An Eye to a Kill: Using Predatory Bacteria to Control Gram-Negative Pathogens Associated with Ocular Infections

Ocular infections are a leading cause of vision loss. It has been previously suggested that predatory prokaryotes might be used as live antibiotics to control infections. In this study, Pseudomonas aeruginosa and Serratia marcescens ocular isolates were exposed to the predatory bacteria Micavibrio aeruginosavorus and Bdellovibrio bacteriovorus. All tested S. marcescens isolates were susceptible to predation by B. bacteriovorus strains 109J and HD100. Seven of the 10 P. aeruginosa isolates were susceptible to predation by B. bacteriovorus 109J with 80% being attacked by M. aeruginosavorus. All of the 19 tested isolates were found to be sensitive to at least one predator. To further investigate the effect of the predators on eukaryotic cells, human corneal-limbal epithelial (HCLE) cells were exposed to high concentrations of the predators. Cytotoxicity assays demonstrated that predatory bacteria do not damage ocular surface cells in vitro whereas the P. aeruginosa used as a positive control was highly toxic. Furthermore, no increase in the production of the proinflammatory cytokines IL-8 and TNF-alpha was measured in HCLE cells after exposure to the predators. Finally, injection of high concentration of predatory bacteria into the hemocoel of Galleria mellonella, an established model system used to study microbial pathogenesis, did not result in any measurable negative effect to the host. Our results suggest that predatory bacteria could be considered in the near future as a safe topical bio-control agent to treat ocular infections.


Introduction
In an era of increasing antibiotic resistance among bacterial pathogens, the search for new antibiotics and novel treatments for infections caused by these organisms is a priority among researchers. One novel treatment is biological therapy using specific bacteriophage for controlling the infecting pathogen [1][2][3]. Another novel treatment that might hold the potential to treat antibiotic resistant infections are predatory bacteria [4]. Recently, several studies have highlighted the ability of predatory bacteria Bdellovibrio spp. and Micavibrio spp. to prey on Gram-negative pathogens. Among the pathogens which were evaluated were bacteria associated with oral infections [5][6][7], gastrointestinal infections [8], zoonotic infection [9], pathogens associated with food processing and spoilage [10][11][12], as well as bacteria linked to systemic infections, burns and wounds [13]. Although the data published so far supports the claim that predatory bacteria could be used to control human pathogens, there is still concern regarding the toxic effects of administering large numbers of Gram-negative bacteria as live antibiotics. Therefore, treatment of local infections where the pathogens are easily accessible to topical or locally injected treatment would be ideal candidates to demonstrate a ''proof of concept'' that infections can be successfully treated with predatory bacteria.
One such local bacterial infection that is treated by direct administration of antibiotic to the site of infection is keratitis, infection of the cornea. Bacterial keratitis can be caused by both Gram-positive and Gram-negative pathogens. Common Gramnegative pathogens associated with keratitis are Pseudomonas aeruginosa and Serratia marcescens [14][15][16]. Bacterial keratitis is usually localized to an area of the cornea and is treated with antibiotic solutions delivered topically to the eye.
The first step in demonstrating that predatory bacteria can successfully treat bacterial keratitis caused by Gram-negative bacteria is showing that the predatory bacteria can kill Gramnegative bacteria isolated from keratitis cases. Secondly, we must show that the predatory bacteria are non-toxic and noninflammatory to human cells. In this study we tested whether predatory bacterial species Bdellovibrio bacteriovorus and Micavibrio aeruginosavorus were able to kill keratitis isolates of P. aeruginosa and S. marcescens, including antibiotic-resistant isolates, in vitro. We also tested whether B. bacteriovorus and M. aeruginosavorus were cytotoxic and inflammatory to human corneal-limbal epithelial cells (HCLE) in vitro. To further test whether these predatory bacteria were pathogenic, we used the Galleria monella pathogenesis model to determine whether B. bacteriovorus and M. aeruginosavorus reduced the viability of the G. monella larvae.

