A cocktail of three virulent bacteriophages prevents Vibrio cholerae infection in animal models

Effective prevention strategies will be essential in reducing disease burden due to bacterial infections. Here we harness the specificity and rapid-acting properties of bacteriophages as a potential prophylaxis therapy for cholera, a severely dehydrating disease caused by Vibrio cholerae. To this end, we test a cocktail of three virulent phages in two animal models of cholera pathogenesis (infant mouse and rabbit models). Oral administration of the phages up to 24 h before V. cholerae challenge reduces colonization of the intestinal tract and prevents cholera-like diarrhea. None of the surviving V. cholerae colonies are resistant to all three phages. Genome sequencing and variant analysis of the surviving colonies indicate that resistance to the phages is largely conferred by mutations in genes required for the production of the phage receptors. For acute infections, such as cholera, phage prophylaxis could provide a strategy to limit the impact of bacterial disease on human health.

C holera is an acute, severely dehydrating diarrheal disease caused by the water-borne bacterium Vibrio cholerae. Cholera remains a substantial global health burden and is endemic to many parts of Africa and Asia 1 . Recent widespread epidemics in disaster-stricken or war-torn countries such as Haiti 2 and Iraq 3 highlight the vulnerabilities of populations to sudden outbreaks. Current recommended preventatives include mass vaccinations with the World Health Organizationprequalified oral cholera vaccine 4 and increased awareness of sanitation and hygiene practices 5 . However, access to clean water is difficult and vaccination campaigns require forethought and time for efficacy; both methods may not be logistically feasible for immediate protection in the event of an outbreak. Household transmission is a major contributor to the rapid spread of V. cholerae within communities. Household contacts of index cases often present with cholera symptoms 2-3 days after the initial patient becomes sick 6 . Therefore, there is currently an unmet need for a clinical intervention to stem the household spread of cholera by use of a rapid prophylactic treatment. Although chemoprophylaxis with antibiotics may effectively reduce cholera burden 7 , the World Health Organization does not recommend this practice due to the development and spread of drug-resistant bacteria (http://www.who.int/cholera/technical/ prevention/control/en/). Moreover, the broad-spectrum action of antibiotics would cause dysbiosis of the resident intestinal microbiota, which could put patients at risk of other intestinal infections.
There has recently been a renewed interest in the use of bacteriophages (phages) for environmental and clinical applications 8 . In contrast to antibiotics, phages are specific in their targets and, because they are replicating viruses, are capable of auto-dosing, a phenomenon where phage replication increases their number and contributes to the dose. Previous attempts to use phages to prevent or treat cholera have produced mixed results. Dutta et al. 9 showed that a single phage type given 1 h before V. cholerae challenge in an infant rabbit model prevented onset of cholera symptoms. Studies by Jaiswal et al. 10 showed that a cocktail of five lytic bacteriophage types given 6 or 12 h before V. cholerae challenge in an adult rabbit model reduced diarrheal severity slightly but did not significantly lower the bacterial load; however, the same phage cocktail could successfully reduce the V. cholerae load when administered 6 or 12 h after challenge. A second study in adult mice also showed promise for treatment of cholera with a phage cocktail 11 . These studies suggest that a prophylaxis approach merits further, in-depth investigation.
We previously isolated three V. cholerae-specific, lytic (virulent) phages ICP1, ICP2 and ICP3 from rice-water stool samples of cholera patients in Bangladesh 12 . The receptors for ICP1 and ICP2 were identified as the lipopolysaccharide (LPS) O1 antigen 13 and the major outer membrane porin OmpU 14 , respectively, which are considered virulence factors of V. cholerae. The receptor for ICP3 is as yet unknown, although we hypothesize at least a partial role for the LPS O-antigen. We have recently shown that these virulent phages impose significant bactericidal pressure on V. cholerae during its natural course of infection in humans 14 . A cocktail comprising phages that target different receptors would reduce the likelihood of multi-phageresistant V. cholerae isolates in the surviving population. Therefore, we hypothesized that a cocktail of the three ICP phages may be used as a prophylaxis treatment to specifically target V. cholerae that transits into the small intestine to prevent signs of cholera in animal models of disease.
