The microbiome of the octocoral Lobophytum pauciflorum: minor differences between sexes and resilience to short-term stress

Abstract

Bacteria associated with marine invertebrates are thought to have a range of important roles that benefit the host including production of compounds that may exclude pathogenic microorganisms and recycling of essential nutrients. This study characterised the microbiome of a gonochoric octocoral, Lobophytum pauciflorum, and investigated whether either sex or environmental stresses influenced the diversity of the associated microbiome through amplicon profiling of the bacterial 16S rRNA gene. Sequences affiliated to Spirochaetaceae and Endozoicimonaceae dominated the microbiome of L. pauciflorum, representing 43% and 21% of the community, respectively. Among the dominant class affiliations, no sex-specific differences were detected, though unassigned sequences were at a 2-fold higher relative abundance in samples from female individuals than from males. These potentially novel sequences contributed to observed differences between sexes as detected by a multivariate analysis at the OTU level. Exposing L. pauciflorum fragments to increased temperature (31°C), decreased pH (7.9) or both stressors simultaneously for 12 days did not significantly alter the microbial community, indicating that the soft coral microbiome is relatively resilient to short-term environmental stress.

soft coral microbiome, Lobophytum pauciflorum, Spirochaetes dominated, sex differentiation, ocean acidification

INTRODUCTION

Tropical coral reefs are among the most diverse ecosystems on earth, hard corals (order Scleractinia) and soft corals (subclass Octocorallia) representing two of the major benthic communities constituting these ecosystems (e.g. Done 1982). The bulk of the reef structure is due to hard corals (Veron 1986), although some soft coral species contribute to its consolidation via production of sclerites (Schuhmacher 1997; Jeng et al. 2011). Despite lacking an extensive exoskeleton, soft corals provide additional structural complexity to reefs, increasing habitat diversity available to fishes and other marine invertebrates (Crossland 1938; Dinesen 1983). Although other aspects of their biology are often overlooked, soft corals have well-developed chemical defence arsenals, and these have been the primary focus of octocoral research (Tursch et al.1974; Coll et al.1982, 1986, 1989, 1995).

The soft coral Lobophytum pauciflorum (Ehrenberg 1834) is a common species in the shallow waters of the Indo-West Pacific (Tursch and Tursch 1982; Verseveldt 1983; Benayahu 2002). Like most soft corals (Kahng, Benayahu and Lasker 2011), L. pauciflorum is gonochoric, with dimorphic polyps. Males and females have been shown to differ in levels of the UV protective mycosporine-like amino acids (Michalek-Wagner 2001) and some secondary metabolites (Fleury, Coll and Sammarco 2006), implying that there are physiological differences between the sexes. The physiological significance of the differences in secondary metabolite levels observed between the sexes is unknown but, as these compounds have cytotoxic and antimicrobial properties (Zhao et al.2013), the concentration differences might be consequences of a differentially regulated immune system. This phenomenon has been observed in Bilateria, where sex-specific differences in immune system expression have been identified as one contributor to physiological differences (Klein 2000; Peng, Zipperlen and Kubli 2005; Love et al.2008; Nunn et al.2009). The immune system may also influence the composition of the microbial communities associated with the soft coral, with an increasing number of studies recognising the importance of microbial constituents within the coral organism, the sum of which is termed the holobiont (Margulis and Fester 1991). Microbial communities associated with hard corals have been the subject of many recent studies; however, despite the abundance of L. pauciflorum, this species and soft corals in general have received little attention at the organismal, molecular or microbial levels (Fan, Chou and Dai 2005; Yan et al.2010a,b, 2011).

The cnidarian innate immune repertoire is complex and vertebrate-like, and includes homologues of many pattern recognition molecules and components of signal transduction pathways whose functions in vertebrates have been extensively studied (Miller et al.2007; Augustin, Fraune and Bosch 2010; Shinzato et al.2011; Palmer, Bythell and Willis 2012). Upon exposure of the Caribbean sea-fan Gorgonia ventalina to a common fungal parasite, a number of likely pattern recognition molecules and candidate antimicrobial peptides were differentially expressed (Burge et al.2013). In marine environments, induction of apoptosis through upregulation of caspase activity is a key indicator of stress (reviewed in Lesser 2006, 2012), and several recent papers focussing on the hard coral Acropora millepora (Sakamaki et al.2014; Sakamaki et al.2015; Moya et al.2016) highlight the presence of mammalian-like caspase repertoires in the common ancestor of cnidarians and Bilateria. Under elevated pCO₂, several caspases were upregulated in Acropora (Moya et al.2012) and increased caspase activity has frequently been documented in corals under thermal stress (Kvitt et al.2011; Pernice et al.2011; Tchernov et al.2011).

