High variability and enhanced nocturnal oxygen uptake in coral reef sponges

Sponges are animals that feed by filtering water through their perforated body. We examined the in situ diel dynamics of sponge metabolism by continuously measuring the oxygen concentrations in the water inhaled and exhaled by undisturbed sponges. A clear daily pattern of oxygen removal was evident for six of the seven species we studied with their nocturnal oxygen removal being almost double the diurnal values (+ 86 ± 57%). Oxygenic photosynthesis by the sponge's symbiotic or endolithic phototrophic microbes may explain some of the diel difference, but significant day–night differences were also observed in three sponge species for which no evidence of photosynthetic activity (tested with imaging pulse‐amplitude‐modulation Fluorometry) was found. Mean oxygen removal (± 95% confidence interval for the mean) per species ranged from 1.7 ± 1 μmol O2 per liter (hereafter: μmol O2 L−1) for the low microbial abundance (LMA) sponge Callyspongia siphonella to 30.5 ± 10.5 μmol O2 L−1 for the high microbial abundance HMA) sponge Theonella swinhoei with considerable variation in oxygen removal across all scales (minutes to hours, within and among specimens). Events of high oxygen removal (> 50 μmol L−1) were regularly observed for five of the seven species and were predominantly nocturnal, occasionally lasting several hours. The high variability in oxygen removal stresses the need for long‐term in‐situ measurements of benthic suspension feeders metabolism.

Sponges (Porifera) are sessile suspension feeders that dominate benthic environments from the equator to the poles. In addition to providing structures and shelter for other organisms, sponge feeding and excretion activities strongly connect seafloor and water column processes (Reiswig 1971;Kahn et al. 2013;Pawlik and McMurray 2020). Sponges specialize in filtering extremely small particles such as micron and submicronsize bacteria and even viruses (Yahel et al. 2003(Yahel et al. , 2005Hadas et al. 2009) and dissolved substances (de Goeij et al. 2008;Mueller et al. 2014;Gantt et al. 2019), that are largely unavailable to other benthic metazoans. Sponges thereby intake and recycle particulate and dissolved organic matter and essential nutrients (N, P) from the water column. Some of this organic matter is assimilated into sponge biomass production that is available to benthic spongivorous organisms (e.g., McMurray et al. 2018, Wooster et al. 2019). Sponge waste is released near the seafloor in the form of particulate detritus (Alexander et al. 2014;) that is consumed by benthic detritivores (de Goeij et al. 2013;Rix et al. 2018; Bart et al. 2021). The waste also comprises dissolved inorganic nutrients that become available to benthic algae and mixotrophs (reviewed by Maldonado et al. 2012;Maldonado 2016;Hoer et al. 2018b), and dissolved organic matter that is consumed by coral reefs microbes (Diaz and Rützler 2001) and nearby sponges (Morganti et al. 2017;Hudspith et al. 2021).
Sponges often form complex symbiotic relationships with diverse communities of microorganisms and are commonly classified into two guilds accordingly: high microbial abundance (HMA)-sponges and low microbial abundance (LMA)-sponges (Hentschel et al. 2006;Gloeckner et al. 2014). HMA sponges harbor dense and diverse communities of bacteria (Hentschel et al. 2003) that can occupy > 40% of the sponge body volume (Wilkinson 1978;Magnino et al. 1999). Microbial concentrations in HMA sponges are commonly 2-4 orders-of-magnitude higher than that of seawater (10 8 -10 10 cells mL À1 ), whereas, in LMA sponges, the concentrations are within the range of natural seawater, but feature a different species composition (Hentschel *Correspondence: razmoskovich@mail.tau.ac.il This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Additional Supporting Information may be found in the online version of this article.
