Bacteriophages (or phages) are the most abundant biological entities on earth, and are estimated to outnumber their bacterial prey by ten-fold 1 . The constant threat of phage predation has led to the evolution of a broad range of bacterial immunity mechanisms, which in turn, result in the evolution of diverse phage immune evasion strategies leading to a dynamic coevolutionary arms race 2,3 . Though bacterial innate immune mechanisms against phage abound, the only documented bacterial adaptive immune system is the CRISPR/Cas (Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated proteins) system, which provides sequence-specific protection from invading nucleic acids including phage 4–11 . Here we show a remarkable turn of events, in which a phage encoded CRISPR/Cas system is used to counteract a phage inhibitory chromosomal island of the bacterial host. A successful lytic infection by the phage is dependent on sequence identity between CRISPR spacers and the target chromosomal island. In the absence of such targeting, the phage encoded CRISPR/Cas system can acquire new spacers to rapidly evolve and ensure effective targeting of the chromosomal island to restore phage replication. Vibrio cholerae serogroup O1 is the primary causative agent of the severe diarrhoeal disease cholera, and lytic V. cholerae phages have been implicated in impacting disease burden particularly in the endemic region surrounding the Bay of Bengal 12,13 . We recently described the isolation of the ICP1 (for the International Centre for Diarrhoeal Disease Research, Bangladesh cholera phage 1) -related, V. cholerae O1-specific virulent myoviruses that are omnipresent amongst cholera patient rice-water stool samples collected at the ICDDR,B from 2001 to 201114 and current study. V. cholerae readily evolves resistance to ICP1 predation through mutations in O1 antigen biosynthetic genes outside the human host; however, this mutational escape comes at a cost as virulence necessitates maintenance of the O1 antigen 15 . This dynamic between predation by ICP1 and virulence of V. cholerae O1, specifically in the context of human infection, provides a unique opportunity for discovery of novel bacterial immunity and phage immune evasion strategies. One bacterial defensive strategy against phages is the CRISPR/Cas system. CRISPR loci consist of an array of short direct repeats separated by highly variable spacer sequences of precise length corresponding to segments of previously captured foreign DNA (protospacers) 4,7,9 . CRISPR loci are found in ~40% and ~90% of sequenced bacterial and archaeal genomes, respectively 8,16 . The CRISPR array is transcribed and the transcript cleaved into small CRISPR RNAs (crRNAs), that in conjunction with the Cas proteins, execute an efficient process of immunity in which foreign nucleic acids are recognized by hybridization to crRNAs and cleaved 4,7,8 . We isolated eleven ICP1-related phages from stools of cholera patients at the ICDDR,B14 and current study, five of which encode a CRISPR/Cas system located between ORF 87 and ORF 88 of the ancestral ICP1 genome 14 . The GC content of this CRISPR/Cas system is the same (~37%) as the rest of the ICP1 genome. The ICP1 CRISPR/Cas system consists of two CRISPR loci (designated CR1 and CR2) and six cas genes (Fig. 1a) whose organization and protein products are most homologous to Cas proteins of the type 1-F (Yersinia pestis) subtype system 17 (Supplementary Table 1). V. cholerae is divided into two biotypes, classical and El Tor, the former of which is associated with earlier pandemics and has since been replaced by the El Tor biotype 18 . The classical strain, V. cholerae O395, has a CRISPR/Cas system belonging to the type I-E (Escherichia coli) subtype 17 , and to date there has not been any description of El Tor strains possessing a CRISPR/Cas system. Thus, the origin of the CRISPR/Cas system in ICP1 phage is unknown. Protospacer-adjacent motifts (PAMs) are type specific, short conserved sequence motifs in the immediate vicinity of protospacers that are required for acquisition and targeting 7,9,11,19 . In contrast to the GG PAM reported for the type I-F CRISPR/Cas systems in bacteria 19 , the protospacers targeted by the ICP1 CRISPR array display a GA PAM (Supplementary Fig. 1). The majority of spacers in the ICP1 CRISPR show 100% identity to sequences within an 18 kb island found in a subset of V. cholerae strains that include the classical