Statistics and data reproducibility

Where applicable, statistical analyses were conducted using unpaired, two-sided Student’s t-tests in GraphPad Prism. Relevant statistical results, including P values and standard deviations, are reported in the figure legends alongside the data. No statistical methods were used to predetermine sample size, and blinding and randomization were not used.

Patient stool samples and isolation of bacteria and phage

Cholera patient stool samples were collected and screened for V. cholerae and phages as described previously³⁴. Stool samples were collected from patients with suspected cholera at the icddr,b Dhaka Hospital and the Government Health Complex in Mathbaria, Pirojpur, under protocol number PR-16083 approved by the icddr,b Ethical Review Committee, with written consent obtained from participants or their guardians. Samples positive for V. cholerae serogroup O1 and/or O139 on VC RapidDipStick test (Span Diagnostics) were de-identified and stored at −80 °C with w/v 20% glycerol. For on-site isolation of V. cholerae, positive samples were enriched in alkaline peptone water (APW, pH 8.4, Difco) at 37 °C for 6–8 h and then cultured overnight on taurocholate tellurite-gelatin agar (Difco). V. cholerae appearing colonies were further confirmed using previously described biochemical and serological methods³⁵. Further purification of V. cholerae and isolation of phages from stool was performed at the University of California, Berkeley. V. cholerae isolates were further purified twice on Luria-Bertani (LB) agar plates and analysed by PCR for PLE and/or whole-genome sequencing (see Supplementary Table 7 for primers). For phage isolation, a panel of V. cholerae hosts (including PLE(−) E7946 and PLE11(+) BFS783) was used to probe for phages from stool. Where possible, phages were isolated and purified on the cognate V. cholerae strain isolated from the stool sample. Bacterial hosts were grown to the mid-log phase, incubated with a small amount of frozen stool sample collected on a pipette tip (and dilutions thereof) and the mixture was plated in 0.5% LB top agar. Single plaques were picked and purified twice on the same host before being analysed by PCR for adi, CRISPR–cas/odn or TMP mutations using the primers listed in Supplementary Table 7 and/or whole-genome sequencing.

Whole-genome sequencing

Genomic DNA from phages and bacteria was purified using Monarch Genomic DNA Purification Kit (New England BioLabs). Phage samples were initially treated with DNase I to remove non-encapsidated DNA. Illumina sequencing (150-base pair by 150-base pair paired end) was performed by the Microbial Genome Sequencing Center or SeqCenter (for all bacteria and most phage), and Nanopore sequencing was performed by the Barker Sequencing Core at the University of California, Berkeley (for a subset of phage isolates). Genomes were assembled using SPAdes³⁶, and for escape phages selected on PLE11(+) V. cholerae, genomes were analysed using BreSeq (v.0.33)³⁷.

Bioinformatic analysis

The PLE genomes were aligned on the basis of gene product identity using clinker³⁸ at a 30% identity cut-off. We performed BLASTn searches against ICP1 genomes using odn, adi and CRISPR–cas from ICP1_2001_Dha_0, ICP1_2006_Dha_E or ICP1_2011_Dha_A as queries. TMP substitutions were called if the sequence differed from ICP1_2006_Dha_E or ICP1_2011_Dha_A. ICP1 CRISPR spacers were manually annotated between direct repeats in the CRISPR arrays. The phage phylogeny was built by comparing whole-genome sequences of 29 phages isolated from this study and 67 isolates from previous work⁶ using tBLASTx analysis from ViPTree³⁹. The intergenomic similarities between phages sequenced in this study were determined at VIRIDICweb using BLASTn parameters ‘-word_size 7 -reward 2 -penalty -3 gapopen 5-gapextended 2’ (ref. ⁴⁰). ICP1 and PLE11 gene products identified in proteomics were analysed for functional predictions using HHPred⁴¹ or extracted from previous work (for ICP1)⁶.

