Functional Analysis of Bacteriophage Immunity through a Type I-E CRISPR-Cas System in Vibrio cholerae and Its Application in Bacteriophage Genome Engineering

ABSTRACT

The classical and El Tor biotypes of Vibrio cholerae serogroup O1, the etiological agent of cholera, are responsible for the sixth and seventh (current) pandemics, respectively. A genomic island (GI), GI-24, previously identified in a classical biotype strain of V. cholerae, is predicted to encode clustered regularly interspaced short palindromic repeat (CRISPR)-associated proteins (Cas proteins); however, experimental evidence in support of CRISPR activity in V. cholerae has not been documented. Here, we show that CRISPR-Cas is ubiquitous in strains of the classical biotype but excluded from strains of the El Tor biotype. We also provide in silico evidence to suggest that CRISPR-Cas actively contributes to phage resistance in classical strains. We demonstrate that transfer of GI-24 to V. cholerae El Tor via natural transformation enables CRISPR-Cas-mediated resistance to bacteriophage CP-T1 under laboratory conditions. To elucidate the sequence requirements of this type I-E CRISPR-Cas system, we engineered a plasmid-based system allowing the directed targeting of a region of interest. Through screening for phage mutants that escape CRISPR-Cas-mediated resistance, we show that CRISPR targets must be accompanied by a 3′ TT protospacer-adjacent motif (PAM) for efficient interference. Finally, we demonstrate that efficient editing of V. cholerae lytic phage genomes can be performed by simultaneously introducing an editing template that allows homologous recombination and escape from CRISPR-Cas targeting.

IMPORTANCE Cholera, caused by the facultative pathogen Vibrio cholerae, remains a serious public health threat. Clustered regularly interspaced short palindromic repeats and CRISPR-associated proteins (CRISPR-Cas) provide prokaryotes with sequence-specific protection from invading nucleic acids, including bacteriophages. In this work, we show that one genomic feature differentiating sixth pandemic (classical biotype) strains from seventh pandemic (El Tor biotype) strains is the presence of a CRISPR-Cas system in the classical biotype. We demonstrate that the CRISPR-Cas system from a classical biotype strain can be transferred to a V. cholerae El Tor biotype strain and that it is functional in providing resistance to phage infection. Finally, we show that this CRISPR-Cas system can be used as an efficient tool for the editing of V. cholerae lytic phage genomes.

INTRODUCTION

Vibrio cholerae is a Gram-negative facultative pathogen that causes the acute diarrheal disease cholera. The current cholera pandemic, the seventh in recorded history, began in 1961 and is caused by V. cholerae O1 of the El Tor biotype (1). This pandemic has affected much of the developing world, including many countries in Asia and Africa and most recently in the Caribbean (2). The sixth cholera pandemic, however, was caused by V. cholerae O1 of the classical biotype. Classical biotype strains declined following the emergence of El Tor strains and are now believed to be extinct, after last being seen in 1990 in Bangladesh (3). The mechanisms underpinning the replacement of the classical biotype and the subsequent evolutionary success of the El Tor biotype remain unknown. Phenotypically, classical and El Tor biotype strains are differentiated by sensitivity to polymyxin B, the ability to cause agglutination of chicken erythrocytes, and sensitivity to some lytic bacteriophages (1). Comparative genomics have demonstrated that classical and El Tor strains differ by ∼20,000 single nucleotide polymorphisms (SNPs) within the core genome (4). In addition to alterations in the core genome, extensive lateral gene transfer events play a critical role in the emergence and reemergence of pathogenic clones of V. cholerae (5, 6). Comprehensive genomic analyses support the hypothesis that classical and El Tor biotype strains are independent derivatives of a common ancestor (4, 7). Members of the two biotypes do have several genomic islands (GIs) in common (7), including Vibrio pathogenicity island 1 (VPI-1), encoding the toxin-coregulated pilus, which is required for intestinal colonization (8); VPI-2; and an island containing genes for O1 antigen biosynthesis. Seventh pandemic El Tor strains harbor two unique GIs: the Vibrio seventh pandemic island 1 (VSP-1) and VSP-2 (9). Chun et al. (7) identified 73 GIs (where a GI was defined as a genomic region composed of at least five open reading frames [ORFs] where transfer is obvious from comparative genomics) present in a collection of 23 V. cholerae strains isolated between 1910 and 2004. One GI, designated GI-24, was shown to be present in the single classical strain (O395) included in the study but absent in El Tor strains (7). On the basis of bioinformatic analysis, GI-24 was described to be a putative prophage harboring clustered regularly interspaced short palindromic repeat (CRISPR)-associated proteins (Cas proteins).

CRISPR-Cas systems provide sequence-specific adaptive immunity against invading nucleic acids. CRISPR-mediated immunity against bacteriophages was first documented in Streptococcus thermophilus in 2007 (10), and since then, many studies have contributed toward a detailed molecular understanding of these sophisticated immune systems and their recent application in genome engineering (recently reviewed in references 11 to 13). CRISPR arrays are composed of short, partially palindromic DNA repeats and alternating invader-derived variable sequences termed spacers. Cas genes, typically located in proximity to the CRISPR array, encode the necessary machinery to direct immunity. CRISPR-mediated immunity occurs following the transcription and processing of the CRISPR locus into smaller CRISPR RNAs (crRNAs). These mature crRNAs complex with Cas proteins to direct the cleavage of target nucleic acid that has complementarity to the crRNA. As an adaptive immune system, the CRISPR-Cas system is capable of acquiring new spacers derived from foreign invaders; new spacers are acquired adjacent to the leader (an AT-rich sequence proximal to the CRISPR locus), providing an immunological record of previously encountered invading nucleic acid (10). CRISPR-Cas systems are classified into 3 main types and 11 subtypes largely on the basis of the cas gene composition (14). Most systems target foreign DNA (15); however, type III systems have been shown to target DNA and RNA (16). In addition to the requirement for crRNA base pairing to the target (referred to as a protospacer), some systems require a protospacer-adjacent motif (PAM) (17, 18). Specifically, type I systems require a 2- to 3-bp PAM downstream of the target, type II systems require a 4- to 5-bp PAM upstream of the target, and type III systems appear to lack the PAM requirement (12, 19). CRISPR-Cas systems have been identified in ∼84% of archaea and ∼45% of bacteria (20). Uniquely, a functional CRISPR-Cas system was recently found in a group of O1-specific V. cholerae phages, called the ICP1-related phages (21). Experimental evidence demonstrated that the ICP1 CRISPR-Cas system acts to degrade a phage-inducible chromosomal island-like element (PLE) that otherwise blocks phage infection through an unknown mechanism. The origin of the ICP1 CRISPR-Cas system (which is a type I-F system) is not known, but it is unrelated to the V. cholerae CRISPR-Cas system identified in GI-24 (which is a type I-E system [21]).

