage, and cannot be exploited by heterologous BstA-encoding phages. BstA allows prophages to defend their host cells against exogenous phage attack, without sacrificing their own lytic autonomy.
Introduction
The perpetual battle between bacteria and their viruses (phages) has driven the evolution of a diverse array of phage defence systems in bacteria (Bernheim and Sorek, 2020; Hampton et al., 2020; Houte et al., 2016; Rostøl and Marraffini, 2019. Conversely, it is increasingly recognised that phages have evolved many mechanisms to subvert these defence systems (Maxwell, 2017; Samson et al., 2013; Trasanidou et al., 2019). Though perhaps the most intuitive form of phage defence involves the direct rescue of an infected cell, for example by the targeted degradation of phage nucleic acids by CRISPR-Cas, or restriction modification systems, many phage-defence systems in fact function solely at the population level. In a mechanism conceptually analogous to the pathogen-stimulated programmed cell death driven by the innate immune systems of higher organisms (Abedon, 2012), population-level defence systems prevent phage infection spreading across entire populations, at the cost of the lives of infected cells.
Population-level phage defence systems are often grouped under the umbrella term “abortive infection” (Abi) (Labrie et al., 2010; Lopatina et al., 2020) and represent diverse mechanisms to prevent phage replication and cause cell death. These include protease-mediated inhibition of cellular translation (Bingham et al., 2000), toxin-antitoxin pairs (Fineran et al., 2009) and cyclic oligonucleotide signalling (Cohen et al., 2019). Recently, it has been proposed that certain CRISPR-Cas systems function through abortive infection (Meeske et al., 2019; Watson et al., 2019). Such mechanistic diversity and prevalence of abortive infection systems in bacteria underscores the selective advantage this strategy imparts in the battle against phages.
However, an important sub-plot in the bacteria-phage conflict is the pervasive existence of so-called “temperate” or “lysogenic” phages within bacterial genomes. Temperate phages are able to stably exist within the bacterial chromosome as latent, vertically inherited islands known as a prophages. Crucially, to find new hosts, prophages must escape from the bacterial genome and return to the lytic life-cycle (spontaneously, or in response to specific molecular cues). The prophage-state imposes unique existential pressures, wherein the fitness of the phage is indefinitely tied to that of the bacterium. Indirectly enhancing their own fitness, prophages frequently encode “moron” or “accessory” loci that modulate the biology of host bacteria (Bondy-Denomy and Davidson, 2014; Fortier and Sekulovic, 2013; Howard-Varona et al., 2017) in a phenomenon likened to altruism (Shub, 1994). Prophage accessory loci are frequently associated with bacterial pathogenesis, and many notorious bacterial pathogens depend on prophage-encoded toxins and virulence factors to cause disease (Brüssow et al., 2004; Fortier and Sekulovic, 2013). However, another trait conferred by prophages that can significantly increase bacterial fitness, is resistance against bacteriophage attack. Indeed, recent work has shown that prophage accessory genes may represent an underexplored reservoir of phage-defence systems (Bondy-Denomy and Davidson, 2014; Dedrick et al., 2017; Snyder, 1995).
Here, we report a novel phage defence system driven by the BstA protein and encoded by prophages of diverse Gram-negative bacteria. When a bacterium harbours a bstA-encoding prophage, BstA protein confers effective population-level phage defence through abortive infection. The bstA locus includes an anti-BstA element, which can suppress the activity of BstA protein to allow the native prophage to undergo lytic replication. We propose that such elegant molecular systems have evolved to allow prophages to defend their host cells from predatory phages, without compromising their own potential for lytic replication.
Results
The BTP1 prophage-encoded bstA gene mediates phage resistance
Salmonella enterica subsp. enterica serovar Typhimurium (S. Typhimurium) strain D23580 encodes the ∼40 kb prophage BTP1 (Figure 1A) (Owen et al., 2017). An operon within BTP1, the gtr locus (gtrACBTP1), has been shown to confer resistance against phage P22 due to chemical modification of the cellular lipopolysaccharide (LPS), which is the receptor for phage P22 (Kintz et al., 2015). Unsurprisingly therefore, we found deleting the BTP1 prophage from strain D23580 (D23580 ΔBTP1) made the strain highly susceptible to infection by phage P22, confirming that resistance to phage P22 is conferred by the BTP1 prophage (Figure 1B). However, inactivation of the gtr locus of prophage BTP1 (D23580 Δtsp-gtrACBTP1) did not restore sensitivity to phage P22 to the level of D23580 ΔBTP1 (Figure 1B), suggesting a second phage resistance system existed on this prophage. Our previous transcriptomic study showed that the bstA gene was highly-expressed during lysogeny, and therefore a candidate for modulation of the biology of the S. Typhimurium cell (Owen et al., 2020). The bstA gene, encoded downstream of the prophage repressor locus, has been implicated in both virulence, and anti-virulence, of Salmonella isolates, but no functional mechanism has been proposed (Herrero-Fresno et al., 2014, 2018; Spiegelhauer et al., 2020). We hypothesised that bstA was the second element in the BTP1 prophage that conferred defence against phage P22.
Consistent with this hypothesis, removal of the bstA gene from prophage BTP1 (D23580 ΔbstA), did dramatically increase susceptibility to phage P22. To confirm that phage resistance was directly mediated by BstA protein, we introduced two stop codons into the beginning of the bstA coding sequence by exchanging 4 nucleotides (D23580 bstASTOP) (Figure 1B). D23580 bstASTOP was highly susceptible to P22 phage, to the same level as D23580 ΔbstA, demonstrating that BstA protein mediates defence against phage P22. Simultaneous deletion of the gtr locus and inactivation of the BstA protein (D23580 Δtsp-gtrACBTP1 bstASTOP) fully recapitulated the susceptibility to phage P22 achieved by deleting the entire BTP1 prophage (D23580 ΔBTP1), indicating that resistance to phage P22 is solely mediated by the bstA and gtr loci in prophage BTP1. We reproduced these findings by replicating phage P22 in liquid culture, demonstrating that reduction of plaque formation by BstA truly reflected phage replication suppression (Supplementary Figure 1A).
To investigate whether the function of the BstA protein depended on other elements on the BTP1 prophage, we constructed an inducible expression system in S. Typhimurium strain LT2. LT2 is the type strain of S. Typhimurium, and is natively susceptible to many phages, including P22. Expression of the BstABTP1 protein in S. Typhimurium LT2 from within a neutral position on the chromosome (LT2 tetR-PtetA-bstA) conferred a high degree of resistance to P22 and other phages including ES18 and 9NA (Figure 1C; Supplementary Figure 1B). Expression of the derivative containing two stop codons at the beginning of the bstA coding sequence (bstASTOP) did not confer any phage resistance, demonstrating again that the effect is mediated by BstA protein (Supplementary Figure 1C). However, BstA did not mediate resistance against all phages tested: Det7, Felix O1, and notably, phage BTP1 (which encodes the bstA gene) were unaffected by expression of BstA, both at the level of plaque assay and replication in liquid culture (Figure 1C; Supplementary Figure 1B). We could not detect any pattern in the characteristics of phages that were sensitive or insensitive to BstA protein that could relate the mechanistic action of BstA protein.
BstA represents a novel family of prophage-encoded phage defence proteins in diverse Gram-negative bacteria
Having established that BstA functions as a prophage-encoded phage defence system, we sought to further characterise the occurrence of this protein in bacteria. We found BstA protein homologs in the genomes of diverse Gram-negative bacteria (Supplementary Table 1) and we compiled a dataset of 72 homologs representative of phylogenetic diversity. The majority (79%) of these BstA homologs co-occurred with phage genes, and therefore were designated as putatively-prophage associated (Figure 2A). No known phage-associated genes were found in the vicinity of 21% (15 of 72) of BstA homologs, which were considered to be putatively prophage-independent. A small subset of BstA homologs were plasmid-encoded, and this group included both putatively prophage-associated and -independent homologs (Figure 2A). Strikingly, in many cases, BstA homologs were located downstream of putative prophage repressor proteins, mirroring the genetic architecture of BstABTP1 (Figure 2B). The BstA protein appears to be highly associated with prophages of Gram-negative bacteria.
Whilst the BstA protein does not exhibit sequence homology to any functionally-characterised proteins, remote homology detection methods revealed a putative KilA-N domain in the N-terminal region (residues 32-147 of BstABTP1)) (Figure 2C). Though poorly characterised, the KilA-N domain is found in proteins from phages and eukaryotic DNA viruses, and contains the helix-turn-helix motif characteristic of DNA binding proteins (Iyer et al., 2002; Medina et al., 2019). A large screen based on genomic proximity to known-phage defence systems previously predicted KilA-N domain containing proteins to play a role in phage defence (Doron et al., 2008). The KilA-N domain derives its name from the kilA gene product of bacteriophage P1 (lethal when expressed in E. coli) but kilA has no known function in phage infection biology (Hansen, 1989), and no bacterial KilA-N domain-containing proteins have been functionally characterised to date.