Scleractinian corals may benefit from their microbial associates in terms of nitrogen fixation (Rohwer et al.2002; Lesser et al.2004), antibiotic production (Ritchie 2006) and mucus recycling and food supply (Wild et al.2004). Highly diverse microbial populations are associated with coral species, although some conserved taxa have been identified (Rohwer and Kelley 2004; Bourne et al.2008). During periods of environmental stress, the microbiome of hard corals has been shown to shift, driven either by a host response and/or changes in the microbiome itself (Bourne et al.2008). For example, during bleaching events the microbiomes associated with corals become more diverse (Bourne et al.2008), and these changes are typically accompanied by shifts from autotrophic to heterotrophic communities (Littman, Willis and Bourne 2011). When Porites compressa was placed under experimental stress, Vega Thurber et al.(2009) observed shifts in the microbial community from being dominated by Cyanobacteria and Proteobacteria towards communities dominated by Bacteriodetes and Fusobacteria, which are typically associated with diseased corals. Shifts in microbial communities away from ‘normality’ might also trigger innate immune responses to fend off potential infections. Recently, studies have also begun to explore the significance of a ‘core microbiome’, a subset of microorganisms that is shared by most individuals of species but is not necessarily composed of the most abundant microorganisms (Ainsworth et al.2015). The stability and ubiquity of these associations suggests that they may be important, and these communities may be particularly significant during times of stress (Shade and Handelsman 2012).

Relatively little is known about the bacterial communities associated with octocorals. Members of the Gammaproteobacteria have been identified as dominant components of the gorgonian microbiome, the majority of 16S rRNA sequences retrieved in gene profiling studies being affiliated with the genus Endozoicomonas (Bayer et al.2013a; La Rivière et al. 2013; Ransome et al.2014; La Rivière, Garrabou and Bally 2015; Kellogg, Ross and Brooke 2016). It has been suggested that this host–microbe association is symbiotic, based on evidence pointing to a long-term partnership with little geographic or seasonal influence (Bayer et al.2013a, La Rivière, Garrabou and Bally 2015). Furthermore, this association seems to be highly specific, each gorgonian species harbouring a specific Endozoicomonas ribotype even on a regional scale (La Rivière, Garrabou and Bally 2015) with the microbiome in related gorgonian species being more similar than those in taxonomically distinct gorgonians (van de Water et al.2016b). In contrast to the dominance of Gammaproteobacteria in some gorgonian microbiomes, Spirochaetes have been found to be dominant in the microbial community associated with the red coral Corallium rubrum in the Mediterranean and co-dominant in deep sea octocorals (Holm and Heidelberg 2016; Lawler et al.2016), and it has been suggested that these organisms may bring about nitrogen fixation (van de Water et al.2016a).

In this study, the microbiome of the soft coral L. pauciflorum was investigated, and the possibility of sex-specific microbiomes was explored. The effects on the microbiome of acute (1 day) and more prolonged (12 days) exposure to elevated temperature, increased pCO₂ and the two stressors simultaneously were also investigated.

MATERIALS AND METHODS

Animal collection and maintenance

Two separate collections of Lobophytum pauciflorum colonies were undertaken, one for each of the experiments described below, in August 2013 and October 2014 from reefs around Orpheus and Pelorus Island (18.57 °S, 146.48 °E; 18.54 °S, 146.49 °E according to the GBRMPA permit G12/35295.1). To ensure that the same species was collected each time, the octocoral-specific msh1 gene was sequenced after PCR amplification of genomic DNA (see extraction method below) (forward primer: 5΄-TTCAACTTAGCAGAGGAAAA-3΄, reverse primer: 5΄-ACATGGCAAATTGGTTAGTG-3΄, initial denaturation at 95°C for 5 min, followed by 35 cycles consisting of denaturation (95°C for 30 s), annealing (50°C30 s) and extension (72°C for 60 s) and a final extension of 5 min at 72°C). The generated sequences were blasted (BLASTn) (Altschul et al.1997) against the NCBI database, and all sequences were assigned to L. pauciflorum.

For the sex-specific microbiome analyses, 17 L. pauciflorum colonies were sampled from two locations around Orpheus Island (Cattle Bay, Little Pioneer Bay) and one location around Pelorus Island and these sampling locations were one factor for the multivariate analysis. The collected colonies were brought to Orpheus Island Research Station (OIRS) and held in a 1000 L flow through tank and acclimated for 1 week prior to sampling of two replicate lobes from each colony. One lobe was immediately snap frozen in liquid nitrogen for total DNA extraction and the second lobe transferred into HEPES buffered seawater with 4% formaldehyde for fixation. After taking tissue samples, the colonies were returned to the ocean to a marked location, before bringing them back to the station one month later to spawn during the annual coral spawning event. This allowed identification of the sex of the individual colonies, where histological sections were not conclusive.

