Phage-encoded contingency loci enable bet-hedging against host defence mechanisms

Abstract

Bacteriophages are the most diverse biological entities on Earth but the processes influencing the evolution of genomic diversity in phages are poorly understood. Here, we show that phage genomes contain contingency loci (CL), hypermutable DNA regions that promote reversible frameshift mutations through DNA polymerase slippage on simple sequence repeats (SSRs). CL have been extensively described in bacteria, archaea, and eukaryotes yet are understudied in phages. We use experimental evolution and genome sequencing to demonstrate that CL in E. coli phages T2 and T4 reversibly generate genomic and phenotypic heterogeneity in progeny that allow them to hedge their bets against host defenses. We find that SSRs are widespread in diverse E. coli phages and vary in abundance across genes with different functions. Collectively, our study describes a previously unappreciated facet of phage replication in which mutagenic SSRs drive genetic diversification and population heterogeneity, allowing phages to exploit hosts despite varying defense mechanisms.

Introduction

Bacteriophages (phages) are viruses that require bacterial hosts to replicate and produce viral progeny. For strictly lytic phages, this process results in bacterial death, thus selecting for bacterial phage defense systems that enable hosts to survive infection^1–5. Phages, however, evolve anti-phage defense systems, locking phages and their hosts in a co-evolutionary arms race. How phage genetic heterogeneity plays a role in this evolutionary arms race is not understood⁶. Our previous discovery of TgvAB, a Type IV restriction modification system that targets glucosylated 5-hydroxy methyl cytosines, demonstrated that T2 coliphages can evade restriction via mutations in their alpha glucosyltransferase gene (agt)⁷. Here, we show that the T2 agt gene and the T4 rIIA gene encode simple sequence repeats (SSRs) that allow these genes to function as contingency loci (CL) to evade TgvAB restriction^8–10 or RexAB suppression^11–14, respectively. CL are hypermutable regions of DNA that mediate high-frequency, reversible, stochastic, and heritable phenotype switching. CL have been widely studied in eukaryotes and bacteria but less described in archaea^15–18. CL have only been described in a Campylobacter phage in which polyG tracts produced phenotypic heterogeneity in cell surface receptors that altered host infection specificity¹⁹. Whether other phages encode CL has not been experimentally validated. Here, we identify mutagenic SSRs in T-even phages and show that SSR representation varies across diverse E. coli phages and phage gene functions, suggesting that these phages may have evolved other CL for bet-hedging to survive encounters with different hosts and environments.

T2 mutants resistant to TgvAB rapidly evolve.

TgvAB is a Type IV restriction system encoded by Vibrio cholerae ^7,20. Heterologous expression of TgvAB in Escherichia coli protects against T2, T4, and T6 phage infection via recognition of the glucosylated 5-hydroxymethylcytosine (5-hmC) on the phage DNA^7,20. Addition of glucose to 5-hmC on phage DNA is carried out by the agt gene encoded in T-even phages²¹. T2 can escape TgvAB detection by acquiring null mutations in the agt gene⁷. However, loss of glucosylation on 5-hmC results in an evolutionary trade-off as unglucosylated 5-hmC is recognized and restricted by the McrA/BC Type IV restriction enzymes^7,22,23 (Fig. 1A).

Fig. 1 |. CL in T2 agt facilitates exploiting hosts with varying restriction systems.

A. Schematic of WT agt⁺ or agt⁻ infection on E. coli with McrA/BC or TgvAB. B. E. coli with ptgvAB and pEV (empty vector) infected with WT T2 at varying multiplicities of infection (MOI’s) at time 0. C. T2 phage isolated from ptgvAB cultures in Fig 1B was used to reinfect E.coli DH10B with ptgvAB and pEV at varying MOI’s at time 0. D. 10-fold serial dilution plaque assay of T2 pre- and post-infection on pEV and T2 pre- and post-infection on ptgvAB on host pEV and ptgvAB. E. Schematic of WT T2 agt gene with the nucleotide sequence of mutants and the surrounding region in agt compared to WT T2 sequence. Regions highlighted in red indicate a nucleotide deletion while yellow indicates a nucleotide insertion and green indicates a substitution in corresponding position. Mutants chosen for the reversion experiment are boxed in red. F. The mutant frequency of T2 Mut 1 (single strand repeat) and Mut 8 (point mutation) is shown with dots representing one biological replicate and the dashed line represents the average of 5 biological replicates (p-value = 0.01 and is indicated by *, two-sided Wilcoxon rank-sum test was performed). G. MG1655 infected with T2 Mut 1 or T2 Mut 8 at an MOI of 0.001 at time 0. The mean and standard error of 3 (B,C) or 5 (G) biological replicates each with 3 technical replicates are shown.

