Collection and maintenance of the corals
A single Pocillopora damicornis colony was collected from a coral nursery located at ~ 8 m depth in the Gulf of Aqaba (Eilat, Israel) and transferred to an aquarium at the Interuniversity Institute for Marine Sciences (Eilat, Israel), where it was supplied with ambient flowing seawater (24 ± 2 °C) and natural light, shaded in order to mimic conditions experienced on the reef (i.e. 300–400 μmol photons m− 2 s− 1 at midday). The coral was fragmented into small pieces (5 × 5 mm) in April 2016 and left to recover for a week in the aquaria before being transported to the Weizmann Institute of Science (Rehovot, Israel). On arrival, the fragments were placed in a custom-built raceway chamber consisting of three separate channels, which were suspended above a temperature-controlled water reservoir. A submersible pump was added to the reservoir to circulate water between the two layers [24]. Separation of the two layers ensured that any water-loss by evaporation was minimal and thus stabilized salinity in the system. Photosynthesis-saturating light levels (150 μmol photons m− 2 s− 1) were provided by alternating blue and white LED strips, which were glued to a Plexiglas shelf positioned 10 cm above the glass raceway. The coral fragments were provided with conditions that matched those in Eilat (temperature: 25 ± 1 °C, pH: 8.1 ± 0.2, salinity: 40, light-dark cycle: 13.5 L/10.5 h D), for 1 week prior to the experiment to allow the fragments time to recover from any stress incurred during transportation. Experimental fragments were selected based on visual confirmation of health (i.e. skeleton covered by tissue, polyps extended and no paling of the coenosarc or excess mucus production). At this point, the temperature in the raceway was increased to 31 ± 1 °C for 3 d to prime the fragments for bacterial infection with Vibrio coralliilyticus [23].
Preparation of the inoculum
The modified V. coralliilyticus strain (YB2), which contains a plasmid encoding for the T3 DsRed fluorescent protein [24] was grown overnight in 15N-labeled growth media containing: 5 g L− 1 15N 98% Celltone powder (Cambridge Isotope Laboratories Inc., Tewksbury, MA, USA), 2 g L− 1 glucose, and 50 μg mL− 1 kanamycin dissolved in filtered seawater (0.22 μm; FSW). 12 h incubation at 31 °C with gentle shaking (150 rpm), resulted in an inoculum density of ~ 108 cells mL− 1 (estimated from flow cytometry counts). The bacterial suspension was centrifuged for 10 min at 3500 rpm. The supernatant was then discarded, replaced with an equivalent volume of FSW and vortexed, before it was returned to the incubator (31 °C, 0 rpm) for a further 4 h. This step, prior to inoculation, was crucial because it enhanced the secretion of zinc-metalloproteases, which are considered potent toxins in the infection process [22, 25, 26]. Importantly, this step did not reduce the 15N-labeling in the bacteria because the pathogens were already in the stationary phase and were thus, no longer dividing. Motile bacteria present in the supernatant, were collected immediately before the start of the experiment and transferred to sterile Corning® cell culture flasks (Sigma Aldrich, St. Louis, MI, USA).
Inoculation in the Microfluidic Coral Infection (MCI) experimental platform
Inoculations were conducted in the state-of-the-art MCI system using specifically-designed microfluidics chambers that were constructed from polydimethylsiloxane (PDMS). A detailed explanation of the system and how the microfluidics chambers are fabricated is provided by Gavish et al. (in revision), but the resulting product is a microchip that measures 5 × 1.5 × 5 cm (L × W × H) and contains four 250 μL volume chambers. Each chamber has an inlet and outlet tube made of polyethylene (ø = 0.8 mm), the latter of which is connected to a peristaltic pump, enabling similar flow rates (2.6 ± 0.8 mL h− 1) to be attained in all of the chambers. The chamber is sealed with an ApopTag® Plastic cover slip and transferred to the temperature-controlled microscope stage of an inverted fluorescence microscope (Olympus IX81, Tokyo, Japan). Temperature (31 ± 0.5 °C) was monitored via a probe, which was inserted directly into the PDMS chip.
Fragments were placed in the system 4 h before inoculation to give them time to acclimate to the conditions on stage.
Images were taken of the coral fragments immediately before the inoculation period to confirm the health of fragments (Fig. 1a-d). Three of the four chambers were designated ‘infection chambers’ and were subsequently exposed to the 108 cells mL− 1 inoculum, while the fourth chamber acted as a control and was exposed to FSW only. The inoculation period lasted 2 h. The inlet flow was then switched to FSW for the remaining incubation. Images were taken at four fixed positions on the coral surface, at 10 min intervals for the duration of the experiment using a Coolsnap HQ2 CCD camera (Photometrics, Tuscon, AZ, USA). Fluorescence was captured in three channels: green fluorescent protein (Ex: 490 nm, Em: 535 ± 50 nm), chlorophyll (Ex: 490 nm, Em: 660 ± 50 nm), and DsRed (Ex: 555 ± 20 nm, Em: 590 ± 33 nm). Between fluorescence imaging, the corals were provided with 250 μmol photons m− 2 s− 1 of white light, which was supplied by the microscopes transmitted light function. Because images were acquired in real time, we were able to visualize the progression of the infection and use the images to make a decision as to when to fix the samples (in 4% paraformaldehyde and 0.1% glutaraldehyde) for subsequent TEM/NanoSIMS imaging. Fragments were thus fixed at different stages of the infection process in line with the occurrence of symptoms of disease, assessed visually by the state of the tissue (confluence, coenosarc tearing, and polyp isolation).