Author Contribution Statement: R.M. designed the research, performed the fieldwork, the laboratory and data analysis, and wrote the manuscript. R.D. participated in fieldwork and helped in data analysis and the writing of the MS. M.I. was involved in the research design, data analysis, and the writing of the manuscript. G.Y. was involved in the research design, fieldwork, data analysis, and the writing of the manuscript. et al. 2006). Weisz et al. (2008) showed that HMA sponges pump at a significantly slower rate (normalized to sponge volume) than LMA sponges. This difference was suggested to be related to the lower density of their choanocyte chambers and the smaller size of the chambers (Poppell et al. 2013). HMA sponges appear to receive nutritional and energy inputs from their diverse microbial communities (Weisz et al. 2007;de Kluijver et al. 2021). These microbes possess diverse metabolic pathways such as sulfate reduction, nitrification, and nitrogen fixation (Schläppy et al. 2010a;Zhang et al. 2019). In addition, the contribution of photosymbionts to sponge metabolism has also been frequently reported, especially in more oligotrophic waters (Wilkinson 1980;Pawlik et al. 2015;Achlatis et al. 2021).
In multicellular animals, metabolism is largely based on the oxidation of organic matter, and the rate of oxygen uptake (respiration) is directly linked to the rate of oxidation of organic carbon. The ratio between the rate of release of oxidized carbon (carbon dioxide) and the intake of oxygen is known as the respiratory quotient (Hatcher 1989;Koopmans et al. 2010), and in different animals, it ranges from $ 1 for diets based on carbohydrates to $ 0.7 for the high protein and fat diets typical of marine microorganisms (Anderson 1995). Consequently, measuring the amount of oxygen removed from the water processed by a sponge provides a direct estimate of the sponge's metabolic capacity and an upper limit for its organic carbon consumption and CO 2 release.
Sponges receive oxygen from the ambient waters that they pump through their elaborated aquiferous systems (Reiswig 1974;Yahel et al. 2007;Schläppy et al. 2010b). High oxygen concentration in the tissue of pumping sponges and a dramatic reduction of oxygen levels when not pumping have been reported (Hoffmann et al. 2008). Pfannkuchen (2009) concluded that a link exists between the supply of oxygen inside the sponge and the sponge's pumping activity. In contrast, Lavy et al. (2016) and Hoer et al. (2018a) witnessed events of extreme oxygen depletion from the water exhaled by the coral reef sponge T. swinhoei and Xestospongia muta respectively, during periods when the sponge was vigorously pumping.
Coral reefs are typified by intense diel cycles of many parameters: as a consequence of the intense diurnal photosynthesis and calcification, the levels of dissolved oxygen and pH in the overlying water, increase, while the levels of CO 2 and alkalinity decrease (Goldshmid et al. 2004;McMurray et al. 2014;Silverman et al. 2012). These processes affect the concentrations of food available to filtering animals and in particular to sponges. There is also increasing evidence of diel variations in the microbial composition and concentrations of dissolved organic carbon in the benthic boundary layer of coral reefs (Kelly et al. 2019;Weber and Apprill 2020).
Previous studies have primarily measured sponge metabolism in laboratory conditions or recorded short-term oxygen removal in situ (Reiswig 1974;Hoer et al. 2018a;Wooster et al. 2019) but see (Lavy et al. 2016;Hoer et al. 2018a, and Table S2). Information regarding the diel dynamics of sponge metabolism in their natural habitat is limited and this lack of knowledge is especially crucial in coral reefs, where the diel cycles are pronounced, and the ecological role of sponges is rising due to a combination of ocean acidification and other anthropogenic effects (Bell et al. 2013;De Goeij et al. 2017;. In the present study, we examined the diel dynamics of sponge metabolism by continuously measuring, in situ, oxygen removal by seven coral reef sponge species together with discrete (short-term) measurements of sponge pumping rates (PR). Our findings have revealed a high variability both within and between sponge species, as well as a marked diel pattern with enhanced nocturnal respiration in six of the seven studied species.