Bacterial strains, and growth conditions
The predatory bacteria used in the study were Bdellovibrio bacteriovorus strains HD100, 109J (ATCC 43826) and Micavibrio aeruginosavorus strain ARL-13 [17,18]. Ten Pseudomonas aeruginosa and nine Serratia marcescens isolates were examined in this study. All clinical isolates were isolated from keratitis patients by Dr. Ritterband at the New York Eye Infirmary and Regis Kowalski at the UPMC Eye Center. Many of these bacteria were fluoroquinolone resistant and previously used in antibiotic efficacy studies [19,20]. Pseudomonas aeruginosa and Serratia marcescens were grown with aeration and maintained in LB media. Predator stocklysates were prepared by co-culturing the predators in the presence of host bacteria suspended in diluted nutrient broth (DNB), a 1:10 dilution of nutrient broth amended with 3 mM MgCl2 and 2 mM CaCl 2 [13]. E. coli ZK2686 and P. aeruginosa UCBPP-PA14 were used as host cells for B. bacteriovorus and M. aeruginosavorus, respectively. The co-cultures were incubated on a rotary shaker at 30C. Fresh predator cultures were prepared as previously described [13,21,22], in brief, 2 ml of overnight-grown host cells (,1610 9 CFU/ml) were added to 2 ml of predatory bacteria taken from a stock-lysate and suspended in 20 ml DNB. The cocultures were incubated for 24 hrs at 30uC to reach ,1610 8 PFU/ml predator's cells. At this point, the lysates were filtered through a 0.45 mm Millex-HV pore-size filter (Millipore, Billerica, MA) in order to remove any residual host cells (harvested predator).

Predation Experiments
Predation experiments were conducted as previously described [13]. Five ml co-cultures were prepared by adding 0.5 ml of washed host cells to 0.5 ml of freshly harvested predator bacteria in DNB media. The cultures were incubated at 30uC for 48 hrs. The capability of B. bacteriovorus and M. aeruginosavorus to prey was evaluated by the reduction in prey cell viability in the predator cocultures. Cell viability was measured by dilution plating and CFU enumeration at 24 and 48 hrs. Each co-culture was conducted twice in triplicate.

Cytotoxicity assays
B. bacteriovorus and M. aeruginosavorus were prepared as described above using 5 ml of washed host cells and 5 ml of freshly harvested predator in 50 ml DNB media. The co-cultures were incubated for 24 and 36 hrs for B. bacteriovorus and M. aeruginosavorus, respectively. Thereafter, the lysates were filtered four times through a 0.45-mm Millex-HV pore-size filter to remove any residual host cells. The filtered harvested lysate was washed twice by centrifugation, 15,000 rpm for 30 min, and resuspended in 2 ml of DNB. Aliquots of the predator preparation was removed and plated on agar plates, to confirm that the samples are free from host cells. Samples were also taken to determine predator concentration using standard double-layered agar method [23]. Purifications were conducted on 3 and 4 separate occasions for M. aeruginosavorus and B. bacteriovorus, respectively.
Cytotoxicity assays were conducted as described [24] with some modifications, Human corneal-limbal epithelial (HCLE) [25] cells were cultured in 24-well plates until they were confluent. HCLE cells were grown in Keratinocyte serum-free medium (KSFM) with L-Glutamine, supplemented with 25 mg/ml BPE, 0.2 ng/ml EGF, and 1 mM CaCl 2 . HCLE cells were seeded without antibiotics to prevent interference of antibiotics in subsequent assays. The plates were incubated in an incubator at 37uC with 5% CO 2 . The wells were washed 3 times using PBS, pH 7.4 (Sigma-Aldrich, St. Louis, MO) and 450 ml of KSFM media was added to each well. Thereafter, wells were inoculated with 50 ml of each predator prep (,0.2-1.1610 9 PFU/well for B. bacteriovorus strains and ,2610 8 PFU/ well for M. aeruginosavorus) or predator free DNB control for maximum viability. Other controls included 0.25% of triton X-100 to measure total killing and 50 ml of DNB washed P. aeruginosa PA14 (,2.5610 7 CFU/well) as a positive control for bacterial cytotoxicity. Cell cultures were incubated for 4 and 24 hrs. After the incubation, aliquots of medium were removed from each well, centrifuged to remove bacteria, and stored at -20uC for subsequent pro-inflammatory cytokine analysis. The cells were then washed three times with PBS. Alamar Blue viability reagent (Invitrogen) in KSFM containing amikacin (10 mg/ml) was added to each well (500 ml/well) to assess cell viability. Fluorescence was measured after 1.5 hrs of incubation using a Synergy 2 microplate reader (Biotek) at 500/27 nm excitation and 620/40 nm emission wavelength. Experiments were conducted four times using B. bacteriovorus and three times using M. aeruginosavorus. Each experiment was conducted in quadruplicate (4 cell culture wells).

Cytokine analysis
HCLE supernatants were collected at 4 and 24 hrs post bacterial exposure. Four biological samples were used for ELISA and also tested on two different days with a different harvest sample set. IL-8 and TNFa ELISAs were run on the 4 and 24hour sample sets according to manufacturer's instructions (for IL-8, R & D Systems H ; for TNFa, Thermo Scientific Pierce Biotechnology). Upon completion of the assay, samples were read according to the manufacturer's instructions on a Synergy 2 plate reader (BioTek). Samples were graphed and statistical analysis was performed using GraphPad Prism 5 using one-way ANOVA with Tukey's post-hoc test.