During liquid culture infections, E. coli expressing TgvAB (ptgvAB) exhibited only transient protection against T2 during the initial stages of growth for all multiplicities of infection (MOI) tested (Fig. 1B). This dynamic differs from other restriction systems in which phage infection of hosts encoding these defenses does not significantly impact growth dynamics^4,5,24,25. We hypothesized that the inhibition of growth by T2 in these cultures was due to rapid selection of TgvAB-resistant T2 mutants. To test this, we collected T2 phages after infecting E. coli (ptgvAB) and reinfected the same strain. These phages now exhibited complete killing of E. coli (ptgvAB), demonstrating rapid evolution of T2 resistance to TgvAB (Fig. 1C, D).

T2 agt functions as a contingency locus

To identify which T2 mutations provide TgvAB resistance, we isolated spontaneous TgvAB-resistant plaques from 15 independently propagated T2 populations by plating wildtype (WT) T2 phages on E. coli (ptgvAB). Illumina whole genome short read sequencing revealed that phages from all 15 TgvAB-resistant plaques had mutations in agt. This result is consistent with our previous findings that null mutations in agt provide T2 resistance to TgvAB⁷. Five mutants had unique single nucleotide polymorphisms (SNPs) in the agt gene while 10 mutants had frameshift mutations of +/−1 in two monomeric adenine (A) single sequence repeats (SSRs) (Fig. 1E). Due to the frequency of mutants via these agt SSRs and the rapid evolution of agt resistant T2 mutants, we hypothesized that these SSRs are CL in agt.

CL are repeat sequences in DNA that have an elevated frequency of reversible mutations relative to the basal rate of DNA polymerase due to DNA polymerase slippage during replication^8,16. CL allow organisms to exhibit genetic heterogeneity, driving bet-hedging strategies to rapidly evolve to survive different selective pressures. To explore whether the agt SSRs exhibit an elevated, reversible mutant frequency, we quantified the agt reversion frequency of T2 Mut 1 (deletion of an adenine in the agt SSR encoded from 288-296 bases into the agt gene) and T2 Mut 8 (an adenine to a guanine single nucleotide polymorphism (SNP)) (Fig. 1E). Equivalent numbers of phages harvested from five independently plaque purified populations of each agt mutant were plated on E. coli strain MG1655 as it encodes the McrA/BC Type IV restriction system that recognizes and restricts agt⁻ T2 (Fig. 1A), thus selecting for agt⁺ T2. Our results from these five independent populations showed that T2 Mut 1 reacquired agt⁺ at a frequency ~1000-fold greater than that of T2 Mut 8 (Fig. 1F). Sequencing of the agt gene of phages from McrA/BC resistant plaques showed WT agt sequence for both the SSR and SNP, demonstrating true reversion for both mutants. Moreover, T2 Mut 8 (SNP) was significantly more restricted by E. coli MG1655 in a liquid culture infection whereas T2 Mut 1 (SSR) more rapidly evolved resistance at all MOIs tested (Fig. 1G, Extended Fig. 1), illustrating the elevated reversion frequency of the mutation in the SSR allows rapid development of resistance to and protection from McrA/BC restriction. Together, these results indicate that the T2 SSR in agt at bases 288-296 is a CL.

T-even phages contain highly mutagenic SSRs

Because CL have an elevated mutation frequency compared to the basal DNA polymerase error rate^8,16, we hypothesized that potential CL could be detected in individual sequence reads generated from Illumina whole genome short read sequencing of phage genomes. This approach quantifies the generation of sequence variation independent of phenotypic selection. To test this hypothesis, we sequenced the T-even phages as they are morphologically, antigenically, and genetically similar^26–28. We propagated five independent WT T2, T4, and T6 populations started from individual plaques on a non-selective host and sequenced to a depth of ~2,000 fold. We identified every SSR (of any base) with a length ≥6 and analyzed individual sequencing reads to quantify the number of variant reads with single-base insertions or deletions in SSRs, which could be due to phage mutations and/or sequencing errors. We limited our analysis to SSRs located in open reading frames (ORF) as insertions or substitutions in these genes will be a frameshift mutation and would most likely produce a null phenotype. Several such SSRs are also encoded in intergenic regions, but these are not discussed further as the functional impact of base insertion or deletion is less clear. This analysis revealed five 7-bp and 149 6-bp A/T SSRs encoded in ORFs in the T2 genome with no G/C repeats ≥ 6-bp (Fig 2A, Supplemental Table 5). Additionally, one 8-bp, five 7-bp and 170 6-bp A/T SSRs are encoded in ORFs in the T4 genome while there are six 7-bp, 166 6-bp A/T and one 6-bp G/C SSRs encoded in ORFs in the T6 genome (Fig 2A, Supplemental Table 6 and 7).

