Zebrafish (D. rerio) strains used in this study were predominantly albino mutants (slc45a2^b4/+) and transgenic fish of the Tg(lyz:DsRED2)nz50 line that produce red fluorescent protein in cells of the myelomonocytic lineage able to migrate to inflammatory sites and phagocytose bacteria (Hall et al., 2007), primarily neutrophils from 50 h post fertilization (hpf) (Clatworthy et al., 2009). In addition, the Tg(fli1a:eGFP) line which produces green fluorescent protein in endothelial cells, was used to visualize the vascular system (Lawson and Weinstein, 2002). Adult fish were kept at a 14/10 h light/dark cycle at a pH of 7.5 and 27°C. Eggs were obtained from natural spawning between adult fish which were set up pairwise in separate breeding tanks. Embryos were raised in petri dishes with E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM MgSO₄) containing 0.3 μg/ml of methylene blue at 28°C. From 24 hpf, 0.003% 1-phenyl-2-thiourea (PTU) was added to prevent melanin synthesis. The albino mutants lack melanised melanophores, and for these PTU treatment was not necessary. Staging of embryos was performed according to Kimmel et al. (1995).

Research was conducted with ethics approval (no. 216/2012) from the animal research ethics committee of the Veterinary Office, Public Health Department, Canton of Zurich (Switzerland). Larvae were maintained up until 7 days post fertilization (dpf), at which time all were euthanized by applying an overdose of 4 g/l buffered tricaine (MS-222, Ethyl 3-aminobenzoate methanesulfonate, Sigma-Aldrich) in accordance with ethical procedures.

Waddlia chondrophila strain WSU 86-1044 (ATCC VR-1470) were grown within Acanthamoeba castellanii strain ATCC 30010 in 25 cm² cell culture flasks containing 10 ml of peptone yeast extract glucose (PYG) broth (Greub and Raoult, 2002). Cultures were harvested after 5 days and filtered through a 5 μm membrane to remove remaining amoebae. The flow-through was centrifuged at 7000 g for 15 min. The resulting pellet of bacterial elementary bodies (EBs) was then suspended in E3 medium for bath immersion or phosphate buffered saline (PBS) for microinjection experiments. The infectivity of W. chondrophila was tested in Epithelioma Papulosum Cyprini (EPC) cells, to investigate the infection process and morphology of the bacteria in these fish epithelial cells (Supplementary Material Figure S1). Subsequently, Inclusion Forming Units (IFU) of the cultures were determined by infecting monolayers of EPC cells in 24-well plates with a 10-fold dilution series of 1 μl of the bacterial suspension at 28°C. After 24 h cells were fixed with 4% paraformaldehyde and subsequently stained with a primary rabbit anti-Waddlia antibody and detected with a secondary goat anti-rabbit-IgG conjugated to a fluorescent AlexaFluor dye. After imaging with a fluorescent microscope, inclusions were counted in Imaris (Bitplane) to calculate a mean IFU value.

Zebrafish embryos between 24 hpf and 4 dpf were incubated in groups of 15 for each time point and each condition in 24-well plates with E3 medium containing 2 × 10⁹ IFU/ml of W. chondrophila at 28°C. Embryos younger than 48 hpf were manually dechorionated prior to immersion. After 4 h of incubation embryos were washed twice in fresh E3 medium and transferred to 6-well plates containing 4 ml of E3 medium per well and further incubated at 28°C. Embryos were then observed under a binocular microscope for signs of disease and survival twice a day. At several time points embryos from each group were euthanized in E3 medium containing 4 mg/ml buffered tricaine (MS-222).

