Sponge bioerosion on changing reefs: ocean warming poses physiological constraints to the success of a photosymbiotic excavating sponge

Excavating sponges are prominent bioeroders on coral reefs that in comparison to other benthic organisms may suffer less or may even benefit from warmer, more acidic and more eutrophic waters. Here, the photosymbiotic excavating sponge Cliona orientalis from the Great Barrier Reef was subjected to a prolonged simulation of both global and local environmental change: future seawater temperature, partial pressure of carbon dioxide (as for 2100 summer conditions under “business-as-usual” emissions), and diet supplementation with particulate organics. The individual and combined effects of the three factors on the bioerosion rates, metabolic oxygen and carbon flux, biomass change and survival of the sponge were monitored over the height of summer. Diet supplementation accelerated bioerosion rates. Acidification alone did not have a strong effect on total bioerosion or survival rates, yet it co-occurred with reduced heterotrophy. Warming above 30 °C (+2.7 °C above the local maximum monthly mean) caused extensive bleaching, lower bioerosion, and prevailing mortality, overriding the other factors and suggesting a strong metabolic dependence of the sponge on its resident symbionts. The growth, bioerosion capacity and likelihood of survival of C. orientalis and similar photosymbiotic excavating sponges could be substantially reduced rather than increased on end-of-the-century reefs under “business-as-usual” emission profiles.

on coral-excavating sponges 9 , which often account for 40-70% and up to >90% of macroborer activity on coral reefs 10 (references therein). Excavating sponges influence seawater carbonate cycling and they play important ecological roles by breaking down and sculpting the reef framework, thereby changing the heterogeneity and availability of space 10 . In contrast to many calcifiers, the abundance, activity and competitive vigour of certain excavating sponges has been observed to increase on perturbed reefs (e.g. refs [13][14][15][16].
As excavating sponges are demosponges that have a siliceous skeleton, their skeletogenesis is unlikely to be as strongly impacted by carbonate saturation changes as that of calcifiers 17 . However, ocean warming, acidification and eutrophication may still affect these sponges for several reasons. Sponge bioerosion proceeds through chemical etching of CaCO 3 chips, which purportedly involves acid regulation, followed by mechanical removal of the chips from the substrate 18,19 . The energetic cost of chemical bioerosion may be reduced in more acidified oceans, as the CaCO 3 dissolution threshold will be more easily met 20,21 . As a result, bioerosion may be enhanced on future reefs 22 , as has been suggested after observing bioerosion patterns in CaCO 3 materials from naturally low-pH waters [23][24][25][26] . Some dominant bioeroding sponges are aggressive space competitors capable of overgrowing living corals 27,28 , and reduced competitive pressure caused by increased weakness and mortality of corals may further elevate their abundances 13,14,29,30 . Moreover, bioeroding sponges are filter feeders with an efficient pumping system 31,32 and they have been observed to thrive in eutrophic waters [33][34][35] . Eutrophication may result in greater access to food and thereby increase energetic availability, which may not only affect sponge abundances and growth, but may also increase their bioerosion rates 36 .
Apart from their heterotrophic filter feeding, certain bioeroding sponges also benefit from photoautotrophic inputs provided by symbiotic dinoflagellates of the genus Symbiodinium 37 . The Cliona viridis species complex consists of such species, which are very competitive and destructive to the CaCO 3 framework 10 . Presumably, this is due to the symbiosis providing greater access to energy and hence promoting greater sponge growth, survival and bioerosion [38][39][40] . In comparison to corals, these sponges are also thought to be relatively resilient to bleaching 10 , which suggests that they are well-positioned to dominate newly available space should corals decline [13][14][15] .
The potential of increased bioerosion by excavating sponges in changing environments implies a growing threat to the three-dimensional framework of future reefs and the organisms that inhabit them 41 . However, the physiological limits to the enhanced performance of excavating sponges remain to be explored. Research on photosymbiotic bioeroding sponges under experimentally elevated partial pressure of CO 2 (pCO 2 ) displayed only little adverse response or accelerated bioerosion (reviewed in ref. 11). Experiments observing effects of elevated temperature have either not shown a strong response 42,43 , or led to bleaching 44 (also BD Ramsby, pers. comm.) or partial necrosis or mortality 45 . Combined exposure of the Indo-Pacific photosymbiotic sponge Cliona orientalis to temperature and pCO 2 anomalies under two future scenarios over eight weeks (Austral spring to summer transition) dramatically enhanced both the growth and bioerosion of the sponge 44 . However, under spring business-as-usual conditions C. orientalis bleached and showed energetic deficiencies 44,46 that are likely to lead to mortality in the longer term over summer (JKH Fang, pers. comm.).
