Ciliary flows in corals ventilate target areas of high photosynthetic oxygen production

tropical corals live in


INTRODUCTION
Coral reefs are among the most diverse and economically important ecosystems on the planet. 1 Despite their importance as ecosystem engineers, the corals providing the foundation of today's reefs are threatened by a multitude of anthropogenic changes acting on many spatial and temporal scales. 2,3 Nutrification of coastal areas in the wake of deforestation, increased terrestrial run-off over the last century, 4,5 and overfishing, dating back to the origin of human expansion, 6 have caused coral-algal phase shifts. 7 Global effects such as ocean acidification, deoxygenation, and warming have caused severe mass bleaching and mortality of corals [8][9][10] contributing to the demise of corals in all reef provinces over the last decades. 11,12 However, although the importance of corals and their threats can be assessed on a broad scale (colonies and reefs), it is the individual coral polyp (usually mm in size) and its symbionts that are directly responding to anthropogenic disturbances. 13 Hence, our ability to estimate the response of coral reefs to environmental stressors requires a deep understanding of polyp-scale physiology and the manner in which it reacts to disturbances.
Tropical scleractinian corals live in symbiosis with dinoflagellate algae (Symbiodiniaceae), an association that allows them to recycle essential elements and thrive in nutrient-poor waters. 14,15 Within the coral tissue, symbiont distribution can be highly heterogeneous, with densities dependent on tissue structure, light, and nutrient availability. [16][17][18] Among the Symbiodiniaceae, chlorophyll concentrations can be highly variable, since the coral holobiont can adapt to reduced light by increasing the symbiont and/or chlorophyll densities in order to optimize light harvesting. 19,20 Symbiont and pigment densities, together with the tissue's optical properties, generate physiological microniches-suggested to play the role of refugia for minor symbiont populations-which might be important for the coral's response to stress. 13,21 Therefore, understanding where the symbionts are located along and across the coral tissue becomes highly relevant, although in vivo studies of Symbiodiniaceae distribution at m-metric resolutions are still scarce.
The relationship between the coral and its Symbiodiniaceae symbionts is highly complex. Although the coral consumes most of the O 2 produced by its symbionts, 22 the surplus diffuses out of the tissue into the coral-water interface-the diffusive boundary layer (DBL). 23 Diffusion acts as a bottleneck for solute exchange and indirectly controls the coral's physiology 24,25 and stress response. 26,27 In the coral Favia sp., spatial variations in O 2 and CO 2 within a single polyp were attributed to coral morphology causing m-metric variations in the thickness of the DBL and differences in the light field. 28 Similarly higher photosynthetic rates were found in the middle parts of the corallite, tentacles, and tissue surrounding the mouth of Galaxea fascicularis, whereas the wall of the corallite and the coenosarc showed lower photosynthetic rates 29 (see also Figure S1 for details on coral structure). This heterogeneity was attributed to differences in symbiont distribution, localized higher photosynthetic activity in response to increased metabolic demand and local supply of solutes through the DBL, 29 but concurrent assessments of symbiont distribution and O 2 are so far lacking. The response of corals to environmental stress, and subsequent bleaching, has also been partly attributed to the DBL controlling the exchange of O 2 and solutes between the coral and the water column. 30,31 Although O 2 exchange in the coral tissue and DBL has long been considered passive by means of diffusion, 24,32 more recent work has shown the importance of ciliary currents enhancing the effective diffusivity at the coral surface. 33,34 Ciliary action was shown to lower excessive O 2 concentrations at the coral surface serving predominantly as a homeostatic control mechanism in coral stress response.