In situ metabolic measurements
Underwater metabolic measurements were taken from the Gulf of Aqaba, Red Sea, with minimal disturbance to the seven species of studied sponges ( Fig. 1). For each species, we measured the oxygen removal in six oscula belonging to 2-6 individuals for a duration of 24-72 h. Measurements were taken at a depth of 5-12 m in front of the Interuniversity Institute for Marine Sciences in Eilat (N 29 30.211 0 E 34 55.068 0 , Gulf of Aqaba, the northern tip of Red Sea) during summer (August 2018 and June 2019). Oxygen concentrations in the water inhaled and exhaled by the sponges were measured continuously (at 10 s intervals) using a custom-made underwater Firesting system (see below). The electronic signal was transferred in real-time to a shore-based computer connected by a cable to the underwater instruments. An additional cable linked an underwater camera to the computer onshore so that the measuring system could be continuously monitored. Each of the studied sponges was visited by divers approximately three times a day (morning, afternoon, and night) to cross-calibrate the optodes (see below) and measure the sponges' pumping rate. It should be stressed that our main focus was the diel and temporal variation in the amount of oxygen removed from each liter pumped by the sponge. The data reported here are not normalized to the sponge pumping rate or to its biomass, and therefore cannot be interpreted as respiration rate.

Studied sponges
The studied sponges are all common in the shallow reefs (from a depth of 10 m and below) in the Gulf of Aqaba. Studied specimens were selected based on ease of access and were studied at their natural reef habitat with no manipulation or physical contact. Sponges were sampled at their natural environment. The selected species display different life strategies (HMA or LMA) and diverse growth forms. Tubular-shaped sponges were represented by the delicate LMA sponge Callyspongia siphonella (Fig. 1a) and the massive and dense HMA sponge T. swinhoei (Fig. 1b), both sampled within small crevices. Encrusting sponges were represented by Mycale fistulifera (Fig. 1c), which usually covers the dead skeleton of branching corals (Meroz and Ilan 1995), and Niphates rowi (Fig. 1d), which tends to cover flat vertical surfaces in very shallow water, down to 5 m (Ilan et al. 2004). The branching sponge, Diacarnus erythraeanus ( Fig. 1e), is common in exposed and well-lit areas. Crella cyathophora (Fig. 1f) is a sponge with a soft and delicate texture, found in all areas of the reef at a wide range of depths on dead corals or artificial structures. Subarites clavatus (Lévi and Stock 1965), is a massive psammophile sponge often found half-buried at the interface between dead rocks and sand (Fig. 1).

Imaging pulse-amplitude-modulation (PAM) fluorometry
To gauge the presence of photosymbionts and estimate the photosynthetic potential of each sponge species we obtained several fresh representative samples of $ 5 square cm from each sponge taxa and transferred them in seawater to the lab, where they were qualitatively examined using an Imaging-PAM (Walz GmbH, Effeltrich, Germany) for the presence of photosynthetic activity by measuring the maximum photochemical efficiency (Fv/fm) as described in Banc-Prandi et al. (2022).

Oxygen removal measurements
Oxygen removal was continuously measured at high resolution (10 s intervals) for periods of > 24 h by comparing the concentration of dissolved oxygen in the ambient water near the sponge surface, with that of the water exhaled from its oscula. Oxygen concentration was measured and recorded in real-time using four robust (3 mm diameter) Firesting O 2 optical sensors (OXROB10-CL4, PyroScience, hereafter optodes), sealed by MacArtney (Denmark) for the underwater operation. The system was deployed in a custom-built waterproof housing (The Sexton, USA), and Arduino-based electronics controlled the communication with the instrument and the data logging. To enable better control of the experiment, the system was connected by a 70-m cable to a computer on the shore. The instrument, with four sensors, was placed a few meters from the studied sponge. The tips of three optodes measured the oxygen in the water exhaled through the oscula while the fourth optode measured the oxygen in the waters adjacent to the sponge surface.
The studied specimens were chosen based on ease of access and osculum size, selecting for larger oscula for which the placement of the optodes is easier. For multiosculated species, we sought specimens with ≤ 3 oscula. Optodes were set to measure the dissolved oxygen concentration every 10 s and mounted on custom-made manipulators that allowed precise positioning of the optode tip a few millimeters into the selected osculum (C. siphonella and the larger specimens of T. swinhoei) or (for most of the sponges) within the exhalent jet, >10 mm above the osculum (Fig. 2). To visualize the exhaled jet and verify the location of the optode tip, we squirted fluorescein-dyed seawater (filtered through 0.2 μm syringe filter) next to the sponge ostia (Video S1).
