Viral recombination systems limit CRISPR-Cas targeting through the generation of escape mutations

Abstract

CRISPR-Cas systems provide immunity to bacteria by programming Cas nucleases with RNA guides that recognize and cleave infecting viral genomes. Bacteria and their viruses each encode recombination systems that could repair the cleaved viral DNA. However, it is unknown whether and how these systems can affect CRISPR immunity. Bacteriophage λ uses the Red system (gam-exo-bet) to promote recombination between related phages. Here we show that λ Red also mediates evasion of CRISPR-Cas targeting. Gam inhibits the host E. coli RecBCD recombination system, allowing recombination and repair of the cleaved DNA by phage Exo-Beta, which promotes the generation of mutations within the CRISPR target sequence. Red recombination is strikingly more efficient than the host’s RecBCD-RecA in the production of large numbers of phages that escape CRISPR targeting. These results reveal a role for Red-like systems in the protection of bacteriophages against sequence-specific nucleases, which may facilitate their spread across viral genomes.

Graphical Abstract

Blurb

Bacteriophages harboring mutations at the target site of RNA-guided Cas nucleases can escape CRISPR immunity. Hossain et al. report that phage λ Red recombination system inhibits the host bacteria RecBCD pathway and introduces mutations that more efficiently mediate escape. These results suggest that phage recombination systems counteract DNA-cleaving bacterial defenses.

INTRODUCTION

Clustered regularly interspaced short palindromic repeat (CRISPR) loci and CRISPR-associated (cas) genes protect bacteria and archaea against foreign genetic elements such as viruses (Barrangou et al., 2007) and plasmids (Marraffini and Sontheimer, 2008). Upon infection, short invader sequences, known as spacers, are inserted in between the repeats of the CRISPR locus (Barrangou et al., 2007). These are subsequently transcribed and processed to generate short CRISPR RNAs (crRNAs) (Brouns et al., 2008; Hale et al., 2008; Tang et al., 2005) that are used as guides by Cas complexes to recognize and destroy complementary protospacer sequences within the nucleic acids of the invading virus or plasmid (Garneau et al., 2010; Gasiunas et al., 2012; Hale etal., 2009; Jinek et al., 2012; Jore et al., 2011). CRISPR-Cas systems can be classified into six different types depending on their cas gene content (Makarova et al., 2020). Types I and II are the most common DNA-cleaving systems and have two target requirements for activity: a protospacer-adjacent motif (PAM) and a seed sequence within the protospacer. In the commonly studied type II-A system of the Gram-positive bacterium S. pyogenes, the PAM is a 5’-NGG-3’ sequence immediately downstream of the protospacer and the seed sequence is located in the 6-8 nucleotides that precede the PAM (Bikard et al., 2012; Jinek et al., 2013). Successful recognition of a target by the Cas9 RNA-guided nuclease leads to the introduction of a double-strand break (DSB) in the protospacer DNA three nucleotides upstream of the PAM (Garneau et al., 2010; Jinek et al., 2013). In the E. coli type I-E system, the RNA-guided Cascade complex recognizes targets with a 5’-AWG-3’ PAM (Westra et al., 2012) upstream of the protospacer and a seed sequence in the 8 nucleotides immediately downstream of the PAM (Semenova et al., 2011). Upon target recognition, Cascade recruits the ssDNA nuclease Cas3 (Westra et al., 2012), which first degrades the non-complementary DNA strand (Mulepati and Bailey, 2013; Sinkunas et al., 2013) and then unwinds and degrades the complementary strand (Mulepati and Bailey, 2013; Redding et al., 2015). These activities not only cut, but also further degrade the target DNA; however, the extent to which Cas3 or other host nucleases are involved in the destruction of the invader’s genome is not known. In both systems, mutations in either the PAM or seed sequences lead to evasion of CRISPR immunity by the invader (Deveau et al., 2008; Nussenzweig et al., 2019; Semenova et al., 2011). A recent study investigating type II-A immunity against T4 phage in E. coli showed that viral escape mutations accumulated during the course of infection (Tao et al., 2018), suggesting that target mutations are unlikely pre-existing but rather introduced de novo after Cas9 cleavage. However, how these mutations are generated is not known.

We investigated how bacteriophage λ escapes cleavage by type II-A and type I-E CRISPR-Cas systems in E. coli. Both the host and its invader possess pathways that can hypothetically repair the DSBs generated by Cas nucleases through homologous recombination (Wright et al., 2018). First, an exonuclease recognizes the free DNA ends generated at the break and degrades each DNA strand asymmetrically. This introduces a 3’ end overhang that is subsequently covered by a ssDNA binding protein that mediates recombination with a homologous DNA template. In E. coli, the main homologous recombination pathway is RecABCD, where RecBCD is the exonuclease and RecA is the recombinase (Dillingham and Kowalczykowski, 2008; Kuzminov, 1999). RecBCD binds to free dsDNA ends at a DSB, then rapidly and processively degrades both strands of DNA until it reaches a chi site, beyond which asymmetric degradation of dsDNA by RecBCD generates a 3’ overhang (Dillingham and Kowalczykowski, 2008). RecA coats the overhang and mediates recombination via strand invasion into an intact DNA molecule with a homologous sequence. Following branch migration, DNA polymerases fill in the gaps and the Holliday junctions are resolved. Phage λ harbors the red operon, which contains three genes: gam, exo and bet. exo encodes the exonuclease (Exo) that asymmetrically degrades free DNA ends to generate 3’ overhangs and bet produces the recombinase of this system (Beta) that coats the ssDNA for recombination via single-strand annealing of two homologous templates (Mosberg et al., 2010). gam encodes a RecBCD inhibitor (Gam) (Kuzminov, 1999) which enables Exo-Beta to drive recombination during phage λ infection (Murphy, 1998). Previous studies found that the Red system is important for phage λ replication (Echolas and Gingery, 1968; Signer and Weil, 1968), to generate long genome concatemers during rolling circle replication (Enquist and Skalka, 1973). Notably, the Red system is not essential and its absence can be rescued by the introduction of an E. coli chi site into the λ genome, which mediates recombination via the host’s RecABCD system (Henderson and Weil, 1975). Therefore, the widespread distribution of phage recombinases in a large number of both temperate and lytic phage genomes (Lopes et al., 2010) suggests they possess additional functions that select for their presence in otherwise highly compact phage genomes (Brüssow and Hendrix, 2002).

Here we found that many spacers mediating the targeting of both type I and type II effector complexes, Cascade-Cas3 and Cas9, respectively, against bacteriophage λ allow the propagation of large numbers of escaper phages with target mutations, and thus provide poor defense. Surprisingly, escaper formation requires Exo-Beta recombination to generate de novo escape mutations. Importantly, both host and phage mutations that enable robust RecABCD recombination and repair of the phage DNA fail to generate escapers as efficiently as the λ Red recombination pathway. Our results define an additional function for the λ Red system, and possibly for other similar phage recombination systems, in driving escape from CRISPR-Cas targeting through the mutagenic repair of DSBs.

RESULTS

Escape mutations within phage λ Cas9 targets are generated during infection

To investigate the impact of recombination systems on the outcome of CRISPR targeting, we decided to study the effects of the best characterized RNA-guided DNA nuclease, Cas9 (Jiang and Doudna, 2017) on the most studied bacterial virus, the λ phage (Wegrzyn et al., 2012). Therefore, we programmed pCas9 (Jiang et al., 2013), a plasmid that carries the type II-A Cas9 RNA-guided nuclease from Streptococcus pyogenes SF370 and its co-factor tracrRNA, with six crRNA guides that target different genomic regions of a virulent mutant of phage λ (λvir) (Figures S1A and S1B). We found that only one of the six spacers mediated robust immunity (spc45, with an efficiency of plaquing, EOP, of ~10⁻³; Figures S1C–S1D). The other five spacers either mediated poor (spc15 and spc26D) or extremely weak (spc9, spc40 and spc14) immunity. DNA sequencing revealed that phages from all plaques analyzed contained escaper mutations that prevent Cas9 cleavage (Figure S1E and Table S1). Escapers of spc9, spc40 and spc45 displayed single-nucleotide modifications in either the target seed or PAM sequences. Escapers of spc14 harbored deletions between short homologous sequences, which is generated through microhomology-mediated end joining (MMEJ), a poorly characterized DNA repair pathway in prokaryotes (Wright et al., 2018). Finally, escapers of spc15 and spc26D contained multiple mutations across the target region that matched the sequence of cryptic prophages present in the chromosome of E. coli MG1655 (Figure S1F), suggesting homologous recombination with λvir to generate escapers. We then sequenced the spc9 target from phages present in plaques formed on lawns of non-targeting bacteria expressing Cas9 programmed with spcNT. Given that Cas9 programmed with spc9 leads to a reduction in plaque formation of less than one order of magnitude, we should expect to detect about one inactivating mutation per ten targets sequenced. This was not the case, as no spc9 target mutations were observed in 32 different plaques (data not shown). Therefore, the high numbers of phage mutants detected during spc9-mediated Cas9 targeting cannot be pre-existing and are likely generated during infection.

The phage λ Exo-Beta recombination system is required to generate escape mutations during infection

Given the mutation frequency for phage λ, calculated to be 7.7 × 10⁻⁸ mutations per base pair (Drake, 1991), the high frequency of escaper mutations we observed for spc9 and spc40 (<10⁻¹) led us to hypothesize that mutations could be introduced during repair of the DSB generated by Cas9, by DNA polymerases that fill in ssDNA gaps across the cleaved sequence (Wright et al., 2018). The main pathway for DNA repair in E. coli is RecABCD recombination (Dillingham and Kowalczykowski, 2008). However, expression of Gam during λ infection will inhibit the RecBCD nuclease from processing DSBs generated by Cas9 in the phage DNA. This allows the phage-encoded Exo-Beta recombination pathway to process DSBs during λ infection. To test if repair via the Exo-Beta pathway could mediate phage escape, we compared plaque formation between λvir and λvir Δexo or λvir Δbet (Figure S2A) and found that deletion of the phage recombination genes severely decreased the number of escaper plaques. However, the reduction in escapers in the λvir Δexo and λvir Δbet phages could be a result of the involvement of this recombination system in the generation of target mutations, but also could be a consequence of the importance of these genes for λ replication (Echolas and Gingery, 1968; Signer and Weil, 1968). To determine the relative replication rates of these mutants, we performed quantitative PCR (qPCR) to measure the levels of phage DNA after 30 minutes of infection (Figures S2B and S2C). Phages carrying the Δexo Δbet double deletion, as well as the elimination of the full Red system (Δred), however, displayed a significant reduction in DNA content (Figure S2C), which could limit the amount of λ DNA inside the host and thus lower the probability of mutagenic repair. Previous work has shown that a burst size defect of λ Δred phages, presumably due to impaired formation of genome concatemers, can be overcome by a mutation that introduces a functional E. coli chi site (5’-GCTGGTGG-3’) in the λ genome (Henderson and Weil, 1975). We wondered whether this modification would normalize the replication levels of the different λvir mutants and therefore we introduced a chi site to generate λvir chi¹ (Figure S1A). First, we looked at the targeting efficiencies of the six spacers used in this study, finding almost identical results for λvir chi¹ to those obtained for λvir (Figures S1B–D). Next we generated a Δexo Δbet double deletion variant of λvir chi¹ and found that it accumulates similar DNA content during infection to the wild-type phage λvir chi¹ (Figure 1A). We then used this phage to test the contribution of Exo and Beta to the generation of escapers of spc9-, spc40- and spc45-mediated Cas9 targeting. We found very similar results to those obtained with the λvir phage (Figures 1B–D and S2D). We also corroborated that the individual gene deletions displayed the same phenotype as the Δexo Δbet double mutant (Figures 1E and S2D). Finally, we complemented the spc9-targeting E. coli host with plasmids expressing Exo or Beta and repeated the infections. The decrease in the number of escapers of the λvir chi¹ Δexo and Δbet phages was rescued to λvir chi¹ levels by the complementing plasmids (Figures 1F, 1G and S2E). Since all of the strains are infected with aliquots of the same viral population, this rescue demonstrates not only that exo and bet are responsible for the increase in the number of λ escapers, but also that escapers are not pre-existing but rather generated during infection. Altogether these results show that the Exo-Beta recombination system from λ phage promotes escape from Cas9 cleavage by increasing the production of phages carrying target site mutations.

Figure 1. The phage lambda Red system is required for the generation of Cas9 escaper phages during type II-A CRISPR-Cas targeting.

(A) Quantitative PCR analysis of viral DNA within E. coli cells, 30 minutes after infection with different λvir chi¹ phages at an MOI of 1. ΔP values were obtained after infection with a non-replicative λ phage lacking the P gene necessary for initiating phage DNA replication. Fold-change values relative to λΔP 15-min time point values are reported. (B-D) Efficiency of plaquing of different λvir chi¹ phages on lawns of E. coli expressing Cas9 programmed with spc9 (B), spc40 (C), or spc45 (D). (E-H) Same as (B) but infecting bacteria carrying the pAM38 vector (E) expressing Exo (F), Beta (G) or Gam (H). Mean ± SEM values of three independent experiments are shown for all measurements. See also Figures S1 and S2 and Table S1.

The phage λ Gam prevents RecBCD degradation of viral DNA

Next, we investigated the function of the third gene of the red operon, gam, during Cas9 targeting. We compared plaque formation by λvir and λvir Δgam on lawns of cells expressing Cas9 and the spc9 (Figure 2A), spc40 or spc45 crRNA guides (Figure S2A). Compared to infection with λvir, targeting of the Δgam deletion mutant with spc9 or spc40 (but not with spc45) reduced plaque formation approximately three orders of magnitude. This reduction depended on the exonuclease activity of the RecBCD complex, as it was lost in E. coli ΔrecB cells expressing RecB^D1080A, a nuclease-deficient version of RecB (Anderson et al., 1999), during spc9 targeting of λvir Δgam (Figures 2A and S3A). Similar to the Δbet and the Δexo Δbet mutants, absence of Gam resulted in a lower DNA content for λvir (Figures S2B and S2C). We therefore repeated the experiment with the viruses containing one chi site, λvir chi¹ and λvir chi¹ Δgam, whose DNA accumulation during infection are equivalent (Figure 1A). Quantification of plaque formation showed that the absence of Gam in λvir chi¹ resulted in a reduction of two orders of magnitude for both spc9 and spc40 targeting (Figures 1B–C and S2D), which was rescued when the inhibitor was expressed from a plasmid in the E. coli host (Figures 1H and S2E).

