Bacteria encode a wide range of antiphage systems and a subset of these proteins are homologous to components of the human innate immune system. Mammalian nucleotide-binding and leucine-rich repeat containing proteins (NLRs) and bacterial NLR-related proteins use a central NACHT domain to link detection of infection with initiation of an antimicrobial response. Bacterial NACHT proteins provide defense against both DNA and RNA phages. Here we investigate the mechanism of phage detection by the bacterial NLR-related protein bNACHT25 in E. coli. bNACHT25 was specifically activated by Emesvirus ssRNA phages and analysis of MS2 phage escaper mutants that evaded detection revealed a critical role for Coat Protein (CP). A genetic assay showed CP was sufficient to activate bNACHT25 but the two proteins did not directly interact. Instead, we found bNACHT25 requires the host chaperone DnaJ to detect CP and protect against phage. Our data support a model in which bNACHT25 detects a wide range of phages using an indirect mechanism that may involve guarding a host cell process rather than binding a specific phage-derived molecule.

Breadth of RNA phage defense by bNACHT proteins

We previously reported that four bacterial NLR-related proteins provided robust phage defense against both DNA and RNA phages [9]: bNACHT02, 12, 25, and 32. Each encode a central NACHT module. bNACHT02 is 39% identical to bNACHT12, and bNACHT25 is 81% identical to bNACHT32. bNACHT02 and bNACHT12 fall within clade 14 of bacterial NLR-related proteins, which all contain short C-terminal NACHT-associated (SNaCT) domains but lack readily discernible effector domains [9] (Fig 1A). bNACHT25 and bNACHT32 are from clade 6, contain NACHT C-terminal helical domain 1 (NCH1) domains, and N-terminal PD-(D/E)XK family DNase effector domains [9,18] (Fig 1A). Nuclease activity of bNACHT25 and bNACHT32 is required for phage protection, as mutating a conserved active site residue of the endonuclease domain (D48A) abrogated defense (Fig 1B).

To investigate the breadth of RNA phages restricted by NLR-related proteins, we challenged E. coli MG1655 expressing GFP as a negative control or bacterial NLR-related proteins with diverse ssRNA phages from the Fiersviridae family (Fig 1C and 1D). All four systems defended against MS2 and MS2-like phages (the Emesvirus genus), however, none of these systems were capable of defending against Qβ or Qβ-like phages (the Qubevirus genus). While MS2 and Qβ have similar genome organizations and infectious cycles [19], they possess little shared sequence identity (38% identical). Within the Emesvirus and Qubevirus genera, the tested phages are highly related (≥92% identical) (S1 Fig) [20].

We confirmed our findings in liquid culture by infecting MG1655 expressing EV or bNACHT25 with either MS2 or Qβ. At a multiplicity of infection (MOI) of 0.2, bacteria expressing EV were lysed by MS2 and Qβ, while cultures expressing bNACHT25 continued to grow after infection with MS2 (S2A–2D Fig). At a MOI of 2, the OD₆₀₀ of bNACHT25-expressing bacteria stopped increasing upon MS2 infection (S2C Fig).

We hypothesized that OD₆₀₀ plateaued when bNACHT25-expressing bacteria were infected due to activation of the endonuclease effector domain and tested this by measuring intracellular DNA degradation following infection with MS2. Plasmid DNA was harvested to represent total host DNA. Infection of bacteria expressing either bNACHT25 or bNACHT32 with MS2 caused DNA degradation indicative of effector activation (Fig 1E), and nuclease activity of both proteins was ablated by introduction of the D48A mutation. Nuclease activity was not observed when cultures were infected with Qβ (Fig 1F); of note, we cannot distinguish whether Qβ evades detection by bNACHT25 or instead encodes a bNACHT25 inhibitor.

Mutations in CP enable MS2 to evade bNACHT proteins

ssRNA phages have some of the smallest known viral genomes, with MS2 only encoding four genes in its 3.6 kb genome: mat encodes maturation protein (MP), which enables attachment to the F pilus and RNA entry into the host; cp encodes coat protein (CP), which forms the viral capsid; lys encodes lysis protein (L), which allows for host lysis; and rep encodes replicase (Rep), which is an RNA-dependent RNA polymerase that replicates the ssRNA viral genome [21,22] (Fig 2A). We sought to determine how MS2 activated bNACHT proteins and generated spontaneous MS2 escaper mutants that evaded these systems. We isolated 28 MS2 escaper mutants that successfully formed plaques in the presence of either bNACHT02, bNACHT12, or bNACHT32 from three separate, clonal parent lineages. We were unable to isolate MS2 mutants that could grow on bNACHT25. Mutant phage genomes were sequenced and variations from parent genomes were determined (Tables 1 and S1).

