Sponge and coral zooxanthellae in heat and light: preliminary results of photochemical efficiency monitored with pulse amplitude modulated fluorometry

Authors

  • Christine H. L. Schönberg,

    1.  Centre for Marine Studies, The University of Queensland, St Lucia, Qld, Australia
    2.  Present address: Carl von Ossietzky University Oldenburg, Faculty 5, Institute of Biology and Environmental Sciences, Department of Animal Biodiversity and Evolution, Oldenburg, Germany
    Search for more papers by this author
  • Ryota Suwa,

    1.  Department of Marine and Environmental Science, Graduate School of Engineering and Science, University of the Ryukyus, Nishihara, Okinawa, Japan
    Search for more papers by this author
  • Michio Hidaka,

    1.  Department of Marine and Environmental Science, Graduate School of Engineering and Science, University of the Ryukyus, Nishihara, Okinawa, Japan
    Search for more papers by this author
  • William Kok Weng Loh

    1.  Centre for Marine Studies, The University of Queensland, St Lucia, Qld, Australia
    2.  Marine Biology Australia, St Lucia, Qld, Australia
    Search for more papers by this author

Christine H. L. Schönberg, Department of Animal Biodiversity and Evolution, Institute of Biology and Environmental Sciences, Faculty 5, Carl von Ossietzky University Oldenburg, 26111 Oldenburg, Germany. E-mail: christine.schoenberg@uni-oldenburg.de

Abstract

Photochemical efficiency (Fv/Fm) was compared between a common symbiotic bioeroding sponge, Cliona cf. orientalis, and a common reef-builder, Acropora palifera using pulse amplitude modulated fluorescence (PAM). The study was conducted on Sesoko Island, Okinawa, where reefs were severely damaged during previous bleaching episodes. Sponge and coral dinoflagellate symbionts were treated with heat and light in a tank experiment, both in hospite (=still within their host) and isolated from their hosts. We found significant differences for photochemical efficiency of holobionts (=host and symbiont together) compared to the isolate symbionts and over time. All symbionts suffered in isolation and displayed stronger reactions to the treatments, and there was evidence for increasing damage despite returning to control conditions. However, because of large variability of the bi-symbiont coral samples and restrictions of the experimental design, our main results remained inconclusive, with no significant differences between sponge and coral samples and between the different stress treatments. Judging the results based on the uniform trends in the subsets of data, the G-clade sponge symbionts appeared to be more stress tolerant than the C- and D-clade coral symbionts, with no treatment effects in hospite and less damage in isolation compared to the coral symbionts, but this is an unconfirmed assumption. Isolated sponge symbionts were very resistant against heat stress, but may have suffered from light stress. In hospite, the latter risk can be countered by the sponge’s 3-dimensional morphology, the endolithic life style that affords shading, and by behavioural adaptation, i.e. the ability to move symbionts away from the source of stress. Overall, C. cf. orientalis symbionts displayed a more stable photochemical efficiency during and after stress than those of A. palifera. Results of this study suggest that with climate change C. cf. orientalis might have a better survival potential than A. palifera, but further investigations are necessary.

Problem

Recent studies on coral reefs are largely concerned with anthropogenically caused damage and effects of climate change. Global warming has resulted in a mounting number of serious bleaching events (e.g.Loya et al. 2001) and is believed to cause lasting damage if not near-extinction of the coral reef environments we presently know (e.g.Hoegh-Guldberg 1999; Risk 1999; Reaser et al. 2000). Bleaching events largely relate to the failure of the symbiosis between corals and dinoflagellates of the genus Symbiodinium (zooxanthellae), and a large number of studies have addressed coral zooxanthellae, investigating their system and phylogeny, distribution and host relationships (e.g.LaJeunesse 2001, 2002). Increasing research effort in that area revealed that different clades of coral zooxanthellae display different tolerance levels against stress (Baker et al. 2004; van Oppen et al. 2005).

In contrast to corals, other zooxanthellate invertebrates are less well-studied, although they may also have vital roles in the persistence and resilience of reefs. Until recently, it remained largely unknown that some sponges contain zooxanthellae. In this context, it is interesting to note that symbiotic dinoflagellates mainly occur in bioeroding sponges of the family Clionaidae that are leading agents of reef bioerosion (e.g.Hudson 1977; Scoffin et al. 1980). Despite having the same kind of symbionts, corals and clionaid sponges play antagonistic roles in their habitat – reef accretion and reef degradation. If environmental stress relating to bleaching events may affect one group more strongly than the other, the balance between calcification and decalcification may ultimately shift.