In the laboratory, the optodes were calibrated in moist air and zero oxygen water according to the manufacturer's instructions (roughly every 10 d). Following laboratory calibration, all optodes were positioned together for cross-calibration, and if the differences between optode readings exceeded 1 μmol L À1 , the calibration was repeated. During sampling, optodes were cross-calibrated underwater three times a day by carefully removing them from the oscula and positioning all four tips together (within 10 cm) in the ambient water, away from the bottom and other organisms, for 30-60 min (180-360 measurements, Fig. S1). The offset of each optode from the common median was calculated as described below.
Sampled sponges were monitored in real-time using a surveillance camera (HIK DS-2CD2432F-IW) mounted in a pressure housing (Astral Subsea) and equipped with a Near IR light source for close-range night vision. The continuous recording allowed us to compare anomalous measurements with surrounding disturbances or rare events in which the optodes were dislodged from the osculum by a passing fish.

Pumping rate measurements
Pumping rate measurements were obtained for each measured osculum at least three times; at the beginning and the end of each series, and at least once in the middle of these measurements using Dye Front Speed (DFS) method (Yahel et al. 2005;Morganti et al. 2019). Briefly, a transparent tube was positioned as close as possible above the sponge osculum, to capture the excurrent jet, and the movement of the dye inside the tube was recorded by a diver using a video camera. Night-time inspection of the sponge and sampling was performed using NearUV flashlights to minimize disturbance to the sponge and other reef organisms (see video S1 for a demonstration of the method during both day and night). In most cases, these resulted in two daytime samples and one nighttime sample per sponge. During each visit, DFS measurements were repeated at least 10 times for each osculum. Only good shots were selected for analysis, with each PR measurement being the average of $ 7 repeated DFS measurements.

Data processing and statistical analysis
Even after laboratory calibration, optode readings deviated from each other by a few micromoles per liter and this offset usually remained constant for at least 24 h under field conditions (e.g., Fig. S1), and for many days in the running seawater aquaria (data not shown). To correct for potential drifts and differences between optodes we calculated the optode-specific median offset (MedOffset i ) during each cross-calibration, as the median of the differences of the DO concentration reading DO i,t of optode i during each time interval t from the median of the four optodes at that time interval MedDO t .
To calculate the corrected DO concentration, we subtracted the MedOffset i calculated during the cross-calibration that preceded each time-series from the DO i,t reading of every optode (i) during each 10 s interval (t).
The instantaneous oxygen removal per liter pumped (ΔO 2 , μmol L À1 ) was calculated for each 10 s interval as the difference between the oxygen concentration in the ambient water (DO AMB,t ) measured by an optode positioned close to the sponge surface (AMB), and the oxygen concentration in the excurrent water (DO EX,t ) as: Optode positions were routinely switched after each crosscalibration session.
Oxygen removal was binned for each osculum into 3-h time intervals using the local daylight-saving time (sunrise in mid-August was at 06:03 and sunset at 19:22) with the 6:00 h bin (04:30-7:30) representing the dawn and the 18:00 h bin (16:30-19:30) representing the dusk. Mean values (μmol O 2 L À1 ) and confidence intervals were calculated for each 3 h bin. Time-series data for each taxon were roughly balanced between diurnal and nocturnal hours.
Due to the limited number of replications within each taxon (2-6 specimens), we used a permutation test (Manly 2007) as an alternative to a standard ANOVA in order to test for significant differences between diurnal and nocturnal ΔO 2 , sponge species, and their interactions. In each test, we compared the observed F ratio to the distribution of 10,000 F-ratio values calculated for a random assignment of the datum attributes (see the R codes we used in GitHub).
Statistical analysis was carried out with the R programming language (version 4.02), in the Rstudio environment (version 1.3.1056) and graph using package ggplot2 (Wickham, 2016).