The phylogeny of bacterial genomes was calculated as described previously¹⁵. Briefly, fastp v.0.23.2 (ref. ⁴²) was used to evaluate the quality of the raw shotgun paired-end sequences. Genetic variants were identified by mapping the raw reads to the V. cholerae N16961 reference genome (National Center for Biotechnology Information (NCBI) accession IDs NC_002505.1 and NC_002506.1) using snippy v.4.6.0 (ref. ⁴³). Phylogenetic analysis was performed using IQ-TREE v.2.2.0 (ref. ⁴⁴) with 1,000 bootstrap and the best fitted evolutionary model was selected using ModelFinder⁴⁵. Spades v.3.15.4 genome assembler was used to generate contigs. Each of the ten previously known PLEs¹³ and PLE11 were used as BLASTn queries against the V. cholerae genomes and annotated in the phylogeny. Lists of the single-nucleotide polymorphisms in the core genome and strains used to build the phylogeny are in Supplementary Tables 8 and 9, respectively.

The structural predictions for TAC^PLE4 and Rta were made using ColabFold⁴⁶ on COSMIC2 (ref. ⁴⁷) and GoogleColab Structural alignments TAC^HK97 (Protein Data Bank (PDB) ID 2OB9), TAC^PLE1 (5IR0) and predicted TAC^PLE4 were done on ChimeraX⁴⁸ using ‘Smith-Watermann ssFraction 0.8008 matrix BLOSUM-45 hgap 10 sgap 10 ogap 4′ parameters. The root mean-squared deviation values for aligned pruned amino acid residues are reported. Putative satellite genomes from non-cholera Vibrio spp.³⁴ and cf-PICIs⁴⁹ were analysed for the presence of TMPs and integrases using BLASTp and HHPred⁴¹. Genome visualizations were generated with R, using the gggenes package.

Bacterial growth and cloning

Bacterial and phage strains are reported in Supplementary Tables 1, 2 and 6. V. cholerae cultures were grown in LB medium at 37 °C with aeration or on LB agar plates. When required, antibiotics were used at the following concentrations: 100 µg ml⁻¹ spectinomycin, 75 µg ml⁻¹ kanamycin, 2.5 µg ml⁻¹ chloramphenicol in LB plates and 1.25 µg ml⁻¹ chloramphenicol in liquid cultures. Escherichia coli cultures were grown in LB medium at 37 °C with aeration or on LB agar plates, and 25 µg ml⁻¹ chloramphenicol was when needed. V. cholerae strains were made naturally competent through previously reported methods⁵⁰ and then were transformed with DNA fragments generated from purified PCR products. To create the laboratory strain of PLE11(+) V. cholerae, a kan^R cassette was inserted downstream of the last PLE11 ORF by means of the natural transformation of clinical strain BFS783. The PLE11:kan^R was then transduced into V. cholerae E7496. Gene deletions in the PLE11 kan^R E7946 derivative were made by replacing the target gene(s) with a frt-spec^R-frt cassette by natural transformation, and where indicated, the spec^R was cured as described previously¹⁰. To generate the plasmid p-rta, the rta coding sequence from PLE11 was cloned into the pKL06.2 vector using Gibson assembly and selected in E. coli XL-1 Blue. The purified plasmid was electroporated into E. coli S17 and conjugated into V. cholerae strains. All deletions and plasmid constructs were confirmed by PCR and Sanger sequencing.

Phage propagation, generation of escapes and phage engineering

For standard plaque assays, phage isolates were propagated on V. cholerae to generate confluent lysis plates, collected in STE buffer (100 mM NaCl, 10 mM Tris-Cl pH 8.0, 1 mM EDTA), treated with chloroform and clarified by centrifugation at 5,000g for 15 min. The aqueous layer containing phages was stored at 4 °C. Escape phages were generated by picking individual plaques from plaque assays with the parental phage, ICP1_2006_Dha_E (for CRISPR(+)) or ICP1_2011_Dha_A (deleted of CRISPR spacers targeting PLE11 (ACMphi232), for Adi(+)), on PLE11(+) V. cholerae (JEF42). The plaques were purified three times, and then genomic DNA was extracted from high-titre stocks and whole-genome sequenced (ICP1_2006_Dha_E), or the tmp was analysed by Sanger sequencing (ICP1_2011_Dha_A derivatives). Phages with engineered mutations were generated by passaging the parental phage on a V. cholerae strain expressing an inducible CRISPR–Cas system with a targeting spacer and a repair template encoding the desired mutation as described previously in ref. ⁵¹. Individual plaques were purified, and the tmp sequence was confirmed by Sanger sequencing.