Earlier accounts of dinoflagellate symbionts in bioeroding sponges include transmission electron microscopy and descriptions of zooxanthellae of Cliona viridis from the Mediterranean, Cliona orientalis from New Caledonia and Cliona varians from the Caribbean (Crumeyrolles-Duclaux 1970; Vacelet 1981; Rützler 1990; respectively). The sponge symbionts occur intracellularly in all studied sponges (e.g.Rützler 1974, 1990; Vacelet 1981). But, as yet, we know little about the physiology of sponge symbionts. Apparently, they play an important role in the nutrition and enhance growth of sponges, bioerosion rates, survival and possibly reproduction (C. viridis: Rosell & Uriz 1992; Rosell 1993; C. varians: Hill 1996; C. orientalis: Schönberg et al. 2005; Schönberg 2006); however, the exact mechanisms are not yet understood. We assume that bioeroding sponges can bleach, but again, we have little information. Acker & Risk (1985) noted that sediment cover can result in pigment loss in these sponges; however, Vicente (1990) observed that bioeroding sponges did not bleach during a heat event in Puerto Rico. Hill & Wilcox (1998) succeeded in experimentally bleaching C. varians by transplanting it to shallower depths. After a recovery period, the sponges contained a clade of zooxanthellae that occurred in a sympatric anemone. Regrettably, the authors did not investigate which clade the sponges harboured before the experiment. Schönberg & Loh (2005) found the Australian C. orientalis to contain G-type Symbiodinium. They agreed with Mariani et al. (2000) that the symbionts of bioeroding sponges are very likely maternally transferred. They further reasoned that the G clade might be comparatively hardy, possibly explaining the perceived bleaching resistance in bioeroding sponges. However, to date, no experimental evidence is available to support this theory.

Our study provides results obtained from an aquarium experiment including symbiotic sponges and corals, thus contrasting bioeroders and reef builders. It compares their stress resistance and investigates host effects. In other words, we assessed the probabilities of the following null hypotheses: (i) coral and sponge symbionts have the same reaction to climate change related stress; (ii) symbionts will show similar stress reactions in hospite and in isolation; and (iii) different sources of stress will cause similar stress reactions in the symbionts.

Material and Methods

Sampling

Live fragments of Cliona cf. orientalisThiele, 1900 and Acropora palifera (Lamarck 1816) were obtained at Bise near Motobu, Okinawa, Japan (26°42′ N, 127°52′ E; Fig. 1) mid-July 2004. Both species were found to be abundant in the area and have prevailed despite previous, serious heating events (e.g.Loya et al. 2001; Omori et al. 2001). The sponge is a member of the aggressive Cliona viridis species complex, species of which have been found in all oceans between 42° N and 25° S. The site was chosen for its easy and public access without sampling restrictions and the occurrence of both studied species.

Figure 1.

 Sample site near Sesoko Island, Okinawa, Ryukyu Islands, Japan. A – Japan with the Ryukyu Islands in the south. B – Okinawa. C – Study area in the north-west of Okinawa. D – Enlargement of study area with sample area at Bise Beach (large arrow) and Sesoko Island (small arrow).

Sampling was conducted while snorkelling in the sheltered bay at Bise Beach, harvesting specimens from very shallow depths of no more than 0.8 m and from horizontal surfaces fully exposed to light. The water temperature was about 28 °C during sampling. C. cf. orientalis occurred in papillate growth and formed small colonies on smooth blocks of calcium carbonate that were lightly covered with algal turf (Fig. 2). Small, about half-fist sized pieces of sponge-inhabited substrate were removed by using hammer and chisel. Branches of A. palifera were broken off from small colonies. All material was contained in plastic bags filled with seawater and transported to the Tropical Biosphere Research Center, Sesoko Station, University of the Ryukyus. At the Station, all fragments were kept in a large indoor, flow-through tank, where they adapted to the same conditions for 2 days.

Figure 2.

Cliona cf. orientalis at Bise Beach, Okinawa, Japan in 0.5 m depth.