Environmental conditions
Environmental conditions at the study site were downloaded from the website of the Israel National Monitoring Program of the Gulf of Aqaba (Shaked and Gening 2019). The water temperature range was 27.0-30.0 C during the August 2018 and 24.0-25.5 C during June 2019. As typical to the oligotrophic summer conditions in the Gulf of Aqaba, ambient chlorophyll concentrations were low, ranging from 0.12 to 0.25 mg L À1 , and nitrate (+ nitrite) and phosphate concentrations were negligible (> 0.1 μmol L À1 , and > 0.01 μmol L À1 , respectively). Ambient ammonium concentrations were also very low, but among the highest measured during the year (0.2 μmol L À1 ; data courtesy of the Israel National Monitoring Program of the Gulf of Aqaba) The average oxygen concentration during daylight in the coral reef (measured within the benthic boundary layer at the height of the sampled sponges, > 0.5 m above bottom) was 207 AE 0.1 μmol L À1 , while the nighttime average was $ 4 μmol L À1 lower (203 AE 0.2 μmol L À1 ).

Oxygen removal
Continuous, ongoing measurements (10 s intervals, for 24-72 h per osculum) of the dissolved oxygen removal (ΔO 2 ) by the sponges revealed a considerable variation across all scales, both within and between oscula, individual sponges, and species, even when the measurements were carried out simultaneously on adjacent oscula of the same individual (Fig. S2). On average, T. swinhoei and D. erythraeanus exhibited the highest values of oxygen removal (30.5 and 19.5 μmol L À1 , respectively, Table 1). Events of high oxygen removal (> 50 μmol L À1 as exemplified in Fig. 3), resembling those reported by Lavy et al. (2016), were frequent in T. swinhoei and D. erythraeanus and appeared in five of the seven studied species (Fig. S2).
Extreme oxygen removal events were predominantly nocturnal. For T. swinhoei oxygen removal > 190 μmol L À1 (i.e., near anoxia of the exhaled water) which lasted > 10 min, were recorded multiple times (Fig. S2). In some instances, we observed a dramatic elevation in the oxygen removal that lasted throughout the night: for example, D. erythraeanus at times removed 30 times more oxygen at night than during the daylight hours (Fig. 3b).

Nocturnal increase in oxygen removal
On average, during night-time the studied coral reef sponges (excluding C. siphonella) removed 86 AE 57% (mean AE 95% CI) more oxygen than during daytime from each liter of reef water they pumped. A similar value (75 AE 56%) was obtained when using median ΔO 2 rather than the mean. These differences were highly significant for six of the seven coral reef sponge species (permutation test, n = 10,000, p < 0.005, Figs. 4a, S3). The largest absolute difference was observed in T. swinhoei for which nocturnal oxygen removal (39 + 16 μmol L À1 ) were 17 μmol L À1 higher than during the daylight hours (22 + 4 μmol L À1 ), corresponding to a 74% elevation in nocturnal ΔO 2 . In species with lower average oxygen removal, such as N. rowi (5.6 + 1.6 μmol L À1 ) and M. fistulifera (4.0 + 1.7 μmol L À1 ), the absolute differences between night and day values were also smaller (5.2, and 2.7 μmol L À1 , respectively). However, in terms of relative differences, nocturnal ΔO 2 in these species was more than double their diurnal values (166%, and 100%, respectively). C. siphonella had the lowest oxygen removal (1.7 AE 1.0 μmol L À1 ), and presented a reverse pattern of oxygen removal with higher oxygen drawdown during the day, reaching a maximum in the afternoon and sharply decreasing toward midnight. However, the absolute differences were very small presenting an amplitude of $ 1 μmol L À1 .
By binning the ΔO 2 values for each osculum into 3-h bins, we identified a clear daily pattern for all species (except C. siphonella Fig. 5), in which a reduction in oxygen removal was usually observed at mid-day for both LMA and HMA sponges. Oxygen removal were maximum at night-time, with intermediate oxygen removal at dusk and dawn.