For microinjections of W. chondrophila into zebrafish larvae bacteria were first harvested from a 5 days old amoebal co-culture as described above. The concentration of the W. chondrophila EBs was adjusted to 1000–2000 IFU/nl in PBS and 0.085% phenol red was added to visualize the injection procedure. Injections were done using pulled borosilicate glass microcapillary injection needles (Science Products, 1210332, 1 mm O.D. × 0.78 mm I.D.) and a PV830 Pneumatic PicoPump (World Precision Instruments). Prior to intravenious injections embryos of 2 dpf were manually dechorionated and anesthetised with 200 mg/l buffered tricaine (MS-222). Afterwards embryos were aligned on an agar plate and injected with 1 nl of the W. chondrophila suspension into the Duct of Cuvier, also known as common cardinal vein. Prior determination of the injected volume was performed by injection of a droplet into mineral oil and measurement of its approx. diameter over a scale bar. For swim bladder injections 4 dpf larvae were treated similarly but dechorionisation was not necessary because larvae were already hatched. 4 dpf larvae were then injected with a volume of 2 nl into the lumen of the swim bladder. After injections, infected larvae were allowed to recover in a petri dish with fresh E3 medium for 15 min. Subsequently, larvae were transferred in 6-well plates in groups of about 15 larvae in 4 ml E3 medium per well, incubated at 28°C and observed under a stereomicroscope twice a day. Samples for Immunofluorescence (IF), Transmission Electron Microscopy (TEM) and quantitative Polymerase Chain Reaction (qPCR) were taken at 0, 12, 24, 36, 48, 60 and 72 h post infection (hpi). Sampled larvae were euthanized with an overdose of 4 g/l buffered tricaine and transferred into different buffers and fixatives for subsequent analyses respectively.

For IF stainings, whole zebrafish larvae were fixed in 4 % paraformaldehyde at 4°C overnight followed by 100 % methanol overnight at −20°C. Samples were rehydrated in 50% methanol for 5 min and subsequently in H₂O for 1 h before blocking in PBDT (PBS containing 1% BSA, 1% DMSO, 0.5% Triton X-100, 2.5% goat serum) for 6–8 h at room temperature. Larvae were then incubated with primary antibody overnight at 4°C. Primary antibody was detected by incubation with a secondary goat anti-IgG antibody conjugated to a fluorescent AlexaFluor dye (Life Technologies) at 4°C overnight. Additionally 4′-6-diamin-2-phenylindole (DAPI) was added to visualize bacterial and host DNA. Stained larvae were prepared for microscopy on objective slides mounted in 1.5% agarose, 50% glycerol and screened under a fluorescence microscope. Positive samples were subsequently screened in more detail for W. chondrophila inclusions by Confocal Laser Scanning Microscopy (CLSM).

For histological examination, whole zebrafish larvae were fixed in 4% paraformaldehyde at 4°C and embedded in cubes of cooked egg white in order to position them correctly for histological sections. These cubes containing the larvae were dehydrated in an ascending alcohol series ending in xylene and afterwards embedded in paraffin. Paraffin blocks were cut in 2–3 μm thin sections, mounted on glass slides and stained using a routine protocol for haematoxylin and eosin (HE) staining.

Overview images were taken with an upright light microscope (Olympus BX61) with both bright field and fluorescence modules. The fluorescence filter cube used was optimized for DAPI/FITC/TRIC. For higher resolution images, 3D-image stacks of whole mount samples were prepared using CLSM (Leica TCS SP5, Leica Microsystems). Various combinations of the fluorophors AlexaFluor dyes 594, 546, 488, GFP, dsRED, and DAPI were sequentially excited in descending series with the 594, 561, 488, and 405 nm laser lines, with emission signals collected within the respective range of wave lengths. Three-dimensional image stacks were collected sequentially (to prevent blue-green–red channel cross-talk) and according to Nyquist criteria and deconvolved using HuygensPro via the Huygens Remote Manager v2.1.2 (SVI, Netherlands). Images were further analyzed with Imaris 7.6.1 (Bitplane, Zurich, Switzerland) and aligned with Adobe Photoshop Elements 12. Fluorescent cells were quantified with the Imaris fluorescent spot counting tool.