Species identification

Taxonomic identity of the sponges was studied under a light microscope viewing spicule-tissue preparations mounted with Gel/Mount (Biomeda, USA) and preparations of cleaned spicules produced according to standard methods (e.g.Schönberg 2000). By recognizing similarities with samples from Australia, the first author identified the samples as C. cf. orientalisThiele, 1900; but there were slight differences in spicule features. Therefore, the present identification will have to be confirmed with more detailed studies. Acropora palifera was identified by examining the skeleton (using Veron 2000 as reference).

Sponge and coral symbionts were classified by molecular analysis. Tissue slurries were incubated with 1% SDS for 1 h at 65 °C and then digested with proteinase K (Sigma) in a final concentration of 0.5 mg·ml−1 at 37 °C overnight. DNA was extracted by means of the phenol-chloroform method as described by Loh et al. (2001). The variable domains D1 and D2 of the Symbiodinium large subunit (LSU) nuclear ribosomal (nr) DNA were amplified using dinoflagellate specific primers designed by Zardoya et al. (1995). Original PCR products from seven host corals and three sponges (Table 1) were directly sequenced. Each sequence was determined from both ends of the rRNA gene using dye-terminators and a 373A DNA automated sequencer (Perkin-Elmer, CA). The sequences derived from this study were deposited in GenBank (see Table 1 for accession numbers). All LSU sequences were aligned with Symbiodinium clades A, B, C, D, E, F, G and H (Table 1) using the program CLUSTAL X (Thompson et al. 1997). Trees were rooted using Gymnodinium beii (AF060900), because previous phylogenetic analyses have shown that this genus typically forms a sister group relationship to Symbiodinium (Wilcox 1998). Phylogenetic analyses of aligned sequences were undertaken with Bayesian and maximum likelihood methods. For both analyses, the program MODELTEST version 3.7 (Posada & Crandall 1998) was used to find the most appropriate substitution model for our data. The model selected was TrN+I+G using the Bayesian Information Criterion. The Bayesian tree reconstruction was implemented in the ‘MrBayes’ program by (Huelsenbeck & Ronquist 2001). Starting from random trees, four Markov chains were run in parallel to sample trees using the Markov Chain Monte Carlo (MCMC) principle. After the burn-in phase (the first 13,500 generations were discarded), every 100th tree out of 106 was considered and results were compared among the four chains to confirm that stationality had been reached. The remaining trees were used to estimate the 50% majority consensus trees and the Bayesian Posterior Possibilities (BPP). Maximum likelihood analysis was performed using the computer program PAUP beta version 4.0b10 (Swofford 1998). The model selected was TrNef+G using the Hierarchical likelihood ratio tests. Maximum likelihood analyses were performed heuristically with 1000 random additions, TBR swapping and Multitrees option. Branch robustness was tested using 100 bootstrap replicates. All phylogenetic trees generated were visualized using TREEVIEW version 1.6.5 (Page 1996).

Table 1.   Sponge and coral Symbiodinium identities and GenBank accession numbers together with reference clades and mean Fv/Fm of experimental holobionts (H) and isolates (I) over time. Samples of three sponge symbionts and seven coral symbionts were successfully sequenced.
host organismsample no.Symbiodinium cladeuse in experimentmean Fv/Fm over timeGenBank accession no.
  1. Coral no. 1 showed two closely spaced bands in the gel and superimposed sequences that could not be analysed, but were probably a mix of C and D Symbiodinium. Samples with asterisks became near-aphotosynthetic.

experimental clades
 Cliona cf. orientalissponge1?heat treatmentH: 0.630
I: 0.517
 
sponge2?heat treatmentH: 0.637
I: 0.477
 
sponge4?controlH: 0.629
I: 0.497
 
sponge5?light treatmentH: 0.569
I: 0.376
 
sponge6?light treatmentH: 0.593
I: 0.291
 
sponge7Gheat treatmentH: 0.655
I: 0.465
EU159437
sponge9GcontrolH: 0.588
I: 0.489
EU159438
sponge10Glight treatmentH: 0.618
I: 0.280
EU159439
sponge11?controlH: 0.623
I: 0.410
 