Certain residues in the BstA protein are highly conserved amongst homologs from diverse members of the Alpha-, Beta-, and Gamma-Proteobacteria (Figure 2C; Supplementary Figure 2). A small number of BstA protein homologs (all found in Cyanobacterial plasmids) only exhibited homology to the N-terminal, putative KilA-N domain. A second small group corresponding to Proteobacteria, Cyanobacteria and a single Bacteroidetes isolate were only homologous to the C-terminal region of BstA (shown at the bottom on the alignment in Figure 2C). Such bipartite protein homology suggests that the BstA protein is composed of two functional domains. Additionally, we observed that the identity of BstA protein homologs did not obviously correlate with bacterial taxonomy. For example, homologs from enteric bacteria closely-related to Salmonella (members of the Gammaproteobacteria) sometimes shared less identity to BstABTP1 than homologs from alpha- and betaproteobacteria, suggesting BstA proteins are frequently horizontally transferred, perhaps consistent with their association with prophages.
We selected two diverse BstA homologs from Klebsiella pneumoniae (48.4% amino acid identity to BstABTP1) and E. coli (41.7% identity) to investigate the phage-resistance function of the larger BstA protein family (the native genetic context of these homologs is illustrated in Figure 2B, and their identity to BstABTP1 is highlighted in the alignment in Figure 2C). We engineered inducible expression systems for the BstA the expression construct we previously validated for BstABTP1 (Figure 3A,B; Figure 1C). Expression of BstAKp and BstAEc in S. Typhimurium LT2 conferred resistance to Salmonella phages at a similar level to BstABTP1, despite these homologs only sharing around 40% identity at the amino acid level (Figure 3A,B; Supplementary Figure 3A). Importantly, BstAKp and BstAEc showed additional activity against phage BTP1 (which encodes bstABTP1), a property that was not shown by BstABTP1.
Finally, to investigate the phage-defence function of BstA against well-characterised coliphages, we expressed BstABTP1 and BstAEc in E. coli. Heterologous expression of BstABTP1 in E. coli strain MG1655 conferred resistance to phage λ, ϕ80, P1 and T7, but did not affect phages T4 and T5 (Supplementary Figure 3B,C). Surprisingly we found BstAEc was slightly less active against coliphages than BstABTP1 (Supplementary Figure 3B,C). We note that replication in liquid culture is a more reliable and reproducible measure of phage susceptibility than plaque assay, and we frequently observed stronger resistance phenotypes liquid replication than by plaque assay (Supplementary Figure 3B,C).
We conclude that BstA represents a novel family of phage-resistance proteins associated with prophages in diverse Gram-negative bacteria.
BstA mediates effective population-level phage defence through abortive infection
Phage resistance systems function via diverse functional mechanisms (Hampton et al., 2020; Rostøl and Marraffini, 2019). We used a microscopy-based approach to understand how BstA mediates phage-resistance. P22 phages were used to infect Salmonella cells with and without native BstABTP1 function, at high multiplicity of infection (MOI) in order to ensure that most cells were infected. We were surprised to observe that independent of BstABTP1 function, all cells lysed within the time course of 3 hours (Figure 4A, Supplementary Video 1), and BstABTP1 function did not appear to confer any direct protection from phage infection. We conducted the same experiment in liquid culture, measuring phage replication and the fraction of surviving cells post phage infection. In cells possessing functional BstA (D23580 Δtsp-gtrAC), phage P22 Δc2 completely failed to replicate (Figure 4B). In contrast, in the absence of BstA function (D23580 Δtsp-gtrAC bstASTOP), the phage replicated >100-fold. However, despite preventing the replication of phage P22, BstABTP1 had no effect on cell survival: independent of BstABTP1 function only 1-2% of cells survived (Figure 4C). We hypothesised that BstA protein must instead mediate phage defence at the population level.
To investigate whether BstA protein mediated population-level phage defence, we conducted a second microscopy experiment, wherein approximately only 1 in every 1000 cells was infected with phage P22. Unlike during liquid culture, in our microscopy setup cells were immobilised on agarose pads, which restricts the movement of phage particles to local diffusion. Using this setup, we tracked the spread of infection as primary infected cells lysed and produced secondary infections. To facilitate tracking of phage replication, we used a reporter phage engineered to encode the red fluorescent protein mCherry within the early lytic operon (P22Δc2 p-mCherry), so that fluorescence signal indicated phage replication (Figure 4D).
In the population lacking functional BstABTP1 (D23580 Δtsp-gtrAC bstASTOP), primary infected cells lysed after around 30 minutes (Figure 4D, Supplementary Video 2). Subsequently, the red fluorescence signal was observed in neighbouring cells revealing secondary infection, followed by cell lysis, a cycle which repeated until all cells in the radius of the primary infected cell had lysed, reminiscent of plaque formation (Figure 4C). The impact of the epidemic of phage infection upon bacterial cells lacking BstABTP1 can be seen clearly in Supplementary Video 2.
In contrast, in the population with native BstA activity (D23580 Δtsp-gtrAC) no fluorescence signal or lysis was observed in neighbouring cells following the lysis of the primary infected cells. Instead, neighbouring cells continued to grow, eventually forming a confluent lawn (Figure 4D, Supplementary Video 2). The lack of subsequent rounds of secondary infection after the primary cell lysis events indicates that viable phage were not released upon cell lysis. Furthermore,
Taken together, these experiments demonstrate that BstA protein inhibits successful phage replication, but does not prevent the death of the infected cell. BstA therefore provides phage defence at the population-level and prevents the spread of phage epidemics. We propose that BstA is a novel abortive infection system: a population-level phage defence system that inhibits phage infection at the cost of cell viability.
BstA protein responds dynamically to phage infection and co-localises with phage DNA
To explore the molecular activity of BstA during phage infection, we constructed a translational fusion of the BstABTP1 protein to superfolder green fluorescent protein (sfGFP), confirming that the translational fusion did not compromise the function of the BstA protein (Supplementary Figure S4). We used time-lapse fluorescence microscopy to observe the dynamics of BstA protein inside individual cells during infection with two phages P22 and 9NA. In the absence of phage infection, BstA protein was distributed diffusely within the cytoplasm of the cells, suggesting no particular sub-cellular localisation (Figure 5A, Supplementary Video 3). However, approximately 20 minutes after infection with phages P22 and 9NA, we consistently observed BstA protein aggregating into discrete foci towards the centre of infected cells (Figure 5B, Supplementary Video 3). Cell lysis occurred approximately 40 minutes after the formation of BstA foci.
We speculated that the dynamic establishment of foci by BstA in response to phage infection was likely to reflect the mechanistic activity of the BstA protein. We noticed that the dynamics of the foci formed by BstA proteins during phage infection were reminiscent of live-cell fluorescence microscopy studies of phage replisomes (Cenens et al., 2013; Trinh et al., 2017). We therefore speculated that the focus of BstA protein in phage infected cells might correspond to the replicating phage DNA. To test this hypothesis, we used a ParB-parS system to track the sub-cellular localisation of phage DNA relative to BstA protein. We inserted a parS site into the P22 phage chromosome, and expressed a ParB-mCherry fusion protein inside cells already expressing BstA-sfGFP. ParB protein oligomerises onto DNA at parS sites, and therefore parS-tagged DNA is indicated by ParB-mCherry foci. We conducted a microfluidic infection experiment to co-locate BstA foci and infecting P22 phage DNA, and we observed that the position of ParB-mCherry foci (corresponding to phage P22 DNA) clearly overlapped with foci formed by BstA-sfGFP (Figure 5C, Supplementary Video 4). The microscopy data suggest that BstA protein interacts with the replicating DNA of infecting phages.
Together the data are consistent with a model where BstA proteins are titrated to sites of phage replication inside infected cells, to suppress the successful multiplication of the phage.
BstA phage resistance systems contain anti-BstA elements (aba) that suppress the activity of BstA
When characterising the sensitivity of different phages to the activity of BstABTP1 using our heterologous expression system (Figure 1C), we observed that BTP1 phage, (which itself encodes the bstABTP1 gene) was not affected by expression of BstABTP1 (schematised in Figure 6A). We hypothesized that BTP1 carries an anti-BstA determinant: a self-immunity factor that allows phage BTP1 to replicate without being targeted by its own abortive infection protein. Consistent with this hypothesis, phage BTP1 became sensitive to BstABTP1 expression when the bstA coding sequence was deleted (BTP1 ΔbstA). (Figure 6B). The self-immunity function of the bstA locus was not affected by the introduction of the double stop codon mutation into the beginning of the coding sequence (as described in Figure 1B), indicating that self-immunity is not mediated by the BstA protein itself, but by an alternative genetic element encoded within the bstA locus (Figure 6B). Here, and for the duration of this report, we define the bstA “locus” as the region including the bstA coding sequence and the 5’ upstream sequence.