The current study assessed the capacity of C. orientalis to erode and survive over the summer on future reefs through a 10-week simulation of both independent and concurrent warming and acidification predicted for the year 2100. To explore the influence of local nutrient availability on the outcome of globally changing climate conditions, supplementation of the diet of the sponges with nitrogen-rich particulate organics was also included as a factor. The diurnally and seasonally variable simulation was based on a "business-as-usual" greenhouse gas concentration trajectory called the Representative Concentration Pathway 8.5 (RCP8.5) 1 . As opposed to previous experiments, the individual and combined effects of temperature, pCO 2 and diet supplementation on bioerosion rates, biomass, oxygen flux, carbon flux and survivorship were studied in an orthogonal design. Such designs permit the unravelling of independent effects, which is crucial to fully understand the physiological mechanisms underpinning an organism's response. The key questions of the current study were: (1) Do simulated warming and/or acidification and/or diet supplementation accelerate or decelerate bioerosion rates and growth of C. orientalis? (2) What changes to the carbon budget of the sponge drive the observed responses? (3) Which of the three factors or which of the combinations is most likely to impact the sponge's survival in future oceans under RCP8.5 emissions?

Experimental design and sample collection. Experiments were conducted at the Heron Island Research
Station on the southern Great Barrier Reef over the Austral summer 2014-2015 using the ocean warming and acidification simulation system 47,48 (see also Supplementary Information). This system allows computer-controlled manipulation of seawater chemistry while mimicking natural diurnal and seasonal variability of the reef. Four different flow-through scenarios were produced in 4 mixing sumps under a fully orthogonal combination of temperature and pCO 2 according to the greenhouse gas concentration trajectory RCP8.5 1 (Fig. 1a and Figs S1 and S2): (a) Baseline simulation, reproducing conditions at the reference site, with present-day (PD) levels of both temperature and pCO 2 . (b) Elevated pCO 2 only, with RCP8.5 pCO 2 levels (set at 572 ± 11 μatm above PD) while maintaining PD temperature levels. (c) Elevated temperature only, with RCP8.5 temperature (3.5 °C above PD) while maintaining PD pCO 2 levels. (d) Concurrent elevation of temperature and pCO 2 to RCP8.5 levels.
Dietary supplementation was superimposed on these scenarios in orthogonal setup, dividing the experiment into supplemented and unsupplemented treatments, with three replicate 40 L experimental aquaria ascribed to each treatment (24 aquaria in total, see text below and Fig. 1a). Seawater supplying the aquaria contained dissolved organics and picoplankton cells up to 10 μm diameter that comprise the main part of sponges' diets, including clionaids 31,49,50 . The water flow to the aquaria (1 L/min in each aquarium) was interrupted for one hour daily and diet-supplemented aquaria received a dose of protein-rich microalgal mix to a final concentration of 25 μL L −1 (non-viable, 9% algal biomass, N-rich High Pro, Reed Mariculture, Campbell, USA). Sunset was chosen as the feeding time, since it follows the period of maximum mucus production by corals triggering bacterial enrichment and increased food supply for reef organisms 51,52 . The sponges in the present experiment were still pumping actively at sunset (confirmed with fluorescent dye) and their oscula were open.
The experiment was performed with Cliona orientalis Thiele, 1900 in encrusting "beta" morphology 10,11 . Apart from Symbiodinium, C. orientalis hosts a low abundance of other microorganisms 53 (such as Alpha and Gammaproteobacteria) 54 . In the current study, treatment responses were addressed at the holobiont level. Samples were collected from C. orientalis individuals inhabiting dead coral substrates at 5 m depth at Harry's Bommie (151.9357°E, 23.4675°S), Heron Island in November 2014. Sponge spicules were examined to confirm the species identity 55 . The sponges were cored to produce standardized cylinders of 35 mm diameter. Each core was then horizontally divided into 15 mm slices. The upper slice contained the sponge tissue plus a ca. 3 mm-thin underlying disc of non-infested CaCO 3 to allow downward growth and expansion (sponge core), whereas the lower slice consisted of entirely sponge-free CaCO 3 substrate (control core). Control cores were bleached for 6 h in 12.5% sodium hypochlorite to remove organic matter, thoroughly washed in distilled water and then reconditioned in running seawater for at least 48 h. Control cores served throughout the experiment to quantify background weight loss due to passive CaCO 3 dissolution, abrasion from cleaning and bioerosion by re-colonizing microborers. To eliminate variation in bioerosion rates caused by substrate properties 56 , only material sampled from six sponges excavating CaCO 3 of similar bulk density was selected (measured as outlined in Supplementary  Information and Fig. S3). The sponge-free CaCO 3 of the selected individuals had a higher initial bulk density (2.30 ± 0.02 g/cm 3 , mean ± SEM) compared to the massive Porites sp. (<1.6 g/cm 3 , e.g. ref. 57) eroded by C. orientalis in previous experiments 20,44 .