Interestingly, MS2 mutants generated on bNACHT32 from two different parent lineages had acquired the same missense mutation in the cp gene, resulting in a G17D mutation (Table 1). Escaper mutant phages encoding CP^G17D evaded bNACHT32-mediated defense compared to their parent strains (Fig 2B and 2C). CP is the major structural protein that forms the viral capsid of MS2 and has been long-studied for its role in RNA binding [23]. The G17D mutation is located on the AB loop of CP, a surface-exposed loop that is not predicted to interact with the RNA genome or participate in the CP dimerization interface [24,25]. Additional mutations in cp were also found in MS2 escaper mutants generated on bNACHT02 and bNACHT12 (S1 Table).

bNACHT32 and the highly similar bNACHT25 were selected for further investigation due to their readily assayable DNase effector domains, which provide unambiguous evidence of activation during phage infection (Fig 1E). We hypothesized that CP activated bNACHT32, leading to effector activation and cell death, and that the G17D mutation in CP reduced the ability of CP to activate the defense system, allowing for MS2 escape. We tested this by inducing expression of CP in the presence of bNACHT32 and measuring colony formation. Significant growth inhibition was observed in the bNACHT32 and CP coexpression condition that was not observed when either was expressed individually (Fig 2D) or when CP was coexpressed with the nuclease-dead bNACHT32^D48A. As suggested by our escaper analysis, CP^G17D did not reduce colony formation to the same degree as wild-type CP when coexpressed with bNACHT32 (Fig 2D).

DNA was purified from cultures expressing CP, bNACHT32, or both proteins combined to visualize effector domain activation. DNA degradation was only observed in the bNACHT32 and CP coexpression condition (Fig 2E), indicating that CP is sufficient to activate the endonuclease domain of bNACHT32 in this genetic assay. The CP^G17D mutant activated bNACHT32 to a lesser extent than wild-type CP (Fig 2E). The decrease in activation was not due to decreased protein expression as CP^G17D was produced in higher amounts than wild-type (Fig 2F). Together, these data suggest that CP plays a role in bNACHT32 activation and the G17D mutations in CP allow MS2 to evade detection.

CP synthesis is sufficient to activate bNACHT25

We next interrogated the qualities of CP that were required for bNACHT activation using a genetic assay. bNACHT25 was used in place of bNACHT32 (these proteins share 81% identity and >96% similarity) for these and future assays because bNACHT25 provided more robust effector activation in response to phage (Fig 1E). Expression of wild-type CP in the presence of bNACHT25 resulted in inhibition of colony formation and DNA degradation (Fig 2G and 2H). Expression of CP^W83R, which is unable to oligomerize beyond dimerization and limits capsid assembly [26], similarly induced cell death and DNA degradation, albeit at a slightly slower rate (Fig 2G and 2H). However, mutation of the cp start codon completely abrogated cell death and DNA degradation phenotypes, demonstrating that translation of CP is required to activate bNACHT25 (Fig 2G and 2H). Expression levels of CP were confirmed for each of these mutants by western blot (S3A Fig). Despite the similar structure to CP from MS2 [27], CP^Qβ only modestly induced DNA degradation and did not impact colony formation (Fig 2G and 2H). These findings suggest a mechanism for why bNACHT25 provides resistance against MS2 but not Qβ or Qβ-like phages.

We next tested if bNACHT25 could be activated by capsid proteins from dsDNA phages that bNACHT25 provides resistance to. The major and minor capsid proteins of phage T4, gp23 and gp24, and the major capsid protein of phage λ, gpE, did not inhibit colony formation in the presence of bNACHT25 (Fig 2G) [28,29]. These findings are, perhaps, unsurprising as CP from MS2 adopts a fold unique to the ssRNA phages that is not related to the HK97 fold capsid proteins from tailed dsDNA phages [30–32].

The other three proteins encoded by MS2 were also interrogated for activation of bNACHT25. Interestingly, significant DNA degradation was observed upon induction of the lysis protein L (S4 Fig). A slight amount of activation was also observed upon induction of genes encoding MP and Rep that resembled expression of CP^Qβ (Figs 2H and S4). The relevance of these data is unclear as expression of L is extremely toxic to the bacteria. Of note, the inducible vector in the genetic assay expressed CP to higher than levels observed during MS2 infection (S3B Fig). Our findings suggest that synthesis of CP is required for activation of bNACHT25, and assembly of higher-order CP oligomers may be playing a role in the activation. Additionally, our data suggest that while bNACHT25 has specificity for recognizing synthesis of certain phage proteins, such as CP and L, it is possible that multiple proteins can activate the defense system during phage infection.

CP does not directly interact with bNACHT25

We next sought to understand the mechanism of CP activation of bNACHT25. AVAST systems encode central STAND NTPase modules that are related to NACHT modules and are activated by binding of phage proteins [13,14,33]. We tested whether CP similarly activates bNACHT25 by direct binding using immunoprecipitation (IP) assays. CP failed to IP from bacteria expressing epitope-tagged, catalytically inactive bNACHT25 (FLAG-bNACHT25^D48A) (Fig 3A). bNACHT25^D48A lacking a FLAG-tag was used as a negative control and the nuclease domain was catalytically inactivated to enrich for active complexes in the bacterial cytoplasm and prevent death of bacteria.