 Acropora paliferacoral1C & D?light treatmentH: 0.695
I: 0.444
 
coral3CcontrolH: 0.670
I: 0.613
EU159440
coral4DcontrolH: 0.697
I: 0.451
EU159442
coral7Clight treatmentH*: 0.416
I*:0.053
EU159441
coral8DcontrolH: 0.637
I: 0.192
EU159443
coral9Dlight treatmentH: 0.649
I*: 0.094
EU159444
coral10Dheat treatmentH: 0.651
I*: 0.112
EU159445
coral12Dheat treatmentH: 0.649
I: 0.513
EU159446
coral13?heat treatmentH*: 0.442
I: 0.340
 
reference clades
 Aiptasia pallida A  AF427453
 Aiptasia pulchella B  AF427457
 Seriatopora hystrix C  AY684268
 Pocillopora elegans D  AF170149
 Acropora sp. D  AF396626
  (free-living Gymnodinium varians) E  AF060899
 Sorites sp. 1 F  AJ621144
 Marginopora sp. G  AJ291536
 Sorites sp. 2 H  AJ621129

Extraction of symbionts

Fragments of sponges and corals to be used in the experiment were rinsed in filtered seawater (0.45 μm) and subsampled for extraction and isolation of the symbionts, largely following the protocol of Bhagooli & Hidaka (2003). Coral tissue was removed from the underlying skeleton by using jets of filtered seawater (0.45 μm) administered with dental waterpiks (Johannes & Wiebe 1970). Sponge tissue was plucked out of chambers and macerated with a razor blade. Each tissue sample was filtered through a 350-μm mesh, broken up further in a potter homogenizer, resuspended and filtered through 180-μm nylon mesh. The filtrate was resuspended again and poured through a 40-μm filter. We found that shaking the plastic bags containing the coral tissue in seawater broke up the thick mucus clogging the filters (LaJeunesse 2002). Filtrates were centrifuged, resuspended to make up 5 ml and zooxanthella cells were counted on a haemocytometer slide to decide on the final concentration (i.e. on a suitable volume of the zooxanthella suspension to be pressed through a millipore membrane filter of 0.45 μm using a syringe with a respective tip attachment). The round filter papers now coated on one side with zooxanthellae were attached to glass microscope slides with rubber bands that held them on two sides. In each experimental situation, we thus had holobionts and matching isolated symbionts (Fig. 3). The extraction was conducted on the evening before the experiment, taking care that the slides with the zooxanthellae were resuspended in seawater as soon as possible.

Figure 3.

 Experimental design for three 200-l tanks, and conditions at measurement periods of the experiment (t1–t3). The temperature in the heat tank was controlled with aquarium heaters (thermometer symbol), the light level in the light tank by halide lamps (sun symbol). A cardboard shield prevented light stress in control and heat tanks (black line). Each tank contained holobionts (H) and isolated symbionts (I) of three sponges (S) and three corals (C), i.e. the entire experiment was based on nine sponges (S1H–S9H) and their respective isolates (S1I–S9I) and nine corals (C1H–C9H) and their respective isolates (C1I–C9I). Each small white box represents a single measurement of Fv/Fm at steady state, resulting in 108 measurements.

Experimental design

We aimed at creating heat and light stress for sponge and coral symbionts in hospite and in isolation, and to compare results with a control situation and over time. The effect of stress was estimated by measuring the symbionts’ photochemical efficiency Fv/Fm with a pulse amplitude modulated fluorometer (miniPAM; Walz GmbH, Effeltrich, Germany; active fibre diameter 5.5 mm). Photochemical efficiency is dependent on the number of active photosynthetic centres and inversely proportional to stress (e.g.Schreiber et al. 2003; Ralph et al. 2005; with variable fluorescence Fv = Fm − F0 and F0 = minimum fluorescence and Fm = maximum fluorescence). For the experiment itself, three closed indoor tanks were filled with natural seawater and kept in a darkened room. Bubble stones connected to air pumps aerated the tanks, while strong powerhead pumps circulated the water. Sponge and coral fragments and isolated zooxanthellae on filter paper were placed in these tanks on the evening before the treatment phase and left to adjust overnight. For each treatment and group, three samples were available, resulting in an overall sample size of 108 individual measurements (Fig. 3). Room lights were used in a 12:12 h light–dark cycle, starting at 7 AM in the morning. During a 3-h treatment phase, one tank was the control tank, one tank was heated with aquarium heaters (heating up took about 15–20 min and was part of the heat treatment period) and one tank was exposed to the light of two metal halide lamps (Eye Clean Ace; respective temperatures and light levels are given in Fig. 3). As the lamps generated much heat, the light treatment tank occasionally had to be cooled down with plastic bags containing ice. Temperatures were constantly checked and maintained. After the treatment phase, the metal halide lights were switched off and the tank water of all three tanks was readjusted to 25–26 °C. The samples were left in the treatment tanks 12 h overnight at about 25 °C and in darkness, which is here termed ‘recovery period’. Treatment light intensities were measured with a LI-250 light meter (LI-COR) after the experiment in the empty tanks.