The sponge cores (n = 12 per treatment) and control cores (n = 12 per treatment) from the six genotypes were labelled and randomly distributed 58 across the 24 aquaria (4 sponge cores and 4 control cores per aquarium), with each of the 8 treatments receiving two cores of each genotype where possible. The contents of each aquarium were then assigned a group label (A through X). After collection, the sponge cores recovered for 2 weeks to allow full healing (open oscula on all core surfaces) before being exposed to a stepwise acclimation to the experimental treatments (progressive weekly steps of 0, 25, 50, 75 and 100% treatment water mixed with ambient seawater). Subsequently, full treatment conditions were maintained for 10 weeks (Fig. 1b). During the experiment, the preassigned groups of cores were rotated every 4th day between replicate aquaria (after water renewal) of the same The orthogonal setup of the experiment produced 8 different combinations using three factors (temperature, pCO 2 and diet, with 2 levels each) and 3 replicate aquaria for each combination (i.e. 2 temperatures × 2 pCO 2 levels × 2 diets × 3 replicate aquaria = 24 aquaria in total). These aquaria were randomly assorted on the experimental table, so as not to bias the amount of light that specific treatments received. A total of 96 sponge cores and 96 controls cores from different sponge-invaded dead coral substrates were distributed across the aquaria (n = 12 per treatment which equals to n = 4 per aquarium). The cores were rotated in preassigned groups between replicate aquaria of the same treatment every 4th day to minimize localized variations in light intensity and other positional effects [59][60][61] . All aquaria were equipped with temperature loggers, light loggers and small wave makers and covered with light filters (Lee Filters #131) to mimic the light spectrum of the collection depth 88 (b) After collection, the sponge cores recovered for 2 weeks to allow full healing before being exposed to a stepwise acclimation to the experimental treatments (progressive weekly steps of 0, 25, 50, 75 and 100% treatment water). Full treatment (100% treatment water) was reached on the 1 st of January (start) and maintained until the 9 th of March 2015 (end). Physiological measurements were performed with a subset of the sponges in the 4 th week after the onset of full treatment conditions ("week 4" of the experiment). treatment to minimize localized variations in light intensity and other positional effects [59][60][61] . Upstream temperature and pCO 2 were continuously monitored in the four mixing sumps, while downstream temperature and pH were continuously monitored in the experimental tanks, with additional weekly pH measurements made inside each of the 24 aquaria (see Supplementary Methods for details). The use of a single sump per experimental treatment is technically a pseudo-replicated design 62 . The flow rate and dimensions of the dark sump however significantly limit the potential for confounding effects (see Supplementary Methods). Replicating sumps would necessitate a counterproductive reduction in sump size, significant increases to the energy budget on an offshore island, and would prove to be impractical for an experiment that aims to deliver in situ diurnal fluctuations for multiple factors in reef water of high quality.
Four weeks after the onset of full treatment conditions (hereafter referred to as "week 4", Fig. 1b), physiological measurements were performed with a subset of the specimens, which were subsequently sacrificed (n = 7 out of 12 per treatment). At this point in time the top surfaces of all cores were still entirely covered by sponge tissue and oscula were open. The remaining specimens (n = 5) were retained until the end of the summer period. During the experiment, aerial exposure of the sponges was prevented and epibionts were removed regularly using a pair of forceps and/or a soft brush, taking care not to damage the sponge tissue.
Bleaching and mortality. Bleaching, partial and complete mortality of the sponge cores were scored throughout the experiment. Bleaching was assessed by the colour change of the sponges from dark brown to pale yellow (Fig. 2a,b) and confirmed upon dissection to eliminate the possibility that symbionts were relocated deeper inside the core 63 . As a measure of photoinactivation during bleaching, the maximum potential quantum . Data are averaged across diet, since diet did not influence Fv/Fm ratios. Survival of C. orientalis decreased over the summer in all heated treatments, both for treatments with unsupplemented (d) and supplemented diets (e). Only complete mortality was considered for this comparison. Crosses represent the arrival at full treatment conditions, the 4 th week thereafter when physiological measurements were performed and the end of the experiment respectively. Arrows indicate the points in time when 100% bleaching of sponge cores in heated treatments were reached. yield of photosystem II (Fv/Fm) of the sponge symbionts was measured once a week at 20:00 h using pulse amplitude modulated (PAM) chlorophyll fluorometry 64 (dark-acclimated; Diving-PAM, Walz, Effeltrich, Germany). The PAM measurements (in combination with the oxygen flux assay, see below) further confirmed the observed bleaching 46,65 . Towards the end of the experiment, some sponge cores (only in the treatments where diet was not supplemented) experienced partial mortality, defined here as the loss of sponge tissue from areas of the core, partially exposing the CaCO 3 substrate but without the appearance of black patches. The assessment of complete mortality was based on blackening of all sponge tissue, decaying odour and absence of oscula 45 . Bioerosion rates. To measure total sponge bioerosion over time, all sponge and control cores were buoyant-weighed 66 at the onset of the full-treatments, at week 4 and at the end of the experiment. The decrease in the buoyant mass (BM) of the sponge cores was calibrated for seawater density and corrected for residual mass loss using the control cores. Changes in BM were converted to changes in CaCO 3 mass of the cores (see Supplementary Information). The upper surface areas of sacrificed sponge explants were estimated using a standard aluminium foil method 67 and total bioerosion rates were expressed as mg CaCO 3 cm −2 day −1 in reference to this surface.