The in silico identification of an intact CRISPR-Cas system does not necessarily mean that it is functioning as an immune system, and experimental evidence demonstrating protection from invading nucleic acid is limited to a relatively few studies (22). Some systems, such as the Escherichia coli type I-E system, are repressed under laboratory conditions (23), and other systems appear to play alternative roles in host physiology (24). The bioinformatic description of a CRISPR-Cas system in V. cholerae (7, 25) warrants functional analysis, particularly given the important role that phages (26, 27) and other mobile genetic elements (6) play in the evolution of V. cholerae. Here, we assessed the prevalence of the CRISPR-Cas system in a large collection of geographically and temporally disparate V. cholerae strains and provide evidence that this system is restricted to classical biotype strains where spacer diversity between strains is consistent with its role in phage defense. We experimentally evaluated the function of the V. cholerae CRISPR-Cas system as an immune system against lytic phage infection. We determined the sequence requirements for system activity both by targeting divergent sequences between related phages and by assessing viruses that escape CRISPR-Cas-mediated resistance. Furthermore, in light of recent advances in engineering phages using CRISPR-Cas (28, 29), we adapt this CRISPR-Cas system for efficient engineering of lytic V. cholerae phage genomes.

MATERIALS AND METHODS

Bacterial strains, phages, and culture conditions.

The bacterial strains, phages, and plasmids used in this study are listed in Table 1. Bacteria were grown at 37°C on lysogeny broth (LB) agar or in LB broth. Medium was supplemented with ampicillin (Amp; 50 μg/ml), kanamycin (Kan; 75 μg/ml), spectinomycin (Spc; 100 μg/ml), chloramphenicol (Cm; 25 μg/ml for E. coli and 2.5 μg/ml for V. cholerae), and/or streptomycin (Sm; 100 μg/ml) when appropriate. For expression of the P_tac promoter, bacteria were grown to an optical density at 600 nm (OD₆₀₀) of approximately 0.25 and induced with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for 20 min at 37°C with aeration. Phage susceptibility was determined using the soft agar overlay method; briefly, phages were allowed to adsorb to V. cholerae for 10 min at room temperature, and then the mixture was added to molten LB soft agar (0.5%) and immediately poured into a petri dish (16 by 15 mm). Plaques were counted after incubation overnight at 37°C. The efficiency of plaquing (EOP) was used to measure the effect of the CRISPR-Cas system as well as the efficiency of the editing templates. The EOPs were determined by dividing the phage titer obtained on the CRISPR-targeting strain by the phage titer obtained on the nontargeting parental strain. In order to characterize escape phages, boiled plaques served as a DNA template for PCR to amplify the region of interest, and this region was confirmed by DNA sequencing.

TABLE 1.

Bacterial strains, plasmids, and phages used in this study

Strain, plasmid, or phage	Relevant characteristicsa	Source or reference
V. cholerae strains
O395	Serogroup O1, classical biotype, Smr, isolated in Bangladesh in 1965	51
E7946	Serogroup O1, El Tor biotype, Smr	52
KS802	E7946 with CRISPR-Cas from O395 ΔlacZ::Spcr	This study
KS916	E7946 with Ptac-CRISPR-Cas from O395, Kanr ΔlacZ::Spcr	This study
A50	Serogroup O1, classical biotype, Sms, isolated in Bangladesh in 1963	4
A57	Serogroup O1, classical biotype, Sms, isolated in India in 1980	4
A68	Serogroup O1, classical biotype, Sms, isolated in Egypt in 1949	4
A111	Serogroup O1, classical biotype, Sms, isolated in 1990	4
A50 ΔCascade	A50 with unmarked deletion of Cascade (Cse1, Cse2, Cas6e, Cas7, and Cas5)	This study
El Tor CRISPR-CasA50	E7946 with CRISPR-Cas from A50, Kanr	This study
El Tor CRISPR-CasA50 ΔCascade	E7946 with CRISPR-Cas from A50 with unmarked deletion of Cascade (Cse1, Cse2, Cas6e, Cas7, and Cas5), Kanr	This study
Plasmids
pMMB67EH	IncQ broad-host-range cloning vector, Ampr	35
pCP-T1	RP4-containing plasmid, harbors CP-T1 protospacer in MCS, Cmr	This study
pCRISPR	Derived from pMMB67EH, harbors the V. cholerae CRISPR array under control of an IPTG-inducible promoter and an insertion site for editing templates, Ampr	This study
pTargetϕorf10	pCRISPR cloned with an anti-orf10 spacer directed against ICP1	This study
pTargetϕorf91	pCRISPR cloned with an anti-orf91 spacer directed against ICP1	This study
pTargetϕcas1	pCRISPR cloned with an anti-cas1 spacer directed against ICP1_2011_A	This study
pTargetϕcas2-31	pCRISPR cloned with an anti-cas2-cas3 spacer directed against ICP1_2011_A	This study
pTargetϕcas2-32	pCRISPR cloned with an anti-cas2-cas3 spacer directed against ICP1_2011_A	This study
pTargetϕcas2-33	pCRISPR cloned with an anti-cas2-cas3 spacer directed against ICP1_2011_A	This study
pTargetϕcsy2	pCRISPR cloned with an anti-csy2 spacer directed against ICP1_2011_A	This study
pTargetϕcas6f	pCRISPR cloned with an anti-cas6f spacer directed against ICP1_2011_A	This study
pTargetϕCR1S2	pCRISPR cloned with an anti-CR1S2 spacer directed against ICP1_2011_A	This study
pTargetϕcas2-33-EdGFP	pCRISPR cloned with an anti-cas2-cas3 spacer directed against ICP1_2011_A and a 1,516-bp editing template to replace cas2-cas3 with gfp (a 2,670-bp deletion replaced by 717-bp gfp gene)	This study
pTargetϕcas2-33-Ed	pCRISPR cloned with an anti-cas2-cas3 spacer directed against ICP1_2011_A and a 922-bp editing template to delete cas2-cas3 (a 2,670-bp deletion)	This study
pTargetϕcas1-Ed	pCRISPR cloned with an anti-cas1 spacer directed against ICP1_2011_A and a 526-bp editing template to delete the 33-bp protospacer in cas1	This study
Phages
CP-TI	Generalized transducing phage	53
ICP1	O1-specific phage of V. cholerae O1, unclassified member of the Myoviridae family	44
ICP1_2004_A	CRISPR-Cas-positive ICP1 isolate	44
ICP1_2005_A	CRISPR-Cas-positive ICP1 isolate	44
ICP1_2011_A	CRISPR-Cas-positive ICP1 isolate	21
ICP1_2011_A ΔS9	Spontaneous deletion of spacer 9 in phage CRISPR array	21