To identify the precise genetic basis of BstA self-immunity, we constructed a library of BTP1 mutant phages, carrying truncations of different lengths from the 3’ end of the bstA locus (Supplementary Figure 5A) and screened these phages for their ability to replicate in the presence of BstABTP1 expression. Self-immunity was preserved in all the mutant phages (i.e. insensitivity to BstABTP1 expression) except the mutant with the largest bstA truncation (BTP1 bstAΔ24), in which just the first 24 bp of the bstA reading frame were intact (Supplementary Figure 5A). A similar truncation mutant containing just the first 34 bp of bstA (BTP1 bstAΔ34) retained immunity to BstA, indicating that the anti-BstA determinant was encoded within the upstream region and first 34 bp of the bstA gene. Consistently, the transfer of bstAΔ34 (the first 34 bp of bstA, including the upstream sequence) to phage P22 (P22 bstAΔ34I), conferred BstA immunity (Supplementary Figure 5B). To identify the minimal sequence required for BstA self-immunity, we constructed further P22 bstAΔ34I-derived phages, successively truncating the transferred sequence from the 5’ end (P22 bstAΔ34I - P22 bstAΔ34V, Supplementary Figure 5B). We discovered that a 63 bp sequence (GCCCGCCACACTTTAACAAGGAAAATCAAATGGTTAATCAGATAAGGTCCATATCACCCCGCC) spanning 29 bp of the upstream region and the first 34 bp of bstA (start codon underlined) was necessary and sufficient to confer the self-immunity (Figure 6C). We designated this 63 bp element ‘aba’, for anti-BstA. Supplying the 63 bp aba sequence on the high-copy number pUC18 plasmid (pUC18-aba) rescued P22 phage replication in the presence of BstA protein, demonstrating that the self-immunity function of aba functions in trans (Supplementary Figure 6A).
The aba element is DNA-based
In the native BTP1 prophage, the aba sequence overlaps the start of the bstA gene, preventing the mutational disruption of the aba element without modifying the BstA protein sequence. We therefore used the simple multi-copy plasmid trans-complementation system (wherein the BstA protein and the aba sequence are independently encoded) to further probe the function of the aba sequence (Supplementary Figures 6A). A notable feature of the 63 bp aba sequence is the presence of a direct “CCCGCC” repeat at the terminal ends, which we hypothesised might be functionally important. Single nucleotide exchange of the CCCGCC→CCCTCC in the first (abamut1) and second (abamut2) repeat abolished the self-immunity function of the aba element on the phage and in trans (Supplementary Figure 6B) showing that the aba terminal direct repeats are important for the aba-BstA interaction.
The aba plasmid trans-complementation system additionally allowed us to interrogate the genetic nature of the aba element, which we hypothesised could be either DNA, RNA transcript or peptide-based. Though three short open reading frames exist within the aba sequence, (Supplementary Figure 6C), non-synonymous mutations of the reading frames did not ablate aba function, suggesting the aba-driven immunity is not mediated by a short peptide. Secondly, we tested whether transcription of aba was necessary for immunity. The aba sequence was cloned in either orientation into the Salmonella chromosome, downstream of the arabinose-inducible PBAD promoter (D23580 ΔΦ tetR-bstABTP1 PBAD-aba; Supplementary Figure 6D) to produce high levels of aba RNA transcripts. High-level transcription of aba RNA did not restore P22 or 9NA plaque formation in the presence of BstA protein, suggesting transcription of aba is not required for anti-BstA activity. However, this experiment also demonstrated that supplying aba as a single copy on the chromosome also did not confer self-immunity (Supplementary Figure 6D), suggesting that aba can only suppress BstA protein when it is encoded on high-copy replicative elements. Further mutational disruption of the aba sequence revealed that the self-immunity function was not robust to mutation at multiple sites in the 63 bp sequence (Supplementary Figure 6E). Collectively, our data suggest that aba-driven suppression of BstA is neither peptide or transcript mediated, and instead support a model where BstA suppression is mediated by aba DNA in trans, perhaps by a copy-number dependent mechanism.
The aba element prevents the bstA-encoding prophage from aborting its own lytic replication
A unique feature of the BstA system, unusual amongst the majority of mechanistically-characterised abortive infection systems, is its frequent occurrence on prophages (latent forms of active phages) (Figure 2A). Prophages must be able to switch to lytic replication, or else the prophage-state becomes an evolutionary dead-end for the phage. It follows therefore, that prophages must be obligately immune to any self-encoded phage-defence systems. Despite this intuitive assumption, self-immunity functions in prophage-encoded phage defence systems have not been previously identified.
We hypothesised that the primary biological role of the aba element is to allow the endogenous bstA-encoding phage to escape BstA-mediated inhibition upon induction from the prophage-state. To test this, we measured the level of induction of prophage P22 in the presence of heterologously expressed BstABTP1 protein (Figure 6D). In the absence of BstABTP1, the P22 prophage generated a titer of ∼4×109 PFU/mL after 5 hours growth with inducing agent (mitomycin C, MitC). However, with BstABTP1 expression, the MitC-induced titer of P22 dropped >300-fold to ∼1×107 PFU/mL, consistent with BstA-mediated inhibition of P22 phage replication. Transfer of the aba sequence to prophage P22 (P22 aba) significantly increased the induced titer in the presence of BstABTP1 to ∼3×109, restoring it to the level seen in the absence of BstA, showing that the aba element rescues phage replication via suppression of BstA.
Consistent with our hypothesis, we found the abamut1 mutation (exchange of a single functionally important nucleotide in the terminal direct repeat) reduced the MitC-induced titer of phage BTP1 ∼14-fold in the presence of BstA. This reduction was rescued by supplying the functional aba sequence in trans on the pUC18 plasmid (Figure 6E), confirming that the abamut1 mutation ablates the function of the aba element.
These intricate experiments demonstrate that an aba element is required for the bstA-encoding prophage to switch from lysogenic to lytic replication. In the absence of aba, the bstA-encoding prophage suffers replication inhibition by its own BstA protein (self-targeting), presumably by the same abortive infection mechanism that inhibits exogenous phage infection.
Possession of a functional aba element, which specifically suppresses BstA activity during prophage replication (perhaps via a copy-number dependent mechanism), is required to allow the prophage to protect its host from phage infection during lysogeny, whilst permitting its own lytic replication upon induction.
Distinct BstA proteins are associated with cognate aba elements
Finally, we determined whether the aba sequence from bstABTP1 could suppress the activity of the BstA proteins of other bacteria. We challenged the P22 bstABTP1 phage (immune to expression of BstABTP1 due to the presence of aba) against expression of BstAEc or BstAKp. The bstABTP1 locus did not protect P22 from the heterologous BstA proteins, indicating that each BstA protein has a cognate aba element (Figure 6F). To further test this hypothesis, we engineered P22 phages to encode either bstAEc or bstAKp (including the respective upstream sequence). Consistent with a cognate BstA-aba interaction, P22 bstAEc became specifically immune to expression of BstAEc; and P22 bstAKp gained specific immunity to BstAKp expression (Figure 6F). Each bstA locus therefore encodes highly-specific self-immunity. We conclude that though BstA proteins are broadly functionally interchangeable in terms of phage-defence activity, each bstA locus contains a cognate aba element, inactive against heterologous BstA proteins. The specificity of the aba self-immunity element means that heterologous bstA-encoding phages cannot bypass BstA-mediated abortive infection, and aba-mediated suppression of BstA is exclusive to the induced prophage.
BstA protein inhibits the replication of phage DNA
Sequence-based analysis of BstA protein homologs suggested that the N-terminal domain may bind DNA (Figure 2C), and fluorescence microscopy showed BstA protein interacting with phage DNA. The replication of DNA is a crucial for phage morphogenesis, as new copies of the phage chromosome are required for packaging into capsids. To test whether BstA protein inhibits phage DNA replication in a manner that can be suppressed by aba, we conducted Southern blot experiments to monitor levels of phage DNA during infection. Using our prophage-negative, inducible BstA-expression strain (D23580 ΔΦ tetR-PtetA-bstABTP1), we first tested the replication of the BstA-sensitive virulent phage, 9NA. In the absence of BstA expression, the level of phage 9NA DNA gradually increased over a 50 minute infection time course, indicating successful phage replication (Figure 7A). No accumulation of phage 9NA DNA was observed in the presence of BstABTP1, suggesting that BstA protein strongly inhibited the replication of phage DNA.
Consistent with the self-immunity function of aba, BTP1 phage DNA replication was not affected by the expression of BstABTP1, unless the aba element was non-functional (BTP1 abamut1) (Figure 7B). Likewise, successful replication of phage P22 DNA in the presence of BstABTP1 only occurred when the phage possessed a functional aba element (Figure 6B).
To confirm that BstA protein inhibits DNA replication, we constructed small phage-derived plasmids (‘phagemids’) based on the phage P22 replication module (Figure 7C). P22 phagemid derivatives that included the functional 63 bp aba sequence (pP22-aba), or the non-functional abamut1 sequence (pP22-abamut1) were made.
The P22 phagemids were transformed into Salmonella cells in the presence or absence of BstABTP1 protein expression. In the absence of BstA, the stable replication of all three P22 phagemids in Salmonella cells generated >106 transformants /ng phagemid. However, expression of BstABTP1 reduced the transformation efficiency of pP22 (lacking the aba sequence) to around 101 transformants /ng. Addition of the aba sequence to the phagemid (pP22-aba) restored the transformation efficiency of the phagemid in the presence of BstA to BstA-negative levels.
We conclude that phage DNA replication is strongly suppressed by BstA, but replication can be rescued by the aba element, presumably by inhibition of BstA protein. As replicated phage DNA is an essential substrate for packaging into phage capsids, the inhibition of DNA replication is likely to prevent the production of infectious progeny phages. We propose that BstA blocks phage replication by inhibiting DNA replication, a process that can be suppressed by the native prophage with the aba self-immunity element.
Discussion
Prophages (latent phages residing within the genomes of bacteria) frequently encode accessory genes that bestow beneficial functions on their host bacteria (Bondy-Denomy and Davidson, 2014). A function that can significantly increase bacterial fitness in many environments is phage resistance, and prophages may represent a large reservoir of uncharacterised phage defence systems (Dedrick et al., 2017; Snyder, 1995).