Our genetic data unambiguously showed bNACHT25 activation upon CP synthesis (Fig 2G and 2H), yet the two proteins did not form a complex (Fig 3A). We therefore hypothesized that bNACHT25 may instead be activated by a host protein that specifically changes during phage infection. To address this hypothesis, we performed mass spectrometry (MS) on tryptic digests of bNACHT25 IP samples to identify interacting proteins. A large cohort of proteins increased in abundance over the negative control (Fig 3B and S2 Table). This analysis also confirmed bNACHT25 was successfully immunoprecipitated, while CP was not enriched (Fig 3B). We confirmed that during MS2 infection, CP did not IP with bNACHT25, while one of the top IP-MS hits, DnaJ, specifically pulled down with tagged bNACHT25 (Fig 3C).

bNACHT25 activation is mediated by the host chaperone DnaJ

To identify the specific host protein required for bNACHT25-mediated sensing of phage, we introduced plasmids expressing bNACHT25 or the bNACHT25^D48A catalytically inactive control into E. coli BW25113 carrying marked deletions of genes encoding each of 15 most enriched, non-essential IP-MS hits that repeated in both of our biological replicates [34]. These strains were then challenged with phage T4. Only a mutation in dnaJ impacted bNACHT25-dependent phage resistance (Fig 3D). These results suggest that DnaJ is required for bNACHT25 to protect against phage.

DnaJ is an Hsp40-family chaperone whose J domain is conserved from humans to bacteria [35]. DnaJ combats misfolded proteins in the cell in two ways. First, DnaJ can bind misfolded proteins and prevent their aggregation [36]. Second, DnaJ can deliver misfolded proteins to DnaK, an Hsp70-family chaperone, and stimulate the ATPase activity of DnaK through the J domain of DnaJ [37]. DnaJ has many substrates and is famously required for replication of phage λ and certain plasmids [38,39].

To understand the role of DnaJ in bNACHT25-mediated phage resistance, we constructed a marked dnaJ deletion (dnaJ::cat) in bacteria expressing VSV-G-bNACHT25 (VSV-G epitope tag introduced in order to visualize bNACHT25 expression) or a GFP negative control from the chromosome. Western blots confirmed dnaJ deletion and that the level of bNACHT25 was unperturbed by loss of dnaJ (Fig 3F). bNACHT25 no longer protected against the phages T2 and T4 in the absence of dnaJ (Fig 3G and 3H). Phage protection provided by bNACHT25 was rescued upon complementation with an inducible vector expressing dnaJ (Fig 3G and 3H). Interestingly, mutations in the J-domain of dnaJ that have previously been shown to disrupt binding between DnaJ and DnaK (D35N) [40] or the ability of DnaJ to stimulate the ATPase activity of DnaK (H33Q) [41] were not sufficient to rescue bNACHT25 phage defense in dnaJ::cat cells (Fig 3E, 3G and 3H). These mutants were expressed to the same level as wild-type DnaJ (S5A Fig). This finding suggests that the cochaperone activity of DnaJ in the DnaJ–DnaK chaperone cycle is required for phage protection by bNACHT25.

Similar results were obtained with VSV-G-bNACHT32 expressed from the chromosome, however bNACHT32 expression was slightly decreased in the dnaJ::cat background (S5B and S5C Fig). Together, these data suggest that bNACHT25, and potentially bNACHT32, require host DnaJ to protect against phage.

We next probed the specificity of phage detection through DnaJ by introducing plasmids expressing bNACHT proteins from clade 6 (bNACHT25 and bNACHT32) and clade 14 (bNACHT11 and bNACHT12) into wild-type and dnaJ::kanR bacteria. While bNACHT25 and bNACHT32 required dnaJ to protect against T4, bNACHT11 and bNACHT12 were unaffected by the disruption of dnaJ (Fig 3I), demonstrating that clade 14 bNACHT proteins have a different activation mechanism. We also investigated the specificity of host protein impacts on bNACHT25 by knocking out other E. coli chaperones, chaperone-related proteases, cold shock proteins, and phage shock proteins. bNACHT25 was expressed in bacteria deleted for clpA/X, lon, hslO, spy, htpG, cspA/B/C, hscA/B, and pspB/C and these strains were challenged with T4. bNACHT25 provided the same level of phage resistance for each of these as was observed in a wild-type background (S6 Fig). These findings confirm the specific role for DnaJ in bNACHT25-mediated phage defense.

Our data showed that phage protection mediated by bNACHT25 requires DnaJ and that bNACHT25 abundance is not impacted by loss of dnaJ (Fig 3F). However, since DnaJ is a chaperone that helps maintain protein solubility and function through interacting with exposed hydrophobic regions, an alternative hypothesis to explain our data is that DnaJ-deficient cells cannot express or fold fully functioning bNACHT25, resulting in dnaJ-dependence. Knocking out dnaJ did not cause a decrease in bNACHT25 abundance in whole cell lysates (S7A Fig). We next tested whether bNACHT25 becomes insoluble in the absence of dnaJ by separating whole cell lysates expressing bNACHT25 into soluble and insoluble fractions using centrifugation and testing for bNACHT25 abundance in the soluble fraction. We found that deleting dnaJ had no effect on the abundance of soluble bNACHT25 (S7A Fig). This result suggests that knocking out dnaJ does not decrease the levels of soluble bNACHT25.

To further show that bNACHT25 can function in the absence of DnaJ, we used a previously-characterized allele of bNACHT25 with a H506L mutation in the NACHT domain predicted to destabilize the ADP-bound, “OFF”-state of the protein, leading to stimulus-independent hyperactivation [9,43]. Induced expression of this hyperactive bNACHT25 allele led to DNA degradation in wild-type and dnaJ-mutant E. coli, with only a very small decrease in activity in early timepoints (Fig 4A and 4B). This mild reduction in activation is not comparable to the full loss of bNACHT25 activation by CP or L, and the hyperactive mutant was expressed to the same level as wild-type (S7B Fig). These data demonstrate that downstream effector activation in a stimulus-independent system can still occur in the absence of dnaJ, confirming that bNACHT25 does not require DnaJ to adopt a functional conformation. These data help to rule out the hypothesis that DnaJ is simply required for bNACHT25 function and instead support a model where DnaJ is required for detection of phage protein synthesis upstream of bNACHT25 activation.