Phage assays

Saturated V. cholerae cultures (optical density at 600 nm (OD₆₀₀) greater than 2) were diluted to OD₆₀₀ = 0.05 and grown to OD₆₀₀ between 0.3 and 0.5 (mid-log) at 37 °C with aeration. Cultures were induced with 1 mM isopropyl-β-D-thiogalactoside (IPTG) and/or 1.5 mM theophylline at OD₆₀₀ = 0.2 when necessary. For spot assays, 200 µl or 300 µl of mid-log V. cholerae cultures were mixed with 0.5% LB top agar (with required antibiotics and inducers, as needed) and poured on 100-mm or 150-mm plates, respectively. Next, 4 ml or 12 ml of this mix was poured onto 100-mm or 150-mm LB agar plates, respectively. Tenfold serial dilutions of phage were prepared, and 3 µl of each dilution was spotted onto the prepared top agar with bacteria, allowed to dry and incubated at 37 °C for 6 h. To quantify plaquing efficiencies, 10 µl of phage dilutions were mixed with 50 µl of mid-log V. cholerae for 7–10 min at room temperature. The phage-bacteria mix was added to 0.5% LB top agar (4 ml) in 60-mm plates. Inducers and antibiotics were added when required. The plates were incubated at 37 °C for 6–7 h and then at room temperature overnight. The efficiency of plaquing was calculated from three biological replicates relative to the permissive empty vector control (for p-rta) or permissive PLE(−) control (for PLE11(+)). Data presented as images represent one of the three biological replicates, and extra replicates are reported in Supplementary Information.

qPCR

PLE11 replication PCRs were performed as described in ref. ¹². Briefly, saturated V. cholerae cultures were diluted to OD₆₀₀ = 0.05 and grown to OD₆₀₀ = 0.3 at 37 °C with aeration. Immediately before infection (t = 0), 100-µl samples were collected and boiled. The remaining culture was infected with phage (ICP1_2006_Dha_E∆CRISPR∆cas2-3 or wild-type ICP1_2006_Dha_E) at a multiplicity of infection (MOI) of 2.5, incubated at 37 °C with aeration for 20 min (t = 20), collected and boiled. PLE11 DNA from boiled samples was measured in technical duplicates by quantitative PCR (qPCR) relative to a standard curve using primers indicated in Supplementary Table 7. The fold change in genome copy was calculated as the amount of PLE11 genome at t = 20 relative to the PLE11 genome at t = 0. qPCR experiments were performed with three biological replicates.

ICP1 propagation on V. cholerae with p-rta or PLE11 derivatives

ICP1 production in the presence of an inducible PLE gene was done as described previously in ref. ¹⁰. An overnight culture of the relevant V. cholerae strain was back diluted to OD₆₀₀ = 0.05 in 50 ml of LB medium and incubated at 37 °C with aeration. For V. cholerae strains containing either pEV or p-rta, 1.25 μg ml⁻¹ chloramphenicol was supplemented in media and at OD₆₀₀ = 0.2, 1 mM IPTG and 1.5 mM theophylline were added for induction, or an equivalent volume of sterile water was added in the uninduced control. All cultures were grown to OD₆₀₀ = 0.3. and then infected with ICP1 strains at an MOI of 2.5. On lysis (roughly 90 min), the culture was treated with 0.25 units ml⁻¹ benzonase and 10% chloroform for 5 min with shaking, centrifuged at 5,000g for 15 min at 4 °C. The supernatant was centrifuged at 26,000g for 90 min at 4 °C. The resulting phage pellet was resuspended in phage buffer (50 mM Tris–HCl pH 8.0, 100 mM NaCl, 10 mM MgSO₄, 1 mM CaCl₂) by rocking overnight at 4 °C, treated with chloroform (1:1) and the aqueous layer was collected for further analysis. For TEM analysis of ICP1_2006_Dha_E_∆CRISPR-∆cas2-3 propagated in the presence of p-rta± inducer (Fig. 2d), one biological replicate was performed; three biological replicates of TEM analysis of this phage propagated in the presence of an induced empty vector or induced p-rta are presented in Extended Data Fig. 3e and Supplementary Fig. 2d. For TEM analysis of ICP1 2006_Dha_E derivatives with TMP substitutions, phages were propagated in the presence of p-rta± inducer (Extended Data Fig. 3f), one biological replicate was performed for each of the three mutant derivatives. For TEM analysis of particles produced by ICP1 CRISPR–Cas(+) on infection of V. cholerae PLE11 or PLE11∆rta, three biological replicates were performed as shown in Extended Data Fig. 9b and Supplementary Fig. 4.