^aSm, streptomycin; Spc, spectinomycin; Amp, ampicillin; Kan, kanamycin; Cm, chloramphenicol.

CRISPR analysis.

Whole-genome sequencing for V. cholerae classical biotype strains A50, A57, A68, and A111 (4) (Table 1) was performed using genomic libraries generated as described previously (30) and sequenced using an Illumina HiSeq 2000 sequencing system. Genomes were assembled with CLC Genomics Workbench software (version 6.8; CLC Bio, Denmark). Per base coverage values for CRISPR-Cas-containing contigs ranged from 51 to 77 times. The CRISPRFinder program (20) was used to identify CRISPRs and extract repeat and spacer information. The CRISPRtionary program (20) was used to assign each spacer as unique or shared between strains for graphical representation in Fig. 1A. The sequence of each spacer was compared with the sequences in the NCBI sequence database using the BLASTN program, and only spacers with 100% matches were included in the alignment of protospacer flanking regions to identify the PAM sequence. Sequence logos were generated using WebLogo (31).

FIG 1.

Genetic context and features of the V. cholerae CRISPR-Cas system. (A) Graphic representation of CRISPR spacers in five V. cholerae classical biotype strains. Strain names are given on the left. Spacers are represented as numbered rectangles, and repeats are not shown. Numbered spacers with targets in known phages are highlighted in gray. Black rectangles represent unique spacers present in the strain indicated. Otherwise, identical spacers shared between strains are shown as rectangles with identical color schemes (combination of character color and background color), whereas different color combinations represent distinguishable spacers. Missing spacers are indicated by squares with crosses. Arrays are oriented with respect to the leader sequence (L) located on the left. (B) The CRISPR-Cas system found in classical biotype strains is part of a 16.9-kb genomic island integrated between genes equivalent to VC0289 and VC0290 in the V. cholerae El Tor strain N16961. The colored arrows indicate the genes of the genomic island, the integrase (int) and cas genes are indicated, and genes encoding hypothetical proteins are green and unlabeled for simplicity. The CRISPR locus is shown as black and gray rectangles. The location of the kanamycin resistance gene that was inserted to select for El Tor biotype transformants that acquired GI-24 via natural transformation is indicated. (C) Genetic architecture of the construct used for overexpression of CRISPR-Cas in V. cholerae El Tor. (D) The PAM sequence of the CRISPR system of V. cholerae classical biotype strains. The PAM sequence was identified after alignment of the flanking sequences of all known targets (100% match) of spacers found in the CRISPR systems of the five classical strains in panel A. No sequence conservation was detected 5′ of the target (data not shown). The sequence logo was generated using WebLogo (31).

Generation of mutant strains.

Splicing by overlap extension (SOE) PCR was used to generate all PCR constructs. Primer sequences are available on request. The kanamycin resistance marker was introduced into the classical strains O395 and A50 near the CRISPR-Cas system using pCVD442-lac (32) (Fig. 1B). V. cholerae E7946 was made competent by growth on chitin as described previously (33), ∼2 μg purified genomic DNA from the kanamycin-resistant classical derivative (described above) was added, and the mixture was incubated at 30°C overnight and then plated onto LB plates plus Kan. The desired incorporation of the entire classical CRISPR-Cas system in E7946 was confirmed by PCR; in the case of the parent strain used in our plasmid-based CRISPR assays, the expected sequence of the incorporated CRISPR-Cas system was confirmed by whole-genome sequencing using genomic libraries generated as described previously (30) and sequenced using an Illumina HiSeq 2000 sequencing system. The kanamycin resistance cassette was removed using cotransformation (34) of the wild-type locus with a selected product which replaced lacZ with an Spc resistance (Spc^r) marker (generating strain KS802). The overexpression construct for P_tac-CRISPR-cas was generated by SOE PCR using a fragment derived from a kanamycin-resistant derivative of a Tn10 plasmid (33) and was introduced via natural transformation to generate strain KS916. All experiments with our plasmid-based CRISPR system were in the background of strain KS802 or KS916 (depending on the presence of the overexpression construct, as indicated below).

Generation of recombinant plasmids.

The low-copy-number vector pMMB67EH (35) was used as a backbone to generate pCRISPR. Gibson assembly (New England BioLabs) was used to remove two BsaI restriction sites in the vector backbone and replace the multiple-cloning site (MCS) downstream of the P_tac promoter with a new MCS harboring only XbaI and SacI restriction sites. A synthetic V. cholerae CRISPR locus including the predicted leader sequence with BsaI and BseRI spacer insertion sites (adapted from the sequence described in reference 36) was obtained through Integrated DNA Technologies. The CRISPR locus was cloned into the modified pMMB67EH (described above) via digestion with XbaI and SacI (New England BioLabs) and subsequent ligation with T4 DNA ligase (New England BioLabs), as specified by the manufacturer. Gibson assembly (New England BioLabs) was used to insert a new MCS for cloning of editing templates to yield pCRISPR. New spacers were constructed by oligonucleotide annealing and phosphorylation with T4 polynucleotide kinase (New England BioLabs). Phosphorylated annealed oligonucleotides were cloned into pCRISPR that had previously been cut with BsaI or BseRI and dephosphorylated with Antarctic phosphatase (New England BioLabs). Expression vectors were transferred into V. cholerae by conjugation with E. coli S17, and Kan^r Amp^r colonies were selected.

Conjugation assays.

