Microbial bioenergetics of coral-algal interactions

Experimental design and sampling

Independent 18.9 L aquaria were filled with artificial seawater. One piece of coral (Favia rotunda) and one mass of algae (Chaetomorpha crassa) were placed in direct contact with one another in each tank at distances of >5 cm from aquaria walls (Figs. 1A and 1B). For each replicate (n = 8) a planar oxygen optode sheet (15 × 15 × 20 cm) was vertically fitted around the coral, algae, and interface (Fig. 1) to provide a two-dimensional colorimetric analysis of the O₂ concentration (Fig. 1C). Polyethylene tubing (5 mm) connected to 21 gauge luer lock needles mounted to the back of the optode sheet served as sampling ports through which water was suctioned directly off the surface of the coral, the algae, and the interface between the two organisms via a 10 mL syringe. The aquaria were equilibrated for 36 h at 26–27 °C with a constant recirculating water flow at 75 L/hr provided by a submersible aquarium water pump (Aquatic Warehouse, San Diego, California, USA) with 12-hour dark/light cycles. Irradiance levels (150–180 µE/m²/s), temperature (26–27 °C), and flow rates (75 L/hr) were relatively similar to shallow tropical reef ecosystems. After preliminary equilibration, pictures were taken of the oxygen optodes, after which 5 mL of water was sampled from each of the three benthic areas (coral, algae, and interface) via the aforementioned sampling ports. Ambient water-column samples were also taken. This sampling scheme was conducted on eight separate biological replicates.

Sample processing

Each sample was divided into aliquots for flow cytometry (1 mL), epifluorescence microscopy (1 mL), and calorimetry (3 mL). Flow cytometry samples were fixed with microscopy grade glutaraldehyde at 2% final concentration and then flash frozen in liquid N₂ before being stored at −80 °C. Microscopy samples were fixed with microscopy grade paraformaldehyde at 1% final concentration, vacuum filtered onto a 0.02 µm Anodisc filter (Whatman inc., Florham Park, NJ, USA), stained with SYBR Gold (5 X final concentration; Invitrogen, Carlsbad, CA, USA), and mounted on microscope slides. Calorimetry samples were weighed and placed in a TAM III isothermal heat conduction microcalorimeter (TA Instuments, New Castle, DE, USA) for 8 h.

Sample analyses

Flow cytometry samples were thawed at 37 °C and pipetted into a flat bottomed 96-well-plate on ice. Analysis was performed with a BD FACS-Canto via the high-throughput sampler unit for enumeration utilizing the methods of Zubkov et al. (2000). For enumeration of autotrophic microbe populations, 100 µL of sample were collected for analysis on standard mode using the 488 nm excitation laser. Bivariate plots were used to analyze the sample for chlorophyll fluorescence (red), which were done in the PerCP-Cy5-5 channel (670 nm long pass filter preceded by a 655 nm long pass mirror) and the phycoerytherin (PE) channel (585/42 band pass filter preceded by a 556 nm long pass mirror). For autotrophic populations, threshold gating was determined by using 0.02 µm filtered seawater. Yellow-green fluorescent microsphere beads (0.75 µm) were used to control for sample volume analyzed. To control for consistency between plates and daily runs, a standard seawater sample collected from San Diego, California was used. This control was used for both autotrophic and total microbe enumerations. One hundred µL sample volume was collected and analyzed for SYBR fluorescence, which was excited by the 488 nm laser and detected in the FITC channel (530/30 nm band pass filter preceded by a 502 nm long pass mirror). Threshold gating for the heterotrophic populations was determined by using unstained representative coral reef water, and to verify the amount of background associated with the instrument. Data was collected on FACSDiva 6.1.1 and analyzed using FlowJo 7.6.5. In order to enumerate heterotrophic microbe populations, the comparative autotrophic populations were subtracted from the total microbial counts.

The oxygen optode sheets were prepared and analyzed following (Haas et al., 2013). Prior to each experiment oxygen optodes were calibrated under identical temperatures (26 °C) with known oxygen concentrations. For this calibration, filtered seawater was dosed with nitrogen gas to obtain eight different concentrations of oxygen ranging from 100% air-saturation to anoxia. Images were taken approximately every 30 µM step until the seawater was anoxic. The resulting oxygen concentrations were constantly measured for comparison using an LBOD101 luminescent oxygen probe.

During experiments, oxygen optodes were imaged using a G11 (Cannon, USA) camera placed at ∼25 cm from the aquarium. All images were captured in RAW format with identical settings of ISO 200, f ∖8 and shutter speed of 1.3 s. Four Rebel Royal Blue light emitting diodes (LED) with a λ-peak of 445 nm (Philips-Luxeon; Philips Lighting, Toronto, Canada) were used as the excitation source in combination with a 470 nm short pass filter (UQG Optics, UK). To prevent the excitation source from contaminating the luminescent signal, a Schott 530 nm long pass filter (UQG Optics, Cambridge, UK) was mounted on the camera lens. All images were taken in the absence of ambient light.