Figure 2. The RecBCD pathway degrades phage lambda DNA without increasing escaper mutations.

(A) Efficiency of plaquing of λvir or λvir Δgam phages on lawns of E. coli expressing Cas9 programmed with spc9, in the presence or absence of RecB nuclease activity. Mean ± SEM values of three independent experiments are shown. (B-C) Normalized NGS reads of λvir DNA, obtained 25 minutes after infection of E. coli expressing Cas9 programmed with spc9 with either λvir chi¹ (B) or λvir chi¹ Δgam (C) phages at an MOI of 5. (D-E) Efficiency of plaquing of different λvir phages, with or without the red genes and harboring different numbers of chi sites, on lawns of E. coli expressing Cas9 programmed with spc9 (D) or spc40 (E). Mean ± SEM values of three independent experiments are shown. (F) Efficiency of plaquing of λvir chi¹ or λvir chi¹ Δred phages on lawns of E. coli expressing Cas9 programmed with spc9, in the presence or absence of RecD. Mean ± SEM values of three independent experiments are shown. (G) Fraction of λvir chi¹ from a total of 12 plaques obtained after co-infection with a 1:1 mixture of λvir chi¹ and λvir chi¹ Δred phages of E.coli hosts harboring pCas9 programmed with a non-targeting spacer (spcNT), spc9 or spc45, in top agar at a MOI of 20. Mean ± SD values of three independent experiments are shown. See also Figure S3.

To investigate the effect of RecBCD activity on the phage genome targeted by Cas9, we performed next generation sequencing (NGS) of cells harboring either spc9 or spcNT infected for 25 minutes with either λvir chi¹ or λvir chi¹ Δgam. When Gam was expressed, inactivation of RecBCD led to a dip in the reads spanning the spc9 target site, indicating the occurrence of only minimal degradation of phage DNA, most likely due to Exo-Beta resection and repair of the DSB and/or the action of other cellular nucleases at this location (Figure 2B). In contrast, the presence of active RecBCD in the host reduced the coverage to ~ 50% of the reads detected in the absence of Cas9 cleavage, across the whole λ genome, most likely due to destruction of the phage DNA by the RecBCD nuclease (Figure 2C). Interestingly, RecBCD activity did not seem to be impaired by the presence of the added chi site, and therefore we suspect that DNA degradation affected the overall replication of the phage as well. Altogether, these data show that the Gam protein protects the dsDNA ends generated by Cas9 cleavage from degradation by RecBCD, enabling Exo-Beta recombination to repair the DSB introduced during type II CRISPR-Cas targeting.

Repair of the λ phage genome by host RecABCD recombination generates a limited number of escapers

Lambdoid phages that do not carry RecBCD inhibitors contain multiple chi sites to prevent RecBCD degradation and promote RecA-mediated recombination (Bobay et al., 2013). To investigate how this pathway compares to Exo-Beta in the generation of Cas9 escaper phages, we performed infections with λvir chi¹ Δred, which harbors a deletion of all three components of the Red system as well as a chi site that stimulates RecABCD recombination. First, we confirmed that the Δred deletion in λvir chi¹ does not significantly impact the replication of the phage (Figure 1A). We then measured escaper frequencies for spc9-, spc40- and spc45-mediated Cas9 targeting and found that in the absence of the Red system the number of escapers was reduced by more than three orders of magnitude for spc9 and spc40 targeting but not significantly for spc45 (Figures 1B–D and S2D). These results suggest that recombination through the RecABCD pathway is less efficient than Exo-Beta to generate escape mutations. We hypothesized that the chances of recombination and mutagenesis could be limited by the presence of only a single chi sequence, and therefore we introduced six additional chi sites into the Δvir Δred genome (λvir chi^2-7 Δred, Figure S1A). The addition of the extra chi motifs did not severely impact viral DNA accumulation during infection with Δvir Δred phages (Figure S3B). Critically, the number of Cas9 escaper plaques formed by λvir chi^2-7 Δred was not significantly different than the number generated by the λvir chi¹ Δred phage (Figures 2D, 2E and S3C). We also investigated escaper formation during infection of E. coli ΔrecD hosts. In the absence of RecD, the RecBC complex lacks nuclease activity and instead unwinds dsDNA ends, constitutively loading RecA and promoting recombination independently of chi recognition (a scenario somewhat equivalent to the presence of multiple chi sites flanking the DSB) (Anderson et al., 1997; Churchill et al., 1999). Using ΔrecD cells, the number of spc9 escaper plaques formed after infection with λvir chi¹ Δred was more than two orders of magnitude higher than in the presence of RecD, but still one order of magnitude lower than in the presence of the Red system (Figures 2F and S3D). Similar results were obtained using λvir chi^2-7 phage (Figure S3E). Therefore, although RecABCD repair is not possible in a wild-type λ infection (Gam expression and the absence of chi sites prevent this), even in engineered conditions that maximize it, the number of escapers are significantly lower than those obtained through the repair of Cas9-generated DSBs by Exo-Beta. These results suggest that the phage-encoded repair pathway is more effective than the host pathway at generating target mutations in the phage DNA, and more potently reduces the efficacy of type II CRISPR-Cas targeting. To test this further, we performed a competition experiment between λvir chi¹ and λvir chi¹ Δred phages on top agar (Figure 2G). Cas9-expressing E. coli cells carrying either spcNT, spc9 or spc45 targeting spacers were infected with a 1:1 mix of the two phages at a high multiplicity of infection (MOI) of 20. We found that, after checking for the presence of the red operon via PCR, all of the 36 plaques recovered after targeting by either of the tested spacers contained λvir chi¹ phages, demonstrating a strong selection for viruses carrying the Red system. Altogether, our results reveal a two-pronged, post-cleavage strategy carried out by the Red system to overcome type II-A CRISPR-Cas targeting against phage λ, which confers a strong selective advantage to the bacteriophages carrying it: (i) Gam repression of RecBCD phage DNA degradation and recombination, and (ii) introduction of target mutations via Exo-Beta repair of Cas9-cleaved viral DNA.

The error-prone DNA polymerase Pol IV is important to generate Cas9 escape mutations

We next sought to investigate whether the nature of mutations differed following RecBCD or Exo-Beta repair of Cas9 DSBs. To test this, we amplified and sequenced the targets of 12 λvir chi¹ or λvir chi¹Δred phage plaques that were able to evade spc9-mediated immunity (Table S1A). We found that in the presence of Exo-Beta repair, 7/12 λvir chi¹ escapers displayed an adenine to cytosine change in position −5 (five nucleotides before the first nucleotide of the PAM) and the other 5/12 contained guanine to adenine mutations in the PAM. In contrast, in the absence of Exo-Beta repair, 10/12 λvir chi¹Δred escapers contained the −5A>C mutation, with only 1/12 harboring a change in the second guanine of the PAM to a cytosine (as opposed to adenine in the case of λvir chi¹ escapers). Similarly contrasting results were obtained for spc40 escaper phages (Table S1B). In the case of spc45 escaper mutations, a trend showing seed mutations in 4/12 wild-type phage escapers (not present in the λvir chi¹Δred escapers in the plaques obtained in the experiment shown in Figure 1D, Table S1C) was extended to 15/36 in the competition experiment of Figure 2G (Table S1D). To investigate the mutation spectrum of the spc9 escapers in more detail, we subjected the target PCR products of hundreds of pooled escaper plaques to NGS, in triplicate. The values obtained via these NGS experiments are reported in Table S2. Corroborating our analysis of individual plaques, most of the escape mutations comprised the seed sequence −5A>C substitution for both phages (Table S2), and changes in the PAM region were markedly different in the absence of Red: 2G>A and 3G>A mutations were much less frequent and the 3G>T mutation was more abundant (Figures 3A and Table S2). Complementation with a plasmid expressing the red system showed that both the decrease in plaque formation (Figure S4A) as well as the mutation pattern of the λvir chi¹Δred phages were reverted to wild-type values (Figures 3B, 3C and Table S2). These results illustrate the efficient rescue of the Δred deletion and demonstrate that not only the number of mutations, but their specific nature, are dictated by Exo-Beta repair after Cas9 cleavage.

Figure 3. Lambda phage that escape Cas9 cleavage introduce specific target mutations that are increased in the presence of Pol IV.

(A-C) Normalized mutant NGS reads of the spc9 targets of λvir chi¹ or λvir chi¹ Δred phages that escape type II-A targeting on lawns of E. coli expressing Cas9 programmed with spc9 and no additional plasmid (A), the pKM208(empty) vector (B) or the same vector expressing the λ Red system (C). Mean ± SEM values of three independent experiments are shown. (D-E) Efficiency of plaquing of λvir chi¹ (D) or λvir chi¹ Δred (E) phages on lawns of E. coli expressing Cas9 programmed with spc9, in the presence or absence of genes encoding different E. coli error-prone DNA polymerases. Mean ± SEM values of three independent experiments are shown. (F) Same as (A) but in the presence or absence of dinB. Mean ± SEM values of three independent experiments are shown. See also Figures S4 and Tables S1 and S2.

To investigate the origin of the mutations, we evaluated the involvement of the error-prone polymerases Pol II, IV and V (Henrikus et al., 2018), since they participate in the synthesis of the DNA required to fill in the gaps present in the recombination products (Wright et al., 2018). We compared the number of plaques obtained after infection with λvir chi¹ of E. coli hosts harboring spc9 but with different mutant backgrounds: ΔpolB (Pol II), ΔdinB (Pol IV) and ΔumuC (Pol V). While the elimination of Pol II and V activity did not impact the generation of spc9 escapers, absence of Pol IV reduced the number of plaques formed by Exo-Beta repair of the λvir chi¹ phage by approximately two orders of magnitude (Figures 3D and S4B). Although the quantification of λvir chi¹Δred phages that escape spc9 targeting was close to our limit of detection, the number of plaques obtained after infection of ΔdinB hosts was much closer to the numbers obtained in wild-type, ΔpolB and ΔumuC cells (Figures 3E and S4B). The distribution of the 2G>A, 3G>A and 3G>T mutations, however, did not change in the presence or absence of Pol IV (Figure 3F and Table S2). In addition, we found that overexpression of dinB from a plasmid (pDinB) did not increase the number of spc9 λvir chi¹Δred escapers (Figure S4C). Altogether, these results indicate that while Pol IV is the most important error-prone DNA polymerase for the increase of phage escapers during Exo-Beta repair of Cas9-cleaved λ DNA, the activity of this polymerase is not sufficient and needs to act together with the Red system to introduce specific escape mutations.

Inefficient Cas9 cleavage increases the generation of escape mutations

Our model for the generation of escaper mutations requires cleavage of a target sequence before it can be repaired by Exo-Beta recombination. To test this, we performed experiments using dCas9, a Cas9 mutant that does not cleave its target DNA but can bind it and interrupt transcription (Bikard et al., 2013; Qi et al., 2013). When programmed with spc40, dCas9 reduced λvir chi¹ plaque formation by more than one order of magnitude (presumably by preventing phage gene expression), similar to the level of immunity it provided in hosts expressing wild-type Cas9. In contrast, we did observe a further decrease in pfu during infection with the λvir chi¹Δred phage (Figure 4A). Importantly, the target sequences of the dCas9 escapers, both in the presence and absence of the Red system during infection, resembled those of the Cas9 escapers in the absence of Red (Table S1B and Table S1E). Collectively, these results demonstrate that Cas9 cleavage is necessary for the generation of escaper phages through the Exo-Beta recombination pathway.

Figure 4. Inefficient Cas9 cleavage is required for the generation of high numbers of escaper lambda phage.

(A) Efficiency of plaquing of λvir chi¹ or λvir chi¹ Δred phages on lawns of E. coli expressing Cas9 or dCas9 programmed with spc40. Mean ± SEM values of three independent experiments are shown. (B) Mutation in the seed sequence of spc45 to generate spc45c. (C) Normalized NGS reads of λvir DNA, obtained 25 minutes after infection of E. coli expressing Cas9 programmed with spcNT, spc45 or spc45c with λvir chi¹ phage at an MOI of 5. (D) Efficiency of plaquing of λvir chi¹ or λvir chi¹ Δred phages on lawns of E. coli expressing Cas9 programmed with spc45 or spc45c. Mean ± SEM values of three independent experiments are shown. (E) Fraction of plaque forming units (PFUs) harboring spc45 target mutations from a total of 8 plaques analyzed per infection at each time point after infection of E. coli expressing Cas9 programmed with spc45 or spc45c with λvir chi¹ or λvir chi¹ Δred phages at an MOI of 1. See also Figure S4 and Table S1.

Repair of a DSB requires recombination with an intact copy of the viral genome. Therefore, under conditions where cleavage is efficient and most uncut viral genomes are eliminated from the host cell, escaper generation through Red-based recombination should be inhibited. We wondered whether this was the case during spc45-mediated immunity, in which escaper generation was severely restricted for both λvir chi¹ and λvir chi¹Δred phages (Figures 1D, 4D, S2A and S2D). We attempted to decrease the cleavage efficiency of Cas9 by mutating the spc45 sequence, introducing a mismatch between the crRNA produced by this spacer (spc45c) and the seed sequence (Jiang et al., 2013; Jinek et al., 2012) of its target in the λ genome (Figure 4B). NGS of DNA extracted from cells 25 minutes after infection with λvir chi¹ showed that indeed, spc45c targeting resulted in an intermediate number of reads across the genome, higher than those for spc45 targeting, but lower than those for non-targeting, conditions (Figure 4C). These differences in phage DNA accumulation correlated with the immunity provided by each spacer (Figure 4D). More important, when we infected cells harboring spc45c with λvir chi¹Δred phage, escapers were reduced by almost three orders of magnitude compared to wild-type λvir chi¹ (Figure 4D). These results corroborate a requirement for a high number of uncleaved phage genomes for the generation of Cas9 escapers through Exo-Beta repair.