Discrete oscula flow rate measurements (OFR, L p osculum À1 h À1 ) revealed no significant difference between nocturnal and daytime OFR (permutation test N = 10,000, p > 0.05), except for specimen A of the sponge T. swinhoei (Figs. 4b, S4). In that specimen, nocturnal OFRs were significantly lower than the diurnal values, and the ΔO 2 daynight differences were much lower than the ΔO 2 differences in specimen B, in which no difference was observed between daytime and night-time OFR.

Discussion
Continuous and prolonged measurements of oxygen concentration, and calculation of its removal (ΔO 2 ) by Red Sea coral reef sponges, revealed an enhanced nocturnal oxygen uptake in six out of the seven species examined, as also reported by Hoer et al. (2018a) for the Caribbean barrel sponge Xestospongia muta. On average, the studied reef sponges extract 10.3 AE 9.9 μmol dissolved oxygen from every liter of water they pump (Tables 1, S1, S2), more than any other suspension feeder of which we are aware of (Table S3). These data are in agreement with short-term and continuous measurements of oxygen removal by reef sponges in the Caribbean and the Red Sea (e.g., Reiswig 1974; Hoer Table 1. Summary of the average oxygen removal (ΔO 2 ), pumping rate (PR), and the average difference in oxygen removal between day and night. Asterisk (*) denotes a significant night-day difference. The ex-current jet speed, oscula area, and F/R ratio (the inverse of ΔO 2 ) are provided for compatibility with the literature (Riisgård et al. 2016 and references therein). n is the number of oscula measured per sponge species. The number of individuals sampled (for multioscular species) appears in parentheses. The photosynthetic activity of the sponge species was measured using Imaging PAM. Averages and the 95% confidence intervals for the average, were calculated using the mean values of the respective parameters for each osculum, and the variability between these means. See Table S1 for details and within osculum variations.   (Table S2). Oxygen removal varied considerably among the sponge taxa, ranging from a few micromoles per liter pumped for the delicate LMA sponges C. siphonella and C. cyathophora (1.7 AE 1.0 and 3.0 AE 1.3 μmol L À1 , respectively) to the removal of tens of micromoles of oxygen per liter pumped by the dense and bulky HMA sponges T. swinhoei and D. erythraeanus (30.5 AE 10.5 and 19.5 AE 17.9 μmol L À1 , respectively). While  Fig. 5 below). Extreme and outlier values are also excluded from the plots (but not from the box plot calculation), for clarity of presentation. In each box, the median is presented as a horizontal line. The mean is presented as "x." The box encompasses the second and third quartiles. Whiskers extend to the nearest point that is < 1.5 times the interquartile range, indicating the robust prediction interval for the median. Asterisk (*) denotes a significant night-day difference. The red frame shows HMA sponges and the blue frame shows LMA sponges.
the underlying mechanisms are yet to be resolved, it is noteworthy that on average, the LMA sponges removed roughly four folds less oxygen in comparison to the HMA sponges ( Fig. 6; Table S2), suggesting that the high density of symbiotic microbes may contribute to the higher efficiency of oxygen removal by the HMA sponges. A limitation of the comparative approach of this study is that the oxygen removal was not normalized to the amount of sponge biomass providing the exhalent current for each species and this mass is likely to vary considerably between (and possibly even within) taxa. For example, Weisz et al. (2008), report that the "tissue" density and the residence time of water (volumespecific pumping rate) in Caribbean HMA sponges are much higher than those of their neighboring LMA sponges. Both of these factors can play a key role in controlling the efficiency of dissolved oxygen uptake. High variability was also observed within each osculum throughout the 24-72 h time series (Figs. 3, S2). The magnitude and pattern of oxygen removal of adjacent oscula in the same sponge also varied considerably. A similar variability was reported by Schläppy et al. (2010b) for Dysida avara (lab and in situ) and Chondrosia reniformis (lab). Large differences were also evident between the diurnal and nocturnal oxygen removal of adjacent oscula (Fig. S3A-G). In some cases, oscula that were concurrently measured exhibited similar behavior (for example oscula B1, B2, B3 of the sponge T. swinhoei, Fig. S2), whereas in other cases, adjacent oscula (e.g., oscula B1, B2 of the sponge D. erythraeanus, Fig. S2) exhibited completely disparate patterns. Similar variability was also reported by Lavy et al. (2016) for T. swinhoei. The lack of synchronicity in dissolved oxygen removal is reminiscent of similar phenomena reported for sponge pumping (Reiswig, 1971).
