Bacteria are susceptible to viral (bacteriophage) predation but have evolved several strategies to resist viral infection. One strategy is abortive infection (Abi) in which an infected bacterial cell dies precociously and thereby concomitantly blocks the production of mature phage progeny (Chopin et al., 2005). This protects clonal siblings in the bacterial population and therefore is akin to an “altruistic suicide.” Abi can be mediated through toxin-antitoxin (TA) systems, which are widespread in prokaryotes. Genetically, TA systems are usually composed of two genes transcribed from a single promoter. The upstream gene encodes an antitoxin that neutralizes the toxin product of the downstream gene. Functionally, TA systems impact on several biological processes such as the formation of persister cells, responses to environmental stress, plasmid stabilization, and phage exclusion through abortive infection (Gerdes et al., 2005).

Some Type III TA systems are bifunctional in that they also confer an Abi phenotype on their bacterial hosts. These TA systems are comprised of a proteinaceous toxin and an RNA antitoxin (Fineran et al., 2009). The first Type III TA system identified (ToxIN_Pa) was encoded by a cryptic plasmid (pECA1039) in Pectobacterium atrosepticum. ToxIN_Pa was originally identified because ToxN shared 31% amino acid sequence identity with the AbiQ protein, which was involved in an abortive infection system in Lactococcus lactis (Emond et al., 1998; Fineran et al., 2009). Recent mutational analysis of the antitoxin and structural characterization of the toxin of the AbiQ system revealed that it is also a member of the Type III TA systems (Samson et al., 2013; Bélanger and Moineau, 2015). Since the discovery of ToxN_Pa, numerous other Type III TA systems have been identified bioinformatically in diverse bacterial genera and they can be chromosomally, plasmid, or phage-encoded (Blower et al., 2012). Type III TA systems have been classified into ToxIN, CptIN, and TenpIN families. The toxic proteins of these families are homologous to ToxN_Pa, while the antitoxin primary DNA sequences vary in length and number of tandem repeats. Interestingly, while ToxIN_Pa exhibited a strong Abi phenotype, ToxIN_Bt from Bacillus thuringiensis did not exhibit an Abi phenotype when challenged with over 100 different phages (Blower et al., 2012).

ToxIN_Pa is composed of an RNA antitoxin, ToxI, which binds to and suppresses the toxic endoribonuclease protein, ToxN. Crystallographic evidence revealed that ToxIN_Pa forms a triangular heterohexameric structure with 3 ToxN proteins in complex with 3 ToxI RNA pseudoknots (Blower et al., 2011). The complex is held in an inactive state under normal cellular conditions. However, during infection with specific phages, the complex is activated by an unknown mechanism, allowing the endoribonuclease to degrade host RNAs, leading to bacteriostasis and subsequent cell death. This phenomenon has features of a prokaryotic apoptosis that manifests itself in precocious death of the virally-infected bacterial host and, consequently, also inhibits progeny phage production. As this outcome restricts or terminates further phage invasion of the clonal bacterial population, the abortive infection process can be viewed as an altruistic suicide (Fineran et al., 2009).

Some phages aborted by ToxIN_Pa have the capacity to evolve spontaneous resistant mutants at low frequency that can circumvent the Abi system. The ToxIN_Pa resistant mutants of ΦTE, a P. atrosepticum phage, were found to overcome abortive infection by an RNA-based molecular mimicry of the ToxI antitoxin. The ΦTE escape phage had expanded a “pseudo-ToxI” region in the viral genome that was similar, but not identical, to the ToxI sequence. This expanded “pseudo-ToxI” region was expressed during phage infection and actively suppressed ToxN, thus allowing the ΦTE mutants to evade abortive infection (Blower et al., 2011). Recently, AbiQ-resistant mutations of four phages were sequenced revealing multiple loci involved in resistance, but how these genes conferred resistance to AbiQ remained elusive (Samson et al., 2013).