DnaJ is required for activation of bNACHT25 by MS2 proteins

We were not able to measure changes to MS2 phage protection because MS2 failed to form plaques in bacteria lacking dnaJ (S8 Fig). However, we hypothesized that if DnaJ was required for MS2 detection, it would also be required for bNACHT25 activation by CP in our genetic assay. Accordingly, coexpression of CP with bNACHT25 resulted in DNA degradation in wild-type bacteria, but this phenotype was completely abrogated in a dnaJ::cat background (Fig 4C). Bacterial cell death on plates (Fig 4D) and in liquid culture (Fig 4E) caused by bNACHT25 and CP coexpression was also dependent on the presence of dnaJ. Expression levels of CP and bNACHT25 were comparable in wild-type and dnaJ::cat strains (S9 Fig).

Our analysis of MS2 proteins that activated bNACHT25 unexpectantly revealed that L, in addition to CP, activated nuclease activity (S4 Fig). L is the single-gene lysis protein of MS2 and was previously shown to bind to DnaJ, and dnaJ mutations limited host lysis by MS2 [42]. We hypothesized that L interactions with DnaJ were responsible for L-mediated bNACHT25 activation. Expression of L in a wild-type background induced strong activation of bNACHT25, and this activation was completely ablated in a dnaJ::cat background (S10 Fig), indicating that DnaJ is necessary for activation of bNACHT25 by both CP and L.

Taken together, these data support a model in which DnaJ is required for transmitting an activation signal from phage proteins to bNACHT25, suggesting that bNACHT25 guards host processes or proteins that specifically involve DnaJ (Fig 4F).

Bacterial strains and culture conditions

E. coli strains used in this study are listed in S3A Table. Unless otherwise indicated, all cultures were grown in 1–4 mL of media in 14 mL culture tubes shaking at 220 rpm at 37 °C. “Overnight” cultures were started from either a single colony or glycerol stock and grown for 16–20 h following inoculation. Culture media was supplemented with carbenicillin (100 μg/mL), chloramphenicol (20 μg/mL), kanamycin (50 μg/mL), and/or tetracycline (15 μg/mL) when applicable for plasmid maintenance or strain verification. Experiments were performed with E. coli MG1655 (CGSC6300) or E. coli MG1655 (CGSC7636), and E. coli OmniPir [9] was used for construction and storage of plasmids. Where indicated, bNACHT sequences were inserted into the lacZ locus of E. coli MG1655 using Lambda red methodology as previously described [9,61].

Bacterial cultures used for cloning, strain construction, indicated colony formation assays, and immunoprecipitation assays were grown in LB medium (1% tryptone, 0.5% yeast extract, and 0.5% NaCl). Strains were frozen for long-term storage in LB supplemented with 30% glycerol at 70 °C. Bacteria used to perform phage propagation, phage infection assays, DNA degradation assays, and indicated colony formation assays were cultivated in “MMCG” minimal medium (47.8 mM Na₂HPO₄, 22 mM KH₂PO₄, 18.7 mM NH₄Cl, 8.6 mM NaCl, 22.2 mM glucose, 2 mM MgSO₄, 100 mM CaCl₂, 3 mM thiamine, Trace Metals at 0.01% v/v (Trace Metals mixture T1001, Teknova, final concentration: 8.3 μM FeCl₃, 2.7 μM CaCl₂, 1.4 μM MnCl₂, 1.8 μM ZnSO₄, 370 nM CoCl₂, 250 nM CuCl₂, 350 nM NiCl₂, 240 nM Na₂MoO₄, 200 nM Na₂SeO₄, 200 nM H₃BO₃)). When a strain with two plasmids was cultivated in MMCG medium, bacteria were grown in carbenicillin (50 μg/mL) and chloramphenicol (10 μg/mL). MMCG and LB agar plates contain 1.6% agar and media components described above.

Conjugation of F′ plasmid into E. coli MG1655

The plasmid F′ was introduced into various strains of E. coli MG1655 via conjugation. Briefly, 400 μL of overnight cultures of donor (E. coli MG1655 + F′ donor grown in LB + tetracycline) and recipient (various E. coli strains grown with proper antibiotics) were pelleted at 10,000 × g for 1 min. Pellets were washed in 20 μL of LB, pelleted again, and resuspended in 20 μL of LB. Resuspensions of donor alone, recipient alone, and mixture of donor and recipient were spotted on an LB plate, which was incubated upright for 1 h at 37 °C. Spots were then struck to single colonies on LB + tetracycline + appropriate antibiotics to select for recipients that had received F′.