Here, we have discovered a novel family of prophage-encoded abortive infection proteins (BstA) which efficiently defend bacterial populations from phage epidemics. BstA protein is constitutively expressed inside cells that carry the prophage, and provides effective population-level phage defence through abortive infection, inhibiting phage replication at the cost of the viability of individual infected cells. Possession of such innate phage defence systems by active prophages imposes an obvious challenge: the prophage must avoid self-targeting by its own defence system when switching to lytic replication.
We realised that the native prophage which encodes BstA required a mechanism to counteract the protein upon induction from the prophage-state, to avoid aborting its own lytic replication. The BstA system solves this problem with the aba element (anti-BstA), a co-encoded short DNA sequence that specifically suppresses the activity of BstA protein upon prophage induction, giving the induced prophage self-immunity to BstA. Theoretically, such a system might leave BstA-expressing cells vulnerable to infection by heterologous BstA-encoding phages, which could use their own aba element to bypass native BstA. This problem is avoided by each BstA protein being suppressed only by its cognate aba element, ensuring that BstA supression is specific to the native BstA-encoding prophage.
Despite having been studied for over 60 years, abortive infection systems remain mysterious, and very few have been characterised more deeply than the level of broad functional mechanism (Labrie et al., 2010). Here, we present a complete high-level picture of the BstA phage defence system, and the corresponding anti-BstA aba element. We are left with two major questions regarding the BstA protein. Firstly, what are the phage determinants for BstA sensitivity? Though BstA was active against approximately 50% of the phages tested, we did not to detect similarities between BstA-targeted and non-targeted phages that could reflect the molecular determinants of sensitivity. It remains possible that rather than responding to a physical phage stimulus, such as phage DNA or protein, BstA protein responds to a cellular stimulus produced by the infection of specific types of phages.
Secondly, what is the precise molecular mechanism by which BstA protein inhibits phage DNA replication? Our data show that phage DNA does not replicate in the presence of BstA. The existence of a putative DNA-binding domain in BstA proteins, and microscopic observation of BstA co-localisation with phage DNA makes it tempting to speculate that BstA interacts physically with phage DNA, for example by occlusion of a replication initiation site. In the well-characterised phages P22 and Lambda, DNA replication initially occurs bi-directionally from the origin of replication, generating circular θ-form intermediates (Weigel and Seitz, 2006). These circular forms act as templates for subsequent rolling-circle replication, generating the long concatemeric phage chromosomes needed as the substrate for the packaging machinery of most tailed phages (Fujisawa and Morita, 1997). We speculate that the timing of BstA focus formation, approximately 20 minutes after phage infection, is consistent with the formation of circular θ-form phage DNA. It is possible that BstA specifically inhibits the switch to rolling circle DNA replication, thereby limiting the availability of linear phage chromosomes for packaging into progeny phage capsids. Such a mechanism has been proposed to explain how phage satellites are able to block replication of the ICP1 phage in Vibrio cholerae (Barth et al., 2020). Furthermore, numerous Abi systems in Lactococcus have been proposed to interfere with phage DNA replicative functions (Chopin et al., 2005), though the molecular mechanisms have not been characterised.
Alongside the mechanistic details of the BstA protein that have yet to be established, little is known about the interaction of BstA with the aba element. Our data show that aba interacts with BstA in DNA form, but the mechanism by which aba DNA suppresses BstA protein is unclear. Our data indicate that multiple copies of the aba element can suppress BstA protein activity in trans, which could explain why aba function is specific to prophage induction (when prophage DNA is replicated to a high copy number). Further study of the BstA-aba system is required to resolve the precise molecular mechanisms by which BstA-encoding prophages, such as BTP1, achieve self-immunity.
We consistently observed that phage-infected cells underwent cell lysis independent of the activity of BstA protein. However we cannot be certain whether BstA protein causes cell lysis actively or passively. Abi systems have frequently been termed “altruistic suicide” systems, which mediate “programmed cell death” in response to phage infection (Abedon, 2012; Shub, 1994). Whilst perhaps a useful conceptual analogy for the strictly population-level effect of Abi systems, this narrative implies that Abi systems actively cause cell death. Though this is likely to sometimes be the case, such as in the CBASS system (Cohen et al., 2019) or toxin anti-toxin based systems (Fineran et al., 2009), Abi can also be achieved by simple disruption of the phage replication pathway. Because phage lysis is generally a temporally programmed event that occurs independently of successful virion morphogenesis (Cahill & Young, 2018), phage-mediated cell lysis can occur in the absence of virion assembly. For example many Lactococcus Abi target aspects of phage replication, such as AbiZ, which may interact with phage holin proteins, to stimulate premature cell-lysis before virion assembly is completed (Durmaz and Klaenhammer, 2007).
It is possible that BstA protein simply inhibits infectious phage particle formation, for example by precluding the formation of concatemeric DNA for packaging, whilst allowing the phage lysis pathway to proceed unperturbed. As we did not observe a difference in the timing of cell lysis for phage infected cells in the presence or absence of BstA during microscopy studies, we prefer a model where the BstA protein allows the phage lytic pathway to proceed as normal to cell lysis, except that infectious progeny phage particles are not released.
An intriguing feature of the BstA phage defence system is its tight association with prophages, and specifically, with the prophage repressor locus. Though we found homologs in diverse Gram-negative bacteria, the genetic architecture of the bstA locus (i.e. lying downstream of, and presumably sharing the promoter of the prophage repressor) was strikingly conserved. The region between the repressor (cI) and n gene of lambdoid phages has previously been identified as a hotspot of mosaic diversity (Degnan et al., 2007). In fact, the corresponding site in phage Lambda harbours the rexAB genes, perhaps the most widely studied prophage-encoded abortive infection system (Snyder, 1995). Despite >60 years of research, the molecular mechanisms governing RexAB activity are poorly understood. RexB is reported to be an ion channel, which triggers loss of cell membrane potential upon activation by the intracellular sensor RexA (Labrie et al., 2010; Snyder, 1995). Whilst not mechanistically comparable to BstA, perhaps the shared synteny of the BstA and RexAB abortive infection systems points to a functional significance of this genomic region, as the cI repressor gene is one of the most highly transcribed prophage promoters during lysogeny.
Though somewhat functionally analogous to toxin-antitoxin systems, to our knowledge no other example of self-immunity mechanisms have been described within prophage-encoded abortive infection systems. However there is some evidence supporting the widespread existence of such mechanisms. For example, it was observed that the activity of Lambda RexB protein can be suppressed by overexpression of the rexB gene relative to rexA. It was speculated, but not shown experimentally, that high levels of RexB might allow phage Lambda to replicate lytically in the presence of RexAB (Parma et al., 1992) i.e. giving the Lambda prophage self-immunity against its own Abi proteins. Our discovery of the BstA and the aba element strongly supports this hypothesis.
The discovery of the BstA-aba system opens unexplored avenues of research into the mechanisms used by prophages to suppress their own phage-defence activities. We anticipate that similar strategies may be widespread and commonplace, perhaps existing within characterised prophage-encoded phage defence systems. Given the huge mosaic diversity of temperate phages, and high prevalence of uncharacterised accessory genes, the reservoir of prophage-encoded phage defence and self-immunity systems is likely vast and largely unexplored.
Methods
RESOURCE AVAILABILITY
Further information and requests for bacterial and bacteriophage strains should be addressed to the Lead Contact.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Bacteria and bacteriophages
The full list of bacterial strains used and constructed is available in Supplementary Table 2. All the Salmonella strains were derived from the African S. Typhimurium strain D23580 (GenBank: FN424405.1) (Kingsley et al., 2009) or the model S. Typhimurium strain LT2 (GenBank: AE006468.2)(McClelland et al., 2001; Zinder and Lederberg, 1952). All the Escherichia coli strains constructed were derived from E. coli strain K-12 substrain MG1655 (GenBank: NC_000913.3) (Riley et al., 2006). The bstA homologs were cloned from E. coli NCTC10963 (GenBank: NZ_CAADJH010000002.1) or from Klebsiella pneumoniae Kp52.145 (GenBank: FO834906.1) (Bialek-Davenet et al., 2014). Bacteriophages (phages), including the temperate phages P22 (GenBank: NC_002371.2) (Pedulla et al., 2003) and BTP1 (GenBank; NC_042346.1) (Owen et al., 2017) and their derivatives, are described in Supplementary Table 2. The genomic coordinates and gene identifiers indicated below refer to the GenBank accession numbers mentioned above.
METHOD DETAILS
Growth conditions and transformation
All suppliers of chemical and reagents are specified in the Key Resources Table. Unless stated otherwise, bacteria were grown at 37°C in autoclaved Lennox Broth (LB: 10 g/L Bacto Tryptone, 5 g/L Bacto Yeast Extract, 5 g/L NaCl) with aeration (shaking 220 rpm) or on LB agar plates, solidified with 1.5% Agar. The salt-free LBO media contained 10 g/L Bacto Tryptone, 5 g/L Bacto Yeast Extract. Pre-cultures were inoculated with isolated colonies from agar plates and grown to stationary phase (for at least 6 hours) in 5 mL LB in 30 mL universal glass tubes or in 50 mL plastic tubes (Greiner).