Data acquisition with PAM and data analysis

Before each PAM measurement, all holobiont fragments and all filter papers with extracted zooxanthellae were dark-adapted for 30 min, either in styrofoam containers covered with aluminium or in a black box for the slides, custom-made to allow sliding open one lid at a time above a given slide, revealing a hole large enough to insert the optic fibre of the PAM for measurements (Bhagooli & Hidaka 2003). We kept the distance to the measured surfaces constant (10 mm).

Initial measurements of rapid light curves (RLC) were conducted in the early morning. Such measurements do not allow gradual adaptation to the rapidly increasing light levels as would be possible under natural conditions, and they were here performed to decide on the experimental light level in the light-stress tank. Because the sponge was assumed to be the hardier organism, the RLCs were obtained from five sponge holobionts, which proved to be well-adapted to a strong light environment (Fig. 4). From the mean of these five curves, light compensation and saturation were calculated using the hyperbolic tangent function (Platt & Jassby 1976; Chalker 1980) and light levels in the light-stress tank were chosen well above the obtained saturation level.

Figure 4.

 Rapid light curve for Cliona cf. orientalis from Sesoko Island, Okinawa sampled from <1 m water depth. Mean values from five specimens, bars are standard deviations.

Dark-adapted, pre-treatment photochemical efficiency (Fv/Fm) of all samples was estimated starting at late noon. Samples were then submitted to treatment conditions for 3 h, during which heat and light were administered and a control situation was maintained (see above and Fig. 3). Dark-adapted Fv/Fm was again measured directly after the treatment period in the late afternoon. We measured Fv/Fm again in the morning after the day of the treatments, i.e. after a 12-h period overnight in a darkened room.

Resulting Fv/Fm values were not completely randomized and, to some degree, inherently correlated to each other (Hurlbert 1984; Heffner et al. 1996), as isolated symbionts stemmed from holobionts used in the same experiment, the access to only three tanks prevented proper replication, and the same samples were measured over time. We, therefore, employed a repeated measures ANOVA for non-independent samples, which adjusted the error terms of the within effects: ‘symbiont state’, ‘organism’ and ‘phase of experiment’ (Table 2). This results in a considerably more conservative decision on whether to reject H0. Because there are also more subgroups considered than in a simpler ANOVA, the between effect ‘treatment’ lost test power as well (Table 2). Our software furthermore provided two adjustments for the error terms of the within effects, the Greenhouse–Geisser adjustment and the Huynh–Feldt adjustment. The former provides more conservative results, less likely to reject H0, and the latter is more liberal, possibly rejecting H0 when it is true, especially with small sample sizes (Abacus Concepts 1989). Unadjusted and adjusted probabilities were similar and led to the same decisions with regard to significances.

Table 2.   Statistical results for a repeated measure model for non-independent samples, testing differences in photochemical efficiency (Fv/Fm, dependent value), comparing the symbiotic sponge Cliona cf. orientalis and the coral Acropora palifera (organism), the reaction of symbionts in hospite and in isolation (symbiont state), before the stress treatment, directly after and after a period of recovery (phase) and in tanks subjected to control conditions, heat and light stress (treatment).
sourceeffectdfFP-valueG–GH–F
  1. df, degrees of freedom; F, test statistics; P, unadjusted probability; G–G, P with conservative Greenhouse–Geisser adjustment; H–F, P with liberal Huynh–Feldt adjustment.

  2. All independent factors have between, all non-independent factors within effects. The model adjusts for error terms of within effects, being more conservative to reject H0.