Plasmid construction

Plasmids used in this study are listed in S3A Table. MS2 and Qβ phage genes were amplified from synthesized cDNA (see below). T4 and λ genes were amplified from 1 μL of boiled, diluted phage lysate. T4 gp23* and gp24* include amino acids 66-520 and 11-426, respectively, which are the regions of gp23 and gp24 that are maintained post-proteolytic cleavage [30]. bNACHT genes were cloned from plasmids previously generated [9]. E. coli genes were cloned from purified E. coli MG1655 genomic DNA. See S3C Table for protein accession numbers.

bNACHT genes along upstream native promoter regions/downstream terminator regions were cloned into the SbfI/NotI sites of the pLOCO2 vector or the EcoRI/HindIII sites of the pBAD30x vector, as previously described [9]. All phage genes were cloned into the NotI and BamHI or BmtI sites of pTACxc (for chloramphenicol selection) or pTACx (for carbenicillin selection) vector for inducible expression with IPTG. Restriction enzymes were purchased from New England Biolabs. DNA sequences were amplified and amended with ≥18 bp homology to their destination vectors using Q5 Hot Start High Fidelity Master Mix (NEB, M0494L). Vectors were constructed using Gibson Assembly with HiFi DNA Assembly Master Mix (NEB, E2621L) as described [62] and transformed into OmniPir via heat shock or electroporation. VSV-G and 3X-FLAG tags were appended to the N-termini of bNACHT25 and bNACHT32 in order to perform western blots and IPs on these proteins. To add epitope tags, tag sequences were included in 3′ sequences of primers that annealed to the gene of interest. Plasmid sequences were verified with Sanger sequencing (Quintara Biosciences or Azenta).

Phage amplification and storage

Phages used in this study are listed in S3B Table. F-dependent RNA phages were amplified using the E. coli MG1655 + F′ host [63,64], and dsDNA phages were amplified using E. coli MG1655. A modified double agar overlay plate amplification was used to generate phage lysates [65]. Briefly, 400 μL of mid-logarithmic phase (OD₆₀₀ = 0.2–0.8) bacterial cultures were combined with 5,000–15,000 plaque forming units (PFU) of phage and incubated for 1 min at room temperature to allow for adsorption. An amount of 3.5 mL of “top agar” (0.35% agar, 10 mM MgCl₂, 10 mM CaCl₂, and 100 μM MnCl₂, 0.01% v/v Trace Metals) was then added to the phage-bacteria mixture which was then plated directly onto an MMCG agar plate. Plates were incubated overnight at 37 °C. To recover the amplified phage from the top agar, 5 mL of SM buffer (100 mM NaCl, 8 mM MgSO₄, 50 mM Tris-HCl pH 7.5, 0.01% gelatin) was applied to the top of the plate and the plate was incubated for 1–6 h at room temperature. Phages were then harvested by transferring SM buffer from the plate to a tube. To increase phage titers, top agar overlay was sometimes scraped and harvested along with the SM buffer. In this case, the SM buffer was first centrifuged at 4,000 × g for 10 min and supernatant was transferred to a new tube. Two to three drops of chloroform were added to all phage stocks followed by vortexing to sterilize the lysate. Phages were stored at 4 °C.

MS2-like (FrSaffron, FrHenna, and FrBlood) and Qβ-like (FrSangria, FrMerlot, FrHibiscus, FrBurgundy) phages were kindly provided by Michael Baym and colleagues. To plaque purify, serial dilutions of these phage stocks were plated as described in the above paragraph, and single plaques were picked using a glass Pasteur pipet and soaked out in 500 μL SM buffer with 1–2 drops of chloroform.

We observed that MS2 and Qβ phage stocks were prone to loss of infectiousness over time at 4 °C, at times inexplicably losing over 10⁴ PFU within 6 months. To maintain phage stocks, MS2 and Qβ were frozen for long term storage according to published methods [66]. Briefly, phage was added to mid-logarithmic phase E. coli MG1655 + F′ in MMCG medium at an MOI of 0.5. The phage and bacteria were incubated for 15 min at room temperature without shaking to allow for adsorption. The mixture was then combined with glycerol (final concentration 15%), flash frozen in liquid nitrogen, and stored at −70 °C. To recover and amplify phages from frozen glycerol stocks, a small amount of stock was scraped and swirled into 100 μL of SM buffer. This mixture was then used in a plate amplification with E. coli MG1655 + F′ as described above.

Phage infection assays estimated efficiency of plaque formation

To quantify efficiency of plaque formation or titer for phage lysates, including determining phage defense, 400 μL of mid-logarithmic phase MG1655 + F′ expressing the defense system from the indicated vector or from the chromosome was combined with 3.5 mL of top agar and plated onto an MMCG agar plate. Once solidified, 2 μL of 10-fold serial dilutions of phage in SM buffer were spotted onto the top agar overlay. Once dried, plates were incubated overnight at 37 °C. The next day, plaques were quantified. When single plaques were not distinguishable/countable, the most dilute spot with visible phage (i.e., a hazy zone of clearance) was recorded as 10 PFU. When no plaques or zone of clearance were visible at the most concentrated spot, plaque formation fell below the limit of detection and 0.9 PFU were recorded for that spot. Phage protection data was reported as PFU/mL ± standard error of the mean (S.E.M) of n = 3 biological replicates.

Phage infection time course in liquid culture

Bacteria containing the indicated plasmid were grown to mid-logarithmic phase in 25-mL MMCG cultures. An amount of 150 μL of cultures OD₆₀₀-normalized to 0.15 were added to wells of a 96-well plate in triplicates. Phage was added to each bacterial culture at the indicated MOI. Final volume of phage added to each well was 5 μL. Beginning immediately after infection, OD₆₀₀ readings were recorded every 2 min over the course of the experiment using the SPARK Plate Reader (TECAN). Bacterial growth curves are representative of n = 3 biological replicates. The mean of 3 technical replicates was plotted for each time point.