The heterogeneities in symbiont distribution, photosynthesis, host activity (respiration), and ciliary flows are therefore likely to result in a complex patterning of the O 2 microenvironment of the coral, with differential O 2 accumulation and/or consumption and contrasting cellular responses to O 2 levels. Although O 2 is essential for all aerobic organisms, excess O 2 can be detrimental for photoautotrophs 35,36 especially under high temperatures. 37 Since the photosynthetic apparatus of the symbiont-Rubisco Form II-has a higher affinity to O 2 over CO 2 , the excess of the former causes photorespiration and the production of O 2 radicals. 38 This not only entails higher metabolic costs 39 but also may cause cellular damage and, ultimately, coral bleaching. 40-42 Moreover, the documented high spatial variability of photosynthetic activity 24,28,29 raises the question of how corals deal with the potential negative impacts of localized hyperoxia. Therefore, understanding the O 2 dynamics along and across the different compartments of the coral, i.e., DBL, tissue, and symbionts, is of paramount importance since it has a direct influence on its health, performance, and ability to respond to its environment. Here, we provide the first simultaneous assessment of the microscale flow field and two-dimensional O 2 distribution in the coral boundary layer under controlled ambient illumination and currents, along with a mapping of Symbiodiniaceae Chla concentrations in the underlying coral tissue. We aim to characterize the Chla distribution inside the coral tissue and relate it with the O 2 and flow dynamics at the coral surface, examining the role of ciliary flows in the transport of O 2 away from the tissue surface.

RESULTS AND DISCUSSION
Chla concentrations in the coral tissue Using hyperspectral imaging technology, 43 we analyzed the areal Chla distribution in the tissue of the coral Porites lutea at a sub-millimeter scale. We observed higher normalized absorbance of 0.8 in the coenosarc, indicating high Chla concentration ( Figure 1B, ROI 1; Figure 1D, yellow areas). In contrast, normalized absorbance was 0.3 at the mouth openings, indicating low Chla concentration ( Figure 1B, ROI 2; Figure 1D, blue areas). As the polyps of Porites tend to remain partially retracted during the day 44 in order to enhance planar density and photosynthetic performance of its symbionts, 45,46 it is no surprise that the coenosarc contains higher Chla concentrations than the area surrounding the mouth. The quantification and distribution of Chla densities in the thin veneer of tissue covering the complex coral skeleton has only recently been made possible by the application of high-resolution optical techniques 47,48 requiring extensive sample preparation 48 or the short-term removal of culturing water. 47 Here, images were taken in a flow-through chamber, without disturbing the coral. We expected the spatial heterogeneities in symbiont distribution to be reflected also in the O 2 DBL's landscape, as the differential diffusion of photosynthesis-derived O 2 has been shown to lead to a patchy distribution of O 2 diffusing to the surface of the coral, 49 with highest values over the coenosarc and lowest values over the polyp mouths.
Flow field and O 2 concentrations at the coral surface In order to relate the heterogeneous Chla distribution inside the tissue with the O 2 dynamics at the DBL, we used particle image velocimetry (PIV) using O 2 -sensitive nanoparticles in an image velocimetry system (sensPIV) 50 ( Figure 2A). We observed a heterogeneous O 2 distribution along the coral surface. However, areas of high O 2 concentrations (hotspots) coincided with areas of low Chla concentration over the mouth openings of the coral polyps, whereas areas of lower O 2 concentrations coincided with areas of high Chla concentration over the coenosarc (Figure 2D). The opposed patchiness suggests the superposition of two independent processes: passive diffusion of O 2 from the sites of production (symbionts) and active transport of water by ciliary currents, as outlined below. Under the flow conditions of our experiments (300 and 1,500 mm s À1 ), cilia on the surface of the coral generated vortical currents ( Figure 2B). At high ambient flow speeds (1,500 mm s À1 ), the average height and width of the vortices were 300 ± 55 mm and 1,500 ± 230 mm, respectively. At low ambient flow speeds (300 mm s À1 ), the corresponding values were 500 ± 50 mm and 1,500 ± 220 mm, respectively ( Figure S2). Within the vortices, the vertical flow velocities reached up to 300 mm s À1 and À200 mm s À1 ( Figure 2B), in line with previous reports. 33,34 The vortices' structure was similar to those found in Pacherres et al. 33 with the upward flow of the vortex facing toward the ambient current, whereas the downward flow was located leeward, suggesting cilia beating against the ambient current ( Figure S3). The upward flank of the vortices transported water from the coral polyp periphery aloft (red areas, Figure 2B), whereas the downward flank was situated somewhat leeward of the polyp mouth opening (blue areas, Figure 2B). As a result, the main part of the vortices was situated on top of the mouth opening. In darkness, we found slightly lowered O 2 concentrations in the downwelling areas suggesting respiration ( Figure 2C) with similar velocity magnitudes inside the vortex area (Table S1). In the light, we found high O 2 concentrations near the coral surface, with highest values over the mouths and lower values in the periphery ( Figure 2D).