Fig. 2 |. Certain SSRs exhibit elevated variation in T-even E. coli phages.

A. Representation of homopolymer SSRs 6-bp or longer located within ORFs in the genomes of phages T2, T4, and T6 classified by whether the repeated base is in an A/T or G/C base pair. B. Variation in Illumina sequencing data for five independent populations of each T-even phage. Counts of reads matching each A/T 6-bp SSR in the reference genome versus matching a variant with a single-base deletion or insertion of the repeat base were analyzed for all SSRs found in T2, T4, and T6 ORFs. SSRs in each phage with significantly different numbers of variant reads than expected from the overall frequency across all such SSRs are shown (p < 0.05, two-tailed binomial tests with Benjamini-Hochberg correction for multiple testing within each phage; exact p-values are provided in Table S8). These may represent hypervariable SSRs with mutation rates higher than the Illumina sequencing error rate. The 6-bp SSR in the T2 agt gene encoded at position 35708 that mutated in our experiments is one of these. Some orthologous genes are predicted to have hypervariable SSRs in more than one of the three T-even phages.

To identify mutagenic SSRs, we analyzed the occurrence of variant reads for all 6-bp A/T SSRs in ORFs from all fifteen replicates. The mean variant read frequency in our sequencing data for these SSRs was 0.00021 (955 out of 4,496,124 reads) and is indistinguishable across the T-even phages (Fig. 2B). This mean variant read frequency aligned with the estimated Illumina error rate (see Methods)²⁹. However, 11 of the 6-bp SSRs in the T2 genome exhibited variant read frequencies significantly above the background rate of variation, including the second agt SSR at bases 888-901 (Fig. 1E, Fig. 2B, Supplemental Table 8). The other 7-bp SSR identified in the agt gene described in Fig. 1 exhibits a higher variant read frequency of 0.00072. We could not determine whether this variant read frequency differs from the background Illumina error frequency because there are only four other 7-bp SSRs in the T2 genome, and both SSR mutation rates and Illumina sequencing error rates are expected to be higher for longer SSRs. Additionally, identification of 6-bp SSRs in T4 and T6 showed that T4 encodes 9 highly mutagenic SSRs while T6 encodes 3 (Fig. 2B).

These highly mutagenic SSRs are encoded in all functional gene categories within T-even phages (Fig. 2B, Supplemental Table 8). Furthermore, we identified four genes in both T2 and T4 that encoded highly mutagenic SSRs; genes 43, 45, and 55, which are involved in replication and gene 5, which plays a role in infection. Additionally, one highly mutagenic SSR was identified in both T4 and T6 in the homologous ORFs annotated as 30.1 and gp197, respectively (Fig. 2B, Supplemental Table 8). These results demonstrate that T-even phages encode highly mutagenic SSRs within various functional categories, some of which are conserved between T-even phages, suggesting these highly mutagenic SSRs could function as CL.

SSRs are widespread in phage genomes

To further ascertain the range of SSRs in phages, we analyzed ORFs in the genome sequences of 41 dsDNA E. coli phages for ≥6-bp SSRs. These phages are a subset of the BASEL phage collection, which contains representatives of the major E. coli phage ICTV genera³⁰, that all have a pairwise average nucleotide identity <90% (see Methods). Within these 41 diverse phage genomes, we found that the density of SSRs varied widely. P2vir had the highest abundance with 2.0 SSRs per thousand base pairs (kbp) while three phages (T7, Bas105 and Bas68) had no SSRs of this length or longer in ORFs (Fig. 3A, top panel). Additionally, we found that G/C SSRs are much less prevalent than A/T SSRs (Fig. 3A, top panel). The overall abundance of SSRs and lack of G/C SSRs did not reflect genome size nor GC content (Fig. 3A, second panel).

Fig. 3 |. SSRs abundance varies across E. coli phages and gene functional categories.

A. Characterization of ≥6-bp SSRs in 41 diverse phages with double-stranded DNA genomes. Phages with Bas# identifiers are from the BASEL collection³⁷. The top panel shows the counts of SSRs per 1000 base pairs (kbp) within genes for each phage. The second panel shows phage genome size and GC content. The third panel shows the fraction of the bases in genes in each genome assigned to eight functional categories. The fourth panel shows the ratio of the actual number of SSRs observed in genes in each phage to a random expectation created by generating 1000 codon-shuffled genomes. Error bars are 95% confidence limits estimated from the resampled distributions. B. Most phage genomes have fewer SSRs than expected relative to the random expectation from codon-shuffled genomes. C. Variation in the degree to which SSRs are depleted in genes in different functional categories for all phage genomes considered together. Violin plots show the distribution of the actual to expected ratio over the 1000 sets of codon-shuffled genomes. Letters indicate categories with significant differences (p < 0.05, two-tailed randomization tests with Holm correction for multiple comparisons).