For electron microscopy, larvae were fixed in a mixed solution of 1% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.5 at 4°C overnight. Samples were prepared by embedding into epoxy resin and for TEM microscopy according to standard procedures. Epoxy resin blocks were screened for larval location using semithin sections (1 μm) which were stained with toluidine blue (Sigma-Aldrich) to visualize tissue. Ultrathin sections (80 nm) were mounted on copper grids (Merck Eurolab AG, Dietlikon, Switzerland), contrasted with uranyl acetate dihydrate (Sigma-Aldrich) and lead citrate (Merck Eurolab AG) and investigated using a Philips CM10 transmission electron microscope. Images were processed with Imaris (Bitplane) and assembled for publication using Adobe Photoshop.

A qPCR system was designed against the 16S rRNA gene from W. chondrophila (forward: 5′- AGTCCGGCTACACCAAGTATGC-3′, reverse: 5′-TGGCGAAGGCGGTTTTC-3′, probe: 5′-FAM-TTCGCTCCCCTAGCTTTCGGGCAT-TAMRA-3′), allowing quantification of the bacterial load of individual infected larvae. The TaqMan qPCR system was designed and validated against quantified serial dilutions of the target sequence cloned into the pCR2.1 plasmid and against total DNA extracts of non-infected larvae. Serial dilutions of pCR2.1 containing the target sequence were used as standards in each run. Total DNA of individual larvae was extracted with a MagNA Pure LC (Roche) robot and eluted in 100 μl elution buffer. Reactions were carried out on a StepOne Plus real-time PCR system (Applied Biosystems). TaqMan Fast Advanced reagents (Applied Biosystems) were used according to the manufacturer's instructions with 5 μl input DNA in a total reaction volume of 20 μl.

Knockdown of MyD88 expression was done by standard microinjection of 1 nl of a 5 mM solution of anti-Myd88 morpholino (5′-GTTAAACACTGACCCTGTGGATCAT-3′, Gene Tools; Bates et al., 2007; Cambier et al., 2014) into the early zygote immediately after fertilization using borosilicate glass microcapillary injection needles (Harvard Apparatus, 30-0019, 1 mm O.D. × 0.58 mm I.D.) and an Eppendorf FemtoJet. A group of control larvae was injected with 1 nl of a control morpholino (5′-CCTCTTACCTCAGTTACAATTTATA-3′, Gene Tools) with no known target in zebrafish at the same concentration in parallel. Levels of MyD88 protein in morphant and control larvae were analyzed by Western Blot at 3, 4, and 5 dpf with a mouse anti-myd88 antibody, detected with a goat anti-mouse-IgG antibody conjugated to horseradish peroxidase (HRP).

In order to first establish whether W. chondrophila can infect zebrafish through an oral or dermal route, embryos and larvae of different stages were incubated in a suspension of W. chondrophila infectious particles or EBs, in a bath-immersion experiment. While younger embryos and larvae of up to 72 hpf (hours post fertilization) stages could not be infected with this method, 4 dpf (days post fertilization) larvae appeared to swallow (Figure 1A) the bacteria and interestingly, succumbed to an infection of their swim bladder which we detected by IF staining (Figure 1B).