Metabolic oxygen and carbon flux. Photosynthetic and respiratory activities of the sponge cores (n = 7 replicates per treatment) were assessed at week 4 of the experiment, using an OXY-10 oxygen meter (PreSens, Regensburg, Germany) as described in detail previously 46 . Sponges were 30-min dark-acclimated before being placed -without exposing them to air-inside sealable chambers (250 ml) filled with respective scenario seawater (filtered to 0.45 μm to remove most microorganisms from the seawater), and non-viable algal mix was added to supplement half of the treatments to the same per sponge concentration as during daily supplementation of the experimental aquaria. The sponges were subjected to a cycle of 30 min darkness followed by 20 min light at 450 μmol quanta m −2 s −1 (Aqua Medic Ocean lights, Bissendorf, Germany), with oxygen levels logged every 15 s. Dark respiration (R dark ) and maximum net photosynthesis (max P net at saturating light intensity) 65 were quantified based on the oxygen depletion or evolution (μM O 2 cm −2 h −1 ) during the dark and light phase of the cycle respectively. R dark reflects changes in holobiont metabolism as well as pumping rates (not measured here). The same procedure was performed with the control cores, which generated minimal non-sponge metabolic rates that were used to correct the sponge results. Corrected photosynthetic values were ascribed to Symbiodinium. The hourly respiratory and net photosynthetic rates were scaled up to daily rates and converted into carbon equivalents (see Supplementary Information).
To measure the uptake or excretion of particulate and dissolved organic carbon by the sponges (POC and DOC respectively), incubations in confined seawater (5 L) were performed 46 . Dietary supplementation of the incubated sponges took place in the same manner as daily in the experimental aquaria. Seawater samples were collected and analysed as detailed in the Supplementary Information. Uptake rates of DOC and POC were expressed as mg C cm −2 day −1 and added to the carbon flux that was calculated from photosynthesis and respiration. In that manner, the daily net carbon surplus (C net ) available to the sponge holobiont was estimated.
Biomass. After completion of the week 4 physiological measurements, analysed sponge and control cores were snap-frozen in liquid nitrogen and stored at −80 °C. The sponge cores were later vertically divided into 4 quarters (A, B, C and D), three of which were subjected to different analyses. To quantify total organic mass of the core, spicular mass of C. orientalis and CaCO 3 mass (M CaCO3 ), a loss after combustion method was performed on quarter A with correction factors applied 66 ; to quantify organic mass of other endo/epilithic organisms that could inhabit the sponge holobiont, an acid decalcification method was conducted with quarter B 66 . The organic mass of C. orientalis was estimated by subtracting the organic mass of other organisms from the total organic mass. All biomass data were standardized to CaCO 3 mass of the whole core (mg (biomass) /g (core) ). Quarter C was used to quantify the bulk density of the sponge core ( Supplementary Information and Fig. S3).
Data analysis. Preliminary analysis did not detect group-specific effects for any of the variables and therefore replicate groups of each treatment were pooled 68 . No single genotype or group of genotypes deviated from the others in a way that would substantially bias the analysis (Fig. S4). Dependent variables were analysed using a factorial, fully crossed analysis of variance (ANOVA) with three categorical factors having two levels each, resulting in a 2 × 2 × 2 matrix: temperature (PD and RCP8.5), pCO 2 (PD and RCP8.5), diet (supplemented and unsupplemented; Fig. 1a). Datasets were tested for normality (Shapiro-Wilk test) and homogeneity of variances (Levene's test). Log or square root transformations were applied if normality was violated, and heteroscedastic datasets were assesed at a reduced alpha level of 0.01 68 (Table S4). Main effects were reported when there were no significant three-way or two-way interactive effects. Differences in the case of significant three-way interactions were explored using simple two-way interactions at each level of the third factor whilst using the error term (sum of squares and degrees of freedom) of the three-way ANOVA 69 . If such simple two-way interactions were significant, pairwise comparisons with Bonferroni adjustments (to reduce type I errors) were performed to determine the effect of a factor at each level of the other factor 69 . Chlorophyll fluorescence data were analysed with a three-way repeated measures ANOVA examining the effects and interactions of the categorical factors over time. Partial and complete mortality of the sponge cores were assessed at the end of the experiment with a two-way ANOVA and a two-proportions z-test. Statistical analyses were evaluated at a 0.05 level of significance (unless otherwise stated) and were performed using SPSS Statistics software (IBM, New York).