We also investigated the effects of the different Cas9 cleavage efficiencies in the dynamics of escaper generation during infection. To do this, we performed infections in liquid media since, as opposed to the genotyping of plaques on bacterial lawns in which non-mutant phages are competed out over the 24-hour plate incubation period, these can reveal short-term variations in the appearance of target mutations. We infected liquid cultures expressing Cas9 programmed with the strong spc45 or the weak spc45c crRNAs, with either λvir chi¹ or λvir chi¹Δred phage. We took culture aliquots at 50-minute intervals after infection (the average duration of a lytic cycle for λvir chi¹, Figure S4D) and plated them onto a non-CRISPR strain to obtain phage plaques (Figure S4E).

We then analyzed the target genotype of 8 plaques per time point (Table S1F–I). After infection with both λvir chi¹ and λvir chi¹Δred phages, spc45 escapers were rapidly detected, with all of the eight sequenced plaques formed by mutant phage by 200 minutes (Figure 4E, Table S1F and Table S1G). In contrast, λvir chi¹ phages harboring wild-type spc45c targets continued to be detected over the nine hours of the experiment (Figure 4E and Table S1H). Importantly, in the absence of λ Red recombination during the weak spc45c-mediated immunity, the rise of escapers was similar to that observed in spc45 cultures (Figure 4E and Table S1I). These results demonstrate that the generation of escaper phages through Exo-Beta recombination during poor immunity is a relatively slow process, probably preceded by multiple rounds of cleavage and repair that regenerate wild-type phages with intact targets that can continue propagating. On the other hand, our observations suggest that the rapid accumulation of escaper mutations in conditions where Exo-Beta repair is limited due to efficient target cleavage is most likely driven by the presence of pre-existing target mutations within the viral population.

The phage λ Red system facilitates the generation of different types of Cas9 escape mutations

Next, we investigated if the Red system was also involved in the generation of the other types of escaper mutations (Figure S1). For spc15 targeting, absence of exo and bet, gam or the full deletion of the red operon reduced escaper formation by approximately one to two orders of magnitude (Figures 5A and S5A). Importantly, we detected qualitative differences in the target sequences of the escapers (Table S1J). Similar results were obtained for spc26D (Figures 5B and S5B; Table S1K). This spacer was also able to provide immunity in hosts expressing dCas9, however the number of escapers of dCas9 targeting was not reduced in the absence of the Red system (Figures 5B and S5B). This result corroborates a requirement for Cas9 target cleavage for the generation of escapers via Exo-Beta recombination of phage and chromosomal DNA. For spc14 targeting, deletion of exo and bet, gam or red decreased the number of plaques obtained by approximately two orders of magnitude (three orders of magnitude in the case of λvir chi¹Δred) (Figures 5C and S5A). Importantly, when we analyzed the spc14 target region of six escapers (Table S1L), we found that while all λvir chi¹ mutants contained the same microhomology-mediated target deletion, absence of Red recombination led to the accumulation of target point mutations in 5/6 of the λvir chi¹Δred escapers. The chromosomal recombination and microhomology-mediated deletions in the spc15 and spc14 targets respectively do not require the introduction of de novo point mutations in the target during repair. Therefore, it is expected that Pol IV, involved in the generation of protospacer mutations during spc9 targeting, should not affect the number of spc15 and spc14 escapers. To test this prediction, we performed infections in ΔdinB mutant hosts and found no difference in the number of spc15 escapers (Figures 5D and S5C) and a mild reduction of spc14 escapers (Figures 5E and S5C). In addition, sequencing of the escaper targets showed minimal differences in the type of mutations (Table S1M and Table S1N). Altogether these results demonstrate that Exo-Beta can also facilitate the repair of Cas9 DSBs through recombination with the host’s chromosomal DNA and through MMEJ to generate escaper phages.

Figure 5. The lambda Red system mediates different types of escape.

(A) Efficiency of plaquing of different λvir chi¹ phages on lawns of E. coli expressing Cas9 programmed with spc15. (B) Efficiency of plaquing of λvir chi¹ or λvir chi¹ Δred phages on lawns of E. coli expressing Cas9 or dCas9 programmed with spc26D. (C) Same as (A) but with spc14 targeting. (D-E) Efficiency of plaquing of λvir chi¹ phage on lawns of E. coli expressing Cas9 programmed with spc15 (D) or spc14 (E), in the presence or absence of E. coli Pol IV DNA polymerase (dinB). Mean ± SEM values of three independent experiments are shown for all measurements. See also Figure S5 and Table S1.

The phage λ Red system enables escape from type I CRISPR-Cas targeting

To date, there are no natural isolates of E. coli that harbor a type II CRISPR-Cas system. Instead, this organism carries type I CRISPR loci (Touchon et al., 2011). In particular, the K-12 strain MG1655 harbors a type I-E CRISPR-Cas system that has been thoroughly characterized (Brouns et al., 2008; Semenova et al., 2011). This system is repressed at the transcription level under laboratory growth conditions (Westra et al., 2010), therefore we used the engineered strain ACT-01 in which the type I-E cas genes encoding Cascade and Cas3 are under the control of an arabinose-inducible promoter on the E. coli chromosome (Caliando and Voigt, 2015). spc9R, which targets phage λ in the same region specified by spc9 (Figure S1A), was introduced under the control of the same promoter on a plasmid to provide crRNAs for the Cascade complex. Compared to the non-targeting control, type I-E targeting reduced the number of plaques by almost two orders of magnitude for the λvir chi¹ phage (Figures 6A and S6A). Absence of exo and bet, gam or red further reduced that number by approximately three to four orders of magnitude relative to the wild-type phage. Importantly, we found that while 7/8 spc9R λvir chi¹ targets contained a G>A substitution in position −4, only 1/8 λvir chi¹Δred escapers harbored that mutation (Table S1O).

Figure 6. The lambda Red system promotes escape from type I-E CRISPR-Cas targeting.

(A) Efficiency of plaquing of different λvir chi¹ phages on lawns of E. coli ACT-01 expressing the Cascade-Cas3 complex programmed with spc9R. (B-E) Efficiency of plaquing of λvir chi¹ or λvir chi¹ Δred phages on lawns of E. coli KD263 harboring spcE4-R (B), spcL1-R (C), spcL4-R (D), and spcL6-R (E). (F) Efficiency of plaquing of λvir chi¹ phage on lawns of E. coli KD263 harboring spcL6-R, in the presence or absence of genes encoding different E. coli error-prone DNA polymerases. Mean ± SEM values of three independent experiments are shown for all measurements. See also Figure S6 and Table S1.

We also explored the effect of Red in the escape from the targeting provided by four “natural” spacers; i.e., that were acquired by the type I-E system of E. coli from large genomic fragments of λ DNA in a previous study, using strain KD263 (Strotskaya et al., 2017). Of these, only spcE4-R provided a mild defense comparable to spc9R (EOP ~ 10⁻², Figures 6B, S6B and S6C); the other three (spcL1-R, spcL4-R and spcL6-R) mediated a substantial decrease of λvir chi¹ EOP (Figures 6C–E, S6B and S6C). When cells carrying any of these spacers were infected with λvir chi¹Δred, the EOP values dropped at least two orders of magnitude compared to the values obtained for phages expressing the Red system (Figures 6B–E, S6B and S6C). As reported for spc9R, analysis of the target sequences of spcL4-R and spcL6-R escapers showed distinct mutations in the presence or absence of Red (Table S1P and Table S1Q, respectively). This did not seem to be the case for spcL1-R and spcE1-R escapers, most of which contained small and complete target deletions, respectively (Table S1R and data not shown). Finally, we investigated the role of the error-prone polymerases in the generation of escapers during spcL6-R-mediated Cascade-Cas3 targeting (Figures 6F and S6D). However, none of the DNA polymerase deletions affected the EOP values of λvir chi¹. These observations demonstrate that the Red system also enables an increase in type I-E CRISPR escapers with a distinct mutational signature. However, as opposed to the evasion of type II-A targeting, Red facilitates the introduction of escape mutations when the immunity provided by the spacer is both weak and strong, and also in the absence of Pol IV.

DISCUSSION

Here we investigated how bacteriophage λ escapes DNA cleavage by type I and II CRISPR-Cas nucleases. We showed that the λ red operon enables the accumulation of escaper phage, and thus increases the propagation of the phage by several orders of magnitude. We propose a model in which this phenomenon is a result of the function of the λ Red system in the repair of viral DNA cleaved by RNA-guided Cas nucleases: Gam prevents both the rapid degradation of phage DNA by RecBCD and repair through E. coli RecA-mediated recombination at chi-like sequences, leaving the DSBs generated by CRISPR cleavage available for repair by Exo-Beta recombination (Figures S6E and S6F). Exo-Beta repair then leads to the accumulation of a greater number of escaper phages harboring distinct target mutations. We found that Exo-Beta also mediates other forms of repair (Figures S1E and S1F) in which recombination occurs with a homologous sequence in the host chromosome, or between short homologous sequences flanking the target site in the viral genome. However, we believe that these examples represent rare cases, since they require either the presence of phage-related chromosomal sequences or the generation of non-deleterious deletions (less common within compact phage genomes (Brüssow and Hendrix, 2002)), respectively.

Most repair scenarios will require an intact copy of the phage target DNA for recombination to re-generate the original sequence. Consistent with this, we found that the Red system is particularly efficient at increasing the number of escapers in conditions of weaker defense, when a considerable fraction of wild-type phage will remain uncleaved. Given that only specific seed and PAM target mutations prevent effective target recognition and/or cleavage, multiple rounds of cleavage and repair are probably necessary before de novo escape mutations become prevalent in the viral population. Indeed, during the first 9 hours of infection of cells expressing Cas9 and the weak-targeting spc45c crRNA (and presumably spc9 and spc40 crRNAs as well) most phages contained intact target sequences (Figure 4E), suggesting that the Exo-Beta system was able to efficiently and accurately repair the DSBs generated during type II-A targeting, allowing the phage to replicate and continue its lytic cycle; phages carrying target mutations only become predominant within the population relatively late during infection. Although the type II-A spacers evaluated in this study were not naturally acquired during the CRISPR response to λ infection, a previous study has shown that weak spacers that cannot provide high fitness to the host are not only acquired but also maintained in the bacterial population long after infection (Heler et al., 2019). While these findings were obtained investigating the type II-A CRISPR-Cas immune response to the staphylococcal phage ΦNM4γ4 (a λ-like phage also belonging to the Siphoviridae family), we presume that weak spacers will also be abundant in hosts harboring this CRISPR system following infection with other phages. In the specific case of phage λ, of the 4 spacers acquired by the native E. coli type I-E system (Strotskaya et al., 2017) that we tested in this study, we found one that provided weak defense. We believe that the relative abundance of spacers that mediate poor targeting could be important to increase the targeting diversity, which helps prevent the rise of escaper phages (van Houte et al., 2016). In addition, phage-encoded anti-CRISPRs can diminish the effectiveness of strong spacers (Borges et al., 2018; Landsberger et al., 2018). For all of these reasons, we believe that the evasion of weak targeting bestowed by Red-like systems could have considerable effects for the phage-CRISPR arms race.

When the CRISPR defense is strong and presumably most copies of the infecting DNA are rapidly cleaved, we obtained different results for Cas9 and Cascade-Cas3 targeting. In the experiment where Cas9 was programmed with the strong spacer spc45, the number of escaper plaques was very low, both in the presence and absence of Red. We believe that rapid cleavage of a large fraction of the phage DNA leaves very little substrates for Exo-Beta recombination and repair. We conclude that in this experiment the phages that evade Cas9(spc45) targeting most likely contain pre-existing mutations.

In support of this hypothesis, the frequency of escapers for spc45 was found to be approximately 10⁻⁶ (Figures 1D and 4D; with the exception of an outlier datapoint in Figure 1D), which is close to the value expected from the mutation frequency previously determined for phage λ, 7.7 × 10⁻⁸ mutations per base pair (Drake, 1991), considering that about 10 base pairs of the target sequence (seed or PAM) can be mutated to avoid Cas9 cleavage. In contrast, in competition experiments in which wild-type and Δred phages were used to co-infect the Cas9(spc45) host, all of the 36 plaques tested contained λ harboring red, and displayed a unique escape mutation pattern. This suggests that, even during strong targeting conditions, some uncleaved template is still available for Exo-Beta recombination and repair, especially during infections at high MOI, which can lead to the introduction of mutations and their spread during a more sensitive assay such as competition. On the other hand, escape from targeting mediated by Cascade-Cas3 programmed with spacers that provide strong defense was significantly increased in the presence of Red. We attribute these different results to the contrasting DNA degradation activities of each of these nucleases, and hypothesize that Exo-Beta can more efficiently recombine the products of Cas3 ssDNA cleavage and processive degradation than the blunt dsDNA ends generated by Cas9.

Regardless of the type of escaper we observed (Figure S1E), our data showed a substantial increase in the number of escapers where Exo-Beta but not RecABCD recombination is functional (red⁺ phages) when compared to infection in conditions where RecABCD but not Exo-Beta repair is active (chi¹ Δred phages). Given that Cas9 remains bound to the cut DNA ends after cleavage (Garneau et al., 2010; Sternberg et al., 2014), one simple explanation for our results could be that while Exo effectively displaces the Cas9 nuclease to initiate recombination, RecBCD cannot do this efficiently, reducing the number of genomes that it can repair. However, during infection with Δgam phages, when both repair pathways are functional, the number of escapers is similar to those obtained for the Δred phage (Figures 1B, 1C and 6A). This result suggests that RecBCD can capture and degrade the free DNA ends more efficiently than Exo. In addition, in the absence of Gam but with expression of Exo-Beta, the host RecBCD nuclease was able to extensively degrade the viral DNA (Figure 2C), a result that also argues against the possibility that Cas9, and even Exo, blocks RecBCD access to free DNA ends. Another factor that could affect the differences in escaper generation between the two repair systems is the different sequence requirements for recombination. While Exo-Beta activity does not depend on specific sequences, RecBCD only initiates recombination upon encountering a chi site, and even then there is only a 40% chance that the complex will successfully switch modes from degradation to repair (Taylor and Smith, 1992). Therefore, even in the presence of multiple chi sites, the efficiency of recombination of RecBCD, and hence the generation of escaper mutations, is lower than the “constitutive” Exo-Beta. However, even in the presence of “constitutive” RecABC recombination (in the host with the recD deletion), escaper generation is improved but still lags behind Exo-Beta (Figures 2F, S3D and S3E), a result that suggests that the DNA repair mediated by Exo-Beta recombination provides a specialized route for phage escape.