Toxicity assay in Galleria mellonella invertebrate infection model
Viability experiments were conducted as described previously [26] with some modifications. Galleria mellonella larvae were obtained from New York worms (New York Worms, Glen Cove, NY). Larvae were in their final instar stages and had equal size and weight (330630 mg). B. bacteriovorus strains, 109J, HD100, and M. aeruginosavorus ARL-13 were grown and concentrated as descried for the cytotoxicity assays. The predators were suspended in phosphate buffered saline (PBS), pH 7.4 (Sigma-Aldrich, St. Louis, MO). Final predator concentration 2610 9 PFU/ml. Five microliters of each sample was injected into the hemocoel of each larva via the last left proleg using a Hamilton 25 ml syringe and 30.5gauge needle. Prior to use, the syringes were sterilized using bleach. The syringes were cleaned with 70% alcohol and distilled water and the needles were changed between every sample. In addition to the predators, worms were also injected with 5 ml of PBS buffer (negative control) and 5 ml of 8610 4 CFU/ml P. aeruginosa PA14 (positive control). After injection, the worms were incubated at 30C and the numbers of live larvae were scored for 11 days. Larvae were considered dead when they display no movement in response to gentle shaking of the dish or touching with a pipette tip. Six petri dishes containing 5 worms were assigned to each experimental and control groups (30 worms total for each sample).

Statistical Analysis
Graphpad Prism 5 software was used to perform statistical analysis. This analysis consisted of One-way ANOVA with Tukey's multiple comparison test.

Results
Predation by B. bacteriovorus and M. aeruginosavorus When exposed to the predators, all of the isolates were found to be susceptible to at least one predator. All S. marcescens isolates were found to be susceptible to predation by both B. bacteriovorus 109J and HD100, with cell reduction ranging from 1.7 log 10 to greater than 5 log 10 , compared to the initial cell concentration and the predator free control ( Table 1). 100% of the tested P. aeruginosa isolates were reduced by B. bacteriovorus HD100. However, only 70% of the isolates were reduced by the 109J strain. Eight of the 10 P. aeruginosa isolates were reduced by M. aeruginosavorus with a greater than 2-log 10 reduction measured for 87% of the predation positive strains (Table 1). It was previously shown that M. aeruginosavorus ARL-13 is able to use P. aeruginosa as prey; however, it is unable to utilize S. marcescens [13,22]. Therefore, in this study, M. aeruginosavorus was not tested on S. marcescens.

Cytotoxicity B. bacteriovorus and M. aeruginosavorus to HCLE cells
As a first step towards judging the suitability of predatory bacteria for ocular infections, we tested whether B. bacteriovorus and M. aeruginosavorus were cytotoxic to HCLE cells in vitro. Bacteria were co-incubated with HCLE cells at an MOI of ,100 with a known cytotoxic P. aeruginosa strain [27] as a positive control, and .800 for each of the predatory bacteria. Bacteria and HCLEs were co-incubated for 4 and 24 hours, then bacteria were removed and HCLE cells were tested for viability using the fluorescent vital stain alamar blue. Whereas P. aeruginosa was highly cytotoxic at both time points, the predatory bacteria were not significantly different from the mock at either 4 or 24 hrs (p.0.05, ANOVA, with Tukey's post-test) (Figure 1).