In this study, we show that orally applied, prophylactic use of the ICP cocktail reduces colonization by V. cholerae in the infant mouse model. The ICP cocktail also prevents the onset of cholera symptoms in the infant rabbit model when administered up to 24 h before V. cholerae challenge. This proof-of-principle study demonstrates the successful use of phage prophylaxis to prevent disease caused by a mucosal pathogen.

Results
The three-phage ICP cocktail kills V. cholerae in vitro. Initial in vitro killing time-course experiments provided support for the hypothesis that the three-phage ICP cocktail was more effective in killing V. cholerae than each phage in isolation (Fig. 1). Cultures of V. cholerae E7946 (AC53) were grown in (i) the absence of phage or in the presence of (ii) ICP1, (iii) ICP2, (iv) ICP3 or (v) the ICP cocktail, at a multiplicity of infection of 1. Bacteria (Fig. 1a) and phage (Fig. 1b) titres were enumerated every hour for 8 h and then again at 12 and 24 h. Although the bacterial population declined in all conditions initially, growth resumed for cultures grown in the presence of ICP1 or ICP3 within 4-6 h, respectively (Fig. 1a). Cells incubated in the presence of ICP2 resumed growth more slowly, but reached the same density as the control within 24 h. In contrast, cells grown with the ICP cocktail did not reach the same density as the no-phage control by the end of the experiment. Phage titres dropped but remained stable for all conditions over the course of the experiment (Fig. 1b). It is not surprising that V. cholerae was able to escape phage predation  over time, given that the starting inoculum was high (5 Â 10 7 colony-forming units (CFU)) and probably contained phageresistant mutants that were positively selected.
The ICP cocktail reduces mouse small intestinal colonization. The ICP cocktail was more effective than each phage alone in killing V. cholerae in vitro. To test whether this was also true in vivo, prophylaxis experiments were performed in the infant mouse model of V. cholerae colonization. We hypothesized that the ICP cocktail would be effective in preventing V. cholerae infection of the infant mouse small intestine. Mice were divided into five groups and received either (i) no phage, (ii) ICP1, (iii) ICP2, (iv) ICP3 or (v) the ICP cocktail. Mice were given phages (between 1 Â 10 6 and 1 Â 10 7 plaque-forming units (PFU)) by orogastric inoculation 3 h before infection with 5 Â 10 5 CFU V. cholerae. After 24 h, mice were killed, and V. cholerae and phages in the small intestine enumerated. As shown in Fig. 2a, the number of surviving V. cholerae cells in the small intestine was reduced by at least two orders of magnitude in all conditions where phage was administered, in comparison with the nontreated control, with ICP3 being the most successful single phage. No V. cholerae were detected in the intestinal homogenates of four out of five mice treated with the ICP cocktail (Po0.01, Kruskal-Wallis test with the Dunn's post-hoc multiple comparisons test). These data show that in comparison with ICP1 or ICP2 alone, ICP3 alone or the ICP cocktail is superior at preventing V. cholerae colonization. Phages were still detected in the small intestine at the end of the experiment (Fig. 2b), suggesting that these phages can survive and persist in the intestinal tract. Although the bacterial count medians of the ICP3 alone and the ICP cocktail groups are similar, we chose to proceed with the ICP cocktail to reduce the probability that phage-resistant V. cholerae isolates would survive.
Next, we aimed to test whether the ICP cocktail could still be effective when administered up to 24 h before V. cholerae infection. First, it was necessary to establish if phages could survive in the small intestine for long periods in the absence of their host bacterium. To do this, mice were dosed with phages (between 3 Â 10 7 and 3 Â 10 8 PFU) by oral gavage and killed after 3, 6, 12 or 24 h. Phages were enumerated from intestinal homogenates using plaque assays (Fig. 3a). After 3 h, the phage titres for ICP1 and ICP3 were still stable, whereas ICP2 titres had dropped B10-fold. For all three phages, titres dropped only 10-to 100-fold after 6-12 h. After 24 h, ICP1 and ICP2 were still retained at B10 5 PFU per small intestine, whereas the titre of ICP3 fell to between 10 2 and 10 4 PFU per small intestine. These data further show that the ICP phages can survive in the intestinal tract for at least 24 h in the absence of their host V. cholerae.