Cultures were typically prepared by diluting the pre-cultures [1:100] or [1:1000] in LB, and bacteria were grown in conical flasks containing 10% of their capacity of medium (i.e. 25 mL LB in a 250 mL conical flask) with aeration. For fluorescent microscopy experiments, bacteria were grown in M9 minimal medium (Sambrook and Russell, 2001) supplemented with 0.4% glucose and 0.1% Bacto Casamino Acids Technical (M9 Glu+).
When required, antibiotics were added to the media: 50 µg/mL kanamycin monosulfate (Km), 100 µg/mL Ampicillin sodium (Ap), 25 µg/mL tetracycline hydrochloride (Tc), 20 µg/mL gentamicin sulfate (Gm), 20 µg/mL chloramphenicol (Cm). Bacteria carrying inducible constructs with genes under the control of the PBAD or Pm promoters were induced by adding 0.2 % (w/v) L-(+)-arabinose or 1 mM m-toluate, respectively. For the strains carrying tetR-PtetA modules, PtetA induction was triggered by adding 500 ng/mL of anhydrotetracycline hydrochloride (AHT, stock solubilized in methanol). For these constructs, the same volume of methanol was added to the non-induced cultures (mock treatment). Chemically-competent E. coli were prepared with RbCl-based solutions and were transformed by heat shock (Green and Rogers, 2014).
For the preparation of electro-competent cells, bacteria were grown in the salt-free medium LBO to an Optical Density at 600 nm (OD600) of 0.4-0.5. The bacteria were washed twice with cold sterile Milli-Q water (same volume as the culture volume) and were concentrated 100 times in cold 10% glycerol, prior to storage at -80°C. When ultra-competent Salmonella cells were required, the bacteria were grown in LBO at 45°C to OD600 0.4-0.5, because growth at high temperature inactivates the Salmonella restriction systems (Edwards et al., 1999). Competent cells (10-50 µL) were mixed with 10-5000 ng of DNA in electroporation cuvettes (2 mm gap) and the reactions were electroporated (2.5 kV) using a MicroPulser electroporator (Bio-Rad). Bacteria were re-suspended in 0.5-1 mL LB and incubated for recovery at 37°C (30°C for temperature sensitive plasmids) with aeration, for at least one hour. Finally, the transformed bacteria were spread on selective LB agar plates and transformant colonies were obtained after at least 12 hours incubation at 30-37°C.
Cloning procedures
All the plasmids and DNA oligonucleotides (primers) are listed in Supplementary Table 2. DNA manipulation and cloning procedures were carried out according to the enzyme and kit supplier recommendations and to standard procedures (Sambrook and Russell, 2001). DNA purity and concentration were measured with a DeNovix DS-11 FX spectrophotometer/fluorometer and using the Qubit dsDNA HS assay Kit.
For all the cloning procedures, Polymerase Chain Reactions (PCRs) were performed with the Phusion High Fidelity DNA polymerase, purified template DNA and primers in the presence of 3 % Dimethyl Sulfoxide and 1 M betaine, when required. Prior to Sanger sequencing of the constructs, PCR reactions were carried out directly from bacteria or phages with MyTaq Red Mix 2X. PCR fragments were analysed by electrophoresis, purified and finally sequenced with the appropriate primers (Lightrun service, Eurofins Genomics) (Supplementary Table 2).
All the plasmids were constructed as detailed in the Supplementary Table 2 and were verified by Sanger sequencing. Insertions of DNA fragments into plasmids were performed by digestion/ligation procedures, using restrictions enzymes and the T4 DNA ligase. In addition, PCR-driven restriction-free cloning techniques were used: overlap extension PCRs (Heckman and Pease, 2007) and plasmid assembly by PCR cloning (Van Den Ent and Löwe, 2006) were performed with chimeric primers, purified template DNA and Phusion DNA polymerase, as described previously (Owen et al., 2020). Cloning reactions were transformed by heat shock into E. coli Top10 (Invitrogen) or S17-1 λpir (Simon et al., 1983). New template plasmids were constructed to insert fluorescent protein encoding genes into Salmonella or E. coli chromosomes, as reported previously (Gerlach et al., 2007). These plasmids carry the oriR6K γ origin of replication of pEMG, the frt-aph-frt (KmR) module of pKD4 linked to gfp+ (pNAW52), sfgfp (encoding for superfolder GFP, pNAW62) or mcherry (pNAW73), amplified respectively from plasmids pZEP09 (Hautefort et al., 2003), pXG10-SF (Corcoran et al., 2012) and pFCcGi (Figueira et al., 2013). A similar template plasmid, carrying the frt-aph-frt-tetR-PtetA module (pNAW55) was constructed and was used to insert the tetR repressor and the AHT-inducible promoter PtetA upstream of genes of interest, as reported earlier (Schulte et al., 2019).
The high copy number plasmid pUC18 was used to clone the different versions of the anti-bstA (aba) fragment: the aba fragments (aba1-aba14 alleles) were amplified by PCR, digested with EcoRI and BamHI and ligated into the corresponding sites of pUC18. Phagemids based on the phage P22 replication module were constructed by EcoRI/KpnI digestion and ligation, as follows: the PR promoter and the cro-c1-orf48-O-P genes of P22 (coordinates 31648-34683) were amplified and circularized by ligation with the aph KmR cassette of pKD4 or with the aba-aph / abamut1-aph modules, amplified from strain SNW617. The ligations reactions were purified and electroporated into ultra-competent SNW555, a prophage-free and plasmid-free derivative of S. Typhimurium D23580. The resulting phagemids pNAW229 (pP22-aph), pNAW230 (pP22-aba-aph) and pNAW231 (pP22-abamut1-aph) were obtained after selection on Km medium.
Phage DNA was extracted from high titer lysates in LBO: nine volume of the phage lysates were mixed with one volume of 10 X DNase buffer (100 mM Tris-HCl, 25 mM MgCl2, pH 7.5) supplemented with RNase A (40 µg/mL final) and DNase I (400 µg/mL final). After 1 hour incubation at 37°C, DNase I was heat-inactivated at 75°C for 10 min and phage DNA was extracted from 500 µL of the nuclease-treated lysates with the Norgen Phage DNA Isolation after Proteinase K treatment, as specified by the manufacturer.
Genome editing techniques
Strain constructions are detailed in Supplementary Table 2. For chromosomal insertions and deletions, λ red recombination was carried out with the arabinose-inducible plasmid pKD46 (for E. coli) or with the heat inducible plasmid pSIM5-tet (for Salmonella), both expressing the λ red genes. Bacteria were grown to exponential phase in LBO, according to the resistance and induction condition of the respective λ red plasmid (Datsenko and Wanner, 2000; Hammarlöf et al., 2018; Koskiniemi et al., 2011) and electro-competent cells were prepared as mentioned above. PCR fragments carrying a resistance cassette were constructed by overlap extension PCR or were directly obtained by PCR from the adequate plasmid. Electro-competent cells (40-50 µL) were transformed with 500-5000 ng of the PCR fragments and the recombinants were selected on selective LB agar plates.
Mutations or insertions linked to selective markers were transduced into Salmonella strains using the P22 HT 105/1 int-201 (P22 HT) transducing phage (Owen et al., 2017; Schmieger, 1972). For E. coli, the transducing phage P1 vir was used (Ikeda and Tomizawa, 1965; Tiruvadi Krishnan et al., 2015). Transductants were grown on selective LB agar plates supplemented with 10 mM EGTA. After two passages, clearance of the transducing phages was confirmed by diagnostic PCR using primer pairs NW_62/NW_63 for P22 HT or NW_392/NW_393 for P1 vir and by a passage on Green Agar medium (Maloy, 1990).To remove the antibiotic cassettes, flanked by FLP recognition target sites (frt), the FLP recombinase expressing plasmids pCP20, pCP20-TcR and pCP20-Gm were used, as previously reported (Cherepanov and Wackernagel, 1995; Doublet et al., 2008; Hammarlöf et al., 2018; Kintz et al., 2015). The inducible tetR-PtetA-bstA modules were constructed by fusing the frt-aph-frt-tetR-PtetA module of pNAW55 to the bstA gene of D23580 (bstABTP1, STMMW_03531), E. coli NCTC10963 (bstAEc, E4V89_RS07420) or K. pneumoniae Kp52.145 (bstAKp, BN49_1470). Each construct carries the native bstA ribosome binding site and Rho-independent terminator. The tetR-PtetA-bstA modules were inserted by λ red recombination into the STM1553 pseudogene of S. Typhimurium LT2 (between coordinates 1629109-1629311), corresponding to STMMW_15481 in D23580 (coordinates 1621832-3). Previously we have shown that the STM1553 and STMMW_15481 genes are not expressed at the transcriptional level (Canals et al., 2019).