Production of pro-inflammatory cytokines following exposure to predatory bacteria
Because the predatory bacteria used in this study are Gramnegative bacteria, we predicted that they may cause adverse inflammatory effects upon exposure to ocular cells. Supernatants of HCLE cells co-incubated for four hrs with B. bacteriovorus and M. aeruginosavorus in the above noted cytotoxicity studies and were analyzed for proinflammatory cytokines IL-8 and TNF-a. These cytokines were chosen because they are expressed by ocular surface cells exposed to bacteria [28,29]. Whereas the positive control, P. aeruginosa, elicited a strong and significant induction of both cytokines, neither IL-8 nor TNF-a was found to be elevated above the mock negative control in HCLE supernatants coincubated with any of the three predatory bacteria (Figure 2A-B). The same pattern was observed after co-incubation of the predatory and positive control bacteria at 24 hrs for IL-8; however, TNF-a levels were undetectable (data not shown).
In vivo effect of predatory bacteria on G. mellonella G. mellonella was recently recognized as a suitable host model system to study microbial pathogenesis and innate immunity [30][31][32][33][34]. The G. mellonella system generally demonstrates a positive correlation between virulence factors found in mammals to those isolated in the insect, emphasizing the ability to utilize the system to bridge between in vitro studies and vertebrates [26]. In order to further evaluate the potential risk of using predatory bacteria G. mellonella worms were exposed to high concentrations of each predator. All of the worms injected with P. aeruginosa PA14 (4610 2 CFU/worm) were nonviable 24 hrs post-injection. However, worms injected with PBS, B. bacteriovorus 109J, HD100 (1.1610 7 PFU/worm) and M. aeruginosavorus ARL-13 (0.9610 7 PFU/worm) were all viable 24 hrs post-injection. 11 days post-injection, the viability of the worms were 96.6%, 100%, 96.6% and 93.3% viability for the control, B. bacteriovorus 109J, HD100 and M. aeruginosavorus ARL-13, respectively. Furthermore, no change in larva pigmentation was observed in the predator-infected worms (Data not shown). In G. mellonella the change in color indicates melanization caused by the host immune response to the microbial  challenge [26]. Thus, based on our finding it could be concluded, that unlike other pathogens, predatory bacteria do not provoke an aggressive innate immune response when injected.