Combining O 2 at the DBL and tissue Chla
The inverse relation between O 2 and Chla distributions is in conflict with the current paradigm of an exclusively diffusive transport of O 2 , which would predict a positive relation between Chla and O 2 . We found the opposite, a redistribution of O 2 by cilia-induced vortical currents dominating O 2 transport in Porites. The vortices were located above the Chla-poor areas, with their upward flow extending into the adjacent Chla-rich areas ( Figure 3). This allows O 2 to be transported laterally from the main production site into the vortex, which itself extends vertically as a bulged interface into the ambient water with lower O 2 . As a result, O 2 accumulation in the Chla-rich tissue is alleviated, and at the same time, the high O 2 concentration in the vortex increases the concentration gradient toward the ambient water and thus the diffusive flux.
Previous microsensor studies of O 2 concentrations at the surface of Favia sp. found peaks of gross photosynthetic production in the coenosarc 24,28 and large variations in the O 2 concentrations in the water column above the mouth. 28 On the other hand, Galaxea fascicularis showed an order of magnitude increase in gross O 2 production over the polyp mouth compared with the coenosarc. 29 Although our observations are in line with some of these early studies, it is important to acknowledge that corals exhibit a high plasticity of forms, tissue properties, symbiont distribution, and Chla content (see introduction), all of which influence the diffusion of solutes along, across, and outside of the tissue. Therefore, further research on the linkages between the different coral compartments, using other coral species, will be needed in order to better understand how coral form and functions are related and the way in which they might optimize coral respond to stressors.  Figure 2). See also Figure S1. (C and D) Normalized Chla absorption (used as a proxy for Chla concentrations) along the dashed line shown in (D). Please notice that O 2 distribution and chlorophyll distribution are not from the same locations but are picked as representative examples. Scale bars are applicable for the respective vertical panels.

Simulations of the boundary layer dynamics of P. lutea
To investigate the relation between tissue Chla and the diffusive and advective fluxes of photosynthetic O 2 , we simulated the effects of diffusion and ciliary flow along the coral surface by means of a simplified planar two-layer model. The lower layer has a thickness of 100 mm, representing the tissue of P. lutea in which the Symbiodiniaceae are located, and transport is limited to diffusion (dotted line in Figure 4A). The heterogeneously distributed symbionts produce O 2 at a constant rate, assuming constant light regime and no oxidative stress. The layer above represents the water column in which the cilia stir the boundary layer under flow conditions. The ciliary flow is driven by a moving boundary with horizontal velocity components that oscillate in both directions along the surface at an amplitude (maximum ciliary flow) that was adjusted to c vel = 150 mm s À1 , similar to previous measurements. 34 The undisturbed flow velocity in the simulation was set to 300 mm s À1 , similar to the measurements.