Events of extreme oxygen removal, ranging from 50 to 220 μmol L À1 , resembling those previously reported by us for T. swinhoei (Lavy et al. 2016), and Xestospongia muta (Hoer et al. 2018a) were observed in several but not all species. Periods of high oxygen removal occurred throughout the day but were more prevalent and longer during the night when near-anoxic oxygen concentrations were recorded in the exhalent water for periods of minutes to hours. These extreme events occurred mainly in the (massive) species T. swinhoei, D. erythraeanus, and S. clavatus. Reports of concurrent measurements of pumping rate (OFR) and excurrent oxygen removal in the same osculum are scarce (Ludeman et al. 2017). Ludeman et al. (2017) reported that ΔO 2 increased with the pumping rate, however, circumstantial evidence (Hoffmann et al. 2005;Schläppy et al. 2010b) suggests that when sponge pumping ceases, the oxygen concentration in the atrium (the void inside the osculum) may become depleted and may even reach anoxia (see also discussion in Lavy et al. (2016) and references therein). It could therefore be postulated that the events of high oxygen removal observed in the time series (Figs. 3, S2) are related to the reduction or cessation of pumping. However, since in most cases, the optodes were positioned just above the oscula (rather than inside it), when the sponge stopped pumping, the ambient flow quickly replaced the water surrounding the excurrent optode, and the ΔO 2 was reduced to zero within seconds to minutes. This phenomenon was verified using dye tracing experiments in the field and under controlled laboratory conditions with a simulated mild ambient flow (data not shown). Findings from laboratory experiments (not shown) also indicated a large reduction in the concentration of oxygen in the exhaled water when the sponge resumed pumping after a period of cessation, as also reported by Hoffmann et al. (2008). These reductions normally lasted only tens of seconds to a few minutes. It was speculated that events of extreme oxygen removal may be related to the anaerobic metabolism of the symbiotic fauna of the HMA sponge (see Lavy et al. 2016, for an in-depth discussion) but, in resemblance to other phenomena of sponge "behavior", we currently lack a mechanistic explanation. The direct comparison of the oxygen concentration in the water inhaled and exhaled by the animals that were used here is a powerful tool for the study of the in situ metabolism of totally undisturbed sponges (and other suspension feeders) in their natural habitat. When coupled with pumping rate measurements, it provides an estimate of the animal respiration rate. However, the method is not without limitations. While for sponges bearing a single osculum, such measurements provide integrated information on the entire animal metabolism. It is, however, much more challenging to comparably interpret the information recorded for a single osculum in a large multioscula sponge. The high variability in oxygen removal throughout the day/night cycle and the prevalence of extreme events, during which the sponge removed up to 220 μmol L À1 for periods of hours, also demonstrate the shortcoming of short-term measurements and stress the benefit of long-term measurements in better representing the actual metabolism of the sponge. In cases in which the lack of suitable technology or logistics hinders the application of prolonged and continuous measurements, many replications of short-term measurements are required at different times during the day in order to avoid over-or underestimation of oxygen removal in sponges.
Sponge respiration rate, or the rate at which an animal consumes dissolved oxygen, can be calculated as the product of the pumping rate (L h À1 ) and ΔO 2 (μmol L À1 ) (Reiswig 1974). Our discrete oscula flow rate (OFR) measurements, which were made concurrently with the ΔO 2 measurements when the optodes were removed for cross-calibration, showed no significant difference between daytime and night-time, and no consistent pattern. In contrast, the differences in diurnal and nocturnal ΔO 2 were highly significant for six out of the seven species examined. Therefore, and until reliable continuous OFR measurements will be made available for small sponges with oscula that cannot accommodate acoustic doppler current meters, we assume that the diel pattern in ΔO 2 described above is representative of the diel pattern of sponge respiration rate.