To date, ΦTE remains the only phage whose ToxIN_Pa-escape mechanism is understood. Thus, the primary aim of this study was to characterize additional resistance mutations of new ToxIN_Pa-sensitive phages and, by their study, perhaps add to the repertoire of known escape loci. Depending on the nature of the relevant mutation(s), the escape locus might provide insight into how phage infection leads to activation of the ToxIN_Pa complex.

In this study, we isolated and characterized Serratia sp. ATCC 39006 (S 39006)-specific phages, ΦCHI14, ΦX20, and ΦCBH8. Comparison of the genomic sequences of spontaneous ToxIN_Pa-escape mutants and their wild type progenitors revealed three different routes of escape: (1) mutation of asiA, encoding a predicted phage transcriptional co-activator in ΦCHI14 and ΦCBH8; (2) mutation of an unknown gene (orf84) in a ΦCHI14 mutant; (3) deletion of a large region (6.5–10 kb) of the phage genome in most mutants of all three T4-family phages.

In this study we isolated three highly homologous T4-family environmental phages of S 39006 that are sensitive to ToxIN_Pa-mediated Abi. Rare spontaneous mutants of these phages were able to circumvent ToxIN_Pa via multiple escape routes. The independent isolation of ΦCHI14, ΦX20, and ΦCBH8 is interesting since pairwise alignment of their genomes with other T4-like phages showed that they are mosaics of one another—whole-genome alignments consist of varying-length stretches of high homology interspersed with stretches of no homology (Nolan et al., 2006; Petrov et al., 2006). This kind of similarity and diversity is typical among T4-like phages (Petrov et al., 2010) yet the relatively low overall sequence similarities indicate that ΦCHI14, ΦX20, and ΦCBH8 may represent a new type of T4-like genome configuration. Furthermore, the isolation of only 3 ToxIN_Pa-sensitive, and very similar phages (despite multiple environmental enrichments yielding more than 300 phages over a 3-year span) suggests a relatively rare incidence of ToxIN_Pa-sensitive viruses among environmental phages of S 39006. However, this rarity may not be surprising because wild type S 39006 does not carry ToxIN_Pa naturally. Furthermore, it is possible that these new phages arose in the environment by propagation on alternative native hosts taxonomically unrelated to S 39006.

We showed recently that a single phage gene product from phage ΦM1 was responsible for activating the ToxIN_Pa system in the natural host, P. atrosepticum (Blower et al., 2017). A phage product might interact with ToxI and degrade it or sequester it away from ToxN to liberate the toxin. Alternatively, a phage product could interact with ToxN, reducing the affinity between ToxN and ToxI. Interactions of the phage product(s) with both ToxI and ToxN are also formally possible. Based on our recent observations with the evolution of ΦM1 to ToxIN_Pa resistance in P. atrosepticum, we expected to find mutations in a single locus in “escape” mutants of ΦCHI14, ΦX20, and ΦCBH8. However, our characterization of the latter “escape” mutants has now suggested a more complicated landscape for potential activation mechanisms operating on ToxIN_Pa.