In E. coli MG1655, the bstA modules were inserted into the glmS-pstS intergenic region (coordinates 3911773-4). To generate Ap and Cm sensitive D23580 strains, the pSLT-BT plasmid-encoded Tn21-like element, that carries the resistance genes (Kingsley et al., 2009), was replaced by the KmR cassette of pDK4 by λ red recombination (deletion coordinates 34307 to 57061, GenBank: NC_013437.1). The resulting large single-copy plasmid pSLT-BT ΔTn21::aph was extracted (Heringa et al., 2007) and electroporated into the strains of interest. After selection on Km medium, the Ap and Cm sensitivity was confirmed and the KmR cassette was flipped out using pCP20-Gm. For scarless genome editing, the pEMG plasmid-based allelic exchange system was used (Martínez-García and de Lorenzo, 2011). The pEMG derivative suicide plasmids were constructed as specified in Supplementary Table 2 and were replicated in E. coli S17-1 λpir. Conjugation of the resulting plasmids into Salmonella and subsequent merodiploid resolution with plasmid pSW-2 were carried out as previously described (Canals et al., 2019; Owen et al., 2017). Some key strains and phages (indicated in Supplementary Table 2) used in this study were whole-genome sequenced (Illumina) at MicrobesNG (Birmingham, UK).
Plasmid deletion in S. Typhimurium D23580
The pSLT-BT, pBT1, pBT2 and pBT3 plasmids (Kingsley et al., 2009) were cured from strain D23580, using the CRISPR-Cas9-based methodology (Lauritsen et al., 2017). A novel CRISPR-Cas9 Km resistant plasmid (pNAW136) was obtained by ligating the CRISPR-Cas9 module of plasmid pCas9 (Jiang et al., 2013) with the unstable origin of replication oriRK2, the trfA replication gene and the aph KmR gene. Anti-plasmid protospacers (30 bp) were generated by the annealing of 5’-phosphorylated primer pairs that targeted the pSLT-BT, pBT1, pBT2 and pBT3 plasmids, designed according to the Marraffini Lab protocol (Jiang et al., 2013). The protospacers were ligated into BsaI-digested pNAW136 with T4 DNA ligase and the resulting plasmids were checked by Sanger sequencing, using primer NW_658.
The resulting plasmids pNAW168 (anti-pSLT-BT) and pNAW169 (anti-pBT1), pNAW139 (anti-pBT2) and pNAW191 (anti-pBT3) were electroporated into D23580-derived strains and transformants were selected on Km plates. After two passages on Km, the loss of the pSLT-BT, pBT1, pBT2 or pBT3 plasmids was confirmed by diagnostic PCR. The absence of the unstable pNAW136-derived plasmids was confirmed by the Km sensitive phenotype of colonies after two passages on non-selective medium.
Phage stock preparation and plaque assays
All phage stocks were prepared in LB or LBO. For Salmonella phages, the prophage-free strain S. Typhimurium D23580 ΔΦ (JH3949) was used as host (Owen et al., 2017). Exponential phase cultures of D23580 ΔΦ were infected with ∼105 Plaque Forming Units (PFU) and infected cultures were incubated for at least 3 hours at 37°C (with aeration). Phages lysates were spun down (4,000 X g, 15 min) and supernatants were filter-sterilized (0.22 µm, StarLab syringe filters). The resulting phage lysates were stored at 4°C in the presence 1% chloroform to prevent bacterial contamination.
Coliphage lysates were prepared similarly with E. coli MG1655 as host. When required, maltose (0.2%), CaCl2 (10 mM) and MgSO4 (10 mM) were added during the infection (λ, P1 vir and Φ80pSU3+). For Φ80-derived phages, the infection temperature was reduced to 30°C (Rotman et al., 2010).
Phage lysates were serial-diluted (decimal dilutions) with LB and virion enumeration was performed by double-layer overlay plaque assay (Kropinski et al., 2009), as follows. Bacterial lawns were prepared with stationary phase cultures of the reporter strains, diluted 40 times with warm Top Agar (0.5 % agar in LB, 50°C). The seeded Top Agar was poured on LB 1.5% agar bottom layer: 4 mL for 8.6 cm diameter petri dishes or 8 mL for 12 × 12 cm square plates.
When inducible PtetA or PBAD constructs were present in the reporter bacteria, 500 ng/mL of AHT or 0.2 % arabinose were added in the Top Agar. When required, antibiotics were added in the Top Agar layer. The bacterial lawns were incubated for 30 min at room temperature with the appropriate inducer, to allow solidification and the expression of the inducible genes. Finally, phages suspensions (5-20 µL) were applied on the Top Agar surface and pictures of the resulting plaques were taken with an ImageQuant Las 4000 imager (GE Healthcare) after 16-20 hours incubation at 30 or 37°C.
Construction of P22 virulent phages
For the generation of obligately virulent P22 phages, a 633 bp in-frame deletion (coordinates 31028-31660) was introduced in the c2 repressor gene by λ red recombination in a P22 lysogen as follows. Two fragments of ∼500 bp, flanking c2, were amplified with primers pairs NW_818 / NW_819 and NW_820 / NW_821. The two amplicons were fused by overlap extension PCR and 1000-3000 ng of the resulting Δc2 fragment were electroporated into P22 lysogens (in the prophage-free D23580 ΔΦ background) carrying the λ red recombination plasmid pSIM5-tet, as described above. The transformation reactions were re-suspended in 5 mL LB and incubated for 2 hours at 37°C with aeration. The culture supernatants were filter sterilized and serial-diluted to 10−2. Ten microliters of each dilution were mixed with 100 µL of a D23580 ΔΦ stationary phase culture and with 4 mL of warm Top Agar. The mixtures were poured on LB agar plates and the plates were incubated for ∼16 hours at 37°C. P22 Δc2 recombinants were identified by the clear morphology of their plaques, compared to the turbid plaques of WT P22. The Δc2 deletion was confirmed by PCR and Sanger sequencing with primers NW_406 and NW_805.
Use of the Δtsp-gtrAC genetic background
Where possible, experiments were carried out with native BstA expression (from its natural locus within the BTP1 prophage), to best recapitulate the natural biological activity of the protein. However, as the gtr locus of phage BTP1 blocks attachment of many phages including P22 and BTP1, to achieve efficient phage infections we consistently used a strain background where the gtr locus has been inactivated (Δtsp-gtrAC). The BTP1 prophage spontaneously induces to a titer of ∼109 PFU/mL in liquid culture (Owen et al., 2017), and in the absence of gtr activity in surrounding cells, free BTP1 phages mediate cleavage of the O-antigen via the putative enzymatic activity of the tailspike protein (Kintz et al., 2015). Consequently, to avoid an unnatural, short LPS phenotype as a result of gtr inactivation in a BTP1 lysogen, we additionally inactivated the upstream gene encoding for the BTP1 tailspike (tsp) (D23580 Δtsp-gtrAC, JH4287). Full details of the construction of this strain can be found in the Supplementary Resource list.
Phage replication assay
Stationary phase cultures of the reporter bacteria were diluted to OD600 0.4 with LB. Aliquots (0.2 mL) were prepared in 1.5 mL tubes and phage stock suspensions were added to a final phage titer of 100-1000 PFU/mL. The infections were carried out at 37°C (30°C for Φ80pSU3+) with shaking for 2-4 hours and were stopped by the addition of 20 µL of chloroform. After a 10 sec vortex, the lysates were centrifuged (20,000 X g, 5 min) and serial diluted. When M9 Glu+ was used, Salmonella strains were grown to OD600 ∼ 0.5 in this medium prior to phage infection.
Phage titer was determined by plaque assay: 10 µL of the dilutions were applied to bacterial lawns of the appropriate reporter strain in technical triplicates. Plaques were enumerated after 16-20 hours of incubation and phage titers (PFU/mL) were calculated for each lysate. To measure the phage input at time 0 (T0), the same volume of stock phage suspension was added to 0.2 mL of bacteria-free LB and the titer was determined as described above. The fold-replication for each phage was calculated as the phage titer of the lysate post infection divided by the input phage titer at T0. When the phage titer in the lysate was lower than the phage input, the replication was considered to be null (“<1-fold). When AHT inducible tetR-PtetA-bstA strains were used, AHT (500 ng/mL) or methanol (mock) were added to the diluted bacterial suspension and phages were added after 15 min of incubation at 37°C with aeration.
For replication assays of the coliphages λ, P1 vir and Φ80pSU3+, E. coli strains were grown to exponential phase (OD600 0.4) in LB and phages were added as mentioned above. To stimulate infection by these phages, maltose (0.2%), CaCl2 (10 mM) and MgSO4 (10 mM) were added during the infection and in the lawns of the reporter E. coli MG1655. All the phage replication experiments presented were carried out at least twice with biological triplicates.
Induction of P22 and BTP1 prophages
D23580 ΔΦ-derived lysogens that carried the different versions of P22 and BTP1 were constructed as detailed in the Supplementary Table 2. For complementation with the pUC18-derived plasmids (ApR), Ap sensitive lysogens were constructed by the inactivation of the Tn21-like element, as described above. The resulting lysogens were grown to stationary phase in LB and the pre-cultures were diluted 1000 times in fresh LB and grown to OD600 0.4-0.5, prior addition of Mitomycin C (MitC, 2 µg/mL). The induced cultures were incubated for 3-5 hours at 37°C with aeration and cultures were filter sterilized and serial diluted. The phage titer was measured by plaque assay on the appropriate host strain lawn with technical replicates, as described above. Phage titers were also measured before MitC induction and the titer of the non-induced culture was subtracted from the induced titer to obtain the final phage titer. All the prophage induction experiments were carried out at least twice with biological triplicates.