For transfer of pCP-T1 from E. coli into V. cholerae, 500 μl of the donor strain and 500 μl of the recipient strain at an OD₆₀₀ of 1 were mixed 1:1 and concentrated 10-fold before they were spread onto a sterile filter (0.8-μm-pore-size cutoff; Millipore) placed on an LB plate. Matings were incubated at 37°C for 2 h. The bacteria were removed from the filter by vortexing in 1 ml LB and then plated onto medium selective for donors or transconjugants. The conjugation efficiency was calculated as the number of transconjugants per donor.

RESULTS

CRISPR analysis and distribution in V. cholerae.

A previous analysis of 23 V. cholerae strains identified a GI (GI-24) predicted to encode a CRISPR-Cas system in the classical biotype strain O395 (7). The genomic organization of GI-24 is presented in Fig. 1B. The average GC content of GI-24 (excluding CRISPR spacers) is 45.6%, only slightly lower than that of the rest of the chromosome (47.8%). More recently, whole-genome sequence data were reported for an additional 159 V. cholerae strains (4). Comparative genomics identified a single phyletic lineage comprised of all V. cholerae classical biotype strains included in the analysis (lineage 1 [L1]). Using this data set, along with whole-genome sequences of >60 additional El Tor biotype strains from Haiti (27) and Bangladesh (37), we found that this CRISPR-Cas system is strictly limited to classical biotype strains (Table 2). Detailed analysis of the spacer composition of classical biotype strains was not possible due to the repetitive nature of the CRISPR array and the relatively low sequence coverage obtained in the previous study (4). To facilitate a further in-depth comparative analysis, we resequenced four classical biotype strains isolated between 1949 and 1990 (Table 1) and compared the CRISPR-Cas systems to the CRISPR-Cas system of strain O395 (isolated in 1965). The five classical strains analyzed had the same 28-bp consensus repeat sequence (GTCTTCCCCACGCAGGTGGGGGTGTTTC) and spacers of conserved lengths (33 bp, with the exception of the terminal spacer in all strains, which was 34 bp). The results of the analysis of the consensus repeat sequence using the CRISPRmap web server (38) were consistent with the designation of this CRISPR-Cas system as a type I-E system (21). The cas genes were 100% identical at the nucleotide level between strains, and no debilitating mutations (nonsense mutations or frameshifts) were observed. The number of spacers in a particular strain ranged from 15 to 50 (Fig. 1A). We observed that spacers at the trailing end (considered the oldest spacers) were highly conserved between strains. Conversely, in three of the five strains, the spacers closest to the leader region were unique (Fig. 1A). The observed spacer diversity at the leading edge is consistent with each of these strains actively acquiring new and diverse spacers. Furthermore, the sequences of all spacers with identified targets (22 of the 78 total unique spacers) matched the sequences of Vibrio phages (including CP-T1 [39], phi2 [GenBank accession number KJ545483], and X29 [GenBank accession number KJ572845]) (Table 3), consistent with its function as a protective immune system against phage infection. We identified the PAM to be a conserved TT motif immediately 3′ of the spacer targets (Fig. 1D), which differs from the 3-bp CTT motif identified in type I-E systems in E. coli and Pseudomonas aeruginosa (18).

TABLE 2.

V. cholerae strains harboring cas genes with homology to those found in the type I-E system found in GI-24 in V. cholerae O395^a

V. cholerae strain	Lineage, biotype
GP8	L1, classical
GP16	L1, classical
A46	L1, classical
A49	L1, classical
A50	L1, classical
A51	L1, classical
A57	L1, classical
A59	L1, classical
A60	L1, classical
A61	L1, classical
A66	L1, classical
A68	L1, classical
A70	L1, classical
A76	L1, classical
A103	L1, classical
A111	L1, classical
A279	L1, classical
A389	L1, classical

^aStrain names and lineage information are from Mutreja et al. (4).

TABLE 3.

CRISPR spacers of different V. cholerae classical biotype strains show identity to known Vibrio phages