Purification of ICP1 and PLE11 virions

ICP1 and PLE11 virions were generated by infecting 200 ml of mid-log culture of PLE(−) V. cholerae and PLE11(+) V. cholerae, respectively, at MOI of 2.5 with ICP1_2006_Dha_E∆CRISPR∆cas2-3. After culture lysis, 0.25 units ml⁻¹ benzonase and 10 ml of micellar mix (4 ml of chloroform, 2 ml of methanol, 25 mM sodium citrate and 10 mM sodium deoxycholate) were added and mixed for 5 min at room temperature. The lysate was centrifuged at 5,000g for 10 min at 4 °C; the aqueous supernatant was centrifuged at 26,000g for 90 min at 4 °C. The supernatant was discarded, and the pellet was recovered in phage buffer by rocking overnight at 4 °C. The resuspended pellet was chloroform treated (1:1), and the aqueous layer containing virions was collected. Next, to separate particles from aggregates and free proteins, the concentrated phage stock was subjected to bench-top gel filtration chromatography as follows: 3 g of BioRad P-10 Gel Fine 45-90 was hydrated in 50 mM Tris-Cl pH 8.0 at room temperature and then stored at 4 °C until the next steps. BioRad Polyprep chromatography columns were packed with roughly 3 ml of the hydrated gel to a 1.5-inch bed height. The gel was equilibrated with 15 ml of phage buffer supplemented with benzonase (0.25 units ml⁻¹) using gentle syringe pressure. The concentrated phage stock (roughly 400 µl) was loaded on the resin under gravity, followed by roughly 3 ml of phage buffer with benzonase. The eluent was collected as 100 µl × 30 fractions, which were screened for virions using spot plate plaque assays for ICP1 and transductions for PLE11. The fractions with high titres were also visually inspected using TEM.

TEM

To stage the virions for imaging, copper mesh grids (Formvar/Carbon 300, Electron Microscopy Sciences) were loaded with 5 µl of samples for 1 min, washed with 5 µl of sterile ddH₂O for 15 s and stained with 1% uranyl acetate for 30 s. The absorbent paper was used to wick liquids between each step gently. The grids were imaged on a Tecnai-12 electron microscope at 120 kV. Uncropped TEM images can be found in Supplementary Fig. 5.

Particle measurements

The dimensions of the tails of ICP1 and PLE11 virions were measured using TEM images analysed using Fiji⁵²: the pixel distance was set to a scale distance (nm) using ‘Set Scale’. Tail sizes were measured in a straight line from neck to base using the ‘Analyze’ > ‘Measure’ option. Particle measurements were performed on three biological replicates of particle purifications, for a total of n = 170 particles measured for each ICP1 and PLE11.