For the temperature and CO₂ stress experiments, fragments of seven L. pauciflorum colonies were collected on a separate occasion from Cattle Bay around Orpheus Island (three colonies) and Pelorus Island (four colonies), transported and cultured in a 1000 L tank with flow through at the Marine Aquaculture Research Facility Unit at James Cook University, Townsville for further propagation. The coral fragments were cut twice so that four small fragments of about 7–10 cm in diameter in total of each colony were obtained: the first time after 1 week of acclimation and the second time after a recovery period of 3 days to allow tissue healing and minimise lesions that could lead to infections. The coral fragments were transported back to OIRS and left to acclimate for 2 weeks prior to experimental manipulation.

Histological preparation and sex identification

Following fixation for 24 h, tissue samples were washed twice with tap water and decalcified overnight in 10% formic acid, rinsed with water to remove the formic acid and cut for transactional and longitudinal analyses prior to storage in 70% ethanol. Tissue samples were embedded in paraffin and three 5 μm thick sections obtained every 50 μm of tissue sample. Sections were stained with haematoxylin and eosin to facilitate gonad recognition. Gonads were identified using a ×10 magnification of a compound stereomicroscope (Olympus, Notting Hill, VIC, Australia). Presence of oocytes and spermatogonia indicated the sex of the colony as female or male, respectively. If gonads were not clearly visible in tissue sections, the sex of the colony was identified during the annual coral spawning event. Eight colonies were identified as females and nine as males.

Experimental design for temperature and CO₂ stress at OIRS

The temperature and CO₂ conditions were chosen for the manipulative experimental stress based on previous studies with the hard corals Acropora aspera and A. millepora and which resulted in apoptosis (Moya et al.2015) and suppressed physiological performance (Ogawa et al.2013). For the CO₂ treatment, natural seawater in a 500 L sump tank was acidified with 100% CO₂ using a pH computer (Aquamedic AT-Control, Bissendorf, Germany) as described in Watson et al.(2012). For the control treatment, a non-acidified 500 L sump tank was used. Following acidification, the seawater was heated with in-line pond titanium heaters (Weinu, Taiwan) to 32°C in two 60 L header tanks that were supplied with seawater from the sump tanks. Each of the 60 L header tanks fed seven 3 L aquaria in which the coral fragments were placed randomly one coral fragment per tank (Fig. S1, Supporting Information), without reference to the colony of origin or treatment (control or temperature, CO₂ or combine temperature and CO₂). The flowthrough into each experimental aquarium was adjusted to 250 mL min⁻¹ to ensure maintenance of experimental conditions. Corals were kept at a 12/12 h day/night cycle with Sylvania Oracle lights (Sydney, Australia) connected to a timer clock.

After transferring the coral fragments into the experimental aquaria, temperature was increased in the respective 60 L header tank from 27°C to 32°C over a 48 h period and pH was decreased over 6 h after the temperature had reached 32°C. Water temperature and pH_NBS (model MW102, Milwaukee, USA) were checked and re-adjusted (at least three times a day) and corals were monitored for visual signs of bleaching or visual signs of disease. The experiment time started when stressors had reached their experimental level of a pH of 7.9 and temperature 32°C, respectively. Samples for total DNA extraction and caspase activity assays were taken 1 and 12 days after the start, snap frozen in liquid nitrogen and stored at –80°C until processing.

Total alkalinity and CO₂ partial pressure

Seawater chemistry analyses were performed using water samples taken daily from the four 60 L header tanks and preserved with mercuric chloride, 0.04% final concentration, 100 mL seawater with 40 μL saturated MgCl₂. Seawater chemistry was analysed following Miller et al.(2012). Briefly, total alkalinity was determined by Gran titration with certified reference material (Batch 136, Dr A. G. Dickson, Scripps Institution of Oceanography). Total alkalinity of header tanks was used to calculate the pCO₂ in experimental tanks based on the pH and temperature measurements taken for individual tanks at time of sampling in CO2SYS (Pierrot, Lewis and Wallace 2006) using the NBS scale as pH scale and all other settings as standard (Table 1). Daily average values for salinity were downloaded from the data centre of the Australian Institute for Marine Science (AIMS) for relay pole 2 due to its close proximity to the seawater intake (http://data.aims.gov.au/aimsrtds/datatool.xhtml?site=9).