Annotated spacera	Sequence	Matching phage(s) (% identity)b
O395:1	CAAATGGCTCAATGCGCGATTGACTACGTTACT	CP-T1 (100), 24 (100)
O395:2	CAATCAACGCACTAGACAACGCCCAAATGAACC	—
O395:3	CAACGCTATGCCCGCTAAAATCAGTAAACAAGA	—
O395:4	TATATTTCGCTCCCGTCTCGTCAACTAAAATCA	—
O395:5	CTTCAAATAAGTAACCAGCCTCTGACGCTGTTA	—
O395:6_A50:4_A57:1	CACCGCTAATCATGGTGGAACGAACGCCATCAA	—
O395:7_A50:5_A57:2	TGATTTTGGAAGTAATGGGAACTGAGCGTTAAG	—
O395:8_A50:6	CCAAAAACCTACGCGGTTTTAAATGGATTCGAC	—
O395:9_A50:7	TATCTTGGTTTTGCAGGTTGTTAATCTCAGCGT	—
O395:10_A50:8	TGATTGGTTCCAGTTTATGACAAGAACCAACAC	—
O395:11	TACATTGGCAAGACGTTTGTTTTTCGCTGTGTA	—
O395:12_A50:9	CGACTTTTGCATCATCGATGTACGGAACGCTAG	—
O395:13_A50:25	CACTGAGATTGCGTGTCGCCGACTTGCGCTTGC	—
O395:14_A50:10	TAGACTATCAATGTGCGCTTGCAAGTCTTTTAA	—
O395:15_A50:11	TTGATCGCTCTGAAATTGTGACTTGTTTTGTTA	—
O395:16_A50:12	TAGATCTTAATTGTTCGCGTTGAATGGGAAATT	—
O395:17_A50:13	CAGATTGTAGATAAGCAGGAGACTGCCCACCAG	—
O395:18_A50:22_A68:1_A111:4	CGGTTGATAAAACGCTGCGTAAGTTTTTCGAAG	—
O395:19_A50:23_A68:2_A111:5	TAAAACTTCATAGATTGTTGCCTCCATTGTTTC	—
O395:20_A50:24_A68:3_A111:6	TACGTCAACGGACAAACCAAAACCGAATGGAAG	phi2 (100), X29 (100)
O395:21_A111:12	TAAGTAACGCTGCTACTGCCCTGAGCTAGTACC	—
O395:22_A50:26_A111:13	TACTGGCCGATGAGGTGGACCGCTACGGCTTCA	X29 (100), phi2 (97)
O395:23_A50:27_A68:7_A111:14	TCACCCAGCACATTACCACCCATGATCAGCGTT	X29 (100), phi2 (97)
O395:24_A50:28_A68:8_A111:15	CGAGTGGGTTTAGGTTGTAGGTTGCACATACGC	—
O395:25_A50:29_A68:28_A111:35	TTCGAAAAGCTATTAGGCGGCATAACCACAGTT	—
O395:26_A50:30_A68:29_A111:36	CAACAAACCAGCCACTTTGCATTTTGTAGCAGA	—
O395:27_A50:31_A57:3_A68:30_A111:37	TGGAAGTTATTTATATTGGACCTGATTGCACGG	—
O395:28_A50:32_A57:4_A68:31_A111:38	TACTTTGGGCCTTGTTTTATGTGCCAGTGCCCG	—
O395:29_A50:33_A57:5_A68:32_A111:39	CCACTATTTCAATAATCGGCGTAGCTCCGACTG	—
O395:30_A50:34_A57:6_A68:33_A111:40	CGTAATCTTCTTTGTCTGAGTAATCCAAAATAC	—
O395:31_A50:35_A57:7_A68:34_A111:41	TCATGGCAGCGATATTTGTTTTACCCTTCATAC	—
O395:32_A50:36_A57:8_A68:36_A111:43	CATCTGGAACGCACTCAAGCGGCAATCCGAAGT	—
O395:33_A57:9_A68:37_A111:44	CAGCAATGTGTTATCCAATGCGAAAGCGCCGTT	—
O395:34_A57:10_A68:38_A111:45	TATTTTGCGTATCACCGAAGCGCTGTAGGTTAT	—
O395:35_A50:37_A57:11_A68:39_A111:46	TACGTCCAGCATTACCGCCGCGCCGTGTCGAGT	—
O395:36_A50:38_A57:12_A68:40_A111:47	TCCTCCCTGCTCATATGCCGTTAAAACTTTCTC	—
O395:37_A57:13_A68:41_A111:48	TCTGATGCGGCAAGCTCTTTCGCTAGCTCGATT	—
O395:38_A50:39_A57:14_A68:42_A111:49	TACAACATCCATGCAAGCGGCAAAGAATACAAA	—
O395:39_A50:40_A57:15_A68:43_A111:50	CCTCTTTAATCGAATTCCATAACGGTGACGTTAA	—
A50:1	TGATAAAACTTACGAATTGTTTATTAGCGATGG	CP-T1 (94), 24 (94)
A50:2	TATAGAGGATCACCATAATATTAACGTAAAAAT	CP-T1 (100), 24 (100)
A50:3	CACTGACTCGCCAAGCTTCGCCACCGCTTCTAG	—
A50:14	CGCTTCTGTAGAGGTGATGGGTCCAAAGATGTT	phi2 (100), X29 (100)
A50:15	TACATTCGTGATATCAGCGGATGCGCTTGGTCA	phi2 (100), X29 (100)
A50:16	CAAGGAAATTTGCTACCAAAAGACGTGATCGAG	phi2 (100), X29 (100)
A50:17	CCAAGCAGCAATAAAAATGGAGAGCAATCCTAT	phi2 (100), X29 (100)
A50:18	CCAAAGCAATTCAACAGAGGCCATCAATCGCCT	—
A50:19	TATTGCATTGAATTGCCCCACTCTTTGACCACA	—
A50:20	TCAAAATACTTCTCGCACATCCTGCAAGCGCTG	—
A50:21	CCGAATACAACGGGCAAGCCTGCGCTAACGTTC	phi2 (100), X29 (100)
A68:4_A111:7	CAGTTACCTGACCAGCAAACAAAGCTTCAGCAG	phi2 (100), X29 (100)
A68:5_A111:8	CGCTTTGCTATAAGCGCGGCATACTTTACCCCG	—
A68:6_A111:9	CCATCCTTTCAGCAAGGGTGAGAAGGCTTGCAA	phi2 (100), X29 (100)
A68:9_A111:16	TCAATTTTCAGTGCAGAGTTTGCAGCCGGAATG	X29 (100), phi2 (97)
A68:10_A111:17	TAAAGCTCTTTATACTCGCCTTGTCCAAACATC	X29 (100), phi2 (94)
A68:11_A111:18	TAAAGCCAAGCTCTAACGCTTCTGAGGCTGTCA	X29 (100), phi2 (94)
A68:12_A111:19	CAACTCACGCGCAGACTATGACGACATACACAG	—
A68:13_A111:20	TTGACTTGGCCGTGAAATCACCAGAACTACCGA	—
A68:14_A111:21	CAGACTCTTTTTTATCGCTAATACCAACTGGTG	—
A68:15_A111:22	TCTATCACCAATAGTTTCAGTCTCTTTTACTAT	—
A68:16_A111:23	TAAAAGAAGGGCTTCAGTAGATGGGCTTTCAGC	—
A68:17_A111:24	CACTGTACTGTCCAACGTGCCCAGTGTAGCGGT	—
A68:18_A111:25	TTGAGATAGAAAAATCAATATGACGCGCAGCAG	—
A68:19_A111:26	TTGGCTGTTTGAGTGTGAGGATATACTTGTCAT	—
A68:20_A111:27	TACGTTTTGGTGCAACCGATTTAAAGCTATCTT	—
A68:21_A111:28	CAGAATTCACAATGTTTTTGCGCTCATCAGGCG	—
A68:22_A111:29	CAACTCGCCTTTTTTCACGATGCCATCATAACT	—
A68:23_A111:30	CAAGTCGCTCTTTACACTCGATGCAGTGACGCG	—
A68:24_A111:31	TTCTACGCCTTGGAGCCGAAAAATTCAGAGAGC	—
A68:25_A111:32	CCAAACCAAGCAAGTTTGGAAGCCTGATACACT	—
A68:26_A111:33	CAAAACCAATGCTAAACTATACCAATAAACATT	—
A68:27_A111:34	CACACAAAACCGTGATACACTTCACATAACGAA	—
A68:35_A111:42	TAAAAACATTGATGCTTGCAGGATGGCTTTTCA	X29 (94), phi2 (94)
A111:1	CAAAAAGCCTCTGTTATGCCTTCTGATTTTTTG	phi2 (100), X29 (100)
A111:2	CAGAGCGTCAGGTCATCCGCAGTTTTGCAAACC	phi2 (100), X29 (100)
A111:3	TGAGAATACGCCTAGCGATACAATGTACCGTAT	phi2 (100), X29 (100)
A111:10	TGCCTATTGTGTATCGGTCCTGACCGCTTTGAT	X29 (100)
A111:11	CAGGCGAAACGTCTCTCGCACGTAAATCAGGTC	X29 (100), phi2 (94)

^aThe CRISPR spacer numbering for each strain is according to the numbering in Fig. 1A.