Exo-Beta repair preferentially enables at least three distinct types of escapers. For spc15 and spc26D escape mutants, the target sequence is modified through recombination with homologous sequences in the E. coli chromosome. Previous studies have demonstrated that Exo-Beta is 100-fold more efficient than the RecABCD pathway in recombining homologous sequences that diverge by 22% (Martinsohn et al., 2008), a finding that could explain the higher number of this type of recombinant escapers in the presence of the phage repair system. In this case, target mutations are introduced during infection through recombination with an already mutated template DNA. In contrast, the other two types of escaper mutations are generated de novo. In the case of the target point mutations (spc9, spc40, spc45c, spc9R, spcL4-R and spcL6-R targets), we believe that processing of DSBs exposed after cleavage of the invading DNA by CRISPR nucleases would induce the SOS system (Mo et al., 2021) and lead to the expression of error-prone DNA polymerases (Maslowska et al., 2019). This seems to be particularly important in the case of Cas9 DSB repair, where we found that Pol IV is required for the generation of 99% of the phage escapers. However, this polymerase did not affect the mutation pattern, suggesting that nucleotide changes are related to Exo-Beta activity, with Pol IV either facilitating or tolerating the incorporation of these changes into the phage genomes. We also found that Exo-Beta can participate in the generation of target deletions via MMEJ during spc14 targeting. The exact molecular mechanisms behind this mode of DNA repair are not fully understood in E. coli. However, since Exo-Beta promotes recombination through the annealing of complementary ssDNA, it is possible that this system could facilitate the annealing of short homologous sequences exposed as ssDNA after Exo resection at either end of a DSB, resulting in a deletion of the sequence in between the repeats (Kuzminov, 1999; Roy et al., 2020).

A previous study proposed that phage concatemers may be useful for the generation of Cas9 escape mutations in T4 phages (Tao et al., 2018), by facilitating rapid recombination of a cleaved DNA molecule with an intact template. This would increase the number of cleavage-repair rounds and therefore the probability of target mutation. In the absence of phage recombinases, there would be less phage DNA accumulation, fewer possibilities to pair cleaved DNA with an intact repair template and thus a lower target mutation frequency. Although the Red system has been implicated in the generation of λ phage concatemers that facilitate replication and packaging (Enquist and Skalka, 1973), several lines of evidence presented in our work suggest that differences in replication are not responsible for our findings. First, the λvir Δexo mutant phage, which displayed even higher levels of DNA accumulation than wild-type λvir without the addition of the extra chi site (Figure S2B), was incapable of escaping spc9 or spc40 targeting with high frequency (Figure S2A). Second, our qPCR analysis showed that upon addition of a chi site in the λvir genome, the wild-type and mutant phages used in this study all have similar DNA accumulation levels (Figure 1A). However, it is not clear how the addition of a single chi motif affects the replication and recombination of these phages. For the Δgam and Δred mutants, it is believed that the added chi sequence allows RecABCD recombination to restore concatemer formation (Henderson and Weil, 1975), which would lead to the observed increase in DNA content (compare Figures 1A and S2C). It is also known that chi motifs interfere with the replication and recombination of lambdoid phages that harbor Red-like recombination systems (Blower et al., 2011; Henderson and Weil, 1975), an observation that explains the reduction in viral DNA levels for the λvir chi¹ and λvir chi^2-7 phages (Figure S3B). Nevertheless, due to limitations in our knowledge of λ replication and recombination, we cannot completely exclude a role for concatemer formation by Exo-Beta in the generation of CRISPR-escape mutations.

While λ uses the Red recombination system for efficient replication (Enquist and Skalka, 1973), approximately half of the available lambdoid phage genomes do not encode red homologues or other dedicated recombinases and contain chi sites that allow them to rely on the host recombination machinery for the formation of concatemers (Bobay et al., 2013). Therefore, Rocha and colleagues have suggested that the presence of dedicated phage recombination systems depends on factors that are unrelated to viral replication (Bobay et al., 2013). For example, on one hand phage recombinases like Beta are more tolerant to sequence divergence than RecA (Martinsohn et al., 2008), allowing more genetic exchange and mosaicism; on the other, recombination systems take valuable space in the viral genome that could be occupied by more important accessory genes. We believe that our findings reveal an additional role for Red-like recombination systems that favors their preservation within phage genomes: to evade CRISPR immunity and other sequence-specific, DNA-cleaving prokaryotic defense mechanisms. Similarly to other lambdoid phages that lack the Red system, the introduction of a chi site into the λ genome restores efficient viral propagation of Δred phages (Henderson and Weil, 1975). In contrast, neither this site, nor the addition of multiple chi sequences or the constitutive recombination activity of RecBC, restores high efficiency of CRISPR escape for phages lacking the Red system. Given that CRISPR-Cas and restriction-modification systems are both prevalent and involve sequence-specific DNA cleavage, we propose that an important driver of the evolution and spread of viral recombinases, which are widely distributed across a large number of both temperate and lytic phage genomes (Lopes et al., 2010), is their ability to repair DNA breaks in a manner that counteracts these defense mechanisms, i.e. through the introduction of escape mutations in the target sequence. Two recent reports support this idea. One showed that a recombination system related to λ Red present in the IncC conjugative plasmid of Vibrio cholerae enables escape from type I CRISPR-Cas cleavage through the introduction of target deletions (Roy et al., 2020). The other report found that the phage T4-encoded UvsX recombinase promotes escape from Cas9 and Cas12a cleavage, also through the generation of deletions (Wu et al., 2021). Importantly, Red systems do not lead to a general surge of mutagenesis, which could be detrimental for both the phage and the host, but rather to a specific increase of mutations at the target site; i.e., only where they matter. Phages can escape CRISPR immunity by expressing anti-CRISPR proteins (Acrs) that, in most instances, have evolved to specifically interact with Cas nucleases to prevent target cleavage (Stanley and Maxwell, 2018). In contrast to Acrs, λ Red, and possibly other phage-encoded Red-like recombination systems, provide a unique and versatile mechanism that counteracts the activity of CRISPR-Cas (and probably other nucleases involved in anti-phage defense) acting after, instead of before, DNA cleavage, to facilitate genetic escape.

STAR Methods

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Luciano A. Marraffini (marraffini@rockefeller.edu).

Materials availability

All materials generated for this study are available upon request from the Lead Contact, Luciano A. Marraffini, with a completed Materials Transfer Agreement.

Data and code availability

• Data and code are publicly available. Raw Illumina FASTQ data for in vivo Cas9 cleavage assays and spc9 escape mutation sequencing available on SRA: BioProject ID PRJNA754816

• Code for all analyses performed publicly available at Github: https://github.com/Marraffini-Lab/Hossain_etal_2021

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and subject details

Bacterial strains and growth conditions

Cultivation of E. coli K-12 MG1655 (Guyer et al., 1981) and all other related strains used in this study were carried out in lysogeny broth (LB) media (BD Difco LB Broth, Lennox, BD 240220 for LB liquid media; BD Difco LB Agar, Miller, BD 244510 for LB agar plates) at 37°C (30°C in certain cases) and, for liquid cultures, with shaking. Overnight cultures were inoculated from single bacterial colonies. Wherever applicable, media were supplemented with chloramphenicol (GoldBio, C-105-100) at 25 μg/ml, spectinomycin (GoldBio S-140-50) at 50 μg/ml, kanamycin (GoldBio, K-120-100) at 50 μg/ml and/or carbenicillin (GoldBio, C-103-25) at 100 μg/mL to ensure plasmid maintenance. E. coli K-12 MG1655 Keio knockout strains were obtained from the Coli Genetic Stock Center at Yale University (Baba et al., 2006). The type I-E CRISPR interference strain E. coli K-12 MG1655 ACT-01 was a generous gift from Chris A. Voigt at MIT (Caliando and Voigt, 2015). The type I-E CRISPR-Cas strain E. coli KD263 and KD263 λ-targeting strains (λ_E4-R, λ_L1-R, λ_L4-R, λ_L6-R) were generous gifts from Konstantin Severinov and Ekaterina Semenova at Rutgers University (Strotskaya et al., 2017). The identity of the spacer within the KD263 CRISPR array was confirmed by amplification with primers JW3100 and JW3101 (see Table S3) followed by Sanger sequencing. E. coli K-12 MG1655 Δ9 (a strain with the 9 E. coli prophages deleted from the MG1655 genome) was a generous gift from Thomas K. Wood at the Pennsylvania State University (Wang et al., 2010). The E. coli K-12 MG1655∷λ lysogen was obtained from J. W. Roberts. Strains and plasmids used in this study are listed in Table S3.

Method details

Plasmid construction

Cloning was performed with rubidium chloride chemically competent E. coli K-12 MG1655 cells. Briefly, 20-200 ng of plasmid DNA (prepared by QIAprep Spin Miniprep Kit, QIAGEN, 27104) was mixed with 50 μL of competent cells and incubated on ice for 20-30 minutes. Cells were heat shocked in a 42°C water bath for 45 seconds and then placed back on ice for 3 minutes. Cells were then resuspended in 300 μL room temperature LB medium and incubated at 30°C or 37°C (depending on plasmid) for 1-2 hours. After the outgrowth period, the entire volume of transformed cells was spread on an LB agar plate with the appropriate antibiotic(s) for plasmid selection and incubated at 30°C or 37°C overnight. The next day, individual colonies were picked and grown in 3 mL LB medium with appropriate antibiotic(s) for plasmid maintenance. To store plasmid strains, 900 μL of overnight culture was mixed with 100 μL dimethyl sulfoxide (Sigma-Aldrich, D2650-100ML) and frozen at −80°C. These frozen stocks were streaked on LB agar plates with appropriate antibiotic(s) to single colonies for use in experiments. For E. coli Keio knockout strains, the ACT-01 strain and E. coli strains with two plasmid combinations that were difficult to transform simultaneously, existing strains were first made electrocompetent and then transformed with plasmid. To make cells electrocompetent, 1.4 mL of 3 mL overnight stationary cultures were spun down in 1.5 mL Eppendorf tubes at 10,000 rpm for 1 minute in a tabletop microcentrifuge at 4°C. After discarding the supernatant, cells were washed 2 times with sterile cold water. To wash cells, pelleted cells were resuspended in 1 mL sterile cold water, spun down at 10,000 rpm for 1 minute in a tabletop microcentrifuge at 4°C and supernatant discarded. Cells were resuspended in 150 μL cold 10% glycerol (Fisher Scientific, G33-500) for use in electroporation. For electro-transformation, 50 μL of cells were mixed with 50–150 ng of plasmid DNA (prepared by QIAprep Spin Miniprep Kit, QIAGEN, 27104). Cells were electroporated using a 0.1 cm gap Gene Pulser electroporation cuvette (Bio-Rad, 165-2089) at 1.8 kV in a Bio-Rad Gene Pulser Xcell system. Electroporated cells were immediately resuspended in 300 μL of room temperature LB medium. Cells were recovered at 30°C or 37°C for 1-2 hours before being plated on LB agar (BD Difco LB Agar, Miller, BD 244510) with appropriate antibiotic(s) and incubated at 30°C or 37°C overnight.

The plasmids and oligonucleotides used in this study are listed in Table S3. Plasmid pCas9∷spcNT was generated in a previous study (Jiang et al., 2013). BsaI (NEB, R3535L) cloning (Jiang et al., 2013) (see section on “Oligo cloning”) was used to make pCas9 spacer carrying plasmids by inserting spc9 (annealed primers AA318-AA319), spc14 (annealed primers AA320-AA321), spc15 (annealed primers AA349-AA350), spc26D (annealed primers AA107-AA108), spc40 (annealed primers AA132-AA133), spc45 (annealed primers AA153-AA154), spc45c (annealed primers AA383-AA384), spc12 (annealed primers JW1191-JW1192), as well as annealed primers JW1370-JW1371, JW1546-JW1547, JW1558-JW1559, JW1401-JW1402, and JW1552-JW1553 into the BsaI site of BsaI-digested pCas9∷spcNT. Plasmid pdCas9∷spcNT was generated in a previous study (Bikard et al., 2013). BsaI cloning (Bikard et al., 2013) was used to make pdCas9 spacer carrying plasmids by inserting spc26D (annealed primers AA107-AA108) and spc40 (annealed primers AA132-AA133) into the BsaI site of BsaI-digested pdCas9∷spcNT.

To construct pAM38 (P_ara cat p15A ori), a linear fragment containing the araC gene and P_ara promoter from pBAD18 (Guzman et al., 1995) was PCR amplified using Phusion High-Fidelity DNA polymerase (Thermo Fisher Scientific, F530L) with primers oAM201 and oAM202, a linear fragment containing the cat gene and p15A origin of replication from pPL2e (Lauer et al., 2002) was PCR amplified with primers oAM203 and oAM204, and a two-piece Gibson assembly (Gibson et al., 2009) was performed. To construct pAM38(gam), pAM38 was amplified with primers JW1484 and JW1485, λvir genomic DNA was amplified with primers JW1567 and JW1568, and a two-piece Gibson assembly was performed. To construct pAM38(exo), pAM38(gam) was amplified with primers AA149 and JW1485, λvir genomic DNA was amplified with primers AA147 and AA148, and a two-piece Gibson assembly was performed. To construct pAM38(bet), pAM38(gam) was amplified with primers AA149 and JW1485, λvir genomic DNA was amplified with primers AA145 and AA146, and a two-piece Gibson assembly was performed.