We next analyzed the abundance of SSRs in different phages and in phage genes with different functions. We first annotated all phage ORFs in these 41 phages using PhageScope³¹. Many of these 41 phages, like Bas1, have genes representing all eight main functional categories, while others, like P2vir, have more genes concentrated in a smaller subset of these categories (Fig. 3A, third panel). To determine whether the SSR abundance in ORFs for each of the 41 phage genomes could be due to chance based on their GC-content or codon usage, we simulated phage genomes that maintained the amino acids sequences of each ORF but shuffled synonymous codons genome-wide and compared the number of SSRs in these randomized sequences to the actual number of SSRs. We found that 32 of the 41 phages encode fewer SSRs than expected while only 2, including λ and P2vir, encode more SSRs than would be expected by chance (Fig. 3A, bottom panel, p ≤ 0.05 for two-tailed randomization tests). Further analysis of the 41 phages showed that G/C SSRs >5 bp and A/T SSRs >6 bp are depleted in their genomes overall, relative to the random expectation from codon-shuffled genomes (Fig. 3B). Our analyses suggest that SSRs of a certain length are generally selected against in phage ORFs, potentially due to their high mutation rates.

We next assessed depletion or enrichment of SSRs relative to the randomized genomes in genes classified into each functional category (see Methods). This analysis revealed that SSRs overall are depleted in all functional categories. However, we find that the replication protein category, which encodes proteins such as polymerases, are significantly more depleted compared to the other functional categories (Fig. 3C). This result is not surprising as these proteins are likely to be essential. On the other hand, we observed that the regulatory gene category is significantly less depleted in SSRs compared to the other functional categories, suggesting this class of genes may encode more mutagenic SSRs to promote rapid genome evolution (Fig. 3C). These results show that SSRs are widely distributed in phages and generally underrepresented compared to expectations but specific gene categories exhibit differential abundance of SSRs.

The T4 rIIA gene encodes a contingency locus

We next examined an SSR of 6 As encoded in the rIIA gene in T4 to further explore the role of CLs in phages. We chose this well-studied gene as it was key in elucidating fundamental aspects of molecular biology including determination of a triplet genetic code and demonstration of homologous recombination^13,32–35. Although rIIA has a rich history, a CL has never been described in this gene. Moreover, the frequency of insertions or deletions in the rIIA SSR measured by Illumina sequencing was not statistically elevated over the baseline Illumina variant frequency. To examine whether the SSR in rIIA exhibits a higher reversion frequency relative to other mutations, we engineered an adenine addition in the SSR (T4 Mut 123) and a substitution of an adenine to a guanine 53 bp upstream (T4 Mut 127), regenerating two previously isolated spontaneous rIIA null mutations^13,36 (Fig. 4A). During superinfection, when phages outnumber hosts, WT T4 activates lysis inhibition (LIN) to delay infection. T4 rIIA mutants undergo rapid lysis, overcoming the delayed infection in E. coli B. However, the loss of rIIA results in an evolutionary tradeoff as these phages are now susceptible to RexAB, a phage defense system encoded by a lambda lysogen^11–14 (Fig. 4B). Thus, we quantified restoration of the rIIA⁺ phenotype in T4 Mut 123 and T4 Mut 127 from five independent populations, each started from unique plaques, by plating on E. coli MG1655 (λ). T4 Mut123, which had a mutation in the SSR, reverted at ~10,000× higher frequency than the SNP mutation T4 Mut 127, confirming this SSR repeat in rIIA is a CL (Fig. 4C). Analogous to our results with agt, the rIIA CL provided increased fitness upon infection of K12(λ) as the RexAB defense system only exhibited transient protection against the rIIA CL null mutant, T4 Mut123, but complete protection against the rIIA SNP mutant, T4 Mut127 (Fig. 4D, Extended Fig. 2). Together, our results show that although rIIA mutations were not detectably elevated above the background Illumina sequencing error rate, this SSR functions as a CL, allowing T4 to generate subpopulations of rapid lysis or RexAB-resistant phages.

Fig. 4 |. Evidence for CL reversion frequency in T4 rIIA.