Table 1.

Summary of seawater chemistry including mean values (± SD) for temperature, pH, total alkalinity and pCO2 for each experimental treatment over the course of the experiment.

Treatment	Temperature (°C)	pHNBS	Total alkalinity (μmol kgSW−1)	pCO2 (μatm)
Control	27.5 ± 0.24	8.11 ± 0.02	2319 ± 10.2	502.8 ± 41.41
Temperature	30.9 ± 0.15	8.08 ± 0.02	2323 ± 4.5	542.3 ± 32.45
CO2	27.4 ± 0.28	7.94 ± 0.01	2323 ± 13.8	779.7 ± 24.70
Temperature + CO2	31.1 ± 0.17	7.91 ± 0.02	2325 ± 7.3	869.9 ± 38.34

Average salinity was 35.51 ± 0.40 psu. Total alkalinity was measured in the four header tanks for each treatment condition. Temperature and pH were measured daily in experimental tanks of each experimental treatment. These daily measured parameters were used to estimate pCO₂ in CO2SYS.

Total DNA extraction and microbiome sequencing

Frozen tissues samples (∼25 mg) were ground to powder in liquid nitrogen and total DNA extracted with the DNase Blood and Tissue kit (Qiagen, Chadstone Centre, VIC, Australia) following the supplier's instruction. To avoid RNA contamination, an additional RNA digest was performed according to the manual's suggestions. DNA concentration was measured using the Qubit^® HS DNA assay (Invitrogen/Life Technologies, Scoresby, VIC, Australia) and integrity verified on an agarose gel. For the sex experiment, all 17 samples were used. For the stress experiment, only samples from six colonies were used, as one fragment of the seventh colony suffered mortality prior to the start of the experiment and therefore fragments from this colony could only be subjected to three treatments; i.e., the seventh colony was not considered for microbial profiling.

To identify the microbial diversity, the V3 and V4 regions of the Prokaryotic 16S ribosomal RNA gene were amplified following the 16S library preparation guide for the Illumina MiSeq system. Briefly, the V3 and V4 regions were amplified with PCR primers (S-D-Bact-0341-b-S-17/S-D-Bact-0785-a-A-21) (Klindworth et al.2013) including the Illumina overhang adapter sequences (forward overhang: 5΄-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-3΄, reverse overhang: 5΄-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-3΄), which resulted in a PCR product of about 500 bp. DNA samples were diluted to 5 ng μL⁻¹ prior to PCR amplification. For each sample, the PCR reaction was performed in triplicate using the following conditions: initial denaturation at 95°C for 3 min, followed by 25 cycles consisting of denaturation (95°C for 30 s), annealing (55°C for 30 s) and extension (72°C for 30 s) and a final extension of 5 min at 72°C After purification, the concentration of each PCR product was determined (Qubit^® HS DNA assay, Invitrogen/Life Technologies, Australia) and triplicates for each sample were combined at an equimolar concentration of 2 nM. Illumina adapter sequences and dual-index barcodes were added to each PCR mix with the Nextera XT Index kit (Illumina, San Diego, CA, USA) in a second PCR reaction with the same conditions, though only eight cycles were used. The prepared libraries were pooled at an equimolar ratio of 2 nM after purification. Libraries were sequences on MiSeq at James Cook University using paired 300 bp reads and MiSeq v3 reagents (Illumina, San Diego, CA, USA). As an internal sequencing control, the library pool was spiked with 10% PhiX control (Illumina, San Diego, CA, USA).

The sequencing run was monitored over BaseSpace, and data were directly uploaded from the MiSeq machine. Sequencing resulted in ∼1.86 million reads passing the filter with on average 38 000 reads per sample, ranging from 16 000 to 75 000 reads.

Microbiome analysis

Sequencing runs of both experiments were performed separately; therefore, the microbial community was also analysed individually for each experiment to avoid biasing the analysis between the two experiments by overrepresenting OTUs in one versus the other. Sequence data were available as demultiplexed reads on BaseSpace. The first 16 bases were trimmed of the forward reads and the first 18 of the reverse reads to remove the primer sequences. Further downstream analysis was conducted in QIIME (Version 1.9) (Caporaso et al.2010) and built-in functions using the default settings in QIIME. Prior to de novo OTU picking using mothur with 97% sequence similarity (Schloss et al.2009; Schloss, Gevers and Westcott 2011), chimeric sequences were removed using USEARCH61 (Edgar 2010). Cleaned sequences were aligned against the SILVA database (SILVA SSU Release 119, July 2014) (Quast et al.2013). As L. pauciflorum is a hermatypic coral containing Symbiodinium, sequences aligned to mitochondria or chloroplasts were removed from the dataset to avoid misinterpretation. Alpha diversity metrics (observed OTUs, predicted species (chao1), Shannon-Wiener diversity index and Simpson's evenness) were generated from OTU tables prior to normalisation using the QIIME pipeline.