^bThe sequence of each spacer was compared with the sequences in the NCBI sequence database using the BLASTN program. Matches are indicated for sequences with 80 to 100% identity. CP-T1, 24, phi2, and X29 are all Vibrio phages. —, the spacer sequence does not match any sequence.

Functional characterization of the classical V. cholerae CRISPR-Cas system.

V. cholerae A50 has a single unique spacer (Fig. 1A, spacer 2) that targets phage CP-T1 with an intact TT PAM and 100% identity over the complete spacer sequence. We tested CP-T1 resistance in this strain in order to assess the contribution of CRISPR-Cas to phage resistance in V. cholerae. V. cholerae A50 was challenged with a high-titer preparation of CP-T1, and the efficiency of plaquing (EOP) compared to that of a nontargeting V. cholerae El Tor strain was calculated. We observed that plaque formation on A50 was below the limit of detection (Fig. 2A), consistent with CRISPR-Cas-mediated resistance. In order to confirm CRISPR-Cas-mediated resistance, we made a clean deletion of the genes encoding the essential targeting complex referred to as Cascade (CRISPR-associated complex for antiviral defense, composed of Cse1, Cse2, Cas6e, Cas7, and Cas5 [40–42]). A50 lacking Cascade was equally resistant to CP-T1, demonstrating that resistance was not CRISPR-Cas dependent (Fig. 2A). Since other phage resistance mechanisms may mask CRISPR-Cas activity in this particular strain background, we tested the conjugation of pCP-T1, a conjugative plasmid harboring the CP-T1 protospacer and PAM, from an E. coli donor into V. cholerae A50. The conjugation frequency of pCP-T1 was approximately 150-fold higher in A50 lacking Cascade (Fig. 2B), demonstrating that the CRISPR-Cas system in V. cholerae A50 was functional.

FIG 2.

CRISPR-Cas function in V. cholerae. (A) The EOP was determined for CP-T1 with each of the strains indicated compared to that with V. cholerae El Tor strain E7946. The dashed line represents the limit of detection of this assay. (B) The conjugation efficiency (number of transconjugants per donor cell) for pCP-T1 from an E. coli donor into the indicated V. cholerae strains was determined. Error bars represent standard deviations.

We next evaluated CRISPR-Cas activity in an El Tor background; we chose to use V. cholerae E7946 because it is CP-T1 sensitive. We hypothesized that we could move GI-24 and the associated CRISPR-Cas system from the classical biotype strain into E7946 via natural transformation, because sequences flanking GI-24 are homologous between biotypes (Fig. 1B; see Materials and Methods). V. cholerae E7946 became resistant to CP-T1 upon acquisition of GI-24 from A50 via natural transformation, and this resistance was dependent upon Cascade (Fig. 2A). These results demonstrate that the CRISPR-Cas system found in classical biotype strains of V. cholerae is functional as a phage defense system under laboratory conditions in an El Tor background.

Targeting of phages with an engineered CRISPR and analysis of sequence requirements.

We constructed a plasmid-based CRISPR system to more easily assess the CRISPR-Cas system function and sequence requirements. We designed a CRISPR array containing the predicted leader sequence with alternating repeat spacer units. The spacer units in this array contain BsaI and BseRI restriction sites to allow insertion of new spacers using annealed oligonucleotides (Fig. 3B). The CRISPR array was cloned into a low-copy-number vector (pMMB67EH) under the control of an IPTG-inducible promoter. To measure the efficiency of targeting of phages with this plasmid, we individually introduced 10 new spacers with targets present in ICP1-related V. cholerae phages, all of which had intact TT PAMs 3′ of the protospacer. Many of the plasmids that we generated targeted ICP1's CRISPR-Cas system since this system is dispensable for plaque formation on V. cholerae lacking PLE (21). Each targeting plasmid, which is designated with the annotation of the target site (e.g., pTargetϕcas1 has a spacer directed against cas1 in the phage), was introduced into the V. cholerae El Tor E7946 strain harboring G1-24. Evaluation of the change in phage resistance in this El Tor background was necessary, as the classical biotype strains are resistant to many of the phages in our collection through unknown mechanisms (data not shown). CRISPR-targeting strains were challenged with phage, and the EOP compared to that of a nontargeting wild-type strain was calculated. The introduction of pTargetϕcas1 resulted in a 16-fold decrease in the EOP compared to that of a nontargeting strain, and this phage resistance phenotype was further enhanced (to a nearly 500-fold decrease in the EOP [Fig. 4]) when the cas genes were overexpressed using an IPTG-inducible promoter (Fig. 1C). Interestingly, even though all target sites had an intact PAM and all sequences were targeted with an identically matched spacer, the EOPs of phages on targeting strains varied tremendously from 10⁻¹ to 10⁻⁵ (Fig. 4). For example, we targeted ICP1_2011_A cas2-cas3 with three different spacers and observed EOPs of 10⁻⁴, 10⁻¹, and 10⁻⁵, depending on the particular spacer used. This indicates that phage mutants capable of forming plaques on targeting strains arise to various extents depending on the specific target, and the frequency at which this occurs may be somewhat unpredictable. These data demonstrate that a plasmid-based system can be used for efficient CRISPR targeting to generate phage resistance in V. cholerae.

FIG 3.

The plasmid-based CRISPR system for generating phage-resistant derivatives of V. cholerae. (A) The essential components of the pCRISPR plasmid include an IPTG-inducible CRISPR array and an MCS with several unique restriction enzyme sites for addition of editing templates. Spacers can be inserted into the CRISPR array between BsaI and/or BseRI sites using annealed oligonucleotides. (B) The oligonucleotide design for each spacer insertion site is depicted.

FIG 4.