A. Nucleotide sequence of mutants and the surrounding region in rIIA compared to WT T4 sequence. The region highlighted in yellow indicates a nucleotide insertion while blue indicates a substitution. B. Schematic of WT rIIA or rIIA null mutant infection on either E.coli B host with or without superinfection or host with RexAB (λ) C. The mutant frequency of T4 Mut 123 (CL) and Mut 127 (point mutation) plated on MG1655(λ) is shown. Dots represent one biological replicate, and dashed line represents the average of 5 biological replicates per T2 mutant (p-value = 0.007 and is indicated by **, two-sided Wilcoxon rank-sum test was performed). D. MG1655(λ) infected with T4 Mut 123 or T4 Mut 127 at an MOI of 0.001 at time 0 is shown. The mean and standard error of 5 biological replicates each with 3 technical replicates are presented. E. When an individual phage replicates in its host, nucleotide repeat mutants arise through polymerase slippage at different mutagenic SSRs, resulting in a phage population with many distinct genomes.

Discussion

Our results indicate that during genome replication of dsDNA phages, DNA polymerase slippage at mutagenic SSRs produces frameshift mutations at rates thousands of times higher than basal mutation rates (Fig. 4E). This process generates genetic heterogeneity, providing a bet-hedging strategy that enables phages to survive diverse selective pressures, such as bacterial phage defense systems. Results from analyzing whole-genome sequencing of various phages for ≥6 SSRs revealed SSRs are found in genes with a wide range of functions that include assembly, lysis, packaging, regulation, and replication. The prevalence of SSRs in genes of various functions suggests this evolutionary bet hedging strategy likely extends beyond evolving to overcome phage defense and extends to other selective pressures such as divergent bacterial hosts or environments. Furthermore, we found that phages were highly varied in the average number of SSRs/kB in their genomes. Interestingly, the three phages with the most abundant SSRs/kB, P1vir, P2vir, and λ, are the only three temperate phages we analyzed, suggesting that mutagenic SSRs might be more beneficial to phages that switch between a lytic and lysogenic state.

CL have only ever been described once in a Campylobacter phage, although it was never designated as a CL. Fletchervirus phages were found to encode 7-11 guanine (polyG) tracts in receptor binding proteins, generating heterogeneity that allows a sub-population to infect Campylobacter when its phase-variable receptor is not expressed¹⁹. Two additional papers identified polyG tracts in a thymine methylase gene of bacteriophages BBP-1³⁷ and in PB1-like phages in a baseplate gene³⁸. Yet whether these tracts provided genetic heterogeneity through elevated mutation rates was never tested.

The mutational variation that arises from phage CL may reflect a history of “Red Queen” coevolution between bacteria and phages, where each side must constantly evolve to counter evolution by the antagonist. By evolving gene sequences that experience elevated rates of frameshift mutations in CL, phages can harness their short life cycle and high numbers of progeny to rapidly and reversibly sample genetic variation to enhance evolution. Previous studies have shown that genetic heterogeneity helps phage survive host diversification by creating variation in their tail proteins, but this heterogeneity was driven by base substitution mutations which occur at a much lower rate and are less likely to exactly revert^39,40. Here, we show diverse E. coli phages encode mutagenic SSRs in many functional categories of genes and define two new CL in the agt and rIIA genes that drive a bet-hedging strategy to overcome phage defense systems.

Materials and Methods

Strains and growth conditions

Bacterial and phage strains used in this study are listed in Supplementary Table 1. The Basel phages were obtained from Alexander Harms. Unless otherwise stated, cultures were grown in Luria broth (LB) at 37°C with shaking at 210 rpm or LB plates at 37°C and supplemented with the following antibiotics for plasmid maintenance as necessary: kanamycin (100 μg mL⁻¹) and/or carbenicillin (100 μg mL⁻¹). T2, T4, and T6 were propagated by streaking on MMB agar (LB + 0.1mM MnCl₂ + 5 mM CaCl₂ + 5 mM MgCl₂ + 0.5% agar) with E. coli DH10B at a 1:1,000 dilution of an overnight culture and incubated overnight. Plates were then flooded with 10 mL of phage buffer (0.1 mM Tris HCl 7.5 pH + 10mM MgSO₄ + 0.4% NaCl₂ + dH₂O) and incubated statically at 4°C for 24 hours. Phage buffer was then filtered through a 0.22 μM filter.