Survival assays
For the survival assays, D23580 Δtsp-gtrAC (JH4287), D23580 Δtsp-gtrAC bstASTOP (SNW431) or D23580 ΔΦ [P22] (SSO-128) were grown in M9 Glu+ to OD600 ∼0.5 and two 0.5 mL subcultures were prepared for each culture. The use of D23580 ΔΦ [P22] in these experiments controlled for the effect of lysis from without due to use of high multiplicity of infection (MOI). The strain is a lysogen for WT P22 phage, and therefore is highly resistant to infection by P22-derived phages. P22 Δc2 was added at an MOI of 5. The same volume of LB was added to the two remaining subcultures (non-infected controls). Samples were incubated for 15 min at 37°C to allow phage attachment. To stop phage development, the cultures were chilled on ice and bacteria were washed with 0.5 mL of cold PBS. All the samples were serial-diluted in PBS to 10−6 and kept on ice. For the measure of survival post-infection, 10 µL of diluted infected or non-infected cultures were applied in technical triplicates on LB agar supplemented with 10 mM EGTA (EGTA was used to minimize secondary infection by free phages). Colony forming Units (CFU) were enumerated and the survival rate, was calculated as the ratio of CFUs in infected cultures divided by the CFUs obtained from non-infected cultures (in %). All the survival experiments were carried out at least twice with biological triplicates.
Phage DNA detection by Southern Blotting
D23580 ΔΦ tetR-PtetA-bstA (SNW576) was grown in 50 mL LB to OD600 ∼0.35. The culture was split in two 20 mL sub-cultures and methanol (mock) or AHT (inducer) were added to each subculture. Bacteria were incubated to induce BstA for 20 min at 37°C and the phage of interest was added at an MOI of 5. Infections were carried out at 37°C with aeration and total DNA was extracted (Quick-DNA™ Universal Kit Zymo) from 1.5 mL of culture at 0, 10, 20, 30, 35, 40 and 50 minutes post Infection. Total DNA (100 ng, according to QuBit quantification) was size-separated (2 hours at 100 V in TAE 1X) on a 0.8 % agarose-TAE gel containing Midori Green DNA staining (4 µL for 100 mL gel). One hundred nanograms of none-infected D23580 ΔΦ tetR-PtetA-bstA genomic DNA were used as a negative control. DNA was fragmented by exposing the agarose gel to UV light for 5 min on a UV-transilluminator. DNA was denatured by soaking the gel in the Denaturation Solution (0.5 M NaOH, 1.5 M NaCl) for 30 min and then in the Neutralization Solution (1.5 M NaCl, 1 M Tris-HCl, pH 7.6) for 30 min. DNA was transferred on a positively-charged Nylon membrane using the capillary blotting method. Phage DNA was detected with DIG labelled dsDNA probes generated by PCR amplification with MyTaq DNA polymerase (Bioline), buffer, phage DNA and primers (0.4 µM each), in the presence of 0.2 mM dATP, 0.2 mM dCTP, 0.2 mM dGTP, 0.13 mM dTTP and 0.07 mM DIG-11-dUTP. For the 9NA probe a 588 bp PCR fragment was generated with primer pair NW_602 / NW_603 and for the P22/BTP1 probe a 725 bp PCR fragment was generated with primer pair SO-22 / SO-23. The DNA probes were heat-denatured at 95°C for 15 min and the DNA-DNA hybridizations were carried out at 45°C for 16 hours in DIG-Easy Hyb buffer. The washing and immunodetection procedures were carried out, as specified in the DIG Application Manual for Filter Hybridization (Roche) and the chemilumiscence signal was detected using an ImageQuant LAS 4000 imager (GE Healthcare). Prior to DNA transfer onto the membrane, the Midori green-stained DNA was visualized under UV and the resulting image was used as a loading control.
Phagemid efficiency of transformation
To avoid a reduction in transformation caused by interference interspecies DNA modification/restriction incompatibilities between E. coli and Salmonella, all the P22-derived phagemids were replicated and extracted from S. Typhimurium SNW555.
Salmonella strains carrying the tetR-PtetA-bstA module were grown in 50 mL LBO culture. When OD600 ∼0.4 was reached, each culture was split into two 25 mL sub-cultures and methanol (mock) or AHT (inducer) were added to each subculture. Bacteria were incubated for BstA induction during 15 min at 37°C. The cultures were incubated on ice for 5 min and bacteria were washed twice with cold water (25 mL) and were concentrated in 0.1 mL of ice-cold sterile 10% glycerol. The OD600 of each electro-competent cell sample was measured by diluting 10 µL of competent cells with 990 µL of 10% glycerol. Cell concentration was adjusted with 10% glycerol for each sample, according to the sample with the lowest OD600. The competent cells (20 µL) were mixed with 10 ng (estimated by Qubit) of the P22 phagemids, pP22 (pNAW229), pP22-aba (pNAW230) or pP22-abamut1 (pNAW231) and the mixture was incubated on ice until electroporation (2.5 KV). Transformation reactions were re-suspended in 1 mL LB or 1 mL LB + AHT (for the bstA-induced bacteria) and were incubated for 60 min at 37°C, for recovery. The transformations were diluted (decimal dilution to 10−5) in LB or LB+AHT and 100 µL of each dilution (including the non-diluted sample) were spread on LB agar Km or LB agar Km+AHT plates. After incubation at 37°C, the number of KmR transformants was enumerated for each transformation and efficiency of transformation was defined as the number of transformants obtained per ng of phagemid. This experiment was performed with biological triplicates and was repeated twice with LT2 tetR-PtetA-bstA (SNW389) and once with D23580 ΔΦ tetR-PtetA-bstA (SNW576), giving similar results.
Microscopy- general
For all imaging experiments, bacteria were sub-cultured in liquid M9 Glu+ media. All images were collected with a wide field Nikon Eclipse Ti-E inverted microscope equipped with an Okolab Cage Incubator warmed to 37°C with Cargille Type 37 immersion oil. A Nikon CFI Plan Apo DM Lambda 100X 1.45 NA Oil objective and a Nikon CFI Plan Apo DM Lambda 20X .75 NA objective were used with Perfect Focus System for maintenance of focus over time. Superfolder GFP, mCherry and SYTOX Orange Nucleic Acid Stain (ThermoFisher) were excited with a Lumencor Spectra X light engine with Chroma FITC (470/24) and mCherry (575/25) filter sets, respectively and collected with a Spectra Sedat Quad filter cube ET435/26M-25 ET515/30M-25 ET595/40M-25 ET705/72M-25 and a Spectra CFP/YFP/mCherry filter cube ET475/20M-25 ET540/21M-25 ET632/60M-25. Images were acquired with an Andor Zyla 4.2 sCMOS controlled with NIS Elements software. For time-lapse experiments, images were collected every 3 minutes (unless specified otherwise) via ND acquisition using an exposure time of 100 ms and 50% or 100% illumination power for fluorescence. Multiple stage positions (fields) were collected using the default engine Ti Z. Fields best representing the overall experimental trend with the least technical artefacts were chosen for publication. Gamma, brightness, and contrast were adjusted (identically for compared image sets) using FIJI(Schindelin et al., 2012). The FIJI plug-ins Stack Contrast (Capek et al., 2006) and StackReg (Thevenaz et al., 1998) were used for brightness matching and registering image stacks.
Microscopy- agarose pads
Agarose pads were prepared with 2% agarose and M9 Glu+ media, and mounted on MatTek dishes (No. 1.5 coverslip, 50 mm, 30 mm glass diameter, uncoated). Cells (D23580Δtsp-gtrAC (JH4287) or D23580Δtsp-gtrAC bstASTOP (SNW431) were grown to log phase (OD600 ∼ 0.4) in M9 Glu+ at 37°C with shaking (220 RPM), and where required, diluted in fresh M9 Glu+ to achieve the desired cell density on the agarose pad. For experiments where all cells were infected (Figure 4A), phage P22 Δc2 was added at an MOI of 5. Phage adsorption and initial infection was facilitated by incubation at 37°C with shaking for 10 minutes. Subsequently, infected cells were pelleted at 5000 x g and resuspended in ice-cold PBS to pause phage development. Two microliters of chilled, infected cells were spotted onto opposite sides of an agarose pad (two strains were imaged on the same pad) and inverted onto the MatTek imaging dish. Experimental MOIs were immediately confirmed by CFU and PFU /mL measurement of the cell and phage preparations. Phase-contrast images using the 100X objective were collected every 3 minutes for 3 hours.
Procedures for experiments involving a subset of infected cells (Figure 4C) were identical, except cells infected with P22 Δc2 p-mcherry were washed an additional 4 times in ice-cold PBS to reduce the concentration of un-adsorbed, free phage. In parallel, uninfected cells were washed once in ice-cold PBS. Infected cells were mixed at a ratio of 1:1000 with uninfected cells of the same genotype before being spotted onto the agarose pad. For these experiments, phase-contrast and fluorescence images (mCherry) using the 20X objective were collected every 3 minutes for 3 hours.