To construct pDS-SPcas9∷spcNT, pDS-SPcas (Addgene plasmid 48645) (Esvelt et al., 2013) was amplified with primers JW488 and JW489, pJM62 (Modell et al., 2017) was amplified with primers JW1187 and JW1188, and a two-piece Gibson assembly was performed. BsaI cloning (Jiang et al., 2013) was used to make pDS-SPcas9∷spc9 by inserting spc9 (annealed primers AA318-AA319) into the BsaI site of BsaI-digested pDS-SPcas9∷spcNT. To construct pDS-SPcas9∷spcNT(CmR), pDS-SPcas9∷spcNT was amplified with primers AA68 and AA69, the cat gene of pCas9∷spcNT was amplified with primers AA66 and AA67, and a two-piece Gibson assembly was performed. BsaI cloning (Jiang et al., 2013) was used to make pDS-SPcas9∷spc9(CmR) by inserting spc9 (annealed primers AA318-AA319) into the BsaI site of BsaI-digested pDS-SPcas9∷spcNT(CmR).

To construct pRecB, pEmpty (pCL1920) (Lerner and Inouye, 1990), was amplified with primers JW1209 and JW1059, E. coli K-12 MG1655 genomic DNA was amplified with primers JW1522 and JW1523, and a two-piece Gibson assembly was performed. To construct pRecB^D1080A, pEmpty (pCL1920) was amplified with primers JW1209 and JW1059, E. coli K-12 MG1655 genomic DNA was amplified with primers JW1522 and JW1524 and also with primers JW1525 and JW1523, and a three-piece Gibson assembly was performed. To construct pDinB, pEmpty (pCL1920) was amplified with primers AA696 and AA697, E. coli K-12 MG1655 genomic DNA was amplified with primers AA694 and AA695, and a two-piece Gibson assembly was performed. To construct pCL1920-P, pEmpty (pCL1920) was amplified with primers JW1209 and JW1059, λvir genomic DNA was amplified with primers JW1480 and JW1481, and a two-piece Gibson assembly was performed. Plasmid pKM208(red) or pKM208 was a generous gift from Kenan C. Murphy at University of Massachusetts, Amherst (Murphy and Campellone, 2003). To construct pKM208(empty), pKM208(red) was amplified with primers AA333 and AA331 and also with primers AA327 and AA332, and a two-piece Gibson assembly was performed.

Plasmid pPD207.846 was a generous gift from Andrew Z. Fire at Stanford University (Fu et al., 2017). To construct pACYC 184-TypeIEspcNT, pCas9∷spcNT was amplified with primers AA396 and AA397, pPD207.846 was amplified with primers AA398 and AA399, and a two-piece Gibson assembly was performed. BsaI (NEB, R3535L) cloning (Jiang et al., 2013) was used to make pACYC 184-TypeIEspc9R by inserting spc9R (annealed primers AA77-AA78) into the BsaI site of BsaI-digested pACYC184-TypeIEspcNT.

To construct pTU18C-dgam, pUT18C (Euromedex) (Karimova et al., 1998) was amplified with primers JW1409 and JW1410, λvir Δgam2 phage genomic DNA was amplified with primers JW1554 and JW1555, and a two-piece Gibson assembly was performed. λvir Δgam2 phage was isolated as an escaper of pCas9:JW1370-1 which harbored a 144 bp deletion spanning the target site containing the λ gam gene. To construct pTU18C-dexo, pUT18C was amplified with primers JW1409 and JW1410, λvir Δexo4 phage genomic DNA was amplified with primers JW1556 and JW1557, and a two-piece Gibson assembly was performed. λvir Δexo4 phage was isolated as an escaper of pCas9:spc12 which harbored a 74 bp deletion spanning the target site containing the λ exo gene. To construct pUT18C-dbet, pUT18C was amplified with primers JW1409 and JW1410, λvir genomic DNA was amplified with primers JW1579 and JW1580 and also with primers JW1581 and JW1582, and a three-piece Gibson assembly was performed. To construct pUT18C-dexodbet, pUT18C was amplified with primers JW1409 and JW1410, λvir genomic DNA was amplified with primers JW1612 and JW1613 and also with primers JW1614 and JW1582, and a three-piece Gibson assembly was performed. To construct pUT18C-dred, pUT18C was amplified with primers JW1409 and JW1410, λvir genomic DNA was amplified with primers JW1612 and JW1613 and also with primers JW1619 and JW1620, and a three-piece Gibson assembly was performed. To construct pUT18C-chiD, pUT18C was amplified with primers JW1409 and JW1410, λvir genomic DNA was amplified with primers JW1548 and JW1549 and also with primers JW1550 and JW1551, and a three-piece Gibson assembly was performed. To construct pUT18C-3chiF+R, pUT18C was amplified with primers JW1409 and JW1410, and λvir genomic DNA was amplified with primers JW1413 and JW1561, JW1562 and JW1377, JW1417 and JW1491, as well as JW1406 and JW1418. The λvir-derived fragments were stitched together by overlap extension PCR and a Gibson assembly was performed.

Gibson assembly

For Gibson assemblies (Gibson et al., 2009), 25-100 ng of the largest dsDNA fragment was combined with equimolar volumes of the smaller fragment(s) in a total volume of 5 μL in nuclease-free water. Reaction mixtures were prepared on ice and mixed with 15 μL of Gibson assembly master mix, pipette mixed and incubated at 50°C for 1 hour in a thermal cycler. Gibson reactions were transformed into rubidium chloride chemically competent E. coli K-12 MG1655 cells by mixing 5 μL Gibson reaction with 50 μL cells and following the transformation protocol for chemically competent cells outlined in the section above.

Oligo cloning

Oligo cloning was used to create a repeat-spacer-repeat CRISPR array with a desired spacer following a protocol previously described by this laboratory (Jiang et al., 2013). Briefly, we used a BsaI restriction digest cloning approach. Parent type II-A CRISPR array-containing plasmids with a repeat-spacer-repeat carried a 30 bp spacer sequence with two BsaI cut sites at either end. To set up the BsaI plasmid digest, we mixed 42 μL of the parent CRISPR plasmid (40-60 ng/μL) with 6 μL BsaI-HF (NEB, R3535L), 6 μL NEB CutSmart buffer and 6 μL nuclease-free water. The restriction digest reaction was incubated at 37°C for approximately 6 hours. Two IDT oligonucleotides comprised the type II-A CRISPR spacer to be inserted into the BsaI cut plasmid CRISPR array: a “top” strand oligo with sequence 5’-AAAC-(30 bp spacer)-G-3’ and a “bottom” strand oligo with sequence 5’-AAAAC-(30 bp spacer reverse complement)-3’. For oligo cloning of the type I-E spacer spc9R into pACYC184-TypeIEspcNT, the top strand oligo had sequence 5’-ACCG-(32 bp spacer)-3’ and the bottom strand oligo had sequence 5’-ACTC-(32 bp spacer reverse complement)-3’. The two oligos were phosphorylated with T4 polynucleotide kinase (NEB, M0201S) in a 50 μL reaction: 1.5 μL 100 μM top oligo, 1.5 μL 100 μM bottom oligo, 41 μL nuclease-free water, 5 μL T4 DNA ligase reaction buffer (NEB, B0202S), 1 μL T4 polynucleotide kinase (NEB, M0201S). The reaction was incubated at 37°C for 1 hour in a thermal cycler. After phosphorylation, oligos were annealed_by adding 2.5 μL of 1 M sodium chloride (Fisher Scientific, S271-3) solution to the 50 μL reaction and incubating for 5 minutes at 98°C and then allowing the reaction to gradually cool to room_temperature (approximately 2 hours). The annealed oligos were diluted 1:10 in nuclease-free water and ligated into the BsaI-digested plasmid in a 20 μL reaction: 10 μL BsaI-digested plasmid, 6 μL nuclease-free water, 1 μL 1:10 diluted annealed oligos, 5 μL T4 DNA ligase reaction buffer (NEB, B0202S), 1 μL T4 DNA ligase (NEB, M0202M). The ligation reaction was performed at room temperature overnight. The next day, 5 μL of the ligation reaction was transformed into 50 μL of rubidium chloride chemically competent E. coli K-12 MG1655 cells.

Strain construction

Transductions with bacteriophage P1 (ATCC 25404-B1) were performed to move the polB, dinB and umuC kanamycin-marked deletions in Keio collection strains into the KD263 strain expressing the λ_L6-R spacer. λ_L6-R cells or variants thereof were grown in 5 mL LB overnight at 37°C. Overnight cultures were resuspended in a 1/2X volume of P1 salts solution: 10 mM CaCl₂ (Fisher Scientific, BP510-500), 5 mM MgSO₄ (Sigma-Aldrich, M1880-500G). 100 mL cells were mixed with several 10-fold dilutions of a fresh, high-titer P1 stock (10⁹-10¹⁰ PFU/mL) and incubated at room temperature for 30 minutes. 1 mL LB + 200 mL 1 M sodium citrate were added to each tube and cells were incubated at 37°C for 1 hour with shaking. Cells were pelleted, resuspended in 50 mL LB and spread onto LB/kanamycin plates. Colonies from the plates that received the lowest amount of P1 phage were re-struck to single colonies to ensure phage removal. Colonies were checked for the presence of the polB, dinB and umuC deletions by PCR with primers JW3077 and JW2096, JW3077 and JW2097 and JW3077 and JW2099 respectively, and the identity of the spacer within the CRISPR array was confirmed by amplification with primers JW3100 and JW3101 followed by Sanger sequencing.

Preparation of phage λvir parental stock

λvir was obtained from Bruce Levin and frozen down at −80°C: 900 μL phage in LB medium with 100 μL dimethyl sulfoxide (Sigma-Aldrich, D2650-100ML). A pipette tip was used to scrape off a tiny portion of frozen phage stock and resuspended in 20 μL LB medium. Serial dilutions of phage stock were prepared from the resuspended phage and spotted on a fresh LB top agar (Invitrogen LB broth base, Thermo Fisher Scientific, 12780-029; 0.5% final concentration of Fisher Bioreagents agar, Fisher Scientific, BP1423-500) lawn of E. coli K-12 MG1655 Δ9, in LB agar supplemented with 10 mM MgSO₄ (Sigma-Aldrich, M1880-500G). The plate was incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day a single phage plaque was picked from the top agar lawn using a P20 pipette set to 15 μL and resuspended in 20 μL LB medium. 5 μL of resuspended phage was spotted on to a fresh top agar lawn of E. coli K-12 MG1655 Δ9 in LB agar supplemented with 10 mM MgSO₄. The next day the phage spot on the top agar lawn was picked with a P20 pipette set to 15 μL and resuspended in 100 μL of an overnight stationary culture of E. coli K-12 MG1655 Δ9. The resuspended phage culture was mixed with 6 mL top agar supplemented with 10 mM MgSO₄ and poured over a LB agar plate. This was repeated twice for a total of 3 top agar plates. The plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day top agar from each plate was scraped off with 5 mL LB medium into a single 50 mL conical tube. The tube was then spun down at 15,000 x g for 10 minutes at 4°C. The phage-containing supernatant was filtered using an Acrodisc 13 mM SUPOR 0.45 μM syringe filter (Pall, 4604) into a 15 mL conical tube. The phage stock was stored at 4°C.

Phage construction

The λvir parental stock was used to construct wild-type and mutant phage stocks. 1 μL of the phage stock was resuspended in 100 μL of an overnight stationary culture of E. coli K-12 MG1655 Δ9 containing a recombinant pUT18C-based plasmid. Each recombinant pUT18C-based plasmid contained a cloned segment of phage λvir DNA with the desired gene or genomic region modified or deleted. The resuspended phage culture was mixed with 6 mL LB top agar (Invitrogen LB broth base, Thermo Fisher Scientific, 12780-029; 0.5% final concentration of Fisher Bioreagents agar, Fisher Scientific, BP1423-500) supplemented with 100 μg/mL carbenicillin (GoldBio, C-103-25) and 10 mM MgSO₄ (Sigma-Aldrich, M1880-500G) and poured over a LB agar plate supplemented with 100 μg/mL carbenicillin. Plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day top agar from each plate was scraped off with 5 mL LB medium into 50 mL conical tubes. Tubes were spun down at 15,000 x g for 10 minutes at 4°C. Recombinant phage-containing supernatant was filtered using an Acrodisc 13 mM SUPOR 0.45 μM syringe filter (Pall, 4604) into a 15 mL conical tube.

Serial dilutions of recombinant phage were prepared and spotted on a fresh top agar lawn of E. coli K-12 MG1655 Δ9 containing a pCas9 plasmid in LB agar supplemented with 25 μg/mL chloramphenicol (GoldBio, C-105-100) and 10 mM MgSO₄ (Sigma-Aldrich, M1880-500G). Each pCas9 plasmid carried a type II-A CRISPR spacer targeting the phage region that was modified or deleted to select specifically for recombinant phage with the desired deletion or modification. Top agar plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. For each recombinant phage, the next day a single phage plaque was picked from the top agar lawn using a P20 pipette set to 15 μL and resuspended in 20 μL LB medium. Serial dilutions of the resuspended phage were prepared and spotted on a fresh top agar lawn of E. coli K-12 MG1655 Δ9 containing the corresponding pCas9 plasmid in LB agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄. The top agar plate was incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day a single phage plaque was picked from the top agar lawn using a P20 pipette set to 15 μL and resuspended in 20 μL LB medium. 5 μL of resuspended phage was spotted on to a fresh top agar lawn of E. coli K-12 MG1655 Δ9 containing the corresponding pCas9 plasmid in LB agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄. The top agar plate was incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day the phage spot on the top agar lawn was picked with a P20 pipette set to 15 μL and resuspended in 100 μL of an overnight stationary culture of E. coli K-12 MG1655 Δ9 containing the corresponding pCas9 plasmid. The resuspended phage culture was mixed with 6 mL top agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄ and poured over a LB agar plate. This was repeated twice for a total of 3 top agar plates. The plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day top agar from each plate was scraped off with 5 mL LB medium into a single 50 mL conical tube. The tube was then spun down at 15,000 x g for 10 minutes at 4°C. The phage-containing supernatant was filtered using an Acrodisc 13 mM SUPOR 0.45 μM syringe filter (Pall, 4604) into a 15 mL conical tube. All phage stocks were stored at 4°C.