Data availability. The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

Results
Treatment conditions. Mean pCO 2 levels across the experiment were 493 ± 20 μatm (mean ± SEM hereafter, unless otherwise specified) for treatments with PD levels and 982 ± 17 μatm for treatments with RCP8.5 levels, thus approximately denoting a doubling from PD to RCP8.5. Peak temperature levels of 28.3 ± 0.1 °C for the PD scenarios and 31.6 ± 0.1 °C for the elevated scenarios [+1 and +4.3 °C above the maximum monthly mean (MMM) respectively 70 ] were reached four weeks after the onset of full treatment conditions (Fig. S1). Details of the experimental conditions are given in the Supplementary Information (Figs S1 and S2 and Tables S1 and S2).
Bleaching of sponge cores. Sponges in the heated treatments began to pale 24 days after the onset of acclimation, when midday temperature was 29.4 ± 0.06 °C (MMM + 2.1 °C) and midday pCO 2 was 452 μatm in the low pCO 2 treatment and 903 μatm in the high pCO 2 treatment. By the onset of full treatment conditions, approximately one third of all the sponges exposed to simulated warming were visibly bleached (Fig. 2a,b). Bleaching of 100% of the sponge population in each heated treatment was reached after approximately the same number of full treatment days, regardless of diet (arrows in Fig. 2d,e). In line with the bleaching observations, maximum potential quantum yield of fluorescence for photosystem II of Symbiodinium decreased significantly in all heated treatments (F = 53.76, p < 0.001; Fig. 2c). Throughout the experiment, bleaching occurred only in the heated treatments.
Bioerosion rates. Total sponge bioerosion corrected for control values ranged between 2.64 to 4.39 mg CaCO 3 cm −2 day −1 over the first 4 weeks of full treatment and was on average 35% higher for sponges with supplemented diets compared to unsupplemented sponges (F (1,48) = 8.07, p = 0.007; Fig. 3a,b; Table S4). Simulated warming lowered total bioerosion rates by 20% (F (1,48) = 4.54, p = 0.038), whereas acidification on its own did not have a strong effect on total bioerosion. Rates of weight loss in the sponge-free control cores were an order of magnitude lower than sponge bioerosion rates (0.28-0.59 mg CaCO 3 cm −2 day −1 ) and did not differ between the treatments.
The bulk densities of the sponge cores at week 4 ranged from 1.67 to 1.82 g cm −3 across the different treatments with the average porosity being 35.90 ± 0.94%. The lowest densities and highest porosities were found in the treatments with both PD temperature and PD pCO 2 , yet these differences were not significant.

Metabolic oxygen and carbon flux. Photosynthesis in week 4 of the experiment exceeded respira-
tion rates only in the absence of simulated warming where the sponges did not bleach (max P net , Fig. 4a,b and Table S4). When the diet was not supplemented, acidification enhanced photosynthesis by 80% (F (1,48) = 5.96, p = 0.018), but this effect was absent when the diet was supplemented. Warming, acidification and diet interactively drove the response of photosynthesis (3-way interaction, F (1,48) = 6.48, p = 0.014). Dark respiration (R dark , Fig. 4c,d) of the sponge cores was affected only by temperature, with sponges under RCP8.5 temperature respiring at a slower rate (F (1,48) = 36.44, p = 0.00).
At PD temperature and pCO 2, supplemented sponges had a lower daily net carbon surplus than unsupplemented sponges (C net , Fig. 5 and Table S4; F (1,48) = 15.19, p < 0.001 and F (1,48) = 20.13, p < 0.0001). Regardless of diet or pCO 2 , only the sponges exposed to warming were net consumers of metabolic carbon i.e. the metabolic demand of bleached sponges surpassed the carbon that could be autotrophically or heterotrophically harvested. . Maximum net photosynthesis (max P net ), dark respiration (R dark ) and dissolved organic carbon (DOC) uptake of Cliona orientalis at Heron Island on the southern Great Barrier Reef receiving an unsupplemented (a,c and e respectively) and supplemented diet (b,d and f respectively) under independent and concurrent simulation of warming and acidification from present-day (PD) summer levels to Representative Concentration Pathway 8.5 (RCP8.5) levels of the year 2100 (mean ± SEM). All parameters were measured 4 weeks after the onset of full treatment conditions. Max P net and R dark rates represent hourly rates of oxygen evolution and depletion, whereas DOC uptake was measured over a 24 h period. The designations inter, temp, or pCO 2 indicate significant 3-way or 2-way interactive effects as specified, or (simple) main effects for temperature or pCO 2 respectively within each panel. When interactive effects are present, the capitalized letters beneath the bars indicate statistical differences. When standardized to sponge biomass, reported R dark rates (c and d) range between 14.46 ± 1.22 and 24.60 ± 1.49 μmol O 2 g −1 h −1 (mean ± SEM).