One “escape” locus found in both ΦCHI14 and ΦCBH8 mutants is asiA, which encodes AsiA, a highly-conserved protein in T4-like phages (Figure 7). In T4, AsiA is an anti-sigma factor that represses host transcription through a process known as σ-appropriation, and, together with MotA, co-activates transcription of phage middle genes (Hinton et al., 2005). AsiA monomers bind tightly to regions 4.1 and 4.2 of σ⁷⁰ and abolish the sigma factor's ability to bind to the −35 promoter sequence of host genes (Hinton, 2010). As a result, AsiA inhibits transcription of bacterial genes with a −10/−35 promoter. In addition, AsiA-bound σ⁷⁰ adopts an altered conformation that allows the T4 transcriptional activator, MotA, to bind σ⁷⁰ and the MotA box present in the promoter region of T4 middle-expressed genes. Thus, the AsiA-σ⁷⁰-MotA complex disrupts the ability of RNA-polymerase (RNAP) to recognize and transcribe host genes but reconfigures the enzyme to transcribe phage middle-expressed genes (Ouhammouch et al., 1995; Lambert et al., 2004). Most of the 6 escape mutants of ΦCHI14 and ΦCBH8 produce an AsiA with either an extended or truncated CTD, while only one mutant has a V13G substitution in the NTD. In AsiA the role of the CTD is controversial and not fully understood, with some studies showing that mutations in the CTD of AsiA affected its function in σ-appropriation and MotA binding (Yuan and Hochschild, 2009; Yuan et al., 2009) while other studies showed no compromise in σ-appropriation of AsiA with a mutated CTD (Pal et al., 2003). As for the V13G mutation in the NTD, the mutated amino acid is homologous to V14 in T4, which was defined as an essential residue for AsiA-AsiA homodimer and AsiA-σ⁷⁰ interactions (Gilmore et al., 2010). Future research will investigate whether AsiA plays a direct role in ToxIN_Pa activation as a phage product or as a part of the AsiA-RNAP-MotA complex, or whether AsiA indirectly activates ToxIN_Pa by disturbing ToxI:ToxN stoichiometry or phage middle gene transcription.

The most frequently found escape locus among the S 39006 phage mutants was the deletion of a substantial region of the phage genomes. The size of the deletion represented 3.8–5.8% of the total genome size. These deletant mutants remained viable on S 39006 indicating that the deleted region is not essential for phage replication, at least under the laboratory conditions used. How the deletion mutations confer resistance is not immediately obvious, given the number of genes affected. As no missense mutations were isolated that mapped to this locus, we presume that two or more genes have to be mutated in this locus simultaneously for phages to escape, and presumably, these genes might be located toward the two ends of the deletion, hence the large deletion size. An interesting hypothesis here would be that the entire deletion region might contain genes that wild type phages acquired by recombinogenic lateral gene transfer (Petrov et al., 2010) in the natural environment (perhaps during mixed viral infections) and these genes, in concert, could be responsible for activating ToxIN_Pa in bacterial hosts. The possibility that the deleted region was acquired through lateral gene transfer is supported by two pieces of evidence: (1) attempts to align the large deletion locus with several other T4-like phages (T4, CC31, ΦR1-RT, ΦS16, PG7, and PEi20) showed very little homology (Figure 5); and (2) most of the 5' borders of the large deletion mutations lie within tRNA genes (Figure 4), whose conserved sequences are readily recombinogenic. It could be argued that this has echoes of the descriptions of Pathogenicity Islands in bacteria (Karaolis et al., 1999; Schmidt and Hensel, 2004) and therefore could even suggest that the large deletion region is a potentially mobile viral genome unit.

The only clue to possible functions of ORFs in the deleted regions comes from ORF145, which is a YadA homolog. YadA was first found in Yersinia species and serves as a virulence factor that mediates Yersinia adherence to epithelial tissue (Hoiczyk et al., 2000). YadA cognate genes have also been found in other phages but it is unknown if they might be transferred to their bacterial hosts to enhance virulence (Moreno Switt et al., 2013). Future work will involve investigating the impact of ORF145 in allowing phage “escape.” Finally, given the important physiological roles for phage-encoded tRNAs in viral morphogenesis, future research into the role that the tRNAs encoded within the deleted regions may play in phage “escape” requires investigation.

In summary, this study has discovered a distinct group of T4-family S 39006 phages that could activate the ToxIN_Pa-mediated abortive infection system. The data on the “escape” mutants of these environmental phages suggest that either ToxIN_Pa can be activated by more than one phage product, or that production of the ToxIN_Pa- activating product involves multiple biological processes and that defects in several could enable phages to circumvent Abi. Future investigations will involve further characterization of the products from the escape loci of these phages, and experiments to try to dissect their physiological roles during the process of abortive infection—and its circumvention.

GS, BC, CA, and XF conceived and designed the experiments. BC, CA, and XF performed the experiments. BC and CA prepared figures and graphs. BC and GS wrote the manuscript. All the authors read and approved the final manuscript.