λvir Δgam was generated by passaging λvir through cells harboring pUT18C-Dgam and plaquing the lysate on a lawn of cells harboring pCas9:JW1370-1. λvir Δexo was generated by passaging λvir through cells harboring pUT18C-Dexo and plaquing the lysate on a lawn of cells harboring pCas9:spc12. λvir Δbet was generated by passaging λvir through cells harboring pUT18C-Dbet and plaquing the lysate on a lawn of cells harboring pCas9:JW1552-3. λvir ΔexoΔbet was generated by passaging λvir through cells harboring pUT18C-DexoDbet and plaquing the lysate on a lawn of cells harboring pCas9:JW1552-3. λvir Δred was generated by passaging λvir through cells harboring pUT18C-Dred and plaquing the lysate on a lawn of cells harboring pCas9:JW1552-3. In each case, red gene mutations were confirmed within an escaper plaque by PCR and Sanger sequencing. Wild-type λvir was generated by passaging λvir through cells harboring pUT18C and plaquing the lysate on a lawn of cells harboring pCas9:spcNT.

For phages containing chi site modifications (λvir chi¹ and λvir chi^2-7), chi-containing phages were first generated using the method described above. The process was then repeated for gam, exo, bet, exo-bet and red gene deletions for chi-containing phages. For wild-type λvir phage stocks, all steps were mirrored to mutant phage generation, except instead a pUT18C empty plasmid was used in recombineering and pCas9∷spcNT was used in subsequent selection steps. λvir chi¹ was generated by passaging λvir through cells harboring pUT18C-chiD and plaquing the lysate on a lawn of cells harboring pCas9:JW1546-7. The chi¹ (chiD) mutation was confirmed within an escaper plaque by PCR and Sanger sequencing. λvir chi^2-7 was generated by passaging λvir through cells harboring pUT18C-3chiF+R and plaquing the lysate on a lawn containing both cells harboring pCas9:JW1558-9 and cells harboring pCas9:JW1401-2. The chi^2-4 (chi3F) and chi^5-7 (chi3R) mutations were individually confirmed within an escaper plaque by PCR and Sanger sequencing.

Final phage stocks were PCR checked and Sanger sequenced to confirm appropriate gene deletions or genetic modifications. Furthermore, serial dilutions of each phage stock were prepared and 3.5 μL dilutions were spotted and dripped on fresh top agar lawns of E. coli K-12 MG1655 on LB agar supplemented with 10 mM MgSO₄. A series of four separate dilutions were prepared for each phage stock on top agar lawns to accurately determine phage titers. Top agar plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day, phage plaques were counted to determine titers. Additionally, for gam, exo, bet, exo-bet and red gene deletion mutants, 16 individual plaques were isolated per phage stock and PCR checked to confirm that they all contained the appropriate gene deletion(s).

The replication-deficient phage λΔP was generated as a lysogen using λ Red recombineering, facilitated by the plasmid pKOBEG-A (AmpR) which expresses gam, bet, and exo from the arabinose-inducible pBAD promoter (Chaveroche et al., 2000). MG1655 cells harboring a λ lysogen (obtained from J.W. Roberts) and the pKOBEG-A plasmid were grown at 30°C overnight in LB supplemented with 100 μg/mL carbenicillin. The overnight culture was diluted 1:50 into 500 mL fresh media and grown until optical density (OD₆₀₀) reached 0.2. The culture was supplemented with 0.2% L-arabinose (Sigma-Aldrich, A91906-100G-A) and grown further to an OD₆₀₀ of 1. Cells were then centrifuged at 4000 x g for 10 mins at 4°C and washed twice in 500 mL sterile ice cold water. After the second wash, cells were centrifuged again and resuspended in 40 mL sterile ice cold water. After a final centrifugation, cells were resuspended in 1.5 mL cold 10% glycerol (Fisher Scientific, G33-500) and 60 μl aliquots were stored at −80°C.

To generate a kanamycin-marked deletion construct for the λ replication gene P (λΔP), the kanamycin resistance gene on pKD4 (Datsenko and Wanner, 2000) was amplified using primers JW1518 and JW1519. This amplicon was then re-amplified with primers JW1520 and JW1521, introducing 50 bp of homology to the λ regions upstream and downstream of P. The second amplicon (~100 ng) was electroporated into electrocompetent, arabinose-induced MG1655∷λ cells harboring pKOBEG-A and transformants were selected on LB-kanamycin plates grown at 30°C. Transformants were re-struck on LB-kanamycin plates grown at 42°C to promote pKOBEG-A plasmid loss. Finally, clones were patched onto a LB-carbenicillin plate to ensure loss of pKOBEG-A.

To generate λΔP, MG1655 cells harboring the λΔP lysogen and the plasmid pCL1920-P were grown into exponential phase and treated with 0.1 μg/mL mitomycin C (AG Scientific, M-1108) for 2 hours. Phage-containing supernatants were filtered using Acrodisc 13 mM SUPOR 0.45 μM syringe filters (Pall, 4604) and used to isolate single plaques on a lawn of MG1655 cells harboring pCL1920-P. Single plaques were used to lyse plates of MG1655 cells harboring pCL1920-P and plate lysates were pooled to create λΔP stocks.

Phage λvir genome sequencing and assembly

λvir phage capsids were digested with 50 μg/mL proteinase K and phage genomic DNA was extracted using the DNeasy Blood & Tissue kit (QIAGEN, 69504). Genomic DNA was sequenced following a previously described method in our laboratory (Meeske et al., 2019). Briefly, isolated DNA was sheared using a pre-split snap-cap 6x16 mm Covaris microTUBE (Covaris, 520045) in a Covaris S220 focused-ultrasonicator and prepared for next generation sequencing using the Illumina TruSeq Nano DNA LT kit (Illumina, 20015964). Paired-end 2 × 75-bp sequencing was conducted using the 150-cycle MiSeq Reagent Kit v3 (Illumina, MS-102-3001) on the Illumina MiSeq platform. Reads were quality-trimmed using Sickle (https://github.com/najoshi/sickle) and assembled into contigs using ABySS (Simpson et al., 2009) (https://github.com/bcgsc/abyss). Finally, contigs were mapped to a reference phage λ genome (GenBank: KT232076.1) using Medusa (Bosi et al., 2015) (http://combo.dbe.unifi.it/medusa). Automated genome annotation was performed using SnapGene and the reference phage λ genome (GenBank: KT232076.1).

qPCR of phage DNA replication

To quantify phage DNA replication within an infected E. coli cell, an overnight culture of E. coli K-12 MG1655 cells carrying pCas9∷spcNT(non-targeting spacer) was diluted 1:50 in 50 mL of LB medium (BD Difco LB Broth, Lennox, BD 240220) supplemented with 25 μg/mL chloramphenicol (GoldBio, C-105-100) and 10 mM MgSO₄ (Sigma-Aldrich, M1880-500G). After 1 hour of growth, OD₆₀₀ was measured and the culture was normalized to OD₆₀₀ = 0.3. 700 μL of culture was dispensed between multiple 1.5 mL Eppendorf tubes, corresponding to three replicates and two timepoints (15 and 30 minutes) for each phage being monitored. These 700 μL cultures were infected with corresponding λvir phages at MOI 1 and incubated at 37°C with shaking for either 15 or 30 minutes. At each timepoint, samples were removed from the incubator and tubes spun down at 15,000 rpm for 2 minutes in a tabletop microcentrifuge at 4°C. Supernatants were removed and cell pellets immediately frozen down at −80°C for DNA extraction later. Additionally, three uninfected tubes were also prepared for DNA extraction as no-phage controls for qPCR.

Total DNA was extracted from frozen E. coli cell pellets using the Promega Wizard Genomic DNA Purification Kit (Promega, A1125) following the protocol for Gram-negative bacteria. Extracted DNA was quantified using the Qubit 1X dsDNA HS Assay Kit (Thermo Fisher Scientific, Q33231) and each sample was normalized to 4 ng/μL or 5 ng/μL. A total of 25 ng DNA was used as input for qPCR, performed using Applied Biosystems Fast SYBR Green Master Mix (Thermo Fisher Scientific, 4385612) on the Applied Biosystems QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, A28567) with primer pairs AA385/AA386 (λvir DNA) and AA387/AA388 (E. coli K-12 MG1655 dxs control). The sequences of the primers are listed in Table S3.

Plaque formation assays in E. coli K-12 MG1655

Overnight cultures were launched from single colonies in 3 mL of LB medium (BD Difco LB Broth, Lennox, BD 240220) with appropriate antibiotic(s). Top agar lawns of E. coli were prepared by mixing 100 μL of overnight culture with 6 mL of LB top agar (Invitrogen LB broth base, Thermo Fisher Scientific, 12780-029; 0.5% final concentration of Fisher Bioreagents agar, Fisher Scientific, BP 1423-500) supplemented with appropriate antibiotic(s) and 10 mM MgSO₄ (Sigma-Aldrich, M1880-500G). Top agar mixtures were poured over LB agar (BD Difco LB Agar, Miller, BD 244510) in 10 cm plates supplemented with appropriate antibiotic(s). Where necessary, 0.2% L-arabinose (Sigma-Aldrich, A91906-100G-A) or 1 mM IPTG (GoldBio, I2481C100) was included in the LB top agar and the LB agar plate. Plates were dried at room temperature, partially open by a sterilizing flame, for 25 minutes for the top agar to solidify. Serial dilutions of phage stock were prepared and spotted on the top agar after drying. Before making serial dilutions, aliquots of phage stocks were uniformly normalized to 1×10⁷ plaque forming units/uL (PFU/uL) for all efficiency of plaquing and plaque formation assays, except Figure S2A where phages were uniformly normalized to 5×10⁶ PFU/uL before making serial dilutions for plaquing. For imaging of plaque assays, 2.5 μL of each phage dilution was spotted on top agar using a multichannel pipette. For quantification of phage titers (and in some cases, imaging of plaque assays) and isolation of single phage plaques for phage DNA sequencing, 3.5 μL of each phage dilution was spotted on top agar using a multichannel pipette and the plate was tilted to allow phage spots to drip down the plate for easier quantification and isolation of single plaques. In either case, plates were incubated at 37°C overnight (30°C in certain cases) after drying at room temperature for 25 minutes or until the plates were completely dry.

Plaque formation assays in E. coli KD263 and KD263-derived strains

Overnight cultures were launched from single colonies in 3 mL of LB medium (BD Difco LB Broth, Lennox, BD 240220). For experiments with KD263-derived deletion mutants, 50 μg/mL kanamycin (GoldBio, K-120-100) was included in overnight LB liquid media (BD Difco LB Broth, Lennox, BD 240220), LB top agar (Invitrogen LB broth base, Thermo Fisher Scientific, 12780-029; 0.5% final concentration of Fisher Bioreagents agar, Fisher Scientific, BP1423-500) and LB agar (BD Difco LB Agar, Miller, BD 244510) plates to select for the mutant strains. For imaging of plaque assays, top agar lawns of E. coli were prepared by mixing 100 μL of overnight culture with 6 mL of LB top agar supplemented with 10 mM MgSO₄ (Sigma-Aldrich, M1880-500G). Top agar mixtures were poured over LB agar in 10 cm plates. Where necessary, 1 mM L-arabinose (Sigma-Aldrich, A91906-100G-A) for induction of Cascade, Cas1, Cas2 and type I-E spacers and 1mM IPTG (GoldBio, I2481C100) for induction of Cas3 were included in the LB top agar and the LB agar plate. Plates were dried at room temperature, partially open by a sterilizing flame, for 25 minutes for the top agar to solidify. Serial dilutions of phage stock, normalized to 1×10⁷ PFU/uL, were prepared and spotted on the top agar after drying. Before making serial dilutions, aliquots of phage stocks were uniformly normalized to 1×10⁷ PFU/uL. 3.5 μL of each phage dilution was spotted on top agar using a multichannel pipette and the plate was tilted to allow phage spots to drip down the plate for clearer visualization of small individual phage plaques. Plates were incubated at 37°C overnight after drying at room temperature for 25 minutes or until the plates were completely dry. For quantification of phage titers (and isolation of single phage plaques for phage DNA sequencing) on bacterial lawns of λ_L1-R, λ_L4-R and λ_L6-R strains, top agar lawns of E. coli were prepared by mixing 100 μL of overnight culture with a prespecified amount of phage PFUs in 6 mL of LB top agar (so as to obtain countable single plaques) supplemented with 10 mM MgSO₄, 1 mM L-arabinose and 1 mM IPTG (1 mM L-arabinose and 1mM IPTG were also included in the LB agar plate). 50 μg/mL kanamycin was included in all the media for λ_L6-R polymerase mutant strains. The following phage PFUs were added in the top agar lawns: 5×10⁵ λvir chi¹ PFUs or 5×10⁷ λvir chi¹ Δred PFUs for λ_L1-R (Figure 6C), 5×10⁶ λvir chi¹ PFUs or 2×10⁸ λvir chi¹ Δred PFUs for λ_L4-R (Figure 6D), 1×10⁵ λvir chi¹ PFUs or 1×10⁷ λvir chi¹ Δred PFUs for λ_L6-R strains (Figures 6E and 6F). Plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. For λ_L6-R strains in the polymerase mutant plaquing assays (Figure 6F), CRISPR non-targeting PFUs were also determined on top agar lawns of wild-type or polymerase mutant λ_L6-R cells by dripping serial dilutions of λvir chi¹ phage stock on top agar lawns without any inducer. For non-targeting KD263 and λ_E4-R strains (Figure 6B), plaques were dripped on lawns (with 1 mM L-arabinose and 1mM IPTG) as described above and used to determine phage titers. For infection of the λ_L4-R strain with λvir chi¹ Δred (Figure 6D), only 1-2 pickable plaques were obtained per infected lawn. So, in order to obtain enough escaper phages for characterization of escape mutations, 8 plates of top agar infections were prepared for picking 8 escape plaques for subsequent escaper sequencing (see below).