Mass spectrometry

Three of the highest concentration fractions from gel filtration were pooled (totalling 4.95 × 10¹² TFU per ml for PLE11 virions and 3.00 × 10¹¹ PFU per ml for ICP1 virions: ICP1_2006_Dha_E∆CRISPR∆cas2-3), and each pool was denatured in Lamelli buffer. Then 40 µl of each sample was run on Any-Kd Mini-PROTEAN TGX Precast gel (BioRad). The gel was stained with GelCode Blue stain (Thermo Fisher) for visualization, and the lanes with samples were cut into 1-mm² pieces and destained. In-gel digestion and mass spectrometry analysis were conducted at the Vincent J. Coates Proteomics/Mass Spectrometry Laboratory at the University of California, Berkeley, USA. Briefly, gel pieces were washed twice with 50% acetonitrile (ACN) and 50 mM ammonium bicarbonate for 15 min with shaking and then dehydrated with 100% ACN for 5 min, followed by air drying for 5 min. Pieces were further dried with 10 mM Tris(2-carboxyethyl)phosphine hydrochloride and 40 mM chloroacetamide for 5 min at 70 °C. The gel was rehydrated in 50 mM HEPES pH 8.0 in a minimal volume enough to cover the surface. For digestion, rehydrated pieces were incubated with 1 µg Trypsin (1:50 dilution) for 1 h at room temperature, then supplemented with a minimal volume of 50 mM HEPES pH 8.0 and incubated overnight at 37 °C. Peptides were extracted stepwise in treatment with 25% and 100% ACN. Peptide extracts were concentrated to 30–60 µl and acidified with formic acid.

The digested peptides were analysed by online capillary nano liquid chromatography with tandem mass spectrometry using a 25 cm reversed-phase column fabricated in-house (50 µm inner diameter, packed with ReproSil-Gold C18-1.9 μm resin, Dr. Maisch) equipped with laser-pulled nanoelectron spray emitter tip. Peptides were eluted at 100 nl min⁻¹ on a Thermo Fisher Easy-nLC1200 using a 140-min linear gradient of 2–40% buffer B (buffer A, 0.05% formic acid in water; buffer B, 0.05% formic acid in 95% ACN in water). Peptides were ionized using a FLEX ion source and analysed on a Fusion Lumos Tribrid Orbitrap Mass Spectrometer (Thermo Fisher Scientific), and data were acquired in Orbitrap mode with parameters as follows: MS1 resolution of 120,000 at 200 m/z and scan range of 350–1,600 m/z. The top 20 most abundant ions were fragmented through collision-induced dissociation with 35% normalized collision energy, activation q of 0.25 and a 2 m/z precursor isolation width. Dynamic exclusion was set to a 30-s repeat duration, 20-s exclusion and a single repeat count. RAW files were analysed with PEAKS (Bioinformatics Solutions Inc.) using semi-specific cleavage at R (Arg) and K (Lys) (up to 4 missed cleavages), a precursor mass tolerance of 15 ppm and fragment mass tolerance of 0.5 Da. Variable modifications included methionine oxidation, and cysteine carbamidomethylation was fixed. Peptide hits were filtered using a 1% FDR, with proteins requiring at least 2 unique peptides and 1% FDR. Label-free quantitation was performed with PEAKS using default parameters, except for selecting the top 2 peptides per protein with a minimum abundance of 10 × 10⁴ and normalization based on total ion chromatogram across technical replicates. Data presented in Fig. 4e are based on one biological replicate (note that the western blot analysis of particles was performed on a separate biological preparation of particles).

Custom antibodies and western blotting

The antibody against the ICP1’s capsid protein (α-capsid^ICP1) was used as described in ref. ¹⁹. Polyclonal antibodies against ICP1’s Gp78 (α-BhuB^ICP1), ICP1’s TMP (α-TMP^ICP1) and PLE11’s TMP (α-TMP^PLE11) were generated in rabbits by GenScript. Given the intrinsic toxicity and challenges in expressing TMPs, regions of least relative disorder were predicted using the modelled structure generated by ColabFold⁴⁶ and sequence-based predictions from DEPICTER2 (ref. ⁵³). The codon-optimized DNA sequence for M252–A338 from ICP1’s TMP and M205–N319 from PLE11’s TMP were cloned in pET30a(+), 6×-His-tagged proteins were purified using Ni-affinity chromatography and used as antigens.