01 treatmentbetween 2 2.0460.2102  
02 subject (group)  6    
03 organismwithin 1 1.9240.21470.21470.2147
04 organism*treatmentwithin*between 2 0.6680.54730.54730.5473
05 organism*subject (group)  6    
06 symbiont statewithin 138.1800.00080.00080.0008
07 symbiont state*treatmentwithin*between 2 1.2090.36200.36200.3620
08 symbiont state*subject (group)  6    
09 phasewithin 2 7.2790.00850.03520.0194
10 phase*treatmentwithin*between 4 2.1120.14210.20130.1737
11 phase*subject (group) 12    
12 organism*symbiont statewithin*within 1 3.3510.11690.11690.1169
13 organism*symbiont state*treatmentwithin*within*between 2 0.0030.99720.99720.9972
14 organism*symbiont state*subject (group)  6    
15 organism*phasewithin*within 2 1.9370.18660.21090.1967
16 organism*phase*treatmentwithin*within*between 4 0.4840.74720.65520.7163
17 organism*phase*subject (group) 12    
18 symbiont state*phasewithin*within 2 3.3230.07110.10770.0788
19 symbiont state*phase*treatmentwithin*within*between 4 1.1900.36410.36770.3656
20 symbiont state*phase*subject (group) 12    
21 organism*symbiont state*phasewithin*within*within 2 2.5630.11830.15910.1398
22 organism*symbiont state*phase*treatmentwithin*within*within*between 4 0.2000.93380.82990.8903
23 organism*symbiont state*phase*subject (group) 12    

Results

If data could have been analysed with a simple ANOVA model, all stated null hypotheses would have been rejected. However, because of the conservative approach of data analysis and considerable variation in photochemical efficiency (Fv/Fm) of the coral samples, our two main null hypotheses could not be rejected as expected. Putative differences between the stress-reactions of sponges and corals and for different sources of stress were masked or non-existent, and significant results were restricted to Fv/Fm of holobionts compared to isolates and Fv/Fm changing over time (Table 2). Because of consistent trends in Fig. 5 our observations are described and interpreted regardless of largely insignificant results, keeping in mind that the findings are not demonstrated beyond doubt.

Figure 5.

 Photochemical efficiency (Fv/Fm) of zooxanthellae in hospite (upper row) and isolated on filter paper (lower row) from Cliona cf. orientalis (black) and Acropora palifera (white). Kind of treatment on the x-axis (control situation, heat treatment, light treatment). Mean values from three samples per bar, error bars are standard deviations. Please note that reduced Fv/Fm of coral holobionts after recovery are caused by only one individual per bar (with Fv/Fm of 0.01 in the heat treatment and 0.00 in the light treatment respectively). As treatment effects were statistically not significant we displayed the means over all treatment tanks in the corners of each box.

Fv/Fm was marginally lower in the sponge holobionts than in the coral holobionts and slightly higher in the isolated sponge symbionts than in the coral isolates (Fig. 5). Mean sponge and coral holobiont Fv/Fm over all treatment tanks were 0.60 and 0.65 before the experiment, 0.60 and 0.66 directly after administering the stress, but 0.65 and 0.53 after a period in restored control conditions, possibly implying some delayed photodamage in the coral samples (Fig. 5 shows this in more detail: as means per tank, treatment and organism group). Conditions were reversed for Fv/Fm in all isolated sponge and coral symbionts per tank with 0.48 and 0.40 before the experiment, 0.43 and 0.29 directly after administering the stress, and 0.36 and 0.25 after a period of recovery (Fig. 5). Whereas two coral holobionts (one per heat and light tanks, Table 1) and three sets of isolated coral symbionts became more or less aphotosynthetic (one in the heat tank, two in the light tank, Table 1), sponge holobionts and isolates never did (lowest Fv/Fm for sponge isolates: 0.21). Nevertheless, overall, sponge and coral symbiont photochemical efficiency did not differ enough to yield significant results regardless of whether they were in hospite or isolated from their hosts (Table 2, lines 3 and 12).

Heat and light treatments had no significant effect on the measured values of Fv/Fm either (Table 2, line 1). However, we still think that the sample organisms were stressed, but this was masked by the strong variation in the coral samples. After a lag period of 12 h and compared to the sponge samples and the controls, the coral holobionts displayed a reduced mean Fv/Fm in the heat- and light-treated individuals, implying the possible occurrence of delayed photodamage in A. palifera, but not in the sponge holobionts (Fig. 5). In the isolated symbionts, putative stress reactions seemed to be more immediate, and light appeared to have a stronger effect than heat, especially in the sponge isolates that did not suffer from heat stress (Fig. 5). Overall, sponges, corals and their isolated symbionts reacted roughly in the same manner to the experimental conditions (Table 2, lines 4 and 7).