Imaging and quantification of plaque formation assays

Overnight plaque formation assays were imaged the next day (~16-24 hours after infection) using the FluorChem HD2 system (ProteinSimple). Plaque assay images were all equally adjusted for brightness and contrast using Adobe Photoshop. Images of plaque formation assays with 3.5 μL drips were manually counted for plaque forming units (PFUs) using Fiji (ImageJ) (https://imagej.net/software/fiji/). Three biological replicates, corresponding to overnight cultures launched from three separate single bacterial colonies, were counted for each strain and phage combination. For quantification of plaques on λ_L1-R, λ_L4-R and λ_L6-R strains, individual plaques were obtained across the entire bacterial lawn (see above) and counted across the entire lawn to determine PFUs. Lawn images were all equally adjusted for brightness and contrast using Adobe Photoshop and single plaques across the entire plate were manually counted using Fiji. These plaque counts were then divided by the total number of plaques used to infect each lawn to calculate the efficiency of plaquing.

Target sequencing of bacteriophage escapers

Six to sixteen phage plaques were isolated and resuspended in 20 μL of LB medium. 2-5 μL of each resuspended plaque was spotted on to a fresh top agar lawn of the original CRISPR targeting strain (from which the escapers were isolated) on LB agar supplemented with appropriate antibiotic and 10 mM MgSO₄ (Sigma-Aldrich, A91906-100G-A). For type I-E spc9R escapers, 0.2% L-arabinose (Sigma-Aldrich, A91906-100G-A) was included in the top agar and the LB agar plate for induction of Cascade, Cas3 and the plasmid-encoded type I-E spacer. For escapers of KD263-derived λ_L1-R, λ_L4-R and λ_L6-R strains, 1 mM L-arabinose (Sigma-Aldrich, A91906-100G-A) and 1 mM IPTG (GoldBio, I2481C100) were included in the top agar and the LB agar plate for induction of Cascade, Cas1, Cas2, Cas3 and type I-E spacers. Top agar plates were incubated at 37°C overnight after drying at room temperature for 25 minutes or until completely dry. The next day phage escaper spots were picked using a 20 μL pipette set to 15 μL and resuspended in 20 μL of Colony Lysis Buffer (Pyenson et al., 2017). Resuspended phage mixtures were boiled at 98°C for 10 minutes in a thermal cycler, and 1-3 μL of the boiled phage mixture was then used as template for PCR amplification with Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific, F530L). The oligonucleotide primers used for protospacer target sequencing are listed in Table S3. PCR products were submitted to Sanger sequencing by Genewiz. For escapers of spc14, isolated escaper plaques were resuspended in 20 μL of LB. Serial dilutions of each resuspended escaper plaque were prepared and then spotted and dripped on fresh top agar lawns of the original spc14 targeting E. coli in LB agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄. The next day single escaper plaques were picked for each original escaper phage and resuspended in 20 μL of colony lysis buffer for use in PCR.

In vivo CRISPR-Cas9 cleavage of viral DNA

To observe CRISPR-Cas9 cleavage of anti-viral targets, overnight cultures of E. coli K-12 MG1655 cells carrying pCas9∷spcNT (non-targeting) or pCas9∷spc9 (targeting protospacer 9) were diluted to an OD₆₀₀ of ~0.05 in LB medium supplemented with 25 μg/mL chloramphenicol (GoldBio, C-105-100) and 10 mM MgSO₄ (Sigma-Aldrich, A91906-100G-A) (Figures 2B and 2C). After 1 hour and 10 minutes of growth, OD₆₀₀ was measured for each culture, and each sample was normalized to OD₆₀₀ = 0.25. Cultures were then infected at MOI 5 with λvir chi¹ or λvir chi¹ Δgam for 25 minutes prior to centrifugation and flash-freezing of cell pellets. All samples were stored at −80°C until ready for genomic DNA purification using the DNeasy Blood &Tissue Kit (QIAGEN, 69504) following the protocol for Gram-negative organisms. Purified genomic DNA was sheared using a pre-split snap-cap 6x16 mm Covaris microTUBE (Covaris, 520045) in a Covaris S220 focused-ultrasonicator and prepared for next generation sequencing using the Illumina TruSeq Nano DNA LT kit (Illumina, 20015964). Paired-end 2 × 75-bp sequencing was conducted using the 150-cycle MiSeq Reagent Kit v3 (Illumina, MS-102-3001) on the Illumina MiSeq platform. For in vivo cleavage assays of pCas9∷spcNT, pCas9∷spc45 or pCas9∷spc45c with λvir chi¹ (Figure 4C), the procedure was carried out as described above except that genomic DNA was isolated from frozen E. coli cell pellets using the Promega Wizard Genomic DNA Purification Kit (Promega, A1125) following the protocol for Gram-negative bacteria.

Competition experiment between λvir chi¹ and λvir chi¹ Δred on top agar

Overnight cultures of E. coli K-12 MG1655 cells carrying pCas9∷spcNT(non-targeting spacer), pCas9∷spc9 or pCas9∷spc45 were launched from single colonies in 3 mL of LB medium with 25 μg/mL chloramphenicol (GoldBio, C-105-100). The next day, each overnight culture was normalized to OD₆₀₀ = 3.5. Top agar lawns of E. coli were prepared by mixing 100 μL of the normalized overnight culture with a 1:1 phage PFU mixture of λvir chi¹ and λvir chi¹ Δred at a total MOI of 20 in 6 mL of LB top agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄ (Sigma-Aldrich, A91906-100G-A). The plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The phage infection mix was stored at 4°C overnight. The next day top agar from each infection plate was scraped off with 5 mL LB medium into a single 50 mL conical tube. The tube was then spun down at 15,000 x g for 10 minutes at 4°C. The phage-containing supernatant was filtered using an Acrodisc 13 mM SUPOR 0.45 μM syringe filter (Pall, 4604) into a 15 mL conical tube. The stored phage infection mix and the filtered phage supernatants were then each used to prepare serial phage dilutions to obtain single phage plaques (for PCR of the λ red locus) on fresh top agar lawns of E. coli K-12 MG1655 cells carrying pCas9∷spcNT in LB agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄. Top agar plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day, individual phage plaques were picked and resuspended in 20 uL of Colony Lysis Buffer (Pyenson et al., 2017). Resuspended phage mixtures were boiled at 98°C for 10 minutes in a thermal cycler, and 0.5 μL of the boiled phage mixture was then used as template for PCR amplification of the λ red locus. PCR was performed in a 20 μL total reaction volume using Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific, F530L) and 1 μM each primer (AA549/AA550) with an annealing temperature of 71°C and an extension time of 1.5 minutes. For each set of phage plaques, a total of 12 phage plaques were first isolated and PCR amplified with AA549/AA550 (Table S3). PCR products were run on a 1.5% agarose gel at 140V for 35 minutes to determine the presence (~2.3 kb) or absence (~0.6 kb) of the phage red operon. These were then counted to determine the total number of red+ (λvir chi¹) and red− (λvir chi¹ Δred) phages. Only plaques that showed a single PCR band on the agarose gel were used in our analyses. In cases where the agarose gel showed both red+ and red− PCR bands for an isolated phage plaque, a new plaque was selected randomly from the serial dilution top agar lawns, PCR amplified with AA549/AA550 and run on an agarose gel as before. The competition experiment was performed a total of three times for three biological replicates. Additionally, the twelve phage plaques that were isolated for each run of the spc45-targeting competition for PCR of the red locus were also checked for potential target mutations in the protospacer through PCR of the spc45 target region using primers AA138/AA139 (Table S3) and subsequent Sanger sequencing by Genewiz with primer AA164 (Table S3).

Next generation sequencing of spc9 bacteriophage escapers

Overnight cultures of E. coli K-12 MG1655 cells or E. coli K-12 MG1655 ΔdinB cells carrying pCas9∷spc9 were launched from single colonies in 3 mL LB medium with appropriate antibiotic(s). Serial dilutions of phage stock (λvir chi¹ or λvir chi¹ λred) were prepared and spotted on fresh top agar lawns of E. coli K-12 MG1655 carrying pCas9∷spc9 supplemented with 25 μg/mL chloramphenicol (GoldBio, C-105-100) and 10 mM MgSO₄ (Sigma-Aldrich, A91906-100G-A). In each case, 3.5 μL of each phage dilution was spotted on top agar using a multichannel pipette and the plate was tilted to allow phage spots to drip down the plate. Plates were incubated at 37°C overnight after drying at room temperature for 25 minutes or until the plates were completely dry. With pKM208 plasmid-containing cells, 100 μg/mL carbenicillin (GoldBio, C-103-25) and 1mM IPTG (GoldBio, I2481C100) were also included in the LB top agar and the LB agar plate, and cultures and plates were incubated at 30°C.

The next day, top agar phage drips containing ~100-1000 plaques (one full drip dilution per top agar plate) were scooped up using an inverted sterile pipette tip and resuspended in 500 μL LB medium in 1.5 mL Eppendorf tubes. Tubes were spun down at 15,000 rpm for 3 minutes in a tabletop microcentrifuge at 4°C. Phage-containing supernatants were then filtered using Acrodisc 13 mM SUPOR 0.45 μM syringe filters (Pall, 4604) into fresh 1.5 mL Eppendorf tubes. To isolate phage DNA as template for PCR, 5 μL of phage supernatant was mixed with 15 μL Colony Lysis Buffer (Pyenson et al., 2017) and boiled at 98°C for 10 minutes in a thermal cycler. PCR of the spc9 target was performed with 5 μL phage lysis input in a 50 μL PCR reaction using Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific, F530L) and barcoded primers with identical 4-9 bp DNA barcodes on forward and reverse primers. The barcoded oligonucleotide primers used for PCR are listed in Table S3. PCR products were run on a 2% agarose gel at 140 V for 25 minutes. Gel-run PCR products were then gel extracted using the QIAquick Gel Extraction Kit (QIAGEN, 28704). Extracted DNA was quantified using the Qubit 1X dsDNA HS Assay Kit (Thermo Fisher Scientific, Q33231) and analyzed for DNA fragment size using the Agilent 2200 TapeStation system with the High Sensitivity D1000 ScreenTape (Agilent, 5067-5584 and 5067-5585). Extracted samples were then pooled together for a total DNA concentration of 10-20 ng/μL, with each sample at equal ng/μL concentrations. The pooled PCR sample was prepared for next generation sequencing using the Illumina TruSeq Nano DNA LT kit (Illumina, 20015964), followed by paired-end 2 × 75-bp sequencing using the 150-cycle MiSeq Reagent Kit v3 (Illumina, MS-102-3001) on the Illumina MiSeq platform.

Liquid culture time course of bacteriophage escape

An overnight culture of E. coli K-12 MG1655 cells carrying pCas9∷spc45 and one carrying pCas9∷spc45c were each diluted 1:50 in 10 mL LB medium supplemented with 25 μg/mL chloramphenicol (GoldBio, C-105-100) and 10 mM MgSO₄ (Sigma-Aldrich, A91906-100G-A). Two separate 10 mL starter cultures were prepared for each spacer for infection with each of two phages: λvir chi¹ and λvir chi¹ Δred. After 1 hour of growth, OD₆₀₀ was measured for each of the four outgrowth cultures, and each was normalized to OD₆₀₀ = 0.3. The normalized 10 mL liquid cultures of spc45- and spc45c-targeting cultures were then infected with either λvir chi¹ or λvir chi¹ Δred at MOI 1. Infection cultures were incubated at 37°C with shaking for 9 hours. For the first 5 hours, 400 μL of culture was removed from each sample every 50 minutes and then every 2 hours for the next 4 hours for a total time course of 9 hours post infection. At each collection point, 400 μL of culture was transferred to 1.5 mL Eppendorf tubes and spun down at 15,000 rpm for 3 minutes in a tabletop microcentrifuge at 4°C. Phage-containing supernatants were then filtered using Acrodisc 13 mM SUPOR 0.45 μM syringe filters (Pall, 4604) into fresh 1.5 mL Eppendorf tubes. After completion of the time course, phage supernatants from each timepoint were used to prepare serial phage dilutions to estimate phage titers on fresh top agar lawns of E. coli K-12 MG 1655 carrying pCas9∷spcNT in LB top agar supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄. Top agar plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day, phage plaques were counted to determine titers at each timepoint. Additionally, eight phage plaques were isolated for each timepoint and each infection culture to identify potential target mutations in the protospacer through PCR of the spc45 target region using Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific, F530L) with primers AA138/AA139 (Table S3) and subsequent Sanger sequencing by Genewiz with primer AA164 (Table S3).

Liquid culture time course of bacteriophage propagation

An overnight culture of E. coli K-12 MG1655 cells carrying pCas9∷spcNT (non-targeting spacer) was diluted 1:50 in 10 mL LB medium supplemented with 25 μg/mL chloramphenicol (GoldBio, C-105-100) and 10 mM MgSO₄ (Sigma-Aldrich, A91906-100G-A). Two separate starter cultures were launched, for infection with two phages: λvir chi¹ and λvir chi¹ Δred. After one hour of growth, OD₆₀₀ was measured for each culture, and each sample was normalized to OD₆₀₀ = 0.3. Cultures were then infected at MOI 0.1 with each phage and incubated for 5 minutes at 37°C with shaking. After 5 minutes, cultures were removed from the incubator and spun down at 4300 x g for 5 minutes at 4°C. Supernatant was discarded and the pellet resuspended in 1 mL LB medium and transferred to fresh 1.5 mL Eppendorf tubes. Tubes were spun down at 8000 rpm for 1 minute in a tabletop microcentrifuge at 4°C. Supernatant was discarded and the pellet resuspended in 1 mL LB medium. Cells were washed in this manner another two times to remove unadsorbed phage. Cells were kept on ice between washes. After the final wash, pellets were resuspended in 10 mL LB medium supplemented with 25 μg/mL chloramphenicol and 10 mM MgSO₄. 100 μL and 500 μL of each resuspended culture was used to determine infective centers and unadsorbed phage titers respectively. Resuspended cultures were placed back in the 37°C shaking incubator for a total of 90 minutes post wash. Phage-infected cultures were sampled at the following time points: 30, 45, 60, 80, 95 minutes post wash (i.e. 35, 50, 65, 85, 100 minutes post infection). At each time point, 400 μL was collected from each phage-infected culture and spun down in 1.5 mL Eppendorf tubes at 15,000 rpm for 3 minutes in a tabletop microcentrifuge at 4°C. 100 μL of phage supernatant was collected and phage-containing supernatants filtered using Acrodisc 13 mM SUPOR 0.45 μM syringe filters (Pall, 4604) into fresh 1.5 mL Eppendorf tubes. Filtered phage supernatants were used to prepare two sets of serial phage dilutions to estimate phage titers on fresh top agar lawns of E. coli K-12 MG1655 in LB agar supplemented with 10 mM MgSO₄. Top agar plates were incubated at 37°C overnight after drying at room temperature for 25 minutes. The next day, phage plaques were counted to determine titers.