Three of the studied sponge species (T. swinhoei, D. erythraeanus, and S. clavatus) host a variety of photosynthetic symbionts (Bewley et al. 1996;Oren et al. 2005). The photosynthetic activity of these sponges was also qualitatively tested in this study using laboratory imaging PAM fluorometry (data not shown). During daylight hours the oxygenic activity of the photosynthetic symbionts is expected to reduce the measured differences in oxygen concentrations between the inhaled and exhaled water and increase the observed difference between day-time and night-time hours. Nevertheless, significant day/night differences were also observed in three out of the four species for which no photosynthetic symbiont is known, and no evidence of photosynthetic activity was detected by the imaging PAM (Table 1). Although these species (N. rowi, C. cyathophora, and M. fistulifera) removed on average much less oxygen from the water they pumped than the massive species (Fig. 4a), their nocturnal oxygen removal was about double that observed during the daytime, similar to the night/day removal ratio in the photosynthetically active species (Table 1). It should be noted that while no photosynthetic symbionts were observed in M. fistulifera, our imaging PAM analysis indicated that the dead branches of the coral skeletons over which M. fistulifera encrusts harbor an active photosynthetic community of endolithic algae (Tribollet et al. 2006;Aline 2008).
Sponges are known for their ability to remove copious amounts of dissolved organic carbon from the water they pump (Yahel et al. 2003;Mueller et al. 2014;Gantt et al. 2019). While a number of reports now demonstrate that LMA sponge can take up DOM (e.g., de Goeij et al. 2008;Bart et al. 2021;Ribes et al. 2023), most studies also indicate that the efficiency by which LMA sponges remove DOM (i.e., ΔDOM, μmol L À1 ) is often considerably lower than HMA sponges (e.g., Morganti et al. 2017;Olinger et al. 2021;Ribes et al. 2023). Six of the studied sponges removed much more oxygen from the water during night-time. A nocturnal increase in organic carbon concentrations, as well as a shift in the composition of the planktonic microbial community above a coral reef, were recently reported by Kelly et al. (2019). A preliminary four days and four nights study that was carried out during the June 2019 expedition, concurrently with the dissolved oxygen measurements from N. rowi, revealed a seemingly similar, although smaller and not statistically significant, nocturnal increase of both dissolved and small (< 100 μm) particulate organic matter concentrations in the reef ambient water (Fig. S5a,b). These day/night differences were reflected, to some extent, in the diet of T. swinhoei (Fig. S5c) that was measured during the same time (using the InEx-VacuSip method of Morganti et al. 2016). Observations of high removal of dissolved organic carbon by HMA sponges are in agreement with the higher ΔDO 2 we report here (Figs. 4, 6). It is likely that changes in the concentration and composition of the organic matter in the reef water will be reflected in the sponges' diet and potentially also in their metabolic rate.
Whether or not changes in food availability should also be reflected in the sponge metabolic rate is a more complex question that cannot be fully addressed here. Briefly, it is well documented that the metabolic rate of most animals studied to date, including suspension feeders such as bivalves and ascidians (Bayne and Scullard 1977;Sigsgaard et al. 2003), increases after a meal or when consuming a higher-quality diet (McCue 2006). It is currently unknown if a similar response (traditionally termed Specific Dynamic Action-SDA, Secor 2009) also occurs in sponges. If positive, and if nocturnal changes in the quality and/or the quantity of food available to the sponges are a widespread phenomenon, SDA may be involved in the nocturnal increase in sponge respiration observed in the Red Sea.
Further studies are required in order to understand the mechanisms that underlie the unusual metabolic patterns exhibited by the studied sponge species, their high temporal variability, and their contribution to the ability of these sponges to effectively cope with their environment.

Data availability statement
All the data and R code used for the analysis are available in Dryd: https://datadryad.org/stash/share/nztl3KdUGSfZavssG3Ng vd40kTCl7Gg_4wzHZtsPeTw