Plasmid construction

Plasmids are listed in Supplementary Table 2 and primers in Supplementary Table 3. All PCR products were amplified using Q5-High Fidelity DNA polymerase according to the manufacturer’s instructions (New England Biolabs). To construct, pJBG122, rIIA was amplified from wild-type T4 Carolina with overlapping ends to pUC19 using oJBG206 and oJBG207. One region of pJBG122 was amplified with site directed-mutagenesis primer oJBG208 (pJBG123) or oJBG216 (pJBG127) with oMJF006 and another region was amplified with site-directed mutagenesis primer oJBG209 (pJBG123) or oJBG217(pJBG127) with oMJF005. All plasmids and PCR products were confirmed by sequencing (Plasmidsaurus). Plasmid sequencing results were mapped to T4 reference genome (NC_000866.4) using Geneious Prime 2023.0.2.

To construct pJBG122, rIIA was amplified from wild-type T4 Carolina with overlapping ends to pUC19 using oJBG206 and oJBG207. pUC19 was linearized using HF_pUC19-MCS_F and HF_pUC19-MCS_R. To construct pJBG123 and pJBG127, one region of pJBG122 was amplified with site-directed mutagenesis primer oJBG208 (pJBG123) or oJBG216 (pJBG217) with oMJF006 and another region was amplified with site-directed mutagenesis primer oJBG209 (pJBG123) or oJBG217 (pJBG127) with oMJF005. Both PCR products were used to create pJBG123 or pJBG127, respectively. Inserts were verified by PCR using primers oJBG191 and oJBG192.

T4 rIIA site directed mutagenesis

To generate T4 Mut 123 and Mut 127, E. coli strain DH10B containing pJBG123 and pJBG127 respectively, were grown overnight in LB at 37°C. Overnight cultures were mixed 1:1,000 with 20 mL of MMB agar + carbenicillin and poured into 150 x 15mm Petri dish. Plates were left to solidify for 30 minutes at room temperature. After 30 minutes, 150 μl of wild-type T4 was spread across the plates. Plates were dried and placed at 37°C overnight. Plates were covered with 10 mL of phage buffer and placed at 4°C overnight. Phage buffer was collected and filtered using a 0.22 μM filter. Lysates were then taken and spread on plates made with an overnight culture of E.coli B at 1:1,000 and mixed with 20 mL of MMB agar and poured into a 150 x 15mm Petri dish and incubated at 37°C overnight. Plates were covered with 10 mL of phage buffer and placed at 4°C overnight. Phage buffer was collected and filtered using a 0.22 μM filter. Lysate was then streaked on plates made with an overnight of E. coli B at 1:1,000 and mixed with 10 mL of MMB agar to obtain isolated plaques. Plaques were then picked based on the large plaque phenotype on E. coli B. T4 Mut 123 and Mut 127 were verified by plating on MG1655(λ) and with PCR (Plasmidsaurus) using primers oJBG191 and oJBG192.

Phage infection and reinfection in liquid culture

Overnight cultures of DH10B ptgvAB and pEV (pJBG078 and pMJF103) grown in LB + kanamycin were diluted to an OD₆₀₀ of 0.01 into 10 mL flasks containing LB + kanamycin. Cultures were grown for ~2 hours to an OD₆₀₀ 0.1. The cultures were distributed into a 96 well plate and wild-type T2 was added at MOIs of 0.1, 0.01, and 0.001. Infection was monitored by measuring OD₆₀₀ every 2.5 minutes for 10 hours at 37°C using a BioTek 800 TS (Agilent) plate reader with continuous, linear shaking. After 10 hours plates were placed in 4°C. Cultures from the wells were then collected and transferred to an AcroPrep Advance 96-well filter plate for aqueous filtration with a 0.2 μM filter. The 96-well filter plate was placed on top of another 96 well plate to collect phage lysate. Plates were centrifuged at 2,250 xg at 4°C. Phage concentration was calculated from infection on DH10B. Overnight cultures of DH10B ptgvAB and pEV grown in LB + kanamycin were diluted to an OD₆₀₀ of 0.01 into 10 mL flasks containing LB + kanamycin. Cultures were grown for ~2 hours to an OD₆₀₀ 0.1. Once OD₆₀₀ = 0.1, cultures were distributed into a 96 well plate. Using phage lysate from the initial E. coli ptgvAB infection, MOIs initially at 0.1, 0.01 and 0.001 were added again at the exact MOI into 96 well plate and measured with the same parameters as initial infection. Experiments were replicated 3 times independently and representative images are shown.

Phage infection in liquid culture

Overnight cultures of MG1655 or K-12 (λ) were diluted to an OD₆₀₀ of 0.01 into 10 mL flasks containing LB. Cultures were grown for ~1.5 hours to OD₆₀₀ 0.1. Once OD₆₀₀ = 0.1, cultures were distributed into a 96 well plate and T2 Mut 1 and Mut 8 were added to MG1655 or T4 Mut 123 and Mut 127 were added to K-12 (λ) at MOIs of 0.1, 0.01, and 0.001. Infection was monitored as described above. Experiments were replicated 3 times independently and representative images are shown.