The swim bladder provides the only possibility to examine infections of air-exposed epithelium in fish and, as such, is a putative model for equivalent epithelia in humans, in particular the lungs. In this context, the spontaneous bath-immersion induced infection of the swim bladder and especially the attachment to the cilia of the epithelium was an exciting finding, although infection via bath immersion carried inherent experimental variability due to varying local concentrations of W. chondrophila and different gulping rates of individual larvae. Therefore, we tried direct microinjection of 10³ W. chondrophila EBs into the swim bladder lumen of 4 dpf larvae (Figure 2A). The resulting infection was comparable to that seen using bath-immersion but under more controlled conditions, confirming that directed injection can be used as a more reproducible method of establishing infection. In addition to live W. chondrophila, two groups of control embryos were injected with either heat-inactivated W. chondrophila or PBS in parallel. Samples were taken at 4, 24, 48, 72, and 96 hpi and screened for the presence of W. chondrophila inclusions by confocal laser scanning microscopy CLSM (Laser Scanning Microscopy) and TEM (Transmission Electron Microscopy). W. chondrophila inclusions were detected in the epithelium of the swim bladder as well as in adjacent tissues at 24–48 hpi (Figure 2B). In haematoxylin and eosin (HE) stained samples, the swim bladder walls of infected larvae were markedly thickened when compared with control animals (Figures 2C,D). The epithelial cells lining the cavity were piled up and were no longer a single cell row, with the cells themselves exhibiting a cuboidal instead of a flattened morphology (Figure 2E). The epithelium was surrounded by a moderate proliferation of fibroblasts resulting in a fibrosis, as has been described in mice for chlamydial lung infections (Jupelli et al., 2013). Neutrophils were also found to have migrated into the epithelial cell layers (Figure 2F). No increase in mortality was observed up until the end of the observation period of 3 days post infection (dpi) and infected larvae showed no altered behavior compared to control larvae.

Chlamydial infection of vascular tissue has been an area of intense interest in the past, and not without controversy in respect to atherosclerosis (recently reviewed by Campbell and Rosenfeld (2014). It is also an aspect which is difficult to follow in vivo in other animal models but which is readily accessible in younger zebrafish larval stages by intravenous injection. We injected 48 hpf embryos intravenously via the Duct of Cuvier, also known as common cardinal vein with 10³–10⁴ W. chondrophila EBs. In comparison to the non-lethal swim bladder infection, a systemic infection caused a dosage dependent mortality rate of injected larvae (Figures 3A,B). The injection resulted in a rapid systemic infection with mortalities up to 100% within the first 24 h for the highest tested dosage (10⁴) and a LD₅₀ within 48 hpi of approximately 5 × 10³. A dosage of 2 × 10³ W. chondrophila EBs was chosen for the following experiments to produce moderate mortalities of between 20 and 30% at 72 hpi. With that dosage larvae showed an increasingly impaired blood circulation between 24 and 48 hpi, resulting in the formation of a pericardial oedema (Figure 4A). Histologically, no tissue alterations could be seen.

To follow the distribution of W. chondrophila in the vascular system we used the transgenic Tg(fli1a:eGFP) line, which expresses green fluorescent protein in all endothelial cells. Tracking the infection with IF using an anti-Waddlia antibody, the maximal number of W. chondrophila inclusions were found at 36–48 hpi, distributed throughout the whole embryo, and with inclusions located both within the vasculature and beyond the bounds of the endothelium (Figures 4B–D). Cell types that were identified to be susceptible to W. chondrophila invasion during a systemic infection were predominantly endothelial cells and phagocytosing innate immune cells. Especially in regions where the blood is flowing more slowly, such as the fine capillary network of the tail muscles and veins, infection of the endothelium was more common.

To investigate the morphological features of W. chondrophila infection in zebrafish larvae, we performed detailed TEM and CLSM analyses of infected larvae after IF staining with an anti-Waddlia antibody and an anti-OxPhosIV antibody to stain mitochondria. W. chondrophila could be found to infect different cells of the zebrafish, predominantly epithelial cells of the swim bladder, phagocytes of the innate immune system and endothelial cells (Figure 5). The chlamydial inclusions inside these cells exhibited features typical for a Waddlia infection, including the transformation from the round shaped smaller (up to 0.5 μm in diameter) metabolically almost inactive but infectious EBs with highly condensed DNA, to the larger (about 0.9 μm in diameter) replicating reticulate bodies (RBs) with finely distributed chromatin inside a bacteria containing vacuole. Further, host cell mitochondria were readily recruited and closely associated with the inclusion, a characteristic for Waddlia infection in vitro (Croxatto and Greub, 2010) and now shown in vivo. While epithelial and endothelial cells usually contained a single perinuclear inclusion, individual phagocytes could harbor several inclusions of dividing bacteria. In TEM images of larvae that were injected into the swim bladder, some bacteria were in close relation to microvilli on the epithelial cell surface while others were found within the cytoplasm of these cells. The morphological structure of these bacteria on the surface resembled as well the typical morphology known for infectious particles of W. chondrophila (Rusconi et al., 2013). The perinuclear bacteria containing vacuoles inside epithelial cells were closely associated with host cell mitochondria and ER and are typical for actively replicating W. chondrophila observed in vitro (Rusconi et al., 2013).