SCiEnTifiC REpoRtS | 7: 10705 | DOI:10.1038/s41598-017-10947-1 Acidification led to reduced C net surplus at PD temperature levels when the sponges were not supplemented (F (1,48) = 28.78, p < 0.0001), reflecting the loss in DOC uptake despite the increase in photosynthetic activity which only contributes carbon during light hours.
Biomass. Regardless of diet or pCO 2, the sponges were able to build or maintain more organic biomass under PD temperature than under RCP8.5 temperature (on average 41.74 and 36.69 mg (biomass) /g (core) respectively, F (1,48) = 13.16, p < 0.001, Tables S3 and S4). Non-sponge organics were found to be on average 1.94 ± 0.3% of total organics (no difference between treatments) and were therefore considered negligible in this study. The spicular mass of the sponge cores formed a large fraction of the sponge biomass (approx. 35% of total sponge biomass) and did not differ between treatments.
Sponge mortality. After physiological responses had been measured at week 4 and during the second half of the experiment, the tissue of all bleached sponges increasingly retracted to a small surface area of the core, despite regular removal of epibionts. Bare parts of the core were colonized by a diverse algal community consisting mainly of crustose coralline algae (CCA), Derbesia sp., Ulva sp., pennate diatoms and free-living dinoflagellates (D Bender-Champ, pers. comm.). In approximately half of the specimens exposed to simulated warming, bleaching resulted in complete mortality (z = 3.65, p < 0.001) (Fig. 2d,e). The first cases of complete mortality co-occurred amongst PD pCO 2 and RCP8.5 pCO 2 treatments 6 weeks after the onset of full treatment, regardless of dietary supplementation. Complete mortality was always preceded by bleaching, occurring between 3-6 weeks post bleaching. Dead sponge cores were inhabited by fungi and cyanobacteria (Spirulina sp.) in addition to the algal community described above (D Bender-Champ, pers. comm.).
Regardless of diet and pCO 2 , all specimens exposed to PD temperatures survived beyond the end of the summer in the experiment, without any signs of bleaching. However, in the unsupplemented treatment the substrate of some of these specimens had become partly exposed by the end of the experiment due to sponge tissue loss (partial mortality, Fig. 6). Here, partial mortality and consecutive algal settlement was found in 2 out of 5 of the PD pCO 2 and 3 out of 5 of the RCP8.5 pCO 2 sponges (z = 1.5806, p = 0.057 and z = 2.0698, p = 0.019, respectively) and was an independent effect of diet. Supplemented sponges retained their tissue, but unsupplemented sponges lost about one third of their tissue cover, resulting in 5.20 ± 0.78 cm 2 live tissue compared to 8.23 ± 0.65 cm 2 in supplemented sponges at the end of the experiment (two-way ANOVA, F (1,16) = 7.96, p = 0.015). The surface tissue of supplemented sponges also appeared healthier and thicker.

Discussion
Exposure of the photosymbiotic sponge Cliona orientalis to RCP8.5 summer projections for the year 2100 at Heron Island led to decreases in biomass and rates of bioerosion, autotrophy, heterotrophy and survival. These responses were driven by resource deprivation of both the host and the symbiont and were primarily caused by impacts of simulated warming, but also acidification, even when additional heterotrophic food sources were provided. Excavating sponges are often regarded as "winners" on disturbed coral reefs under projected future conditions 11 . However, based on the combined results of the current experiment, under "business-as-usual" CO 2 Figure 5. Daily net carbon surplus (C net ) of Cliona orientalis at Heron Island on the southern Great Barrier Reef receiving an unsupplemented (a) and supplemented diet (b) under independent and concurrent simulation of warming and acidification from present-day (PD) levels to Representative Concentration Pathway 8.5 (RCP8.5) summer levels of the year 2100 (mean ± SEM). The carbon surplus incorporates autotrophic inputs (photosynthetically fixed carbon), heterotrophic inputs (net uptake of dissolved and particulate organic carbon) and respiratory output (carbon used for maintenance, growth etc.) over a 24 h period after 4 weeks of exposure to full treatment conditions. The response of C net to the treatments was driven by 2-way interactive effects (2-inter) for each combination of the 3 factors. The capitalized letters beneath the bars indicate statistical differences.
emissions C. orientalis and possibly other similar photosymbiotic excavating sponges are expected to suffer losses at the end of the current century comparable to those projected for scleractinian corals 22 .