The interaction of the flow field and the beating of cilia leads to vertical flow speeds of 80-100 mm s À1 , which are slightly lower than the maximum vertical speeds from the experiments (up to 300 mm, Figure 1B) and the velocities seen in the upward zones of the vortices in Shapiro et al. 34 At the coral surface, the opposing horizontal velocities lead to the formation of two stagnation points: at one stagnation point, the currents converge, driving the flow aloft into the boundary layer; at the other, the currents diverge where the downward flow returns to the coral surface and is deflected sideways. The upward and downward flows drive the vortices lining the surface of the coral. The horizontal extent of the vortices is largely determined by the boundary conditions, i.e., the wavelength of the oscillating boundary, which was adjusted to d = 1,200 mm, similar to the average size of the calyx of the coral (see details of calyx size in Figure 3B). The specific shape of the vortices leads to two distinct regions within the coral boundary layer, which are determined by the location of the stagnation points along the surface of the coral ( Figure  In the simulations, O 2 produced diffuses from the coral tissue in proportion to the observed Chla distribution ( Figure 2D). For simplicity, the heterogeneity of O 2 production is simulated by a sinusoidal curve whose spatial extent was adjusted to match the experimentally obtained results: highest O 2 production was situated outside the vortex structures, whereas lowest O 2 production was situated underneath the vortex (Figure 3). The produced O 2 diffuses through the tissue before reaching the water column. Along the tissue surface, diffusive exchange fluxes are highest where O 2 gradients between tissue and the ambient water are strongest. The strongest O 2 gradients occur where O 2 is produced at a higher rate and ambient water with low O 2 is directed downward at high velocities ( Figure 4A We used the model to test how the location of the vortices above the coral surface affects the O 2 distribution in the DBL and inside the tissue of the coral. O 2 accumulates in the vortices independent of the relative location of vortices and O 2 production sites ( Figure 5A). Consequently, the O 2 concentration in the DBL does not reflect the O 2 concentration and production rate in the tissue directly below but is reshaped by the interacting vortical and ambient flow fields. In general, the vortices help to decrease the O 2 concentration in the underlying tissue, as described by Pacherres et al. 33 At ciliary flow conditions similar to the ones in the experiments, we found the spatially averaged O 2 concentration in the tissue to be reduced by 53%, compared with no ciliary flow conditions. However, the efficiency of this reduction depends on the relative location of vortices and hotspots of O 2 production, which has a particular impact on harmful excess O 2 concentrations in the tissue ( Figures 5B and 5C). To exemplify this, we calculated the tissue area that is affected by a critical O 2 concentration of more than 300 mmol L À1 above ambient levels, a threshold based on Sforza et al. 51 ( Figure S5). When vortex location and production hotspots are separated (phase shift: 207 ), the area with critical O 2 concentrations is minimized to 20.7% ( Figure 5C). When vortex location and production hotspots are aligned (phase shift: 43 ), the affected area is nearly doubled to 38% (Video S1). Similar results are found for O 2 concentration thresholds ranging from 200 to 350 mmol L À1 above ambient levels (Table S2). It is important to mention that the reallocation of O 2 from the tissue into the vortex does not imply an increased net mass transfer as long as the O 2 production in the coral tissue is independent of the O 2 concentration.
In summary, the simulation shows that the observed O 2 and Chla patterns are not directly connected, but the result of the diffusive flux interacting with the flow field generated by the cilia (Figure 3). The results further indicate that positioning the vortices adjacent to, rather than above, the O 2 production sites selectively enhances the ventilation of O 2 -producing symbiont patches, which helps avoid photosynthetic inhibition. Article Implications for the coral's response to environmental stress Ciliary vortices localized adjacent to specific sites of excess O 2 production seem beneficial for the coral in the face of oxidative stress, especially under weak currents, which have been found to exacerbate bleaching. 31,52 It has been suggested that photosynthetic activity by the symbionts may be inhibited by O 2 accumulation through photorespiration and the formation of reactive oxygen species (ROS). 40,53,54 Photorespiration occurs when O 2 is used as substrate instead of CO 2 , resulting in the loss of energy, 55 whereas ROS are formed through photoreduction of O 2 and are believed to directly damage the photosystem II of the chloroplast. 37,56 Our results show that the O 2 concentration in the coral tissue can not only be modulated by ciliary flows as such but also depend on the relative location of the vortices with respect to areas of O 2 production, further reducing photosynthetic inhibition by the reallocation of O 2 . The generation of this heterogeneous O 2 landscape is of particular importance when taking into consideration that the ability of an organism to adapt to environmental changes, e.g., climate change, highly depends on its life history exposure to short-term and shortscale environmental fluctuations. 44,57,58 It is known that adjacent colonies of the same coral species exposed to the same water temperatures can present different bleaching responses, 59 and more so, different regions within the same colony may show differential bleaching severities. 13 Our findings suggest that microscale heterogeneities in both the internal arrangement of symbionts and the external micro-currents generated by the ciliated surface of the coral represent a composite buffer to mitigate oxidative stress. We postulate that this buffer is an important pre-requisite for coral resistance to bleaching in the face of global warming.