Engineered V. cholerae CRISPR confers resistance to phages. The sensitivity of each strain to a phage with the target sequence is represented as a histogram of the efficiency of plaquing, which is the ratio of the plaque count of a pCRISPR-targeting V. cholerae strain to that of the nontargeting strain. The phages used for these experiments were ICP1 (gray bars) and ICP1_2011_A (black bars). Data represent the means and standard deviations from three independent experiments. Induction of V. cholerae cas genes was achieved through an IPTG-regulated P_tac promoter upstream of the cas genes; all experiments were performed in the presence of 1 mM IPTG.

In E. coli, the type I-E CRISPR-Cas system has been reported to require both an intact PAM and perfect base pairing with a noncontiguous 7-bp seed sequence immediately adjacent to the PAM (43). To evaluate the sequence requirements of CRISPR-Cas immunity in V. cholerae, we took advantage of our collection of phages whose whole genomes had been sequenced (44) and in which the targets of several of the spacers are heterogeneous between phage isolates. In agreement with previous results (43, 45), we found that a single substitution within the PAM was sufficient for a phage to display an escape phenotype (Fig. 5, isolate ICP1_2011_A with pTargetϕorf91). Phages with two or three substitutions within the seed region displayed an escape phenotype (Fig. 5). Previous results in E. coli indicate that four or five substitutions within the protospacer but outside the seed region can be tolerated (43); conversely, we found that CRISPR interference was abolished with a phage harboring five mutations outside the seed region (Fig. 5, isolate ICP1_2005_A with pTargetϕcas6f). This result is consistent with that of more recent high-throughput analyses of sequence requirements which indicate that even two mutations clustering in close proximity to certain locations outside the seed region can disrupt base pairing with the crRNA (45). A single mutation outside the seed region was well tolerated and still resulted in a CRISPR-imposed decrease in plaque formation but appeared to weaken CRISPR-Cas function (Fig. 5, isolate ICP1_2004_A with pTargetϕcas6f).

FIG 5.

Mutations in the protospacer or PAM decrease CRISPR-Cas-mediated phage resistance in V. cholerae. Mutations are shown schematically beside the resulting sensitivity of each strain to the phage indicated. The EOP was determined for strain-phage combinations in which the target sequence was present (solid bars) or in which the target sequence was mutated (striped bars). Note that the PAM is immediately proximal to the protospacer; a space is present for clarity. Error bars indicate standard deviations.

Phage mutants capable of forming plaques on targeting strains can arise due to SNPs and deletions (17). The characterization of phage mutants (particularly those with SNPs) that arise under these selective conditions can provide additional insight into the sequence requirements of this CRISPR system. ICP1 escape mutants on a strain with pTargetϕorf10 were observed at a frequency of 6.4 × 10⁻² (Fig. 4). The protospacer and flanking regions for 7 of these mutants, each of which displayed an EOP of 1 on pTargetϕorf10-containing cells, were sequenced. The majority of escape mutants had SNPs: 3 of 7 mutants had an SNP within the seed sequence of the protospacer, and 2 of 7 mutants had an SNP in the PAM (Fig. 6A). Two of 7 mutants had deletions of up to 172 bp (out of the 285 bp encoding orf10), strongly suggesting that orf10 is nonessential for ICP1 under these conditions. In conjunction with the results of the analysis of the sequence requirements using our collection of phages whose whole genomes had been sequenced (Fig. 5), these results demonstrate that both positions in the PAM are essential for interference in this system.

FIG 6.

Mutants that escape CRISPR interference arise through point mutations and deletions. (A) Chromatograms of DNA sequences of ICP1 escape mutants that formed plaques on a strain with a spacer directed against orf10 demonstrate that 5/7 escape mutants had point mutations (highlighted) and 2/7 harbored deletions that removed the CRISPR target (indicated by a blue rectangle). WT, wild type. (B) PCR analysis of the CRISPR locus of ICP1_2011_A ΔS9 mutants on a strain with a targeting plasmid directed against the phage CRISPR spacer 2 (highlighted in black). The sequenced CRISPR arrays of all 11 escape phage are schematized below the PCR analysis gel.

We hypothesized that escape phages would readily form by deletion of the protospacer if it were flanked by repetitive sequences that could undergo homologous recombination. To this end, we sequenced the CRISPR loci of 11 ICP1_2011_A escape mutants that formed plaques on a strain targeting the phage's CRISPR spacer (pTargetϕCR1S2). As predicted, numerous escaping phages were isolated at a frequency of 1.3 × 10⁻¹ (Fig. 4). All escape mutants showed a complete loss of the targeted spacer, which was likely achieved through recombination of the different repeats in the CRISPR array, rendering phages with unique CRISPR arrays (Fig. 6B).

Genome engineering of lytic V. cholerae phage genomes.

Our analyses of phages that escape CRISPR interference indicate that this system can be used to isolate phage mutants with point mutations or larger deletions, as previously reported (29), and that this ability may be particularly useful if the protospacer is flanked by repeats that can undergo recombination (Fig. 6B). Recently, two groups have applied CRISPR-Cas to genetically engineer lytic phages (28, 29). Although their methods differed from ours, the underlying principle is that a desired editing template can be provided in trans and CRISPR-Cas provides a selective pressure to identify recombination events (Fig. 7A). Efficient systems for introducing mutations through homologous recombination in lytic phages did not previously exist due to the lack of selectable markers (29). To this end, we engineered a second MCS into pCRISPR (Fig. 3A) to allow the cloning of editing templates. As a proof of principle, we targeted ICP1 cas genes for mutation since these genes are dispensable for plaque formation on V. cholerae strains lacking PLE (21). The plasmid pTargetϕcas1-Ed was constructed to include an editing template (with ∼250 bp of homology on either side of the desired mutation) that would delete the entire 33-bp protospacer targeted by pTargetϕcas1. Strains expressing either pTargetϕcas1 or pTargetϕcas1-Ed were challenged with ICP1_2011_A to compare the number of escape mutants obtained with or without the editing template. We obtained a 10-fold increase in the EOP of ICP1_2011_A in the presence of the editing template (Fig. 7B). Sequencing analysis of cas1 from 8 escape phages on pTargetϕcas1-Ed confirmed that all phages had incorporated the desired 33-bp deletion provided by the editing template. We also created pTargetϕcas2-3³-Ed with an editing template to generate an in-frame deletion (>2 kb) of cas2-cas3 and pTargetϕcas2-3³-EdGFP with an editing template to replace cas2-cas3 with the gene for green fluorescent protein (gfp). Unexpectedly, the presence of these editing templates (which share arms of homology) had an undesired negative impact on V. cholerae growth (data not shown). In infection assays with ICP1_2011_A, the presence of the editing templates led to a striking ∼10,000-fold increase in the EOP (Fig. 7B). PCR and sequencing analysis of escape mutants on pTargetϕcas2-3³-Ed demonstrated that 7/12 had the expected deletion. Similarly, PCR and sequencing analysis of escape mutants on pTargetϕcas2-3³-EdGFP demonstrated that 4/8 had the expected replacement of cas2-cas3 with gfp. We suspect that the ∼50% efficiency of obtaining the desired mutants is a consequence of the toxicity of the editing template; the unedited phages remained sensitive to targeting by pTargetϕcas2-3³ (data not shown), indicating that plaque formation occurred despite intact CRISPR-Cas targeting, perhaps due to the loss of the plasmid. These data indicate that genome engineering of lytic V. cholerae phages can be performed using our plasmid-based CRISPR-Cas system.