Microscopy- microfluidic infection
The CellASIC ONIX2 system from EMD Millipore with B04A plates was used for microfluidic imaging experiments (Figure 5). Phages used in microfluidic infection experiments shown in 5B (P22 HT or 9NA) were stained with SYTOX Orange Nucleic Acid Stain according to the protocol previously described (Valen et al., 2012). Stained phages washed 4 times in 15 mL M9 Glu+ media using Amicon Utra-15 centrifugal filter units. After staining, the titer and viability of phages were immediately assessed by plaque assay, and once stained, phages were used for no longer than 2 weeks. For use in the microfluidic experiments, SYTOX Orange strained phages were normalised to a titer of approximately 1010 PFU/mL. Cells (D23580 bstA-sfgfp, SNW403) were grown to early exponential phase (OD600 ∼ 0.1) in M9 Glu+ at 37°C with shaking (220 RPM) before being loaded into CellASIC B04A plates using the pressure-driven method according to the manufacturer protocol for bacterial cells. The slanted chamber of the plate immobilises the cells, but allows media to flow continuously. Firstly, cells were equilibrated with constant M9 Glu+ media flow for approximately 30 minutes. Secondly, stained phages suspended in M9 Glu+ media were flowed over the cells until the majority of cells were infected (typically 10-30 minutes). In the case of P22 HT phage (which exhibits inefficient adsorption to D23580 bstA-sfgfp due to the gtr locus of prophage BTP), phages were continuously flowed. Finally, M9 Glu+ media was flowed over the cells for the duration of the experiment. Microfluidic experiments typically lasted 5 hours, after which time uninfected cells outgrew the chamber. Phase-contrast and fluorescence images were collected every 1.5 minutes for the experiments in Figure 5B.
For the microfluidic imaging experiments shown in Figure 5C, strain SVO251 (S. Typhimurium D23580 ΔΦ STM1553::(PtetA-bstA-sfgfp-frt) ΔpSLT-BT ΔpBT1 pAW61 (PBAD-parB-mcherry) was used. This strain contains the bstA-sfgfp fusion contrast under the control of the PtetA promoter. However, this strain lacks the tetR gene, and therefore expression of bstA-sfgfp is constitutive (not inducible). Additionally, this strain is cured of two natural plasmids that contain native partitioning systems (pSLT-BT and pBT1), and there for might interfere with the correct function of the ParB-parS system used for phage DNA localization. The ParB-mCherry fusion protein is expressed from the pAW61 plasmid (ApR) under the control of the PBAD promoter (induced by L-arabinose). Strain SVO251 was grown in M9 Glu+ supplemented with 100 μg/mL ampicillin to maintain the pAW61 plasmid and 0.2% L-arabinose to induce expression of ParB-mCherry. The same supplemented media was used in the microfluidic chamber. Cells were grown to ∼OD600 0.1 before loading into the CellASIC B04A plate as described above. After 15 minutes growth, phage P22 Δpid::(parS-aph) [which contains one parS site along with a kanamycin resistance locus, aph, in place of the non-essential pid locus (Cenens et al., 2013)] diluted to a concentration of 108 PFU/mL (in M9 Glu+ amp100 0.2% L-ara) was flowed into the chamber. Phase contrast and red and green fluorescence images were collected every 2 minutes for 4 hours.
BstA protein homolog analysis
BstA protein homologs were identified using tblastn (database: non-redundant nucleotide collection) and the HMMER webserver (Potter et al., 2018) (database: Reference Proteomes). The dataset of BstA protein homologs was manually curated to reflect the diversity of taxonomic background harbouring homologs. To analyse the genetic context of BstA homologs, the sequence region 20 Kb either side of the homolog (40 kb total) was extracted (BstA 40 kb neighbourhoods). To produce homogenous and comparable annotations, each region was re-annotated using Prokka 1.13 (Seemann, 2014). Additionally, the resulting annotated amino acid sequences were queried against our custom BstA profile-hmm and the Pfam 31.0 database (El-Gebali et al., 2019) with hmmerscan (Eddy, 1998), and the highest scoring significant hit per ORF was considered for the results shown in Figure 2. All the code is available in https://github.com/baymLab/2020_Owen-BstA.
Pairwise identity of homologs in Figure 2B to BstABTP1 was computed using the EMBOSS Needle webserver (Needleman and Wunsch, 1970). BstA homologs were designated “putatively-prophage associated” if annotated genes in the 40 kb neighbourhood contained any instance or the word “phage” or “terminase”. For categorisation in Figure 2C, homologs were classed as having “high confidence association” if instances of gene annotations including the aforementioned key words occurred both before, and after, the BstA gene within the 40 Kb neighbourhood (i.e., to account for the possibility that a prophage-independent homolog could co-occur next to a prophage region by chance). Homologs classed as having “low confidence association” had at least one instance of genes whose annotations included “phage” or “terminase” either in the upstream or downstream 20Kb, but not both. Plasmid status was determined from information in the sequence records. The HHpred webserver was used to annotate the putative KilA-N domain (Zimmermann et al., 2018). All homolog neighbourhoods, homolog alignments and sequences is available to download https://github.com/baymLab/2020_Owen-BstA.
QUANTIFICATION AND STATISTICAL ANALYSES
The phage replication, survival rate, efficiencies of transformation and of lysogeny were calculated as mentioned above. The numerical data were plotted and analyzed using GraphPad Prism 8.4.1. Unless stated otherwise in the figure legends, data are presented as the mean of biological triplicates ± standard deviation. The unpaired t-test was used to compare the groups and statistical significance is indicated on the figures. P values are reported using the following criteria: < 0.0001 = ****, 0.0001 to 0.001 = ***, 0.001 to 0.01 = **, 0.01 to 0.05 = *, ≥ 0.05 = ns.
KEY RESOURCES TABLE
Figure legends & Supplementary Materials
Supplementary Table 1: Details of all BstA homologs used in the analysis presented in Figure 2.
Supplementary Table 2: Details of all oligonucleotide sequences, plasmids, bacterial strains and phages used in this study.
Supplementary Video 1: Cells natively expressing BstA (D23580 Δtsp-gtrAC, JH4287) or possessing a mutated BstA locus (D23580 Δtsp-gtrAC bstASTOP, SNW431) were infected with the obligately virulent P22-derivate phage, P22Δc2, at an MOI of 5 (to increase the likelihood of infecting all cells). Infected cells were imaged every 3 minutes on agarose pads. Regardless of BstA function, almost all cells were observed to lyse (indicated by loss of defined cell shape and phase contrast).
Supplementary Video 2: Cells natively expressing BstA (D23580 Δtsp-gtrAC, JH4287) or possessing a mutated BstA locus (D23580 Δtsp-gtrAC bstASTOP, SNW431) were infected with the obligately virulent P22-derivate phage, P22Δc2 p-mcherry, at an MOI of 5 (to increase the likelihood of infecting all cells). Due to the mcherry insertion, red fluorescence corresponds to phage replication. Infected cells were mixed 1:1000 with uninfected cells. Cells mixtures were imaged every 5 minutes on agarose pads for 6 hours. In the BstA+ cells, primary infected cells lysed, but did not stimulate secondary infections of neighbouring cells, and eventually formed a confluent lawn. In BstA-cells, primary lysis events caused secondary infections (neighbouring cells showing red fluorescence and subsequent lysis) causing an epidemic of phage infection reminiscent of plaque formation.
Supplementary Video 3: Cells natively expressing BstA translationally fused to sfGFP (D23580 bstA-sfgfp, SNW403) were grown in a microfluidic growth chamber and imaged every 1.5 minutes. Fluorescently labelled phages P22 HT (left) or 9NA (right) were then added to the cells, and can be seen adsorping to cells as red fluorescent puncta. For purposes of comparison, timestamps are synchronised to the point at which phage are first observed. Typically around 20 minutes after initial observation of phage infection, BstA proteins formed discrete and dynamic foci within the. Cells then proceeded to lyse.
Supplementary Video 4: Cells heterologously expressing BstA translationally fused to sfGFP, and ParB translationally fused to mCherry (D23580 ΔΦ ΔpSLT_BT ΔpBT1 STMMW_15481::[PtetA-bstABTP1-sfgfp-frt] pAW61, SVO251) were grown in a microfluidic growth chamber and imaged every 2 minutes. Phage P22 Δpid::(parS-aph) were then added to the cells. ParB protein oligomerises onto DNA at parS sites, and therefore parS-tagged DNA is indicated by ParB-mCherry foci (red). Translocation of infecting phage DNA is indicated by the formation of red foci within the cells, which is rapidly followed by formation of green, BstA foci. The green and red foci appear to physically overlap. Infected cells proceed to lyse. The same timeseries is presented as separate channels (phase contrast, GFP, mCherry), and a composite merge.
Acknowledgements
We are grateful to present and former members of the Hinton and Baym laboratories, except Lizeth Lacharme Lora, for helpful discussions, and Paul Loughnane for his expert technical assistance. We thank the Bollback Lab (Liverpool) for the gift of pCas9 and P1vir, Allison Lab (Liverpool) for MG1655 and T4, Casadesus Lab (Seville) for 9NA and Det7, Ansaldi Lab (Marseille) for T5, Penadés Lab (Glasgow) for ES18, and Klumpp Lab (Zurich) for FelixO1. We also acknowledge all the other kind labs who responded to our call for Salmonella and E. coli phages. We thank Benoît Doublet for the gift of pCP20-Gm. We thank the David Van Valen lab, Francois St-Pierre and Paul Wiggins lab for λFS135 and pAW62.
Research in this publication was supported by a Wellcome Trust Senior Investigator award (to JCDH, grant number 106914/Z/15/Z), NIGMS of the National Institutes of Health (to MB, award number R35GM133700) and the David and Lucile Packard Foundation (SVO, NQO, and MB).