Extrapolating measured present-day summer rates of substrate removal by C. orientalis to an annual mean resulted in a total bioerosion rate of 12.4 kg CaCO 3 m −2 year −1 when the diet was not supplemented, which is consistent with earlier results from similarly dense substrates 56 . Our study demonstrated a 35% stimulation of total sponge bioerosion by the availability of food, purportedly also allowing faster expansion as would be expected in nutrient-rich environments such as inshore reefs 35 . Our results do not imply a sustained increase in bioerosion performance by photosymbiotic sponges in future oceans however, mainly due to observed adverse impacts of simulated warming. We suggest that bioerosion rates were mainly reduced by the failure of the photosynthesis of the symbionts, highlighting the importance of the symbiosis to supply energy for bioerosion [38][39][40] . Previously documented temperature effects on bioerosion of Cliona species have shown variable responses 42,45,71,72 . Our results are in accordance with the reduction of bioerosion rates under warming reported from a short-term experiment using the same species 45 . Simulated acidification co-occurred with heterotrophic carbon losses in our study, which may have resulted in the loss of a clear response through bioerosion as opposed to previous studies: Bioerosion by C. orientalis significantly increased under similar acidification in a 3-day closed system experiment (chemical bioerosion measured through alkalinity changes) 45 , in a 10-day flow-through experiment (total bioerosion measured through buoyant mass changes) 20 and in a 8-week flow-through experiment over Austral spring (both of the above measures) 44 . We hypothesize that in our study over Austral summer, the longer-term losses to the sponge's carbon budget under acidification may have lessened the energy invested into bioerosion, thereby decoupling acidification effects from erosion enhancement. We measured bioerosion through buoyant mass changes, but identifying chemical and mechanical rates in future studies could better elucidate potential acidification impacts.
Bleaching of C. orientalis tissues has previously been shown under concurrent simulation of RCP8.5 warming and acidification 44 . Here, we provide evidence that our observation of bleaching was caused by warming above the mean summer water temperature. Whether the bleaching is a result of oxidative stress and mechanisms similar to those established for the cnidarian-Symbiodinium partnership 73 remains to be explored, but our results show that the bleached sponges experience reduced holobiont productivity and increased mortality, as is known for bleached corals 73 . The decrease of photosynthetic activity translates into reduced access to resources, yet even though bleached specimens had lower biomass and no symbiont population, they maintained a considerable respiratory demand (metabolic needs increase inherently with temperature rise), which would have represented an additional resource drain. To date, bleaching of Cliona spp. has been considered a rare event in or ex situ 10 . In October 2015 a natural bleaching event was reported for the first time for a clionaid sponge in the lower Florida Keys 74 , but the sponges were only partially bleached and appeared to survive and recover (M Hill, pers. comm.). An unknown encrusting Cliona sp. exhibited impaired photosynthesis during a heating event in March 2013 75 (MMM + 1 °C or more) in depths down to 15 m near Onslow, NW Australia (CHL Schönberg, pers.comm.). In our experiment, the sponges began to pale prior to the full establishment of the 2100 climate scenarios, indicating that at least partial or occasional bleaching could become more common in natural populations under similar warming events before the end of the current century.
Acidification stimulated higher photosynthetic rates in unbleached specimens that were not diet-supplemented, possibly because less resources had to be invested into the conversion of HCO 3 − to CO 2 for use by the Rubisco enzyme 76 . No such effect was found when the sponges received a supplemented diet since they may have been less dependent on autotrophic inputs, despite stable respiration that could otherwise provide an alternative source of CO 2 to the symbiont 76 . With regards to organic carbon uptake, an increase in heterotrophic feeding of the bleached sponges could have served to compensate for the loss of autotrophic carbon. However, sponges in the heated treatments had reduced rather than increased carbon uptake rates. Filter feeding is an energetically costly process 77 that could be rendered unsustainable in the bleached sponges. Negative effects of temperature on sponge feeding have been reported for the Great Barrier Reef (GBR) sponge Rhopaloeides odorabile, with filtration efficiency, pumping rate and choanocyte chamber density and size reduced at 31 °C (MMM + 2 °C) 78 . POC uptake in our sponges was only insignificantly reduced, yet choanocyte functioning and water pumping may also facilitate DOC uptake 79 , which could explain the losses observed here under simulated warming. Acidification similarly reduced uptake of DOC, which may point towards a trade-off between autotrophy (due to stimulation of photosynthesis under elevated pCO 2 ) and heterotrophy. Alternatively, prolonged high levels of H + may have had deleterious effects on the filter-feeding capacity of the sponge, for example by affecting mitochondrial ATP recycling which is crucial to the flagellar beating 80 . Further investigations are necessary to confirm what caused the observed decrease in carbon uptake under acidification.
In treatments with either warming or acidification, dietary supplementation slightly ameliorated the effect of the other factors; the sponges took advantage of the elevated concentration of organics and more organic carbon was incorporated. Sponges in present-day conditions did not respond in the same way. Compared to supplemented sponges that were not in need of food, unsupplemented sponges may have been relatively starved (only particles up to 10 μm in diameter were retained in unsupplemented tanks) and were therefore filtering more actively (i.e. taking up more carbon from the seawater) when assessed over a 24h-period at the height of summer.