Knockout strains used for testing bNACHT25-mediated phage protection

Gene deletion mutants described in Figs 3D and S6 were constructed by replacing the indicated gene with the kanamycin resistance cassette (kanR), as previously described (Keio collection) [34] and obtained as glycerol stock duplicates. Strains were streaked to single colonies on LB + kanamycin, cultivated overnight, and transformed with plasmids expressing bNACHT25 or bNACHT25^D48A via electroporation.

MS2 escaper mutant generation

Escaper mutants were generated from 3 independent, wild-type parent MS2 phage stocks that were separately plate amplified on E. coli lacking a defense system. To generate escaper phages, approximately 5 × 10⁶ PFU of each parent phage was used to infect E. coli carrying plasmids expressing either bNACHT02, bNACHT12, bNACHT25, or bNACHT32 using the modified double agar overlay assay described above. Six single plaques were picked from each amplification (“primary escapers”), soaked out in 500 μL of SM buffer, and re-amplified on bacterial lawns expressing the corresponding defense system to repurify and confirm escaper mutants (“secondary escapers”). bNACHT02 and bNACHT12 primary escapers were diluted 1:1000 to generate single secondary escaper plaques. bNACHT32 primary escapers required higher concentrations of phage lysates to obtain secondary escapers plaques (undiluted or 1:100 dilutions), and no bNACHT25 primary escapers generated plaques upon repurification. Single secondary escaper plaques were picked and soaked out in SM buffer for storage. The resulting 28 escaper phage lysates were plate amplified and escapers of interest were spot plated onto bacteria expressing the corresponding bNACHT system to confirm escaper phenotype.

Phage genome sequencing and escaper analysis

RNA genomes of MS2 and Qβ were purified as previously described [9]. Briefly, phage lysates were treated with DNase I, and RNA was extracted using the PureLink RNA Minikit (Invitrogen) according to the manufacturer’s instructions. On-column DNase treatment was omitted, and RNA was eluted in 30 μL of nuclease-free water. RNA phage cDNA was synthesized using the Invitrogen SuperScript III First-Strand Synthesis System as previously described [9]. MS2 cDNA was then PCR amplified in 3 overlapping fragments using OneTaq polymerase using previously described primers [67]. Amplified MS2 genome was prepared for Illumina sequencing using a modification of the Nextera kit protocol as previously described [68]. Samples were sequenced using the Illumina MiSeq V2 Micro 300-cycle kit (CU Anschutz Genomics and Microarray Core). Using the Map to Reference feature in Geneious software, reads were mapped to the MS2 reference genome (NCBI Genome accession NC_001417). Under the Trim Primers option, the Nextera Trimming Oligo (AGATGTGTATAAGAGACAG) was trimmed from reads; otherwise, default settings were used.

The Geneious feature “Find Variations/SNPs” was used to identify variants in escaper genomes. Variants were identified as escaper mutations if they were present in ≥70% of reads and were not present in parent genomes.

bNACHT25 (DNA degradation) activation assay

Bacteria containing the indicated plasmids were grown to mid-logarithmic phase in 30 mL MMCG containing appropriate antibiotics. If applicable, phage was added at an MOI of 2, or 500 μM IPTG or 0.2% (v/v) arabinose was added to induce gene expression. Following infection/induction, 2 × 10⁹ CFU were harvested from each strain by pelleting at 4,000 × g for 10 min at 4 °C. Plasmid DNA was extracted using standard miniprep protocol (Qiagen) and eluted in 60 μL DNA elution buffer (10 mM Tris-HCl pH 8.0). Twenty microliters of eluted DNA was combined with 5 μL of 6× Loading Dye (final concentration 4 mM Tris-HCl, 3% Ficoll-400, 12 mM EDTA, 0.04% Orange G, pH 8) and run for 30 min at 130 V on a 1% agarose gel (1% agarose, 40 mM Tris, 20 mM acetic acid, 1 mM EDTA, 0.02% v/v SYBR Safe DNA stain). Gels were imaged using an Azure Biosystems Azure 200 Bioanalytical Imaging System.

Colony formation assays for measuring growth inhibition

To test for bacterial growth inhibition, 5 μL of 10-fold serial dilutions of indicated cultures were spotted onto the appropriate agar plates. For testing bNACHT32 activation in Fig 2D, overnight cultures in MMCG were spotted onto MMCG plates carbenicillin (20 μg/mL) and chloramphenicol (4 μg/mL) with or without 500 μM IPTG. In all other cases, overnight cultures in LB were spotted onto LB plates containing appropriate antibiotics and either 0.2% glucose for uninduced conditions or 500 μM IPTG for induced conditions. Once dry, plates were incubated overnight at 37 °C. The next day, colonies were quantified. In instances where single colonies were not distinguishable at the most dilute spot with visible bacteria (i.e., hazy spot of bacteria), 10 colony forming units (CFU) were recorded. When no colonies were visible at the most concentrated spot, 0.9 CFU were recorded. Colony formation data was reported as CFU/mL ± S.E.M of n = 3 biological replicates.