The zebrafish innate immune system reacts to W. chondrophila with a strong recruitment of inflammatory cells to the infection site, as we already observed in our histology analysis (Figure 2).

This recruitment of inflammatory cells could be confirmed by using larvae of the transgenic Tg(lyzC:dsRed^nz50) line, whose neutrophils express red fluorescent protein, easily assessable by fluorescence microscopy. Since the zebrafish swim bladder is an enclosed compartment, usually devoid of immune cells, we used our swim bladder infection assay in order to quantify the induced recruitment of neutrophils following microinjection. In these larvae, large numbers of neutrophils could be detected clustered inside the swim bladder (Figure 6B), compared to PBS injected control larvae (Figure 6A). We quantified the recruitment by counting the fluorescent neutrophils using the image analysis package of Imaris (Video S1). The results show further a significantly increased response of neutrophil invasion to live W. chondrophila compared to heat-inactivated bacteria (Figure 6C). Phagocytosed W. chondrophila show the ability to survive and replicate inside neutrophils (Figures 6D,E).

In vitro studies have previously shown that W. chondrophila is susceptible to antibiotics of the tetracycline group, but is resistant to β-lactam antibiotics (Goy and Greub, 2009). Therefore we compared the effect of tetracycline and ampicillin on W. chondrophila infections in vivo by using our systemic infection assay, for which we compared the survival rates of differently treated larvae. In order to additionally quantify and compare the bacterial burden of individual larvae during the infection, we furthermore developed a specific quantitative PCR targeting the two-copy W. chondrophila 16S rRNA gene.

While treatment with tetracycline significantly (p < 0.01) reduced W. chondrophila replication in vivo and increased the survival rate (Figure 7), treatment with ampicillin had a small but non-significant impact on the bacterial load compared to untreated larvae. Formation and distribution of W. chondrophila inclusions in untreated and ampicillin treated larvae were similar. In tetracycline treated larvae inclusion formation was strongly attenuated. Furthermore heat-inactivated bacteria were quickly cleared from the system.

The signaling molecule MyD88 has been shown to play a central role for pathogen recognition, immune reaction and survival of infected individuals during Chlamydia infection in several in vivo studies. Now we wished to address how MyD88 mediated signaling affects the infection with W. chondrophila in our zebrafish model. To investigate whether MyD88 signaling has a function during W. chondrophila infection in zebrafish, we knocked down MyD88 expression by injection of a specific morpholino at the 1-cell stage. The resulting knockdown lasted for up to 4 dpf as determined by Western Blot (Figure 8A). Survival rates and bacterial loads were compared between control and MyD88 knockdown larvae during systemic infection. Our results show, that MyD88 morphant larvae exhibited a 20% lower survival rate compared to control larvae with a similar initial injection dose of 2 × 10³ W. chondrophila EBs (Figure 8B). Furthermore, the bacterial load of MyD88 morphant larvae at 36 hpi was higher compared to control larvae (Figure 8C).

Additionally, to compare the total numbers of neutrophils during a systemic infection between morphant and control fish we injected transgenic Tg(lyzC:dsRed^nz50) larvae either with sterile PBS, heat-inactivated W. chondrophila or live W. chondrophila (10³ EBs) at 2 dpf. Subsequently, total numbers of neutrophils within individual larvae were counted with the cell counting tool in Imaris. At 36 hpi, these numbers were found to be significantly depleted upon the systemic infection with W. chondrophila compared to PBS injected larvae, although no differences between MyD88 morphant and control fish were evident (Figure 8D).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.