To test whether the ICP cocktail remained protective when given several hours before exposure to V. cholerae, infant mice were inoculated with the ICP cocktail 6, 12 or 24 h before challenge with between 5 Â 10 5 and 9 Â 10 5 CFU V. cholerae. A control group did not receive phage. The data presented in Fig. 3b show that the 6 h prophylaxis was most successful; the number of surviving V. cholerae cells fell at least three orders of magnitude, in comparison with the no-phage group, with no detectable V. cholerae for four of the seven mice that received phage (Po0.0001, Kruskal-Wallis test with the Dunn's post-hoc multiple comparisons test). The number of V. cholerae cells was also significantly lower for the 12 h group (Po0.001), with four of these animals having no detectable V. cholerae at the end of the experiment. The number of V. cholerae cells was also reduced two orders of magnitude in the 24 h group. Furthermore, virulent phages were still detectable 24 h post-infection (Fig. 3c).
To determine whether the ICP cocktail could limit colonization when mice are given a higher challenge dose, infant mice were inoculated with 1 Â 10 8 CFU of V. cholerae, a dose B200-fold higher than the experiment shown in Fig. 3b,c. The data presented in Fig. 3d show that the ICP cocktail was effective in reducing V. cholerae colonization of the small intestine by at least two orders of magnitude when administered at either 6, 12 or 24 h before challenge. Phages were also still detected at 24 h post infection at similar levels in all three phage-treated groups (Fig. 3e). In concert, these data support the hypothesis that a phage prophylaxis approach can be used to prevent V. cholerae colonization of the small intestine.
V. cholerae could be isolated from intestinal homogenates of several of the mice dosed with the ICP cocktail (Figs 2 and 3). Owing to the complexity of infection within the gut, it is possible that these bacteria survive because either they did not encounter the ICP phages or these cells carry genetic mutations that allow them to escape phage predation. To determine the phage resistance profile of surviving cells, V. cholerae isolates from the  Table 1 and detailed in Supplementary Data 1. All isolates from mice that received the ICP cocktail 6 or 12 h before challenge were sensitive to all three ICP phages. The majority of isolates from mice given the ICP cocktail 24 h before challenge were sensitive to all three ICP phages; however, a small number showed differing ICP resistance phenotypes.
To uncover the genetic basis for resistance, 24 V. cholerae isolates were analysed by whole-genome sequencing followed by variant analysis [15][16][17] . It was shown previously that slipped-strand mispairing in the poly-A tracts of O-antigen synthesis genes can result in abnormal O-antigen and confer ICP1 resistance 13 . Consistent with this, the mutations in ICP1-and ICP3-resistant isolates were found in O-antigen synthesis genes located on chromosome 1 of V. cholerae between open reading frames VC0240 (gmhD) and VC0269 (manA) 18 (Supplementary Data 2). Although LPS mutations were a common source of resistance in these isolates, this is not of major concern given that strains carrying such mutations have previously been shown to be avirulent [19][20][21] . We have also shown previously that mutations in ompU and toxR confer ICP2 resistance 14 . In agreement, mutations in ICP2-resistant isolates were found in open reading frames VC0633 (ompU) or VC0984 (toxR) (Supplementary Data 2). Eight of the sequenced strains revealed no mutations to explain their phage-resistant phenotypes, most likely due to incomplete genome sequence coverage.
The ICP cocktail provides protection in the infant rabbit. To further assess the ability of the ICP cocktail to prevent cholera, experiments were undertaken in infant rabbits. In contrast to infant mice, infant rabbits infected by V. cholerae develop the profuse secretory diarrhea associated with cholera. Sickened animals suffer from dehydration and weight loss and show an accumulation of caecal fluid containing a high bacterial load 22 . First, to establish whether phages could be retained in the rabbit d e intestinal tract in the absence of V. cholerae, animals were dosed with the ICP cocktail alone (3 Â 10 8 PFU). After 3 or 24 h, the animals were killed and phages enumerated from intestinal homogenates. Even after 24 h, 10 6 phages could be recovered from the intestine (Fig. 4a).