Quantification and statistical analysis

qPCR phage DNA replication analysis

For qPCR phage DNA bar graphs in Figures 1A, S2B, S2C and S3B, fold-change values relative to λΔP 15-minute (Figures 1A, S2B and S2C) and 30-minute time point values (Figure S3B) are reported. Error bars represent the standard error of the mean of three biological replicates.

Efficiency of plaquing assay analysis

For efficiency of plaquing assays in Figures 1, 2, 3, 4, 5, 6 and S1, efficiency of plaquing is calculated as the plaques formed by the phage on a CRISPR targeting lawn divided by the plaques formed by the same phage on a CRISPR non-targeting lawn (spcNT or no inducer with Type I-E spc9R). Error bars represent the standard error of the mean of three biological replicates. For quantification of plaques on λ_L1-R, λ_L4-R and λ_L6-R strains (Figures 6C–E), individual plaques were obtained across the entire bacterial lawn with 1 mM L-arabinose (Sigma-Aldrich, A91906-100G-A) and 1 mM IPTG (GoldBio, I2481C100) (see earlier sections) and counted across the entire lawn to determine plaque forming units. These plaque counts were then divided by the total number of plaques used to infect each lawn to calculate the efficiency of plaquing. Error bars represent the standard error of the mean of three biological replicates. For quantification of plaques for the λ_L6-R polymerase mutants’ plaque assays (Figure 6F), individual plaques were obtained across the entire bacterial lawn with 1 mM L-arabinose and 1 mM IPTG (see earlier sections) and counted across the entire lawn to determine plaque forming units. These plaque counts were then divided by the total number of plaques used to infect each lawn to determine the efficiency of plaquing. Plaques were also counted from drips of serial phage dilutions on lawns of wild-type or polymerase null mutants with no inducer (CRISPR non-targeting) and the total plaques formed on these non-targeting lawns of wild-type or mutant cells were used to calculate an adjusted efficiency of plaquing for each spcL6-R strain graphed in Figure 6F. Error bars represent the standard error of the mean of three biological replicates.

Statistical analyses of plaquing assays were performed using GraphPad Prism 9. Where necessary, a t-test was used to compare means. In each case, an unpaired parametric t-test was performed with the assumption that samples come from populations with a Gaussian distribution and the same standard deviation.

In vivo Cas9 cleavage assay data analysis

Illumina paired-end sequencing reads were aligned to phage genomes using the Burrows-Wheeler Aligner (Li and Durbin, 2009). The resulting .sam files were parsed and visualized with custom Python scripts.

Competition experiment on top agar data analysis

Individual plaques were PCR amplified with AA549/AA550 (Table S3) and run on an agarose gel (see earlier section) to determine the presence (~2.3 kb) or absence (~0.6 kb) of the phage red operon. These were then counted to determine the total number of red+ (λvir chi¹) and red− (λvir chi¹ Δred) phages. Only plaques that showed a clean single PCR band on the agarose gel were used in counting. In cases where the agarose gel showed both red+ and red− PCR bands for an isolated phage plaque, a new plaque was selected randomly from the serial dilution top agar lawns, PCR amplified with AA549/AA550 and run on an agarose gel as before. Error bars represent the standard deviation of three biological replicates.

Next generation sequencing of spc9 escape mutations data analysis

Forward and reverse sequencing reads of spc9 target PCR products were extracted from raw MiSeq FASTQ files and quality filtered with a Phred quality cutoff of 10 using Python v3.8 on PyCharm CE. For reads passing quality filter, corresponding forward and reverse reads were compiled into a text file for downstream data analysis using a custom Python script. Reads were first organized into individual lists according to their primer barcodes. Reads were selected only if both forward and reverse reads contained the same barcode to avoid issues with barcode switching between samples during library preparation for next generation sequencing. Only sequencing reads from the bottom strand 5’ end of the PCR products were selected since only these contained the full spc9 protospacer and PAM sequence. Barcode and reverse PCR primer sequences were subtracted from each bottom strand read. Each read was then mapped to the spc9 protospacer region reverse complement reference sequence and base pair mismatches between the reference sequence and each sequencing read was recorded and enumerated. For each sample barcode, a list of mutations was generated with the following information: bp position in protospacer/PAM, wild-type base, mutated base, fraction of total mutations (count of the specific mutation divided by the total number of mutations recorded), normalized mutation count (count of the specific mutation divided by the total number of sequencing reads analyzed). This information was then outputted using Python to a separate Microsoft Excel file for each sample barcode. The major escape mutations, ones with the largest normalized mutation counts (graphed as normalized mutated reads), were collected into a Prism file for further graphical and statistical analyses using GraphPad Prism 9.

Next generation sequencing of spc9 escape mutations statistical analysis

For the spc9 escape mutation bar graphs in Figures 3A–C, 3F, S4A, S4C and S4D, the number of reads with a specific mutation is normalized to the total reads analyzed for each sample to give the values for normalized mutated reads. Error bars in Figures 3A–C and 3F represent the standard error of the mean of three biological replicates. Figures S4A, S4C and S4D show the individual replicates, each corresponding to a top agar lawn of a separate bacterial colony. Statistical analyses in Figures 3A–C and 3F were performed using GraphPad Prism 9: t-tests were used to compare mean values for each PAM mutation. In each case, an unpaired parametric t-test was performed with the assumption that samples come from populations with a Gaussian distribution and the same standard deviation.

Supplementary Material

Table S1

Table S1. Sequences of CRISPR escaper phages generated during type II-A and type I-E targeting. Related to Figures 1, 3, 4, 5 and 6. The infection host, the λvir chi¹ phage variant, the targeting CRISPR spacer, and the protospacer and PAM sequence of each escaper phage plaque is specified. In the case of dCas9 escapers (tab E), the deactivating Cas9 amino acid substitutions are specified. Single point mutations in the protospacer or PAM sequences are highlighted in red. For spc14 escaper phages, microhomology-mediated end joining (MMEJ) deletions lengths are specified, with the microhomology repeat sequences mediating the deletions provided in brackets (tabs L and N). All plaques were obtained and sequenced from top agar lawns of CRISPR targeting bacteria, except for the spc45 competition (tab D) and the spc45 and spc45c infection time courses in liquid cultures (tabs F-I) for which phage lysates or culture supernatants were collected at the specified time points post infection and then enumerated on top agar lawns of CRISPR non-targeting bacteria to obtain single plaques for protospacer target sequencing.

Table S2

Table S2. Related to Figure 3. Normalized mutation reads obtained after next generation sequencing of spc9 escaper phages.

Table S3

Table S3. Related to STAR Methods. Tables of bacterial strains, plasmids, bacteriophages, oligonucleotides and CRISPR spacers used in this study.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
E. coli K-12 MG1655	(Guyer et al., 1981)	N/A
E. coli K-12 MG1655 ΔrecB	(Baba et al., 2006)	N/A
E. coli K-12 MG1655 ΔrecD	(Baba et al., 2006)	N/A
E. coli K-12 MG1655 ΔpolB	(Baba et al., 2006)	N/A
E. coli K-12 MG1655 ΔdinB	(Baba et al., 2006)	N/A
E. coli K-12 MG1655 ΔumuC	(Baba et al., 2006)	N/A
E. coli ACT-01	(Caliando and Voigt, 2015)	N/A
E. coli KD263	(Strotskaya et al., 2017)	N/A
E. coli λ_E4-R	(Strotskaya et al., 2017)	N/A
E. coli λ_L1-R	(Strotskaya et al., 2017)	N/A
E. coli λ_L4-R	(Strotskaya et al., 2017)	N/A
E. coli λ_L6-R	(Strotskaya et al., 2017)	N/A
E. coli ΔpolB λ_L6-R	This paper	N/A
E. coli ΔdinB λ_L6-R	This paper	N/A
E. coli ΔumuC λ_L6-R	This paper	N/A
E. coli K-12 MG1655 Δ9	(Wang et al., 2010)	N/A
E. coli K-12 MG1655∷λ	Gift from J.W. Roberts	N/A
E. coli K-12 MG1655∷λΔP	This paper	N/A
λvir parent stock	Gift from Bruce Levin	N/A
λvir Δgam2	This paper	N/A
λvir Δexo4	This paper	N/A
λvir	This paper	N/A
λvir Δgam	This paper	N/A
λvir Δexo	This paper	N/A
λvir Δbet	This paper	N/A
λvir Δexo Δbet	This paper	N/A
λvir Δred	This paper	N/A
λvir chi1 parent stock	This paper	N/A
λvir chi1	This paper	N/A
λvir chi1 Δgam	This paper	N/A
λvir chi1 Δexo	This paper	N/A
λvir chi1 Δbet	This paper	N/A
λvir chi1 ΔexoΔbet	This paper	N/A
λvir chi1 Δred	This paper	N/A
λvir chi2-7 parent stock	This paper	N/A
λvir chi2-7	This paper	N/A
λvir chi2-7 Δred	This paper	N/A
λΔP	This paper	N/A
P1	ATCC	ATCC 25404-B1
Chemicals, Peptides, and Recombinant Proteins
BD Difco LB Broth, Lennox	Fisher Scientific	Cat#DF0402-08-0 (BD 240220)
BD Difco Dehydrated Culture Media: LB Agar, Miller (Luria-Bertani)	Fisher Scientific	Cat#DF0445-07-6 (BD 244510)
LB Broth Base (Invitrogen)	Thermo Fisher Scientific	Cat#12780-029
Fisher BioReagents Agar, Powder/Flakes	Fisher Scientific	Cat#BP1423-500
Chloramphenicol	GoldBio	Cat#C-105-100
Spectinomycin Dihydrochloride Pentahydrate	GoldBio	Cat#S-140-50
Kanamycin Monosulfate	GoldBio	Cat#K-120-100
Carbenicillin (Disodium)	GoldBio	Cat#C-103-25
IPTG	GoldBio	Cat#I2481C100
L-(+)-Arabinose	Sigma-Aldrich	Cat#A91906-100G-A
Magnesium Sulfate Heptahydrate	Sigma-Aldrich	Cat#M1880-500G
Sodium Chloride (Crystalline/Certified ACS), Fisher Chemical	Fisher Scientific	Cat#S271-3
Calcium Chloride Dihydrate (White Crystals to Powder), Fisher BioReagents	Fisher Scientific	Cat#BP510-500
Sodium Citrate Dihydrate (Granular/Certified), Fisher Chemical	Fisher Scientific	Cat#S279-500
Glycerol (Certified ACS), Fisher Chemical	Fisher Scientific	Cat#G33-500
Dimethyl Sulfoxide	Sigma-Aldrich	Cat#D2650-100ML
T4 Polynucleotide Kinase	NEB	Cat#M0201S
T4 DNA Ligase	NEB	Cat#M0202M
T4 DNA Ligase Reaction Buffer	NEB	Cat# B0202S
BsaI-HF	NEB	Cat#R3535L
Mitomycin C	AG Scientific	Cat#M-1108
Phusion High-Fidelity DNA polymerase	Thermo Fisher Scientific	Cat#F530L
Applied Biosystems Fast SYBR Green Master Mix	Thermo Fisher Scientific	Cat#4385612
Critical Commercial Assays
QIAprep Spin Miniprep Kit	QIAGEN	Cat#27104
QIAquick Gel Extraction Kit	QIAGEN	Cat#28704
DNeasy Blood & Tissue Kit	QIAGEN	Cat#69504
Wizard Genomic DNA Purification Kit	Promega	Cat#A1125
Qubit 1X dsDNA HS Assay Kit (Invitrogen)	Thermo Fisher Scientific	Cat#Q33231
High Sensitivity D1000 ScreenTape	Agilent	Cat#5067-5584
High Sensitivity D1000 Reagents	Agilent	Cat#5067-5585
TruSeq Nano DNA LT Library Prep Kit	Illumina	Cat#20015964
MiSeq Reagent Kit v3 (150-cycle)	Illumina	Cat#MS-102-3001
Deposited Data
Next-generation sequencing files (Illumina NGS)	Sequence read archives (SRA)	BioProject ID PRJNA754816
Oligonucleotides
See “Oligonucleotides” in Table S3	This paper	N/A
Recombinant DNA
See “Plasmids” in Table S3	This paper	N/A
Software and Algorithms
Fiji	(Schindelin et al., 2012)	https://imagej.net/software/fiji/
Photoshop CC	Adobe	https://www.adobe.com/products/photoshop.html
Illustrator CC	Adobe	https://www.adobe.com/products/illustrator.html
SnapGene	Insightful Science	snapgene.com
Python v3.8	Python Software Foundation	https://www.python.org/downloads/release/python-380/
Sickle	Joshi and Fass, GitHub, 2011	https://github.com/najoshi/sickle
ABySS	(Simpson et al., 2009)	https://github.com/bcgsc/abyss
Medusa	(Bosi et al., 2015)	http://combo.dbe.unifi.it/medusa
PyCharm CE	JetBrains	https://www.jetbrains.com/pycharm/download/#section=mac
Burrows-Wheeler Aligner	(Li and Durbin, 2009)	http://bio-bwa.sourceforge.net/
GraphPad Prism 9	Insightful Science	https://www.graphpad.com/scientific-software/prism/
Custom code used to analyze in vivo Cas9 cleavage assay Illumina NGS data	This paper	https://github.com/Marraffini-Lab/Hossain_etal_2021
Custom code used to analyze spc9 escape mutation frequency Illumina NGS data	This paper	https://github.com/Marraffini-Lab/Hossain_etal_2021

Highlights.

• The λ Red recombination system enables phage λ to escape type I and II CRISPR nucleases

• Repair of cleaved viral DNA by λ Red leads to the introduction of escaper mutations

• Escaper generation is rare when the host’s RecBCD system repairs the viral DNA