Efficiency of plaquing (EOP) assays

Overnight cultures of DH10B ptgvAB and pEV were mixed 1:1,000 with 20 mL MMB + kanamycin and poured into 150 x 15mm Petri dishes. Wild-type, post, and pre T2 stocks were diluted 1:10 up to a 10⁻¹² dilution. 5 μl of each dilution was spotted on DH10B ptgvAB and pEV plates. Plates were dried before placing at 37°C overnight. Experiments were replicated 3 times independently and representative images are shown.

Mutant frequency assay

T2 Mut 1 and T2 Mut 8 were struck out on overnight cultures of ptgvAB mixed 1:1,000 with 10 mL of MMB agar + kanamycin and poured into 100 x 15mm Petri dish. 5 isolated plaques were picked and resuspended in 150 μl of distilled H₂O (dH₂O). Overnight cultures of DH10B were mixed 1:1,000 with 20 mL of MMB agar and poured into 150 x 15mm Petri dish. Plates were left to solidify for 30 minutes at room temperature. After 30 minutes, 150 μl of each isolated plaque was spread, dried and placed at 37°C. Plates were covered with 10 mL of phage buffer and placed at 4°C overnight. Phage buffer was collected and filtered using a 0.22 μM filter.

T4 Mut 123 and T4 Mut 127 were struck out on overnight cultures of E. coli B mixed 1:1,000 with 10 mL of MMB agar + kanamycin and poured into 100 x 15mm Petri dish. 5 isolated plaques were picked and resuspended in 150 μl of distilled H₂O (dH₂O). Overnight cultures of MG1655 were mixed 1:1,000 with 20 mL of MMB agar and poured into 150 x 15mm Petri dish. Plates were left to solidify for 30 minutes at room temperature. After 30 minutes, 150 μl of each isolated plaque was spread, dried and placed at 37°C. Plates were covered with 10 mL of phage buffer and placed at 4°C overnight. Phage buffer was collected and filtered using a 0.22 μM filter. EOP of all isolated phage populations for T2 Mut 1 and Mut 8 were determined on MG1655 and DH10B. EOP of T4 Mut 123 and Mut 127 were determined on MG1655 and K-12 (λ).

Using EOPs we performed plaque-forming unit quantification assays for all independent phage lysates. PFU quantification assays were performed as follows: overnights of the appropriate cultures were mixed with the lowest concentration of coliphage (180 μl of culture with 20 μl of the appropriate phage) where plaques were still visualized. Culture and phage were then mixed 1:1,000 with 20mL of MMB agar into 100 x 15 mm Petri dishes. Once plates solidified, they were incubated at 37°C overnight. Plaques were counted the next day. The following equation was used to calculate mutant frequency, m = ((D1/C1)*(C2/D2)) where D1 is the dilution factor from non-selective host lysate, C1 are PFU counts on non-selective host, D2 is the dilution factor from selective host lysate and C2 are PFU counts on selective host and m is the mutant frequency. Experiments for each individual population were replicated 3 times independently and representative images are shown.

Illumina sequencing

To extract phage DNA, 500 μl phage lysates suspended in phage buffer and incubated at 98°C for 10 minutes to burst open the phage capsid head and inactivate any residual DNase and RNase from the bacterial culture. Samples were further purified using the Wizard Genomic DNA Purification Kit (Promega) following the protocol for isolation of genomic DNA from Gram-negative bacteria starting at step 3, under Lyse Cells. High-titer phage lysates (>10⁶ PFU μl⁻¹) were used for extraction. DNA extracted samples were analyzed using short read whole genome Illumina sequencing (SeqCoast) using the 200 Mbp/1.3 million reads service with 150 bp paired end reads. Sequencing was performed using either a standard SBS flow cell or a XLEAP flow cell on a NextSeq 2000. The mutations identified in the 15 T2 agt null mutations we isolated upon selection on E. coli (ptgvAB) are provided in Supplemental Table 4.

SSR variation in T-even phage sequencing data

Illumina sequencing results for five populations grown from individual isolated plaques for each of the T-even phages (15 total samples) were analyzed using breseq⁴¹ with the brefito pipeline⁴². Default options were used to align reads to the corresponding reference genome: T2 (NC_054931.1), T4 (NC_000866.4), and T6 (NC_05407.1). Then, reads matching the reference sequence for SSRs of 6-bp and variant homopolymers matching these SSRs expanded or contracted by one base were counted using the breseq tabulate-CL utility command in strict mode, which requires an exact match to the five bases on either side of the SSR for a read to be counted for either the reference or a variant. There were few G/C SSRs, so we restricted our subsequent analysis to A/T SSRs in open reading frames.