Liquid culture time course for measuring growth inhibition

Bacteria containing the indicated plasmids were diluted to an OD₆₀₀ of 0.1 in 30 mL MMCG containing appropriate antibiotics and cultivated shaking at 37 °C. When cultures reached mid-logarithmic phase (t = 3 hr), 500 μM IPTG was added to induce gene expression. OD₆₀₀ measurements were taken at the indicated timepoints over the course of the experiment. Growth curves are reported as the mean ± S.E.M of n = 3 biological replicates.

Construction of dnaJ knockout strains

The E. coli MG1655 dnaJ gene was replaced using Lambda red recombination with cat, a gene encoding chloramphenicol acetyltransferase, as described previously [61]. Briefly, cat and its promoter were amplified from pKD3 with overhangs containing homology to 50 bp immediately upstream and downstream of the dnaJ gene. E. coli with chromosomally inserted KanR-VSVG-bNACHT25, KanR-VSVG-bNACHT32, or KanR-GFPmut3 carrying the pKD46 plasmid were grown to mid-logarithmic phase in LB + 0.2% arabinose to induce the Lambda red system. These bacteria were then made electrocompetent, transformed with gel-purified PCR products containing the cat insert via electroporation, and recovered for 2 h at 30 °C. Cultures were then plated on LB+ chloramphenicol to select for acquisition of cat. PCR and western blot were used to confirm the replacement of dnaJ with cat (Figs 3F and S5C).

Western blots

To measure protein expression via western blot, bacterial strains expressing indicated gene(s) were grown to mid-logarithmic phase. When applicable, 500 μM IPTG or phage at an MOI of 2 was added, and cultures were grown for an additional 20 min or indicated timepoints before extracting whole cell lysates. To harvest, 5 × 10⁸ CFU were pelleted and when testing for CP expression after infection with MS2, pellets were washed with 1 mL of sterile water and re-pelleted. Pellets were all resuspended in 50 μL of LDS buffer (106 mM Tris-HCl pH 7.4, 141 mM Tris Base, 2% w/v lithium dodecyl sulfate, 10% v/v glycerol, 0.51 mM EDTA, 0.05% Orange G). Samples were then incubated at 95 °C for 10 min and centrifuged at 20,000 × g for 3–5 min to remove debris. Samples in LDS were loaded at equal volumes on a 4%–20% SDS–PAGE gel, run for approximately 1–2 h at 100–160 V, and transferred to PVDF membranes charged in methanol. Membranes were blocked in Licor Intercept Buffer for 1 h at room temperature and incubated with primary antibodies diluted in Intercept buffer overnight at 4 °C with rocking. αVSV-G (Rockland, RRID:AB_217930) was used at 1:10,000 to detect VSV-G-bNACHT25 and VSV-G-bNACHT32, αCP (Millipore, RRID:AB_2827507) was used at 1:10,000 to detect CP, αFLAG (Sigma, RRID:AB_259529) was used at 1:10,000 to detect 3× FLAG-bNACHT25^D48A, and αDnaJ (Enzo, RRID:AB_2039063) was used at 1:2,500 to detect DnaJ. αRNAP (BioLegend Cat# 663,006, RRID:AB_2565555) was used at 1:5,000 as a loading control for testing protein levels.

Membranes were washed 3 times for 10 min each in TBS-T (Tris-buffered saline, 0.1% Tween-20) then incubated in Licor infrared (800CW/680RD) αRabbit/Mouse secondary antibodies diluted 1:40,000 in TBS-T for 1 h at room temperature. Blots were visualized with the Licor Odyssey CLx.

Immunoprecipitation (IP) assays

Immunoprecipitation assays were performed as described previously [69]. Briefly, E. coli MG1655 expressing 3× FLAG-bNACHT25^D48A or bNACHT25^D48A from one plasmid and inducible CP from another plasmid were grown in 25 mL of LB with 500 μM IPTG until they reached mid-logarithmic phase. OD₆₀₀-normalized cultures (approximately 1 × 10¹⁰ CFU) were then centrifuged to 4,000 × g for 10 min at 4 °C. The resulting pellet was resuspended in 4.5 mL lysis buffer (400 mM NaCl, 20 mM Tris-HCl pH 7.5, 2% glycerol, 1% triton, and 1 mM β-mercaptoethanol). Cells were lysed by sonication and then centrifuged at 20,000 × g and 4 °C to remove cellular debris. Fifty microliters of each soluble lysate was saved as “Input.” Samples were then mixed with 30 μL magnetic beads covalently linked to the αFLAG M2 antibody (Sigma) overnight at 4 °C with end-over-end rotation. After 3 washes in lysis buffer, IP beads were resuspended in 50 μL LDS. SDS–PAGE and western blots were performed as described above.