Alternatively, localized vortices might as well be redistributing O 2 to areas where the tissue has less symbionts (and therefore has less local supply of O 2 for the host tissue). During the day, coral symbionts produce an excess of O 2 that by far surpasses the metabolic requirements of a healthy coral. 60 However, measurements of O 2 concentrations inside the gastric cavity of corals have shown extremely low O 2 concentrations even during the day. 61 Whether the coral utilizes ciliary flows as an external transport mechanism linking zones of O 2 production with zones of O 2 consumption would require further exploration.
The observed localized transport mechanism is unlikely to be restricted to O 2 but should also be effective for other solutes. With regard to the seawater carbonate system, vortex transport of DIC and protons might be of significant importance to coral calcification and should be considered in studies of coral physiology under climate change. 8,62 Since coral calcification depends on pH, light, and currents, 63,64 vortex-altered proton transport in the DBL, 33 along with the coral's capacity to upregulate its inner calcifying fluid, 65,66 might help explain the sometimes contradicting, species-specific response of corals to ocean acidification. Research is needed to assess potential vortex effects on coral calcification and coral resilience to ocean acidification.
The ability of corals to enhance mass transport in specific areas (here: the coenosarc) ( Figure 2C) might have implications not only in the coral's response to its environment but also in its relationship with bacteria and viruses that inhabit its surface. 67,68 The community of prokaryotic and eukaryotic microbes, viruses, and archaea comprising the coral microbiome is extremely diverse 69 and has been associated not only with diseases 70 but also with the capacity of corals to resist environmental stress by the acquisition of nutrients, defense against pathogens, horizontal gene transfer, etc. [71][72][73] The ability of cilia to generate chemical microenvironments, such as seen here for the O 2 hotspots, could provide an advantage to certain coral-associated microbes by enhancing the transport of the (C) Percentage of tissue area that is above a critical threshold of 300 mmol L À1 above ambient O 2 concentration. The red and gray bars indicate the minimum and maximum, respectively. See also Figure S5 and Table S2. Article metabolic compounds necessary for their development. Further research will be needed in order to unveil the consequence of the observed microenvironment creation and the repercussions it might have upon coral-microbe relations and coral health.

Conclusions
Altogether, this study demonstrates that the spatial arrangement of local interactions between ciliary flows and physiological processes (photosynthesis) can give rise to local heterogeneities in the chemical landscape and solute exchange. It extends the coral's ability to respond to environmental stressors well into the coral's boundary layer, where the flow field can be shaped according to local needs. A 3D exploration of the coral DBL, taking into consideration the different elements that intervene in its shape and characteristics, can be an important next step to reveal further insights into the complex coral-water interface and better understand the role of the DBL in the coral's relationship with its environment.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interests. small fragments (2 cm long, 2 cm wide) bearing 60 to 80 polyps were cut out from 4 of the source colonies and allowed to heal for at least one month in the same culturing tanks their original colonies were kept. Survival of the fragments was >90%.
Porites was chosen as a model organism as it occurs over a wide depth range from surface down to 50 m 74 being an important reef building coral along different regions. 75 Moreover, it has been shown to experience bleaching at 15 m 76 making it an ideal subject to test our hypotheses.

METHOD DETAILS
Experimental set-up For all experiments, the light-sheet-microscopy based flow chamber set-up was used, as in Pacherres et al. 33 (Figures 1A and 2A). The coral fragment was placed in the chamber filled with filtered artificial seawater (FASW) (0.2 mm pore size) from the coral's culturing tank. There, it was allowed to acclimate to the chamber conditions for at least an hour under dark and low flow conditions (300 mm s -1 ) before experiments started. Temperature of the water inside the chamber was monitored every 30 seconds (Pt100 sensor, PyroScience, Germany). For all experiments, temperature was kept constant at 25 C ± 0.5. Flow rate inside the chamber was modulated using a water pump (REGLO Z, ISMATEC, Germany). Experiments were conducted at 2 flow rates: 300 mm s -1 and 1500 mm s -1 measured at 2 mm above the coral tissue. In the case of experiments requiring a light source, a fiber optic lamp (Schott 1500, USA) was used in order to maintain constant illumination of 180 mmol quanta m -2 s -1 (LI-COR LI-192, USA). In order to standardize respiratory differences due to digestion, all experiments were performed after 2 days of starvation.
