Injections of Predatory Bacteria Work Alongside Host Immune Cells to Treat Shigella Infection in Zebrafish Larvae

1The rise in antimicrobial-resistant (AMR) Gram-negative bacterial infections in hospital patients has prompted an urgent search for novel antibacterial agents. One candidate group is the naturally predatory bacteria Bdellovibrio bacteriovorus, which invade and kill a wide range of Gram-negative bacterial pathogens. Invasion is followed by pathogen rounding, after which the prey pathogen dies. Bdellovibrio replicate within the dead pathogen, which persists as a stable “bdelloplast” structure until being lysed 3–4 hr after invasion. Prey lysis releases mature Bdellovibrio progeny that can seek further bacterial victims (Figure 1A).

2Given that Bdellovibrio are ubiquitous in nature, it is likely that these bacteria are already being harmlessly ingested in food or water. Indeed, low levels of Bdellovibrio 16S rRNA have been detected in gut samples from healthy children. Although no equivalent studies have been conducted in adults, there is no known association of Bdellovibrio with disease. Previously, non-injected administrations of Bdellovibrio have been shown to reduce pathogen numbers by oral administration versus Salmonella enteritidis in the gut of chickens and by external application against eye infection in cattle. Emerging global antibiotic resistance calls for new injected therapies to target infected wounds and body compartments. The capability of injected Bdellovibrio to treat bacterial infections has not been tested, nor has the interaction of predatory bacteria with leukocytes, and successes here could dramatically expand the therapeutic scope of predatory bacteria against life-threatening infections. It is, therefore, important to consider the extent to which injected Bdellovibrio survive in the presence of a vertebrate immune system and whether leukocyte interactions aid or prevent pathogen killing in vivo. Measuring these parameters is crucial to develop the therapeutic potential of Bdellovibrio.

3The transparent zebrafish larva provides a unique opportunity to quantify and visualize in vivo both the spread or restriction of bacterial infection as well as bacterial interactions with immune cells. In particular, the zebrafish hindbrain ventricle is highly amenable to imaging, enabling us to follow injected Bdellovibrio ± pathogenic bacteria over time (Figure S1A). To first test for any pathogenic effects of Bdellovibrio inside a vertebrate host, we injected the hindbrain ventricle of zebrafish larvae at 3 days post-fertilization (dpf) with 1–10 × 104 plaque-forming units (PFUs) of mCherry B. bacteriovorus HD100 alone. This assessment was also essential to determine whether injected Bdellovibrio alone would survive in the zebrafish long enough to successfully prey on pathogen infections. Live-cell imaging showed a gradual reduction in fluorescence from mCherry-Bdellovibrio following injection, although some bacteria were clearly observed in vivo after 24 hr post-infection (hpi) (Figure 1B). For quantification of the survival of Bdellovibrio in vivo, injected larvae were homogenized and predatory bacteria enumerated over time post-infection (Figure 1C). Consistent with observations from live-cell imaging, we detected a steady decline in Bdellovibrio numbers from 2 hpi, and by 48 hpi the bacteria were largely eliminated. Similar results were obtained using a 10-fold higher dose of Bdellovibrio or in larvae injected 2 dpf when the immune system is less developed (Figures S1B–S1D). These results reflect the obligatory prey-dependent lifestyle of Bdellovibrio. They are able to survive within zebrafish for extended periods of time but are unable to replicate in the absence of prey bacteria, giving an opportunity for Bdellovibrio to act therapeutically before ultimately being cleared. Moreover, Bdellovibrio-injected zebrafish larvae displayed normal morphology and locomotive activity, with no signs of developmental toxicity or reduced viability (Figures 1D and S1E). These results raise no concern for the use of Bdellovibrio as a therapeutic agent in vivo, and its natural clearance could be viewed as a beneficial property of a limited treatment.