Mass spectrometry (MS) analysis

An IP was performed as described above with the following alterations. To increase protein yields, 100 mL of induced, logarithmic-phase culture was processed for each sample. Following incubation of bacterial lysates with αFLAG beads, beads were washed 3 times with 20 mM ammonium bicarbonate. Dry beads were then subjected to on-bead trypsin digest followed by analysis on a Thermo Obitrap Q-Exactive HF-X using nanoLC–MS MS, as previously described [69]. Peptides were mapped to the proteome of E. coli MG1655 (https://uniprot.org/proteomes/UP000000625) in addition to MS2 CP and 3× FLAG-bNACHT25^D48A sequences. Two biological replicates of the IP were performed and analyzed by MS. Data from replicate 1 can be found in S2A Table and replicate 2 can be found in S2B Table.

bNACHT25 solubility assay

E. coli MG1655 wild-type or dnaJ::cat expressing VSV-G-bNACHT25 from the bacterial chromosome were grown to mid-logarithmic phase in 25 mL of LB medium. OD₆₀₀-normalized cultures (approximately 1 × 10¹⁰ CFU) were then centrifuged to 4,000 × g for 10 min at 4 °C. The resulting pellet was resuspended in 4.5 mL lysis buffer (400 mM NaCl, 20 mM Tris-HCl pH 7.5, 2% glycerol, 1% triton, and 1 mM β-mercaptoethanol). Cells were lysed by sonication, and 1 mL of this material was saved as “Whole Cell Lysate” (WCL). The remaining lysate and then centrifuged at 21,000 × g and 4 °C for 30 min to remove insoluble components of the lysate. The supernatant was saved as “Soluble” (S). Western blots were performed as described above.

Accession numbers

See S3C Table for complete list of protein accession numbers. Briefly, bNACHT02: WP_021557529.1, bNACHT12: WP_021519735.1, bNACHT25: WP_001702659.1, bNACHT32: WP_057688292, bNACHT11: WP_114260439.1, CP (MS2): YP_009640125.1, CP (Qβ): BAP18764.1, maturation protein (MS2): YP_009640124.1, lysis protein (MS2): YP_009640126.1, replicase (MS2): YP_009640127.1, gp23 (T4): AAD42428.1, gp24 (T4): AAD42429.1, gpE (λ): AAA96540.1, DnaJ (E. coli MG1655): NP_414556.1.

S2 Fig. bNACHT25 protects against infection by MS2 but not Qβ in liquid culture.

(A–D) Growth curves of E. coli expressing either GFP or bNACHT25. OD₆₀₀ measurements began immediately following infection with the indicated phages at the indicated MOI. Data are representative of n = 3 biological replicates. The mean of n = 3 technical replicates for a representative experiment is shown. The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003203.s002

(XLSX)

S3 Fig. CP expression levels.

(A) western blot analysis of bacterial lysates generated from E. coli expressing the indicated CP alleles and GFP from the chromosome. Bacteria were harvested 20-min post-induction with IPTG. (B) western blot analysis of E. coli lysates generated from strains expressing bNACHT25 harvested at the indicated timepoints following infection with MS2 at MOI of 2 or induction of CP with IPTG. For (A–B), data are representative images of n = 3 biological replicates.

https://doi.org/10.1371/journal.pbio.3003203.s003

(PDF)

S4 Fig. Analysis of MS2 genes with bNACHT25.

Visualization of plasmid integrity in E. coli expressing either GFP or bNACHT25 on the chromosome coexpressed with the MS2 protein indicated via an inducible plasmid. Plasmid DNA was harvested at indicated timepoints post-induction with IPTG. Data are representative images of n = 3 biological replicates.

https://doi.org/10.1371/journal.pbio.3003203.s004

(PDF)

S5 Fig. DnaJ is required for bNACHT32 phage protection.

(A) western blot analysis of cell lysates generated from E. coli with the indicated genotypes. p-dnaJ expresses dnaJ to WT levels without adding inducer. Data are representative images of n = 2 biological replicates. (B) Efficiency of plating of the indicated phage on WT or dnaJ::cat E. coli MG1655 expressing either GFP or VSV-G-bNACHT32 (V-bNACHT32) from the chromosome. Data plotted as in Fig 1B. Chloramphenicol acetyltransferase (cat). (C) western blot analysis of cell lysates generated from E. coli with the indicated genotypes. Data are representative images of n = 3 biological replicates. The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003203.s005

(PDF)

S6 Fig. Screening bNACHT25 phage protection in E. coli chaperone mutants.

Efficiency of plating of phage T4 on E. coli BW25113 mutants containing plasmids expressing bNACHT25 or bNACHT25^D48A. Not determined (n.d.) indicates the strain was unable to grow with one or both plasmids in the assayed conditions. Data plotted as in Fig 1B. The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003203.s006

(PDF)

S7 Fig. Soluble bNACHT25 is not decreased by the loss of dnaJ.

(A) western blot analysis of Whole Cell (WCL) and Soluble (S) lysates obtained from E. coli WT or dnaJ::cat expressing VSV-G-bNACHT25 from the bacterial chromosome. (B) western blot analysis of cell lysates generated from E. coli BW25113 expressing VSV-G-bNACHT25 with the indicated genotypes. All data are representative images of n = 3 biological replicates.

https://doi.org/10.1371/journal.pbio.3003203.s007

(PDF)

S9 Fig. CP and bNACHT25 expression in wild-type and dnaJ::cat cells.

Western blot analysis of E. coli lysates from the indicated genotypes, timepoints, and conditions. Data are representative images of n = 3 biological replicates.

https://doi.org/10.1371/journal.pbio.3003203.s009

(PDF)

S10 Fig. Activation of bNACHT25 by MS2 L protein requires DnaJ.

Visualization of plasmid integrity in E. coli with the indicated genotype at the indicated timepoints post-induction of L with IPTG. bNACHT25 is on the chromosome and L is on a plasmid. Data are representative images of n = 3 biological replicates.

https://doi.org/10.1371/journal.pbio.3003203.s010

(PDF)