To determine whether the ICP cocktail could block infection by V. cholerae, phages (between 4 Â 10 9 and 8 Â 10 9 PFU) were orogastrically administered to two groups of infant rabbits either 3 or 24 h before challenge with 5 Â 10 8 CFU V. cholerae. A control group did not receive phage. Animals were monitored for signs of cholera, specifically weight loss ( Supplementary  Fig. 1) and the presence of rice-water stool (also referred to here as cholera-like diarrhea). The no-phage control group showed cholera-like symptoms 12-14 h post infection, in line with previously published observations 22 , and were killed at this point. The caecum of each infected rabbit was distended and accumulated B0.5-1 ml of fluid, which is further indicative of V. cholerae infection. Approximately 10 9 -10 10 CFU were enumerated from caecal fluid (Supplementary Table 2) and intestinal homogenates (Fig. 4b) from this group.
V. cholerae were not detected for four of the seven rabbits that were dosed with ICP cocktail 3 h before infection (Fig. 4b), indicating that these animals cleared the infection. Between 10 6 and 10 8 CFU of V. cholerae were obtained from intestinal homogenates from the remaining three rabbits in this group, giving a 10-to 100,000-fold decrease in comparison with the nontreated group (Fig. 4b, Po0.001, Kruskal-Wallis test with the Dunn's post-hoc multiple comparisons test). Between B10 5 and 10 9 CFU of V. cholerae were enumerated from the intestines of animals dosed with phage 24 h before infection (Fig. 4b), presenting a 10-to 100,000-fold decrease in bacterial load when compared with animals that received no phage (Po0.05). It is important to note that despite the bacterial load observed, there was no evidence of rice-water stool or significant weight loss for rabbits in either of the phage-dosed groups up to 20 h after V. cholerae challenge (Supplementary Fig. 1). On dissection of these animals, 0.1 ml of caecal fluid was obtained from one rabbit, which was dosed with phage 24 h before V. cholerae infection. The caeca of the other animals showed no signs of fluid accumulation (Supplementary Table 2). Between 10 6 and 10 9 PFU could be collected from the intestinal homogenates of phagetreated animals (Fig. 4c), indicating that phages persisted in the intestine over the course of the experiment. These observations suggest that the ICP cocktail protected against the signs of cholera over the duration of the experiment.
To determine whether V. cholerae that survived phage prophylaxis were sensitive or resistant to the ICP phages, V. cholerae isolates from each of the intestinal homogenates were randomly selected for colony purification and their phage resistance phenotype measured by EOP assays (summarized in Supplementary Table 3 and detailed in Supplementary Data 3). Thirty-nine per cent of isolates were resistant to both ICP1 and ICP3. No isolates were resistant to all three phages. To determine the genetic basis for resistance, whole-genome sequencing and variant analysis were conducted on 36 isolates, chosen due to their differing resistance phenotypes (Supplementary Data 4). Results were consistent with mutations found in ICP1-and ICP3-resistant isolates from the mouse experiment in Fig. 3. For 33 of these strains, there were mutations in O-antigen synthesis genes. For three strains, no mutations related to known phageresistance strategies were detected. We conclude from these data in two animal models that the leading source of phage resistance is mutation of the genes required for production of the phage receptors.
Collectively, these data suggest that the phage cocktail (i) is efficient at killing V. cholerae cells in vivo, (ii) reduces or prevents V. cholerae colonization of the intestine and (iii) provides protection against the onset of cholera-like diarrhoea in the infant rabbit model of cholera infection. Furthermore, isolates shed by the animal that were resistant to the ICP phages are expected to be avirulent based on past studies of O1-antigen mutants 13 , although virulence of these isolates has not yet been tested.

Discussion
Here we show that prophylactic administration of a V. choleraespecific phage cocktail is protective against cholera, by reducing both colonization and cholera-like diarrhoea in two animal models. We have shown that the ICP cocktail can successfully kill V. cholerae in vitro, prevent colonization of the infant mouse and preclude the onset of cholera-like diarrhoea in the infant rabbit. V. cholerae were isolated from intestinal homogenates of some phage-treated animals; however, the surviving population is smaller than observed from a wild-type V. cholerae infection (Figs 2-4). Moreover, we hypothesize that these surviving isolates may be unable to elicit acute infection, as many of the mutations that confer resistance to the ICP phages affect genes such as those involved in LPS synthesis (VC0240 to VC0269), which have been previously shown to be important for virulence. Crucially, at the end of these experiments, the population also remains sensitive to at least one of the three ICP phages. We suggest that the effectiveness of the ICP cocktail in killing V. cholerae cells is at least in part reliant on the concentration of phages that are in the intestine at the time of V. cholerae infection. The complex architecture of the intestinal tract probably presents a challenge to the phages' ability to access V. cholerae cells. The details of the interplay between the ICP phages, V. cholerae and the host remain to be elucidated. Future optimization of the composition and dose of the cocktail could further improve its effectiveness. The ratio of each phage in the cocktail could be optimized and additional phages could be added to improve V. cholerae killing. Furthermore, as complete reduction of the bacterial load was only observed at shorter prophylaxis times, optimization of the timing for each dose is needed before human use. It is likely that an individual will be repeatedly exposed to V. cholerae; therefore multiple doses may be necessary to control the infection to a point where it can be completely eradicated.