Experiments under dark conditions were performed after one hour of acclimating the coral to the chamber. After those images were taken, light was turned on for 5 minutes, to allow the coral to build a representative concentration boundary layer, before capturing the images for the light experiments.
Experiments with arrested cilia were initially planned by adding sodium orthovanadate. However, we were surprised to find that some corals adapted to the addition of the chemical as cilia activity ceased to fully arrest, suggesting an environmental memory. 77 We therefore decided against continuing with sodium orthovanadate to avoid introducing dose-response and other inconclusive treatment effects into this study. The effect of arrested cilia activity on the oxygen dynamics of the coral DBL has been shown in previous studies. 33,34 Hyperspectral imaging Chla concentrations of Symbiodiniaceae inside the tissue of the coral colony were obtained using hyperspectral line-scanning. For that, hyperspectral imaging optics (Wasatch, USA) were used. Briefly, a line is recorded through a long-distance microscope lens (Optem Fusion) and directed through a diffraction grid. The resulting light spectrum is aligned with a camera CCD chip (FLIR, blackfly, Canada) along a pixel row of 1288 px. in the visible light range, this results in a spectral resolution of approx. 1 nm. Using a micromanipulator (PyroScience, Germany) the camera system was slowly moved parallel to the coral surface at a speed of 50 mm s -1 to the focused line along the coral fragment while continuously recording images with an exposure time of 400 ms. This resulted in a 3D tensor or hypercube, i.e. each pixel of the image consists of a spectrum ranging from 400 -700 nm. Horizontal and vertical resolution was 42 and 6 mm per px, respectively. The set-up was controlled using serial-interfaces and the Spinnaker Software Development Kit (FLIR Spinnaker SDK) through an in-house developed Python script (Python Software Foundation). The raw images were processed using Matlab (2018b).
Chla pigments reflect green light and absorb light in the blue-violet and orange-red range with peaks at 450 -475 nm and around 650 -680 nm, respectively. In our hyperspectral line-scanning set-up, reflectance spectra were measured, implying that low spectral intensities correspond to a high absorbance and, conversely, high spectral intensities correspond to low absorbance. Therefore, to estimate the absorbance spectra induced by the Chla, we subtracted the integrated wavelengths 650 nm -660 nm from the reference light spectrum. To compensate for lower light intensity and 3D structures at the surface of the coral, we normalized the resulting absorbance spectrum by Chla unaffected wavelengths between 550 nm and 560 nm. Therefore, the resulting normalized absorbance value indicates the amount of orange-red light that is absorbed from the incident light, which depends on the amount of Chla in the tissue of the coral.
During the experimental procedure, the coral fragment was placed in the flow-chamber ( Figure 1A) and illuminated using a homogenous white light (Schott 1500, USA), intensity reached 160 mmol quanta m -2 s -1 (LI-COR LI-192, USA). The chamber was placed on top of a white surface, which later served as reference.

SensPIV measurements
In order to identify the relative influence of cilia activity on the 2D structure of the O 2 DBL at the coral surface, O 2 distributions were compared under light and dark conditions (180 and 0 mmol quanta m -2 s -1 respectively). To resolve the O 2 and flow field around the coral fragment, we used O 2 sensitive nanoparticles, 78 which have proven ideal for mapping O 2 concentrations at relevant scales. 50 The stock solution contained 2 mg mL -1 which was further diluted by a factor of 200. Illumination was achieved using a LED pulsing system (LPS3, iLA_5150, Germany) connected to LED light sheet optics similar to the one described by Willert et al. 79 The light sheet was approx. 1 mm thick and intensity reached 4500 mmol quanta m -2 s -1 (LI-COR LI-192, USA) at a wavelength of around 468 nm. For each experiment one hundred images were captured using a Grasshopper3 camera (FLIR, USA) recording at 20 frames per second (fps) with an exposure time of 50 ms. A long distance custom-made microscope lens (based on the Optem FUSION system, Germany) with a yellow-orange long-pass filter (515 nm, LP515, Midopt, USA) was used for all experiments. To avoid photoinhibition of the symbionts from the intense illumination, light exposure was limited to 5 s for each experiment and LED light was only active during 50 ms illumination of the camera chip. The recorded images were post-processed using custom-built Matlab (MathWorks, R2018b) algorithms to obtain the O 2 concentration signal and a map of the particle movement inside the chamber.