Stress was more clearly implied by the appearance of the tanks and specimens after the recovery period. It appeared that some symbionts had been expelled into the tank water and that there were different reactions to the treatments: (i) the control tank had clean, clear water and the organisms looked healthy; (ii) the water in the heat treatment tank was slightly murky, there were brown patches underneath two coral holobionts and the sponge holobionts looked slightly yellow; and (iii) the water in the light treatment tank was slightly murky, two coral holobionts had white patches and the third looked dead, while the sponge holobionts looked healthy.

Regardless of the organism group, holobionts showed a significantly higher photochemical efficiency than their isolated symbionts (Tables 1 and 2, line 6 and Fig. 5), with mean Fv/Fm of 0.62 and 0.44 before the experiment, 0.63 and 0.36 directly after administering stress, and 0.59 and 0.31 after a period of recovery (means over all treatment tanks). The combination of the above factors did not yield a significant result (Table 2, line 12).

Mean Fv/Fm significantly declined over the duration of the experiment (Table 2, line 9), displaying the strongest photochemical efficiency before the experiment with 0.62 for holobionts and 0.44 for the isolates, still exhibiting comparatively stable values directly after the administration of stress, reaching 0.63 in the holobionts and 0.36 in the isolates, and with the most reduced values after the 12 h period of recovery, i.e. 0.59 and 0.31, respectively (means from sponge and coral samples over all treatment tanks, Fig. 5). How strong Fv/Fm decreased over time was statistically neither dependent on whether heat or light was administered, nor on host organism or the state of the symbionts (Table 2, lines 10, 15, 18). However, while especially the sponge holobionts experienced virtually no adverse consequences in the entire experiment, isolated coral symbionts may have been more strongly affected and displayed clearly more variability than the isolated sponge symbionts (Table 1, Fig. 5).

Any other combination of experimental factors did not yield any more significant results, thus not revealing any other interaction effects between them (Table 2, lines 13, 16, 19 and 21–22). This and the lack of significant differences between corals versus sponges and heat and light treatments versus the control situation may have been caused by conflicting trends in the different data sets and the large variability in data derived from coral samples. While sponges contained a single symbiont clade, G, A. palifera samples were dominated by either clade C or D or contained a mix of both (Table 1; see also tree submitted at TreeBASE 1998). No clear patterns were apparent to distinguish clade C and D in their behaviour with regard to stress (Table 1). In contrast, clade G Symbiodinium always retained higher Fv/Fm values in isolated symbionts than the coral clades, with sponge isolates reaching over 80%Fv/Fm of the holobionts and coral isolates only 61%. This pattern remained the same later in the experiment, but at lower values.

Discussion

The results of the present study were largely unsupported by statistics; however, we maintain the opinion that the sponge symbionts reacted differently to administered treatments than the coral symbionts, and that heat and light caused stress in comparison to control conditions. Observed trends were very stable over subsets of the study, implying differences that could not be confirmed (Fig. 5). Stress and possibly symbiont expulsion could indirectly be inferred by the murkiness of the water in the heat and light tanks after the experiment, the colour of the holobionts and the brown patches under the putatively most stressed individuals. Lack of statistical evidence was partly caused by the non-independence of samples that necessitated the use of a conservative, repeated measure model ANOVA; otherwise, the study results would have been significant. In addition, the large variability of the coral samples strongly masked putative patterns. As Acropora palifera harbours two horizontally transferred clades of Symbiodinium, C and D (Table 1; Kojis 1986; Omokawa 2006), photosynthetic behaviour can be more varied than in single-symbiont hosts. D-type symbionts are thought to be more heat-resistant than C-type symbionts (e.g.Baker et al. 2004; van Oppen et al. 2005). D-type symbionts of A. palifera may increase in density during the warm season (Chen et al. 2005), but some of the present samples were still dominated by the clade C (Table 1). We also had evidence that some of the samples may have contained a more even mix of both clades (coral1, Table 1), which makes variability of results even more likely. In this context, it needs to be mentioned that Symbiodinium identifications were conducted after the experiment. If the heat and light treatments caused a shift in the composition of Symbiodinium in the used specimens, the results refer to the situation after such a shift, and most of the successfully sequenced coral samples were of clade D (Table 1).