Discussion
In this study, we have demonstrated that M. aeruginosavorus ARL-13 is able to prey on clinical isolates of P. aeruginosa isolated from ocular infections. This finding is in line with earlier reports regarding the host specificity of this predator and its ability to attack P. aeruginosa [13,22,35]. Our data also suggest that both B. bacteriovorus 109J and HD100 are capable of using S. marcescens as a host. This finding is in agreement with a study reporting a 3 log 10 reduction in cell viability of a non clinical isolate of S. marcescens following predation by B. bacteriovorus 109J [13]. In addition, B. bacteriovorus 109J and HD100 were able to prey on P. aeruginosa. However, the ability of the HD100 strain to attack was broader than that of 109J, preying on 100% and 70% of the isolates, respectively. The narrower ability of B. bacteriovorus 109J to prey on P. aeruginosa is aligned with a recent report in which B. bacteriovorus 109J was able to reduce only 1 out of 4 P. aeruginosa examined strains [5], as wall as earlier findings showing a limited ability of some B. bacteriovorus to prey on P. aeruginosa [8,36]. The different host-and intra-species strain-specificity demonstrated by Bdellovibrio spp. and Micavibrio spp. is well documented. Furthermore, predation ability was found to be specific to the B. bacteriovorus strain used, with different B. bacteriovorus strains demonstrating unique host specificity [8,13,22,[35][36][37][38]. As the mechanisms that define the predator's host specificity are not fully known, we could only speculate on the reason why certain bacteria are recognized as a host while others are not.
A major concern that needs to be addressed when evaluating the potential use of predatory bacteria as topical live-antibiotics are the risks associated with applying Gram-negative microorganisms to human cells. To this end, we have conducted cytotoxicity assays in which HCLE cells were exposed to high concentrations of the predators. Our data indicate that predatory bacteria are significantly less cytotoxic than the control P. aeruginosa. Low cytotoxicity was observed even after an extended exposure period. Although only a small number of studies regarding the safety of using predatory bacteria were conducted, the current data does support the claims that predatory bacteria could be considered safe. In a review article published by Dwider at al [4] the authors cited a study in which Bdellovibrio was injected into mice rabbits and guinea pigs and were found to be non pathogenic to the animals [39]. In a separate study conducted by Lenz and Hespell [40], the investigators attempted to grow the predatory bacteria B. bacteriovorus 109J, Bacteriovorax stolpii UKi2 and Peredibacter starrii A3.12 in the presence of eukaryotic cells. It was concluded that predatory bacteria are unable to grow on hamster kidney cells, mouse liver cells and bovine mammary gland cells. Furthermore, the predators did not grow within rabbit ova, following injection, nor were they able to grow in media containing rabbit ova extracts. Thus, it seems that mammalian cells could not be used by the predators as prey and could not support predator proliferation in the absent of a Gram-negative host. The inability of the predator to grow and establish itself in the intestinal microflora was shown in a study in which Bdellovibrio strain MS7 was fed to Channel catfish, northern leopard frogs, and mice. The predator viability also declined when inoculated into rabbit ileal loops. As Bdellovibrio could not proliferate in vivo it reduces the risk of permanent establishment within the mammalian host, rendering the predator, in the study, as nonpathogenic [41].
In a recent study conducted at the University of Nottingham [9], the in vivo effect of Bdellovibrio in a poultry vertebrate model was examined. It was found that B. bacteriovorus HD100, which was orally administered to chicks, caused no negative health effects on the birds. Furthermore, the authors were not able to recover viable Bdellovibrio from the gut flora, fecal matter or drinking water of the predator-inoculated birds, concluding that the risk of spreading predatory bacteria during treatment is low. Bdellovibrio treatment was also found to improve the well being of the birds colonized with Salmonella Enteritidis in the therapeutic trail.
An additional concern of applying live predatory bacteria is the risk of inadvertently causing inflammation which in could inhibit wound healing and increase the risk of tissue damage. In this study we have demonstrated that exposure to high doses of the predators did not elevate the production of the proinflammatory cytokines IL-8 and TNF-alpha by HCLE cells. Experimental evidence supports that neutrophils attracted by the bacteria-induced inflammation are a major cause of scarring and tissue damage associated with vision loss in keratitis [42]. The low cytotoxic activity of B. bacteriovorus HD100 LPS and its reduced ability to induce TNF-alpha and IL-6, compared to an E. coli control, was previously reported in a study using a human macrophage cell line [43]. The authors attributed their findings to the unique structure of the B. bacteriovorus LPS Lipid A molecule. Unlike Lipid A from many Gram-negative bacteria that contain negatively charged phosphate groups, the B. bacteriovorus Lipid A molecule has a-Dmannose residues which reduced its affinity to LPS receptors thereby lowering inflammation. Our data confirm that B. bacteriovorus HD100 and 109J do not enhance proinflammatory cytokines production. Our data also demonstrate, for the first time, that as reported for B. bacteriovorus o, M. aeruginosavorus also does not enhance inflammation.
Although, our data show that predatory bacteria have little or no adverse effect when applied to human cell cultures, we were interested to evaluate the effect of predatory bacteria in vivo. To address whether predatory bacteria a tolerated by eukaryotic cells in vivo, a G. mellonella microbial pathogenesis model was selected. Although, we did not measure the viability of the predators within the worm over time, we might still conclude that injecting relatively high doses of predatory bacteria do not provoke any measurable toxic or harmful effect to the worm. G. mellonella is recognized as a suitable host model system to study the pathogenesis of both bacteria and yeast and was used to examine pathogenic attributes of many human pathogens including pathogens associated with eye infections [26,30,32]. These studies established the use of G. mellonella for a variety of applications such as: examining microbial pathogenicity and lethality, evaluating microbial growth and proliferation, isolating virulence factors, and inspecting putative virulence mechanisms [30,[32][33][34]44]. Since the innate immune systems of mammals and insects have several features in common [31], G. mellonella could also be used as a model system for studying the host innate immune response to microbial infection as well as identifying microbial virulence factors that mediate the immune response [33,45].
The potential use of predatory bacteria as a bio-control agent to treat eye infections was first suggested some 40 years ago. In a study conducted in 1972 [46] the ''pro-biotic'' ability of E. coli and B. bacteriovorus to impact the pathogenesis of Shigella flexneri in animal modules was examined. It was shown that the B. bacteriovorus was able to reduce the severity of keratoconjunctivitis induced by S. flexneri in a rabbit keratoconjunctivitis model. The simultaneous inoculation of Bdellovibrio with S. flexneri was able to prevent the development of the infection. The rate of development of typical keratoconjunctivitis was also decreased when Bdellovibrio was administered within 48 hrs of the initial S. flexneri infection. In a more recent study, the ability of B. bacteriovorus 109J to inhibit growth of and reduce the adherence of Moraxella bovis to Madin-Darby bovine kidney (MDBK) cells, used to mimic bovine keratoconjunctivitis, was confirmed [47]. The ability of Bdellovibrio to survive and prey in human fluids was also demonstrated in an experiment in which B. bacteriovorus 109J was abele to significantly reduce biofilms of Aggregatibacter actinomycetemcomitans in the presence of human saliva that contains many of the same antimicrobial compounds as do tears [5].
In conclusion, our work demonstrates that predatory bacteria have the ability to attack ''real-life'' Gram-negative human pathogens associated with ocular infection. Furthermore, in vitro studies had revealed that the presence of high concentrations of predatory bacteria don't appear to be harmful to human cells. These findings, coupled with the ability of predator bacteria to prey in conditions that might be encountered in the eye, emphasize the potential use of applying predator bacteria as a topical agent to treat eye infections caused by pathogens which are resistant to traditional antimicrobials. The efficacy of predatory bacteria to control infection using in vivo models of ocular infection should be the focus of future studies. The long-term goal is to develop a topical predator bacteria product, which might include a single or multispecies predator bacteria mix and could be used alone or in concert with traditional antimicrobial therapies.