4Given that Bdellovibrio have no adverse effects on zebrafish, we tested their efficacy as an antibacterial agent against a streptomycin- and carbenicillin-resistant strain of Shigella flexneri M90T. Shigella is a Gram-negative enteroinvasive pathogen that is responsible for 163 million illness episodes and over 1 million deaths annually. Similar to many other Gram-negative pathogens in hospital patients, cases of antibiotic-resistant Shigella are rising. Bdellovibrio invade Gram-negative pathogens without a simple receptor-based recognition system, making Bdellovibrio resistance genetically complex for prey to acquire. Previous studies have used the zebrafish as a model organism to study the infection biology of S. flexneri M90T infection in vivo. We developed this model of Shigella infection to incorporate the additional injection of B. bacteriovorus and test its therapeutic potential. Here, 30–90 min after an otherwise lethal (at 48–72 hpi) hindbrain inoculation of GFP-Shigella M90T (>5 × 103 colony-forming units [CFUs]), we injected 1–2 × 105 PFUs of mCherry-Bdellovibrio into the hindbrain ventricle of zebrafish larvae. In the absence of Bdellovibrio injection, zebrafish larvae showed increasing fluorescence of GFP-Shigella (Figure 1E; Movie S1). Strikingly, larvae injected with Bdellovibrio showed diminishing fluorescence of GFP-Shigella in regions contacting mCherry-Bdellovibrio (Figure 1E; Movie S1). Consistent with this, Shigella enumeration demonstrated that zebrafish larvae injected with Bdellovibrio were able to control Shigella replication significantly better than those infected with Shigella alone (Figure 1F). Moreover, Bdellovibrio could rescue zebrafish from lethal Shigella infection, increasing survival by ∼35% at 72 hpi (Figure 1G).

5To confirm susceptibility of Shigella to Bdellovibrio predation, we performed co-incubation assays in vitro, and we observed mCherry-Bdellovibrio invade GFP-Shigella and, subsequently, induce prey rounding to form bdelloplasts (Figure 2A). In live-cell counting assays of viability in buffer, Shigella numbers were reduced >4,000-fold by Bdellovibrio treatment (Figure 2B). Plate reader assays measuring optical density of Shigella and fluorescence of mCherry-Bdellovibrio confirmed that reduction in Shigella numbers was followed by a rise in Bdellovibrio numbers after an ∼3 hr predatory cycle, during which they invade and grow within Shigella and then emerge (Figures 2C and 2D).

6To investigate Bdellovibrio predation of Shigella in vivo, we imaged the zebrafish tail muscle after sequential GFP-Shigella and mCherry-Bdellovibrio co-injections. Shigella are rod-shaped bacteria (∼0.5 × 2.0 μm); however, Shigella in the presence of Bdellovibrio were mostly rounded (∼1.0 × 1.0 μm), suggesting their invasion by Bdellovibrio inside zebrafish (Figure 2E). To investigate this further, we followed in vivo predator-prey interactions at the level of single cells (Figure 2F). Remarkably, confocal microscopy inside live zebrafish confirmed that mCherry-Bdellovibrio invade individual GFP-Shigella to form green rounded bdelloplasts containing red Bdellovibrio over a similar time frame to that seen in vitro (Movie S2). To test for Bdellovibrio predation of Shigella at the whole-animal level, we quantified Bdellovibrio over time after sequential GFP-Shigella and mCherry-Bdellovibrio co-injections (Figure 2G). As expected from experiments performed both in vitro (Figure 2D) and in vivo (Figure 1C), in the absence of Shigella, prey zebrafish larvae showed decreasing numbers of Bdellovibrio over the 24 hr following infection (Figure 2G). In contrast, zebrafish larvae infected with Shigella and Bdellovibrio showed significantly increased numbers of Bdellovibrio at 5 hpi, indicating predatory replication inside Shigella in vivo. Thus, live bacterial predation is occurring within zebrafish. To dissect the efficacy of Bdellovibrio therapy in the context of innate immunity, we next tested how leukocytes can affect the efficacy of Bdellovibrio predation in our zebrafish infection model.