To analyse the presence of listed proteins in purified virions, samples were prepared in Lamelli buffer, denatured and loaded onto SDS–PAGE in duplicate. The proteins were transferred to 0.22 polyvinyl difluoride membrane using TransBlot Turbo (BioRad) at 1.5 V for 7 min, and the blot was blocked with 0.5% skim milk in Tris-buffered saline with Tween (TBST) (20 mM Tris-Cl pH 7.6, 150 mM NaCl 0.1% Tween 20). The blot was cut at 50 kDa and 25 kDa. The pieces larger than 50 kDa were probed with α-TMP^ICP1 and α-TMP^PLE11 (diluted 1:1,500), blots from 50 kDa to 25 kDa were probed with α-capsid^ICP1 (diluted 1:1,500) and pieces smaller than 25 kDa were probed with α-BhuB^ICP1 (diluted 1:3,000). The primary antibodies were diluted in TBST with 2% bovine serum albumin. Blots were incubated in primary antibodies overnight at 4 °C, washed three times in TBS (20 mM Tris-Cl pH 7.6, 150 mM NaCl) and incubated in goat anti-rabbit secondary antibodies in TBST with 2% bovine serum albumin for 45 min, and then washed twice in TBS. Blots were developed in ECL chemiluminescence reagent (BioRad) and imaged on BioRad ChemiDoc. Data presented in Fig. 4d are based on one biological replicate (note that the mass spectrometry analysis of particles was performed on a separate biological preparation of particles). Uncropped western blot images can be found in Supplementary Fig. 5.

For detection of Rta-3xFLAG, overnight cultures of V. cholerae PLE11 or PLE11:rta-3xFLAG were diluted to OD₆₀₀ = 0.05 in 50 ml of LB medium and grown to OD₆₀₀ = 0.3, 25 ml of the cultures were retrieved into equal volumes of chilled methanol. The remaining 25 ml of the cultures were infected with either ICP1_2006_Dha_E CRISPR–Cas(+) or ICP1_2006_Dha_E_∆CRISPR∆cas2-3 at an MOI of 2.5. At 16 min postinfection, the cultures were collected into equal volumes of chilled methanol. Cells were gathered at 5,000g for 10 min at 4 °C and washed with 1× PBS. Cells were lysed in buffer containing 50 mM Tris-Cl, 5 mM SDS and 1 mg ml⁻¹ lysozyme. Total protein estimation for all samples was done using Pierce BCA protein assay kit. For western blot analysis, 70 µg of total protein was loaded onto SDS–PAGE gel and, after transfer, the blot was cut below 25 kDa, then probed with α-FLAG (GenScript, diluted 1:2,000) and further probed with goat anti-rabbit secondary antibodies as described above.

Transduction assays

Donor strains of V. cholerae with PLE marked with a kan^R gene downstream of the last PLE gene were grown to saturation. The donor strains were back diluted to OD₆₀₀ = 0.05 in 2 ml of LB medium, grown to OD₆₀₀ = 0.3 and infected with ICP1_2006_Dha_E_∆CRISPR∆cas2-3 at MOI of 2.5. After 5 min, the unbound phage was washed off, and the infected cells were resuspended in fresh media. After roughly 20 min, the lysates were collected, treated with 20 µl of chloroform and centrifuged at 5,000g for 15 min at 4 °C. As necessary, 1.25 µg ml⁻¹ chloramphenicol was added to the media for the growth of donor strains but was excluded from the washing and resuspension steps. Strains harbouring pEV or p-rta were induced at OD₆₀₀ = 0.2 with 1 mM IPTG and 1.5 mM theophylline. In parallel, the recipient strain, V. cholerae(∆lacZ::spec), was grown for 6–7 h and supplemented with 10 mM MgSO₄ right before adding the lysate for transductions. Next, 20 µl of supernatant was mixed with 180 µl of the recipient. Tenfold dilutions of the mix were prepared in LB medium and incubated at 37 °C with shaking for 20 min. All dilutions were then plated on LB agar plates containing kanamycin and spectinomycin, and the resulting colonies were counted to calculate transducing-forming units per millilitre of the donor lysate. The transduction efficiency was calculated relative to wild-type PLE, wild type with empty vector or empty vector controls as applicable. Transduction efficiency was calculated from three biological replicates.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.