FIG 7.

The V. cholerae CRISPR-Cas system can be used to engineer lytic V. cholerae phages. (A) Concept of phage genome editing using the CRISPR-Cas system. The CRISPR-targeting construct is cointroduced with an editing template that can recombine with the phage target sequence to permit phage escape and, thus, plaque formation. (B) The EOP was determined for strain-phage combinations in which targeting occurs in the absence (black bars) or presence (open bars) of an editing template. The lengths of the up and down arms of homology used in the editing template are indicated below the green and yellow bars, respectively. Chromatograms for representative engineered phages with deletions or insertion sizes are shown below the wild-type sequence for each editing construct. Lowercase letters highlighted in gray represent the wild-type sequence that was deleted by the editing construct.

DISCUSSION

The CRISPR-Cas system in classical biotype strains is part of a larger genomic island (GI-24), which we successfully mobilized into an El Tor background via chitin-induced natural transformation. In the El Tor background, the CRISPR-Cas system retained its predicted function in phage defense and showed the expected dependence on the Cascade complex (Fig. 2A). Natural transformation has previously been implicated as a mechanism responsible for the emergence of different serogroups of V. cholerae via transformation with ∼30-kb serogroup-specific gene clusters (46). Interestingly, phage predation has been implicated as a potential source of selective pressure promoting serogroup conversion in aquatic reservoirs. Given that the CRISPR-Cas system in GI-24 can protect against phage infection and the El Tor and classical biotypes were found together in environmental reservoirs for decades (3), it raises the question of why El Tor variants with CRISPR-Cas have not been selected for. Intriguingly, the only sequence information available for pre-sixth pandemic V. cholerae strains is from a second pandemic strain from 1849 which did have a GI-24 including the CRISPR-cas region, although the spacer content of the CRISPR array could not be discerned (47). This observation raises the possibility that seventh pandemic V. cholerae strains uniquely lack CRISPR-Cas compared to V. cholerae strains responsible for earlier pandemics. The evolutionary advantage of having CRISPR-Cas to destroy invading lytic phages is obvious; however, foreign DNA (for example, DNA from lysogenic phages, plasmids, and conjugative elements) can be tremendously beneficial owing to its tendency to encode virulence factors and genes for antibiotic resistance. An inverse correlation between the presence of a type II CRISPR-Cas system and antibiotic resistance was found in Enterococcus faecalis and Enterococcus faecium (48), suggesting that CRISPR-Cas is a barrier for acquiring beneficial mobile traits in these and other important human pathogens (49). Not surprisingly, V. cholerae relies on the transfer of accessory genetic elements for its evolutionary success as a pathogen. Two notable examples include the introduction of cholera toxin, which is encoded by a lysogenic filamentous phage (50), and the SXT/R391 integrated conjugative elements, which have been key drivers of antibiotic resistance since the early 1990s (4). The potential barrier to sampling such elements imposed by CRISPR-Cas may have had too great a cost, and the system may have been lost in seventh pandemic isolates, despite the need for protection from prevalent lytic phages in the host and the environment (26, 27).

Our use of a plasmid-based CRISPR system in V. cholerae has provided insight into the protospacer sequence requirements of this type I-E system. In selecting for phage escape mutants, we found that mismatches in particular positions in the protospacer and flanking PAM abolish efficient targeting (Fig. 6A). Using our collection of phage isolates, we obtained evidence that mismatches between the crRNA and the protospacer can result in intermediate resistance (Fig. 5). These more subtle effects may not have been observed by looking solely at phage escape mutants. Engineering of the CRISPR array on a low-copy-number plasmid and inserting new spacers using annealed oligonucleotides allows one to generate particular phage-resistant isolates relatively quickly and at a low cost. Selection of a spacer for insertion into pCRISPR is based on in silico identification of a target of interest that is flanked by a TT PAM immediately 3′ of the target, which in this study we demonstrated is essential. The V. cholerae CRISPR-Cas system can also be used as tool to select for phages that have lost the protospacer, particularly when sequences facilitating homologous recombination surround the target (Fig. 6B). The plasmid generated in this work allows cloning of an editing template that facilitates the generation of precisely edited V. cholerae phages (Fig. 7B). We successfully generated a small deletion (33 bp) and a large deletion (>2.6 kb), as well as a gene replacement, using editing templates with arms of homology ranging from ∼250 to 575 bp. Novel genes without a predictable biological function are often very abundant in phage genomes; for example, 75% of the 230 predicted ORFs in the ICP1 genome lack homology to known proteins (44). In order to understand this largely unexplored reservoir of novel genetic information, efficient genetic tools are required. Recently, CRISPR-Cas has successfully been employed to manipulate lytic phage genomes with success rates of ∼40% (28) to 100% (29). Owing to this success in E. coli (28) and Streptococcus thermophilus (29), we used similar principles in designing our genome editing strategy for lytic V. cholerae phages. Kiro et al. (28) used a two-step technique in which they amplified the wild-type phage in the presence of an editing template and then performed a subsequent round of infection on a selective (CRISPR-targeting) host. By performing the targeting and editing in a single round of infection, our system allows the more rapid isolation of the desired recombinant phage. The frequent isolation of escape phages in the presence of specific targeting spacers may pose a challenge in incorporating a desired editing template; however, we designed our engineering plasmid to have up to two unique spacer insertion sites (Fig. 3), anticipating that, if required, we could increase phage resistance by targeting multiple positions within a desired target (10), thereby facilitating the isolation of desired recombinant phages.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.