7Time-lapse microscopy in the hindbrain showed that, by 6 hr following injection, the initially dispersed mass of mCherry-Bdellovibrio clustered and formed dynamic punctae of ∼10 μm. Similar observations were made after caudal vein injection of Bdellovibrio into the bloodstream of zebrafish larvae (Figure S2A). These observations suggest that Bdellovibrio can reside within leukocytes. Although little studied, some features of Bdellovibrio have been predicted to allow a degree of “silent running” in the immune system of a vertebrate. Bdellovibrio are characteristically small, 0.25 × 1.0 μm, comprise a single-membrane-sheathed flagellum, and have a modified mannosylated lipopolysaccharide (LPS) outer membrane. Furthermore, Bdellovibrio gene expression and surface protein production are significantly lower outside of bacterial prey hosts than inside. To test for innate immune detection of Bdellovibrio in vivo, we injected larvae with mCherry-Bdellovibrio, and quantified the expression of the pro-inflammatory cytokines interleukin 1 β (il1b) and tumor necrosis factor α (tnf-α) by qRT-PCR. Increased expression of both il1b and tnf-α was detected 4 hr after larval inoculation of Shigella, Bdellovibrio, or Shigella + Bdellovibrio combined (Figure S2B). Importantly, the cytokine response from Shigella + Bdellovibrio together is not additive beyond Shigella alone, demonstrating that Bdellovibrio is not solely stimulating a further immune response to help clear pathogenic bacteria.

8Cytokine signaling during zebrafish infection is typically accompanied by an active immune cell response. The innate immune system of zebrafish is highly homologous to that of humans, and responses are mediated by neutrophils and macrophages. Leukocytes do not typically reside in the hindbrain during steady-state conditions, making this site ideal to study directed leukocyte migration in response to injected bacteria. To test Bdellovibrio interactions with leukocytes, we used Tg(mpx:GFP)i114 transgenic larvae with GFP-neutrophils and Tg(mpeg1:Gal4-FF)gl25/Tg(UAS-E1b:nfsB.mCherry)c26 transgenic larvae with mCherry-macrophages. Consistent with a stimulated cytokine response (Figure S2B), imaging of Bdellovibrio hindbrain injections in zebrafish with these fluorescent leukocytes revealed that both neutrophils and macrophages localize to the site of injection (Figures 3A and 3B; Movie S3). Despite detection of Bdellovibrio by innate immune cells in vivo, quantification of leukocyte recruitment to the larval head, via image analysis, revealed only a slight increase over PBS controls in neutrophils (1.5-fold ± 0.1) and macrophages (1.2-fold ± 0.1) in the hindbrain at 6 hpi (Figures 3C and 3D).

9Analysis by confocal microscopy of Bdellovibrio-leukocyte interactions within live zebrafish confirmed that these bacteria are engulfed by both neutrophils and macrophages (Figures S2C and S2D; Movie S4). To assess the role of leukocytes in Bdellovibrio clearance, we performed experiments in immunocompromised zebrafish larvae using an antisense morpholino oligonucleotide targeting Pu.1, a zebrafish transcription factor driving myeloid gene expression. Larvae depleted of leukocytes, via prior injection of Pu.1-targeting morpholino into the one- to eight-cell-stage embryo, were injected at 3 dpf in the hindbrain with 1–2 × 105 PFUs of Bdellovibrio. In agreement with a role for leukocytes in the clearance of Bdellovibrio in vivo, significantly more predatory bacteria were recovered from larval homogenates of Pu.1 morphants as compared to controls (Figures 3E and S2E). Survival of control or Pu.1 morphants injected with Bdellovibrio was not significantly different from each other (Figure S2F), highlighting that prolonged exposure to Bdellovibrio is not detrimental to the health of an immune-compromised animal.