Current phage therapy research is largely focused on treating ongoing infections. Our results highlight the potential of phage therapy in preventing infections. The data presented herein suggest that the premise of using a phage cocktail to prevent cholera warrants further investigation. We anticipate that a rapidacting phage prophylaxis approach would be best used in at-risk individuals, such as the household contacts of individuals who display cholera symptoms. By limiting spread within households, the overall burden of the disease could be reduced. The application of phages as prophylactic treatments for mucosal pathogens represents a fast and specific means by which to restrict the impact of bacterial infections on human health.

Methods
Growth conditions and strains. Bacterial strains and phages used in this study are listed in Table 1. V. cholerae strains were grown in Luria-Bertani (LB) broth supplemented with 100 mg ml À 1 streptomycin (Sm). The ICP cocktail comprises the Vibrio phages ICP1, ICP2 and ICP3 in equal number. These phages were previously isolated from Bangladeshi cholera patient rice-water stool samples 12 . ICP1 was isolated on strain AC4741, whereas ICP2 and ICP3 were isolated on AC53, an Sm-resistant isolate of E7946. All experiments were carried out using AC53 V. cholerae O1 El Tor strain E7946. Strains AC4653 and AC2846 were used in plaque assays as negative controls. ICP1 and ICP3 cannot plaque on AC4653, whereas ICP2 cannot plaque on AC2846.
Bacteriophage preparation. High-titre stocks of the ICP phages were prepared by growth on agar plates followed by polyethylene glycol (PEG) precipitation. Briefly, each phage was grown with the appropriate V. cholerae strain in soft agarose (LB broth supplemented with 0.3% agarose) overlays. Once confluent, overlays were incubated with STE buffer (100 mM NaCl, 10 mM Tris and 1 mM EDTA) overnight at 4°C with gentle rocking to elute phage. The STE-phage solution was clarified by centrifugation, sterile-filtered and incubated with 1 Â PEG (4% PEG 8000, 0.5 M NaCl) at 4°C for 1-3 days, to allow for phage precipitation. Phages were collected by centrifugation at 10,000 g for 15 min at 4°C and the phage pellet re-suspended in STE buffer. Phages were titred by plaque assay, as previously described 10 .
In vitro phage-killing assay. Overnight cultures originating from single colonies of V. cholerae were diluted back to an OD 600 of 0.05 in 50 ml LB supplemented with 100 mg ml À 1 Sm and grown at 37°C with aeration. After 15 min, phages were added to each culture. At the indicated time-points, samples were collected to measure (i) CFU per ml and (ii) PFU per ml.
Infection of infant mice with V. cholerae. All animal experiments were in accordance with the rules of the Department of Laboratory Animal Medicine at Tufts University and the Institutional Animal Care and Use Committee. Four-and 5-day-old CD-1 infant mice (Charles River Laboratories) were infected. Each group of mice was comprised of animals from at least two different litters. Both male and female animals were used in this study. There was no observable correlation between sex of the animal and outcome of each experiment. Numbers of mice used in each condition in each experiment are listed in the legends of Figs 2 and 3. These mice were not tested for the presence of resident V. cholerae phages as the animals were not previously exposed to V. cholerae. As such, we do not predict that they will harbour any V. cholerae-specific phages in their microbiome.