Absolute O₂ concentrations were determined by measuring the red pixel intensity (oxygen-dependent dye) to green pixel intensity (oxygen-independent antenna dye) using the methods of Haas et al. (2013). Briefly, each RAW file was converted into two16-bit TIFF images (i.e., red and averaged green) (RawHide v0.88.001; My-Spot Software, USA). The pixel information from the red and green channel images were imported into MATLAB and further analyzed using the image toolbox. The red and green intensity values were obtained for each pixel and used to calculate the pixel intensity ration.

Isothermal calorimetry was conducted using a TAM III multi-channel isothermal heat-conduction microcalorimeter. Samples were placed in the machine within an hour of the sampling process. Samples were lowered into the measurement position after a 30 min equilibration period. Measurements on temperature (K), cumulative heat (J), and instantaneous heat flow (W) were taken continuously for each channel for 8 h at 299 K. Heat and heat flow were subsequently normalized by the total mass (g) of the respective sample.

Statistical analyses

All tests were conducted with an alpha of 0.05 (95% confidence level). A one-way analysis of variance (ANOVA) followed by a Student’s T-test post hoc analysis were used to test for significant differences in microbial power outputs (µW), microbial abundances (cells per mL ×10⁶), heterotroph to autotroph ratios, and dissolved oxygen (µM) by treatments (i.e., an effect of coral, algae, interface). The data was further analyzed with linear regression by treatments over time compared to power output, normalized heat (J/g) compared to heterotroph to autotroph ratio, and normalized heat (J/g) to microbial abundance (cells per mL ×10⁶). To determine if there was a significant effect of treatment area on the heat production over time, we used and ANCOVA test for the analysis of covariance of power output between treatment groups over time. All statistical analyses were performed using JMP 10 software (SAS Software). Our statistical results are listed below and in Tables  S1–S4.

Macroorganism-associated microbial thermodynamics

Calorimetry demonstrated that microbial thermodynamic power output (interface: 33.652 ± 1.65600B5 J; coral: 28.552 ± 1.65600B5 J; algae: 26.534 ± 1.77000B5 J; water-column: 19.4359 ± 1.91200B5 J; mean ± s.e.m. n = 8) was significantly different between the four treatments (One-way ANOVA: F_3,28 = 10.7653p < 0.0001) with the interface-associated microbial community having a significantly higher power output than either the coral- or algae-associated communities alone (Fig. 2 inset; Post hoc t-test: interface-coral p = 0.0390, interface-algae p = 0.0070) (Table S1). Furthermore, both coral- and algae-associated microbial communities had significantly higher power output than the microbial communities in the surrounding water (Fig. 2 inset) (Post hoc t-test: coral-water p = 0.0014, algae-water p = 0.0116) and in the 0.02 µm filtered seawater control (Post hoc t-test: p < 0.0001). Integration of the total heat production over time (Fig. 2 and inset) demonstrates a significant effect of the sampling area on the heat production over time (ANOCOVA: F_3,55938 = 12, 118 p < 0.0001). The interface community not only produces significantly more total heat over the eight hours in the calorimeter, but also has a significantly higher rate (5.13 ± 0.00347 µJ/s) of heat production (i.e., power) at all points in time (Fig. 2 inset) than the microbial communities associated with coral (4.50 ± 0.003702 µJ/s) or algae (4.14 ± 0.002416 µJ/s; mean ± s.e.m.).

This data demonstrates that there are small-scale spatial differences in the way that energy flows though macroorganism-associated microbial communities. More specifically, the microbial assemblages associated with the coral-algal interface have a higher thermodynamic power output than the microbial communities associated with either macroorganism alone. The higher microbial power output at the interface indicates that the community metabolism at the interface was faster and less efficient, as it used more energy per unit time, and dissipated more of that energy as heat.

Microbial community trophic structure

Flow cytometry analysis revealed that the interface microbial community was significantly more heterotrophic (32.946 ± 4.3231) than the communities associated with coral (18.4932 ± 4.6695) or algae (15.6103 ± 4.3231; mean hetrerotroph: autotroph ratio ± s.e.m.) (Fig. 3A) (One-way ANOVA: F_3,26 = 3.4106 p = 0.0345, Post hoc T-test: interface-coral p = 0.0328, interface-algae p = 0.0094) (Table S2). This shift in community metabolism to a more heterotrophic microbial consortium at the interface correlates significantly with the total heat output of the microbial communities (Fig. 4A) (R²:0.427; ANOVA, F_1,26 = 18.6651 p = 0.0002). Whereas, total microbial abundance was not significantly different between sampling sites (Fig. 3C and Table S4; one-way ANOVA: F_3,26 = 0.5631 p = 0.6444) (Table S3), and was not a strong predictor of heat production (Fig. 4B) (R²:0.047; ANOVA, F_1,27 = 1.7697 p = 0.2997), indicating that the shift towards a more heterotrophic community metabolism is responsible for the increased power at the coral-algal-interaction interface.