10To understand better the relative contributions of Bdellovibrio and the host immune system to Shigella clearance, we performed infection studies in zebrafish larvae in which leukocytes were depleted by Pu.1 morpholino. Control and Pu.1 morphants were injected with lethal hindbrain doses of Shigella as before and treated with PBS or Bdellovibrio (Figure S3A). Enumerations of Shigella from larval homogenates showed that treatment with Bdellovibrio reduced Shigella numbers in both immune-compromised and immune-competent zebrafish larvae (Figure 4A), but survival was significantly greater in immune-competent larvae (Figure 4B). Remarkably, these results show that maximal therapeutic benefit of Bdellovibrio against a Gram-negative bacterial infection is ultimately the product of eukaryotic leukocytes working cooperatively with prokaryotic predators.

11The Bdellovibrio predatory process, which we have shown to be effective in vivo, could possibly allow treatment of infection sites in humans. We propose that, following Bdellovibrio application, an early predatory-killing phase can reduce pathogen burden to a level manageable for clearance by the vertebrate innate immune system (Figure 4C). Predation will release small pathogen-derived fragments into the vertebrate, albeit in a digested state, after predator enzyme action. The limited immune stimulation by the injected Bdellovibrio (and possibly these pathogen fragments) is not detrimental to the well-being of zebrafish larvae. This limited immune stimulation may contribute to the clearance of Shigella, but active predation with predator-prey encounters in vivo occurs as early as 30 min post-injection (Figure 2F; Movie S1), time points before extensive leukocyte infiltration. Moreover, Bdellovibrio have replicated extensively in vivo at 5 hr, in a process dependent on Shigella killing (Figure 2G).

12Predator-prey interactions in our experiments occurred early, when a significant percentage of pathogens would be outside of leukocytes. Bdellovibrio engulfed by leukocytes inside zebrafish were detectable by their mCherry fluorescence within the vacuoles of those leukocytes for up to 6 hr (Movie S4). This suggests that predatory bacteria may persist intracellularly within immune cells. Future studies beyond this work will be testing whether intracellular Bdellovibrio are able to access, invade, and kill Gram-negative pathogens, which themselves invade leukocytes, such as Shigella.

13In conclusion, these results highlight the first successful use of Bdellovibrio in vivo as an injected antibacterial therapy, improving survival in live infected animals. The zebrafish infection model reveals host recognition and clearance of Bdellovibrio within days following treatment, a feature that provides a useful limitation to an applied therapy. Most importantly, we show that injected Bdellovibrio persist in vivo sufficiently long enough with predatory capacity to reduce numbers of pathogenic bacteria, before themselves being removed by immune action of the host.

14In our study, the prokaryotic predator Bdellovibrio works together with the host immune system, which would otherwise be overwhelmed by a Gram-negative infection. These biological experiments suggest that when tackling pathogenic AMR bacterial infections in a human medical setting, active predation and any associated/limited immune-stimulatory side effects can be beneficial as long as patient physiology and well-being can be supported. Future experiments will allow us to characterize the host immune response in more detail, determine how predators can be prepared with modified immune-stimulatory properties, and examine how multiple doses of predators can be applied in more long-lived infections. The data in this study represent key milestones in future use of Bdellovibrio as a “living antibiotic” in vivo, and they warrant further research into the development of predatory bacteria as an antibacterial agent for infected sites or wounds in higher vertebrates and, ultimately, humans. The strength of such prokaryotic-predator: eukaryotic-leukocyte combinations is an important therapeutic consideration as we move forward in responding to new Gram-negative bacterial threats.

15S.M. and R.E.S. acted jointly as senior authors of this collaboration; in line with journal policy, S.M. has been designated as lead contact as the majority of the work by all authors was performed at his lab. A.R.W., S.M., and R.E.S. designed the research and carried out analysis along with C.M. A.R.W., R.E.S., M.M.-M., C.M., C.L., and S.K. performed experiments with supervision and assistance from S.M. R.T. constructed the fluorescent Bdellovibrio strains, and A.R.W., C.M., M.M.-M., and S.M. carried out zebrafish husbandry. A.R.W., S.M., and R.E.S. wrote the manuscript with assistance from C.M. and input from all authors.