strain O395 isolated in India in 1964, El Tor strain MJ-1236 isolated in Bangladesh in 1994, and several El Tor strains collected at the ICDDR,B between 2001–2011 (Supplementary Table 2). The 18 kb island resembles the phage inducible chromosomal islands (PICIs) of Gram-positive bacteria, including the prototype Staphylococcus aureus pathogenicity islands (SaPIs) 20,21 . SaPIs are induced to excise, circularize and replicate following infection by certain phages. They use varied mechanisms to interfere with the phage reproduction cycle to enable their own promiscuous spread 21 and this can protect the surrounding bacterial population from further phage predation. The organization of the V. cholerae 18 kb island targeted by the ICP1 CRISPR/Cas system is similar in length, base composition, and organization to that observed in the SaPIs subset of PICIs, with an integrase homologue at one end and a GC content lower than that of the host species (37% compared to 47.5%). We therefore refer to the 18 kb element as the V. cholerae PICI-like element (PLE) (Fig. 2). To address the functional relevance of the ICP1 CRISPR/Cas system, we focused on the interaction between the paired ICP1_2011_A phage and the V. cholerae O1 El Tor strain (harboring PLE1) that were isolated from the same stool sample (for simplicity hereafter referred to as ICP1 and V. cholerae PLE+). ICP1 has two CRISPR spacers (8 and 9) (Fig. 1b) that have 100% identity to sequences within the V. cholerae PLE (Fig. 2 and Supplementary Table 2). Using the standard soft agar overlay method, we found that ICP1 can plaque efficiently on V. cholerae PLE+ (Fig. 3b). We used northern blot analysis to confirm that ICP1 crRNAs are transcribed and processed during V. cholerae infection (Supplementary Fig. 2). To test whether targeting of the PLE by the ICP1 CRISPR/Cas system impacts phage fitness, we eliminated spacer 8 and 9 targeting. Spacer 8 targeting was disrupted by introducing silent mutations into its target within the PLE, generating V. cholerae PLE(8*) (Fig. 3a). We then infected this strain with a spontaneous ICP1 spacer 9 deletion mutant, referred to as ICP1(ΔS9). ICP1(ΔS9) was blocked for plaque formation on V. cholerae PLE(8*); however, it maintained wild type plaquing efficiency on V. cholerae PLE+ (Fig. 3b). Importantly, V. cholerae PLE(8*) is sensitive to plaque formation by ICP1 (Fig. 3b), which still harbors one spacer (S9) targeting the PLE. These results demonstrate that ICP1 CRISPR/Cas must target the PLE for destruction in order to effectively infect and form plaques, and that a single spacer that targets the PLE is sufficient to facilitate successful phage replication. A mutant in which PLE ORFs 7–20 were deleted was susceptible to infection by ICP1(ΔS9) with wild type plaquing efficiency (Supplementary Fig. 3). This demonstrates that an intact PLE is required to inhibit ICP1 in the absence of CRISPR targeting. These results, in conjunction with the observation that PLE1 circularizes following ICP1 infection (Supplementary Fig. 4), further support our designation of the 18 kb island as a PICI-like element. It has been well documented in the type I-E (E. coli) system that CRISPR interference requires an intact PAM and a fully complementary seed region (a noncontiguous 7 bp sequence immediately adjacent to the PAM) 22 . To address the sequence requirements of the ICP1 CRISPR/Cas system we constructed a series of point mutations in the spacer 8 target in V. cholerae PLE that span the PAM, seed region, and remainder of the target sequence and determined their impact on immunity. In accordance with previous results, we found that single mutations within the PAM or the first four positions in the seed region immediately adjacent to the PAM abolish ICP1 CRISPR/Cas immunity (Supplementary Fig. 5). Interestingly, mutations of increasing distance from the PAM showed a concordant decreasing effect on immunity. Up to five mismatches outside of the seed region of the target are known to be tolerated in the type I-E system 22 , and similarly we found that three and five mutations outside of the seed region were tolerated, however, eight mutations were not (Supplementary Fig. 5). In experiments where the ICP1 CRISPR/Cas system could not target the V. cholerae PLE and therefore plaque formation was greatly reduced, we observed phage escape mutants at frequencies that were dependent upon the host strain on which the phage had been previously