We only found clade G Symbiodinium in Cliona cf. orientalis from Okinawa, the symbionts being very closely related to those found in Australian C. orientalis (Schönberg & Loh 2005; see also tree submitted to TreeBASE). The G-type sponge symbionts were more stable in their photosynthetic behaviour than those of the corals, with much less variation than in the coral samples, no discernible change in the holobionts and no heat effect in the isolates (Fig. 5). Considering the significantly stronger photosynthetic efficiency of holobionts compared to isolated symbionts and with respect to changing environmental factors, sponge symbionts may have a few advantages over coral symbionts. In contrast to corals, sponges are 3-dimensional, providing some auto-shading in deeper tissue layers (Schönberg et al. 2005). Moreover, bioeroding sponges are endolithic and further sheltered and shaded by the substrate. With the additional ability to shift the symbionts between different layers of tissue depending on the ambient light (Schönberg & Suwa 2007), the sponges are extremely well forearmed against light stress. Therefore, if their symbionts are as susceptible to light stress when isolated, as suggested by the present data (Fig. 5), the host can strongly reduce the effect. If they are inherently more heat-tolerant than coral zooxanthellae (Fig. 5), they may have an important advantage over many coral symbionts in a climate change scenario. This supports earlier thoughts on the hardiness of G-type symbionts (Schönberg & Loh 2005) and may in part explain why bioeroding sponges can survive bleaching events that are detrimental to other reef organisms (Vicente 1990). The present data represent the first experimental approach to study stress tolerance in G-type Symbiodinium, but only generated indirect evidence that needs to be confirmed by future studies.

Keeping in mind that the study findings are not well supported and only refer to one species of sponge and one species of coral, they still need to be discussed in the context of possible future trends. Heat-susceptible C-type symbionts are common in corals throughout the Indo-Pacific (e.g.Baker et al. 2004), i.e. such coral holobionts may fall victim to future bleaching events and succumb to mass mortality or at least to sublethal damage, which may prevent sexual reproduction (e.g.Omori et al. 2001). In contrast, symbiotic bioeroding sponges appear to be coping very well by having comparatively more stress-tolerant symbionts (Table 1, Fig. 5) and by being able to shift symbionts away from the source of stress, if needed (Schönberg & Suwa 2007). In case of environmental stress, the sponges would thus benefit in the long run by colonizing clean substrate available after coral death and by experiencing less space competition from weakened neighbouring organisms (but see Marquez et al. 2006). In the Caribbean and on the central Great Barrier Reef, bioeroding sponges have become more abundant, which appears to be related to disturbance events (Rützler 2002; C.H.L. Schönberg, unpublished observations). In these cases, the symbiotic sponges in particular have become more prevalent. On strongly damaged reefs of Okinawa, a field survey showed that within the group of bioeroding sponges, those with dinoflagellate symbionts dominated (C.H.L. Schönberg, unpublished observations). Not only sponge abundances are likely to increase with changed climate conditions, but bioerosion rates may also be affected. Sponges erode chemo-mechanically by etching cup-shaped grooves into the substrate and removing lentil-shaped, silt-sized fragments (Pomponi 1980; Schönberg in press). Chemical reactions are enhanced with rising temperature. Symbiotic sponges also display higher erosion rates in light compared to shade (Hill 1996; Schönberg 2006; but see Zundelevich et al. 2007), an effect that may now be intensified with decreasing global dimming after more stringent clean-air regulations (Wild et al. 2005). Moreover, sponges are filter feeders and coastal areas are increasingly eutrophic. A large number of studies found bioeroding sponge occurrences – and often of aggressive species – to be proportional to marine anthropogenic eutrophication (e.g.Rose & Risk 1985; Holmes 1997, 2000; Holmes et al. 2000; Ward-Paige et al. 2005). It would seem reasonable to predict that bioeroding sponge abundances and sponge erosion will increase over the next decade.

Acknowledgements

We acknowledge friendly assistance from staff of the Sesoko Station, Tropical Biosphere Research Center and colleagues at the University of the Ryukyus. The first author is grateful for receiving a scholarship from the Japanese Society for the Promotion of Science (JSPS) to visit Okinawa and to conduct this research.

Ancillary