We analyzed variant versus total (variant plus reference) read counts for each 6-bp A/T SSR, summed across all five replicate samples for each phage, using binomial models in R. We did not find any evidence of differences in the overall frequencies of variant reads for these SSRs among the three T-even phages (p = 0.168, likelihood-ratio test comparing binomial generalized linear models with and without phage as a predictor). Thus, we assigned one overall background SSR variant rate, which is likely dominated by the rate of sequencing errors, to all phages based on summing read counts across all SSRs. To identify outlier SSRs with rates that were significantly different than the background rate of variation, we performed two-tailed binomial tests for each SSR comparing its actual variant and reference counts to this overall rate. We corrected the resulting p-values for multiple testing within each phage using the Benjamini-Hochberg procedure. Results for all 6-bp SSRs in the T-even phages are provided in Supplemental Table 8.

SSR abundance in E. coli phages and gene categories

BASEL collection and reference phage genome sequences were downloaded using GenBank accessions provided in Humolli et al 2025³⁰. We down selected to a core set of phages with less redundancy by removing those with similar genomes until 41 remained that all had a pairwise average nucleotide identity <90% as calculated using FastANI (v1.34)⁴³. GenBank accession numbers and information about these 41 phage genomes are provided in Supplemental Table 9. We transferred annotations of phage protein categories from the GenBank portion of PhageScope³¹ to proteins in the 41 analyzed BASEL genomes via BLASTP (v2.15.0 +) searches⁴⁴, taking top matches and requiring 70% query covered and an E-value of ≤ 10⁻⁶ to assign a category to a protein. Because there were few proteins in the “immune” and “tRNA_related” categories, we combined those with “hypothetical” and “unsorted” proteins into one “other” category. SSRs located within genes in each phage genome were identified using a Python script and annotation information loaded using BioPython⁴⁵.

We assessed depletion or enrichment of ≥6-bp SSRs by comparing the actual counts of SSRs per genome or gene functional category to the results of randomization tests in which we shuffled codons across all genes in a genome. This procedure holds genome-wide GC-content, codon usage, and gene lengths constant. We estimated confident limits and p-values from 1000 total randomized versions of every phage genome. For enrichment or depletion of SSRs per phage, we added 0.5 pseudocounts when calculating ratios to deal with observations of zero SSRs in some phages. We did not add pseudocounts for the functional category analysis. For genes assigned to multiple functional categories, if any were “other”, we ignored that assignment. Otherwise, we counted all functions when analyzing the representation of SSRs. The gene’s length was divided by the number of categories to which it was assigned for plotting the fraction of gene bases assigned to different functional categories in a phage genome.

Statistics & Reproducibility

No statistical methods were used to predetermine sample size, and no data points were excluded from the analysis unless explicitly stated. Randomization and blinding were not used on samples in this study because experimental groups were defined by phage and host genotypes and all outcomes were measured objectively. For infection curves, each biological replicate had three technical replicates and represents independently performed experiments. Number of biological replicates performed are indicated in the corresponding figure legend. Data is presented as mean and standard error of all biological and technical replicates. For mutant frequency assays, each biological replicate represents independently performed experiments. Statistical analyses for mutant frequency assays were performed using two-sided Wilcoxon rank-sum tests. SSR frequency analysis of T2, T4 and T6 was performed using two-tailed binomial tests with Benjamini-Hochberg corrections for multiple testing within each phage. Ratio of actual number of SSRs observed in 41 diverse phages with double stranded DNA genomes to a random expectation was calculated by generating 1000 codon-shuffled genomes with 95% confidence limits estimated from the resampled distribution.

Extended Data

Extended Fig. 1 |. Contingency loci in T2 agt protects against varying restriction systems.

MG1655 infected with T2 Mut 1 or T2 Mut 8 at an MOI of 1 and 0.1 at time 0. The mean and standard error of 5 biological replicates each with 3 technical replicates are presented.

Extended Fig. 2 |. Evidence for CL reversion rate in T4 rIIA.

MG1655(λ) infected with T4 Mut 123 or T4 Mut 127 at an MOI of 1 and 0.1 at time 0. The mean and standard error of 5 biological replicates each with 3 technical replicates are presented.

Data Availability

All data generated in this work is available at FigShare identifier 10.6084/m9.figshare.32680731. All plasmids, bacterial strains, and phages are available upon request. Whole genome sequences are available at NCBI reference PRJNA1478164.

Code Availability

All code generated in this work is available at FigShare identifier 10.6084/m9.figshare.32680731.