Infant mice in phage-treated groups were orogastrically dosed with phage diluted in 2.5% sodium bicarbonate. Infant mice in cholera-treated groups received ca. 10 5 CFU ('normal' infectious dose) or ca. 10 8 CFU ('high' infectious dose) of V. cholerae diluted in 2.5% sodium bicarbonate. As phages were given at least 3 h before bacteria, any multiplicity of infection calculated would be inaccurate. Instead, the titres of bacteria and phage inoculated into animals were calculated from the input materials and the ranges specified in the figure legends. Mice were killed 24 h post infection, the longest infection time permitted in our protocol. Small intestines were dissected and homogenized in LB broth supplemented with 20% glycerol. Cells were serially diluted and plated on LB agar supplemented with 100 mg ml À 1 Sm and CFU per small intestine was calculated. To extract phage from small intestine homogenates, an aliquot of the intestinal homogenate was treated with chloroform and centrifuged at 10,000 g for 5 min. The supernatant was collected and used in plaque assays, to allow for calculation of PFU per small intestine. To assess the resistance profile of V. cholerae cells that survive phage prophylaxis, up to ten colonies were randomly picked per animal for use in EOP assays with ICP1, ICP2 and ICP3.
Infection of infant rabbits with V. cholerae. Three-day-old New Zealand White infant rabbits (Charles River Laboratories) were used in this study. Each group of rabbits comprises animals from at least two different litters. Both male and female animals were used in this study. There was no observable correlation between sex of the animal and outcome of each experiment. Infant rabbits in phage-treated groups were orogastrically dosed with phage diluted in 2.5% sodium bicarbonate. Three hours before infection with V. cholerae, infant rabbits were pre-treated with ranitidine hydrochloride to reduce stomach acidity (Caraco Pharmaceutical Laboratories) by intraperitoneal injection (2 mg per gram body weight). Infant rabbits were infected with 5 Â 10 8 CFU V. cholerae AC53 diluted in 2.5% sodium bicarbonate. As for the mice experiments, exact titres of bacteria and phage inoculated were calculated from the input material and a range is detailed in the figure legends. Animals were weighed at the start of the experiment and periodically throughout the course of the infection. Percentage body weight was calculated by dividing body weight at the end of the infection period by body weight at the start. Animals were killed 12-20 h post infection. Infant rabbits that suffer from cholera typically lose 10-15% of their body weight within 12-14 h and are killed at this time. Infant rabbits that lost o10% of their body weight are killed 20 h post infection, in accordance with our Institutional Animal Care and Use Committee protocol. This time restriction is necessary, as infants cannot be placed back with their mother after V. cholerae infection due to culling behaviour and so are without food for the infection period.
After dissection, the intestines of each rabbit were homogenized in LB broth supplemented with 20% glycerol. Caecal fluid, if present, was collected with a 1 ml syringe. Caecal fluid and intestinal homogenates were serially diluted and plated on LB agar containing 100 mg ml À 1 Sm for calculation of CFU per intestine. Phages were collected from homogenates as described for infection of infant mice. Surviving V. cholerae colonies were colony-purified and assessed by EOP assays for resistance to ICP phages. EOP assays. EOP assays were performed for isolates that survived phage predation. Each phage was titered on the isolate, on AC53 and also a phage-resistant strain as a negative control (DwbeL AC4653 for ICP1 and ICP3; DompU AC2846 for ICP2). The EOP was calculated by dividing the titre of the phage on the animal isolate by the titre of the phage on AC53. The limit of detection was 1 Â 10 -6 . Isolates were described as resistant when the EOP waso1 Â 10 À 6 , sensitive when the EOP was41 Â 10 À 1 or partially sensitive when the EOP was between these values.
Sequence analysis of phage-resistant isolates. V. cholerae genomic DNA was extracted using a DNeasy Blood & Tissue Kit (Qiagen). Whole-genome libraries were prepared for single-end 150-bp sequencing using the Nextera XT DNA Library Preparation Kit (Illumina). Sequencing was conducted at the Tufts University Core Facility using an Illumina HiSeq 2500. Genomes were assembled using CLC Genomics Workbench 8 software and aligned to the V. cholerae O1 N16961 reference genome 13 . To determine the mutations that may confer phage resistance, variant analysis was performed on mapped reads with a frequency threshold of 51%. Results were compared with AC53 variants 14 to remove those found in the wild-type inoculum. A surviving isolate that was determined to be sensitive to all three ICP phages was sequenced and resulting variants were also removed from resistant isolates variant analyses.
Data availability. The authors declare that the data supporting the findings of this study are available within the paper and its Supplementary Information files.