Biological oxygen demand

Planar oxygen optodes (Gregg et al., 2013) provided a fine scale 2-dimensional assessment of the O₂ concentrations associated with the coral-algal interactions. Figure 1C presents a representative picture of the oxygen concentrations visualized in a typical coral-algal interaction experiment. Two dimensional oxygen concentration assessment demonstrates a general trend of an O₂ decline at the coral-algal interface relative to the oxygen levels associated with coral, algae, and the watercolumn (Figs. 1C and 3B; O₂ concentrations: interface: 39.8299 ± 18.7 µM ; coral: 82.422 ±  20.907 µM; algal: 63.4961 ± 20.91; water-column: 90.3468 ± 18.7µM; mean ± s.e.m.) (Table S4), in accordance with previous observations by Haas et al. (2013). While lowered O₂ concentrations at the interface were not statistically significant, they are likely biologically significant, as the interface is the only area in which O₂ concentrations were significantly below the 70 µM O₂ concentration at which >50% of coastal marine organisms die from hypoxia (Vaquer-Sunyer & Duarte, 2008).

Nonequilibrium thermodynamics and ecology

This study demonstrated that power output at the coral-algal interaction interface is significantly higher than in the communities associated with either of the single organisms. This was evident through higher microbial heat production rates (Fig. 2) at interfaces compared to areas distal from them. These measurements support prior observations that coral-algae interfaces foster microbial communities and metabolic profiles unique from those associated with either coral or algae (Barott et al., 2012c; Haas et al., 2013). This study further identifies small-scale, spatial alterations in the thermodynamics of microbial communities associated with specific areas across the coral-algal interface. Specifically, the data confirms a yield-to-power switch (as suggested by Haas et al., 2016) occurring at the interface between coral and algae. A better understanding of the thermodynamics of coral-algal interactions may provide bioenergetic data that can be used for modeling reef ecosystems to predict the systems’ trajectories and possibly devise plans for ecological remediation.

Bioenergetic mechanisms of the DDAM feedback model

At the coral-algal interface, algae exude Dissolved Organic Carbon (DOC) in close proximity to coral, which causes a shift in the microbial community towards more heterotrophic metabolisms (Fig. 3A). As heterotrophic microbial metabolism is favored by these interaction events, the microbial community metabolism is shifted towards more copiotrophic consumers relative to the amount of autotrophic producers (Fig. 3A). This affects coral reef systems in two ways. First, copiotrophs have greater net oxygen consumption, which causes localized hypoxia at the coral-algal interface (Figs. 1C and 3B), (Barott et al., 2012b; Wangpraseurt et al., 2012; Haas et al., 2013). Second, this microbial community shift towards more heterotrophic consumers means that a greater portion of the available energy will be utilized by the microbial fraction of the ecosystem leaving less energy for higher trophic levels, a process referred to as microbialization (McDole et al., 2012; Haas et al., 2016).

Microbialization is a metric of the proportion of energy allocated to the microbial fraction of an ecosystem (i.e., a measure of the trophic cascading from microbes to macro-organisms) (McDole et al., 2012). One cause of microbialization on coral reefs is over-fishing (Jackson et al., 2001), which reduces grazing pressure on algae. Releasing macroalgae from grazing begins the DDAM positive feedback loop where DOC released by the un-grazed algae enriches for a copiotrophic, bacterial community, ultimately leading to coral mortality and more benthic space for algae. Our data suggests that a possible underlying mechanism in the DDAM loop (Kline et al., 2006; Barott & Rohwer, 2012) and the subsequent microbialization of coral reef ecosystems is the higher power output of the microbial communities at the coral-algal interaction interface. The increase in power—that is, energy used per unit time—at the interaction interface (Fig. 2) stems from an augmentation in net heterotrophy at the interface (Figs. 3A and 4A). This switch to a microbial community with higher thermodynamic power coupled to the increase in heterotrophic metabolisms, such as oxidative respiration, creates localized areas of hypoxia (Figs. 1C and 3B), which damages the coral tissue, ultimately leading to coral mortality (Fig. 5).

Micro-scale and whole-reef-scale dynamics

Field studies have shown that increased macroalgal cover is significantly correlated with a net increase in heterotrophic microbial communities (Dinsdale et al., 2008; Kelly et al., 2012) and a shift toward faster, lower yield metabolisms (Haas et al., 2016). Our data suggest that the increase in power and net heterotrophy observed on a reef-wide-scale could be due to the small-scale (on the order of 10 µm–1 mm) spatial dynamics occurring at the coral-algal interface. That is, the altered community metabolisms and bioenergetics could possibly stem from the increased number of coral-algal interactions that occur as algae begins to increase in benthic abundance. This raises an interesting question: to what extent are the activities observed at the coral-algal interface driving the large-scale bioenergetic dynamics observed on the reef as a whole? Future work should focus on understanding the ways in which the small-scale activities at the interface affect whole reef dynamics. In silico modeling may provide a better understanding of how small-scale interaction events affect the overall bioenergetics of reef ecosystems.