Nanoparticle readings were later compared and calibrated with O 2 profiles, performed before each experiment, using an electrochemical sensor of 10 mm tip diameter (Unisense, Denmark). The profile was recorded 3 min before the images were captured. Values were recorded using the SensorTrace-PRO software (Unisense, Denmark). For each experiment, microelectrodes were 2-point calibrated in O 2 -free (bubbling pre-filtered seawater with N 2 gas for 10 min) and air-saturated FASW of known salinity and temperature. 24 The tip of the sensor was carefully placed at the surface of the coral. A micromanipulator (Unisense, Denmark) was programmed to move the sensor up in 20 mm steps. The range of the vertical profile was 1000 mm. At each step, dissolved O 2 was measured one time with a sampling interval of 2 s.
To directly relate the O 2 production of the coral colony with the Chla concentration inside the coral tissue we carefully pinpoint the area of the coral observed through the camera and later extracted the Chla concentrations from the same area out of the hyperspectral images.

Model formulation
The cilia beating of corals generates a complex flow pattern, which is reproduced using numerical simulations which were performed in Comsol (Comsol Multiphysics, V5.6). The flow field is given by the Stokes equations: Where u is the velocity vector, p the pressure, m dynamic viscosity and V the gradient-operator. The continuity equation for incompressible fluids reads: The O 2 distribution was calculated solving the stationary advection-diffusion equations: R H is the volumetric O 2 production rate (see below), D is the diffusion coefficient for O 2 in sea water at 20 C: 2.1 10 -9 m 2 s -1 . According to the model proposed by Shapiro, 34 the cilia-induced currents were simulated assuming oscillating horizontal velocity components at the coral surface (slip boundary condition): where x is the horizontal distance, c vel the maximum ciliate beating velocity and d is the characteristic length scale of the vortices, see text for values and explanation. Volumetric rates of O 2 production in the tissue were simulated based on the assumption that the clustering of chlorophyll a at certain areas of the tissue follows a sinus curve: Where R C is the production rate, m is the parameter for the relative phase shift. The mathematical model allowed us to simulate O 2 concentrations inside the DBL and in the tissue in the presence of a heterogeneous O 2 production rate by the coral.

Limitations of the model and topography effects
We applied a simplified 2D transport-reaction model in order to investigate the effect of the ciliary flow on the O 2 concentration inside the tissue and within the coral boundary layer. The model approach does not consider 3D effects or changes in the topography of the coral, which might become substantial when investigating the turbulent flow field along an entire coral colony see for example Bartzke et al. 80 However, on the scales of the investigated coral fragments, the topographical features where an order of magnitude smaller (21 -123 mm) than the observed vortices (typically around 1 mm) and were not a substantial factor within the experiments. It remains a task for future studies to investigate how O 2 is re-distributed along an entire coral colony and how topography and ciliary flow act in tandem for coral ventilation. Further, in the modelling approach we investigated the effect of the ciliary flow by comparing the tissue volume below and above a specific O 2 -threshold of 300 mmol L À1 ( Figure S5). This O 2 -threshold was inferred from an experiment in which a microsensor was placed directly above the coral tissue and the light intensity was increased until O 2 production was affected (see Figure S5). However, to test the sensitivity of the obtained model results, we tested varying thresholds between 200 mmol L À1 and 400 mmol L À1 . Overall, the pattern of maxima and minima along the different phase shifts persisted. Only the difference between the maximum and minimum varied between 6-19 % (compared to 18% for 300 mmol L À1 , see Table S2).