TEM and NanoSIMS imaging
Coral fragments were rinsed thoroughly in Sörensen sucrose phosphate buffer (0.1 M phosphate at pH 7.5, 0.6 M sucrose, 1 mM CaCl2) and decalcified in 0.5 M ethylenediaminetetraacetic acid (EDTA at pH 8) for 3 d at 4 °C. The remaining tissue was micro-dissected into single polyps using a binocular microscope. Polyps were post-fixed for 1 h in 1% osmium tetroxide, dissolved in distilled water. A series of washes (4 × 10 min) in distilled water followed, before the samples were dehydrated in a stepwise series of ethanol washes (3 × 10 min at 50, 70, 90, and 100%, respectively), and embedded in Spurr’s resin. One polyp per fragment was selected at random for processing and thin (70 nm) and semi-thin sections (500 nm) were cut using a 45° diamond knife (Diatome, Hatfield, PA, USA). Thin sections were stained with 4% uranyl acetate and Reynold’s lead citrate solution and imaged using a Philips CM 100 transmission electron microscope, located in the Electron Microscopy Facility (EMF) at the University of Lausanne (Switzerland). Initially we were unsure where the V. coralliilyticus would be localized and how abundant the pathogens would be in the tissue, so we created multiple high-resolution montages. These sections then were gold-coated and the same areas were imaged using a NanoSIMS 50 L ion microprobe.
In the NanoSIMS, secondary ions were obtained by bombarding the sample with a beam of 16 keV Cs+ primary ions, focused to a spot-size of about 150 nm. The secondary ions 14N12C− and 15N12C− were counted in individual electron-multiplier detectors at a mass resolution power of about 9000 (Cameca definition), which is sufficient to resolve all potential interferences in the mass spectrum. Isotopic images (50 × 50 μm in size), were generated by rastering the primary beam across the surface of the sample, controlling the dwell time spent on each pixel (5 ms), the number of pixels (256 × 256), and the number of layers (5) for each image. Four tissues were analysed in each polyp: the oral epidermis, the oral gastrodermis, the aboral gastrodermis, and the mesenterial filaments (the majority of which consist of gastrodermis tissue; [27]). It was not possible to analyze the calicodermis, because this tissue layer was not preserved in sections. Between 5 and 14 images were obtained per tissue per coral fragment (n = 73 images in total). High-resolution images, typically 12 × 12 μm2, of specific, highly 15N-enriched sub-cellular structures were also obtained with a lateral resolution of ~ 100 nm. The software L’IMAGE (created by Dr. Larry Nittler, Carnegie Institution of Washington) was used to produce drift-corrected 15N-enrichment maps. All 15N-enrichment levels are expressed in the delta-notation:
$$ {\updelta}^{15}\mathrm{N}\ \left({\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} \right)=\left(\left({\mathrm{R}}_{\mathrm{sample}}/{\mathrm{R}}_{\mathrm{control}}\right)-1\right)\times 1000, $$
where Rsample is the 15N/14N ratio measured in the sample, and Rcontrol is the measured ratio of a sample with natural 15N/14N ratio, prepared and analysed in an identical manner. For easy comparison, a scale from 0 to 4000 was applied to the δ15N (‰) images. This image, in conjunction with the 12C14N− image, was used to draw regions of interest (ROI) around the tissue(s) present. The average δ15N (‰) was calculated for each tissue. The same method was used to define ROIs around 15N-hotspots (areas enriched above background levels) present in the tissues. We defined a “hotspot” as a ROI with a δ15N > 300 and a size > 10 pixels. The density of hotspots was subsequently calculated by dividing the number of hotspots by the area of tissue, and expressed as the number of hotspots per μm2.
Statistical analysis
The tissue enrichment data was log-transformed to achieve normality (Kolgomorov-Smirnov, p > 0.05). The importance of time (ordinal factor: 2.5, 6, or 22) and tissue (nominal factor: oral epidermis, oral gastrodermis, aboral gastrodermis, and mesenterial filament) were analysed using a two-way analysis of variance (ANOVA). A Tukey’s honestly significant difference post-hoc test was used to identify where the differences lay in the event of a significant interaction being found. Analysing the hotspot density data was complicated by the number of images that contained zero hotspots (40 out of 111) and the high variability between images (which ranged from zero to 0.039 hotspots per μm2). The data could not be transformed to achieve normality and did not meet the criteria for the homogeneity of variance either, so a non-parametric Kruskal-Wallis test was used to compare structures at different time-points. In the event of a significant difference being found, a Nemanyi post-hoc test was used to identify where the differences lay.