propagated. When ICP1(ΔS9) was grown on a PLE+ host prior to plaquing on V. cholerae PLE(8*), the efficiency of plaquing (EOP, which is the plaque count on the mutant host strain divided by that on the wild-type host strain) was 3.5 × 10−5. The CRISPR loci from ten independent ICP1(ΔS9) escape mutants were sequenced, and in all cases, a new spacer was present at the leader end of the CRISPR CR1 array. Furthermore, the new spacers had 100% identity to sequences within the PLE (Fig. 2), and all newly integrated spacers target the PLE with the conserved GA dinucleotide PAM sequence (Supplementary Fig. 1b). The experimentally acquired spacers target both the coding and noncoding strands (Supplementary Table 3) although most (nine out of ten) target the coding strand. The pre-existing spacer (S8) (although mismatched in these experiments) also targets the coding strand; these data are in support of recent evidence that the DNA strand from which new protospacers are incorporated is heavily biased towards the existing protospacer orientation 23,24 . In contrast to when phage were propagated on a PLE+ host prior to plaquing on V. cholerae PLE(8*), phage escape mutants were detected at a much lower frequency (EOP=1.1 × 10−8) when ICP1(ΔS9) was grown on a V. cholerae PLE− host. This shows that new spacers targeting the PLE are incorporated into the CRISPR array during ICP1(ΔS9) infection of the PLE+ host (the immunization process), and that an immune host possessing an untargeted PLE can subsequently be employed to select for new ICP1 CRISPR acquisition events that confer targeting and thus restore phage replication. These results demonstrate that the ICP1 CRISPR/Cas system is fully functional as an adaptive immune evasion system that benefits the phage. ICP1 has evolved to effectively target the V. cholerae PLE with an adaptive immune evasion system that has never before been shown to function in bacterial viruses. During ICP1 infection of V. cholerae PLE+, PLE circularizes (Supplementary Fig. 4) and inhibits ICP1 through an unknown mechanism. In order to successfully replicate, ICP1 uses the CRISPR/Cas system to target the PLE for destruction and because host cell death and DNA damage is inherent to lytic phage infection, CRISPR-mediated DNA cleavage of the PLE does not negatively impact ICP1 infection. Sequencing data has been used to identify putative CRISPR arrays within a Clostridium difficile prophage 25 , and more recently in metagenomic data sets of free viruses 26,27 . However, there is currently no evidence for expression or function of these putative arrays. We show that the ICP1-encoded CRISPR/Cas system actively and autonomously functions to inhibit host immunity and thereby permit lytic infection. The implications of this finding, in conjunction with the previous observations regarding the presence of CRISPR loci in other phages 25–27 , suggest that the use of the so-called bacterial adaptive immune system by these bacterial predators may be an underappreciated immune evasion strategy in the unfolding phage versus host coevolutionary arms race. METHODS Phage (ICP1_2011_A and ICP1_2006_E) and V. cholerae were isolated from cholera rice-water stool samples and propagated as described 14,15 . Genomic libraries were generated for phage and host strains as described 28 and sequenced using an Illumina HiSeq2000. A V. cholerae O1 El Tor isolate collected at the ICDDR,B in 2006, which was sequenced in this study and found to not harbor a PLE, was used as the PLE- host for propagation experiments. We used the CRISPRFinder program 16 to identify CRISPR loci. WebLogo 29 was used to generate sequence logos for identification of the PAM. Point mutations were constructed using splicing by overlap extension PCR and introduced using pCVD442-lac as previously described 15 . The PLE1 deletion construct (missing 8.6 kb including ORFs 7–20) was constructed using the FLP/FRT site-specific recombination system 30 . ICP1(ΔS9) was identified by screening for alterations in the CRISPR array by PCR following growth on V. cholerae PLE+. RNA was purified using the mirVana kit (Ambion) at the indicated times and run on 12% polyacrylamide urea gels. Northern blots were prehybridized in Ultrahyb-oligo (Ambion) and hybridization was carried out at 37°C overnight using 32 nt 5′ end-labeled DNA probes (generated with [γ-32P]ATP and T4 polynucleotide kinase) complementary to spacers 8 and 6. Supplementary Material 1