The assumption of resource deficiency is also supported by the partial mortality found towards the end of the experiment in unbleached sponges that did not receive a supplemented diet.
All bleached sponges had lower organic biomass, which will in part explain the reduction in total bioerosion rates and respiration in the heated treatments. Bleached C. orientalis in previous experiments increased biomass 44,46 , but this may have been due to a lag effect after an initial stimulation of growth earlier in spring when the symbionts were still present (JKH Fang, pers. comm.). The duration and the severity of our experimental warming went beyond physiological thresholds, and bleached sponges suffered biomass losses due to reduced autotrophy and heterotrophy. Supplementary feeding with protein-rich organics was expected to stimulate growth, since the extent to which the sponge holobiont can utilize photosynthetically transferred or heterotrophically attained carbon beyond its respiratory needs in the oligotrophic reef waters may depend on the availability of commonly limiting nutrients such as nitrogen 81 . No such effect was observed, but instead the resources gained from the extra nutrition may have been redirected towards bioerosion activity. Alternatively, the sponges may have not been nitrogen-limited in the first place, yet the partial mortality observed in the unsupplemented treatments is contraindicative.
Even though spiculogenesis is considered energy-demanding 82,83 , the mass of siliceous spicule was not affected by any of the treatments, which confirms previous results 17,44 . The formation of one demosponge spicule may take approximately one week and an even longer lag period needs to be considered when looking for environmental effects on spicules 84 .
A daily net carbon surplus was only realized under present-day temperatures, and that surplus was significantly reduced with acidification, since organic carbon uptake decreased. Overall, the physiological measurements revealed carbon deprivation in the bleached sponges, which explains their subsequent mortality. Heterotrophy by itself does not appear to be sufficient for the energetic needs of C. orientalis (also shown by ref. 46). We conclude that the symbiosis between the sponge and its dinoflagellates is neither facultative nor merely beneficial to bioerosion performance alone, but it is vital to the survival of the holobiont in several ways.
For marine sponges, interactive effects of environmental factors such as warming, acidification and food availability remain largely unknown and vary between species depending on the natural conditions that they are adapted to 45,85 . The lack of and the need for related studies has recently been highlighted in order to quantify present trends of carbonate budgets and to provide modelled trajectories for different scenarios to facilitate management 11 . Our experiment assessed the effects of simulated warming, acidification and diet supplementation on the bioerosion efficiency and survival of C. orientalis separately and in combination. Overall, warming appears to be a more important factor in determining the physiological thresholds of the sponge, and increased bioerosion rates observed in previous short-term acidification-only experiments 11 may not realistically reflect future developments under coexisting ocean warming and acidification. In non-bioeroding sponges from the GBR, the effects of RCP8.5 warming were also shown to be physiologically more important than acidification effects 86 , but Caribbean sponges exposed to similar warming and acidification for 24 days remained largely unaltered 85 . Our study suggests that future climate conditions may temporarily incur increased bioerosion rates at intermediate levels of environmental change, but that ultimately escalating environmental conditions under presently predicted "business-as-usual" fossil fuel usage will cause photosymbiotic clionaids to fail along with other benthic organisms. This supports a parabolic rather than a linear response of bioeroding sponges to future change 11 . Bioerosion and biomass maintenance are not the only energetically-costly activities of C. orientalis; other costly processes such as reproduction or competition for space 39 could also be impacted by climate change, thereby possibly further reducing the likelihood of survival in this species.
Climate simulations on coral reef assemblages from the southern GBR have provided evidence that firstly, there has been little adjustment of corals to changes over the past 100 years, and secondly, 100 years from now corals are unlikely to calcify or survive over RCP8.5 summers (when temperatures and seawater pCO 2 concentrations are at their seasonal highest) and beyond 47 . To date, excavating sponges were often considered more tolerant to future changes than scleractian corals 10 . However, our study stresses that there are limits especially to the temperature but also the pCO 2 conditions that C. orientalis can tolerate, and that, despite subtle benefits from higher food availability, future summers can be expected to have adverse effects on the bioerosion capacity, general physiology and ultimately survival of the sponge. Keeping in mind the limitations of extrapolating from simulations to the field, we nevertheless cannot support that excavating sponges such as C. orientalis are still likely to be "winners" in 2100.
We acknowledge that the present results may only be relevant for excavating sponges that are symbiotic with Symbiodinium and that other excavating sponges might show different response patterns. This remains an essential area to explore if we are to gain a complete understanding of the implications of a warmer, more acidic and more eutrophic ocean for these key reef organisms. It is also important to note that our study did not identify the temperature threshold for the loss of C. orientalis from coral reefs like those of Heron Island and the southern Great Barrier Reef, or how well C. orientalis is performing as climate change intensifies in the interim, causing mass coral mortality events 87 and thus increasing the availability of bioerosion substrate.