Sequence data from sex-specific and stress experiments were analysed separately. The sex-specific microbiome sequencing resulted in 121 369 OTUs with 2 263 516 counts. For the stress experiment, a total of 34 947 OTUs were observed with 1988 523 counts. However, the counts varied from 81 204 to 218 374 counts per sample in the sex experiment and from 3248 to 119 032 in the stress experiment. For statistical analysis, only OTUs with more than 15 counts in total were considered, which reduced the number observed OTUs to 2578 and 3509, respectively. Alpha diversity metrics for the whole microbial community (Table S2, Supporting Information) identified comparable diversity and evenness between the two data sets.

To gain an insight into the most abundant ribotyptes associated with L. pauciflorum, sequences of the 20 most abundant OTUs were extracted from each data set and blasted (BLASTN) (Altschul et al.1997) against the NCBI database. A more restricted subset of the microbiome was analysed using the compute_core_microbiome function in QIIME (Caporaso et al.2010). Due to varying definitions of the core microbiome and based on previous studies (Philippot et al.2013, reviewed in Hernandez-Agreda, Gates and Ainsworth 2016), a liberal representation of 51% was chosen for the restricted microbiome analyses of the sex and the stress experiment. This implies if an OTU was not present in at least 51% of the samples than it was considered as individual variability of colonies. The 16S rRNA sequences generated in this study were stored in NCBI Sequence Read Archive with the accession number SRP092008.

Caspase activity measurements

To investigate the host response to increased temperature, low pH and the combination of increased temperature and low pH, caspase activity was studied. Frozen tissue samples (∼15 mg) were homogenised in 1.5 mL (1:100 (w v⁻¹)) aliquots of hypotonic extraction buffer (25 mM HEPES, 5 mM MgCl₂, 1 mM EGTA and protease inhibitors (complete protease inhibitor cocktail, Roche, Australia; 1 pellet in 10 mL buffer)) using a FastPrep^®-24 (MP Biomedicals, Seven Hills, NSW, Australia) at program 3, twice with a 5 min interval on ice until tissue was completely homogenised. Subsequently, samples were centrifuged for 15 min and 13 000 g at 4°C. Caspase enzyme activity was measured using a Caspase-Glo 3/7 assay kit (Promega, Alexandria, NSW, Australia). Briefly, 50 μL of supernatant and 50 μL reagent were mixed in a white-walled 96-microwell plate (NUNC F96) and incubated at 25°C for 1 h. Light emission was measured in a Tecan GENios Pro Microplate Reader (Männerdorf, Switzerland) in the luminescence mode. To control for reagent quality on subsequent days and comparability between different kits, one reference sample was prepared as 0.5 mL aliquots and deep frozen at –80°C. One of these aliquots was measured once a day. Caspase activity was measured in relative light units (RLU) and normalised to protein concentration. Caspase activity was calculated as RLU mg protein⁻¹. Protein concentration was determined using the Qubit^® Protein assay (Invitrogen/Life Technologies, Scoresby, VIC, Australia).

Statistical analysis

Sequencing read counts were normalised using the relative log expression method proposed by Anders and Huber (2010) in edgeR (Robinson, McCarthy and Smyth 2010). Following normalisation, counts were standardised by species and site and a multivariate correspondence analysis was performed with vegan (Oksanen et al.2013). A multivariate correspondence analysis was chosen instead of a principle component analysis as the data did not meet the requirements for neither normality nor homogeneity of variance. The 20 most abundant OTUs were extracted from normalised read counts and used to generate a heat map with gplots (Warnes et al.2009) and RColorBrewer (Neuwirth 2011). Box and CA plots were produced using ggplot2 (Wickham 2009). Relative abundance of different taxa was calculated for each sample before calculating the average for each sex or sampling time point. Differences in relative abundance between sexes and sampling time points were tested using a two-sided Student's t-test for taxa with normal distribution or Mann-Whitney U test for all others in R.

The relationship between sampling time and caspase activity and colony of origin and caspase activity was assessed by trying to fit a linear mixed effect model and linear models using generalised least squares with the nlme package (Pinheiro et al.2014) in R.

RESULTS

The microbial community associated with Lobophytum pauciflorum

A total of 2578 OTUs were identified in the sex experiment in association with female and male L. pauciflorum, derived from 2161 800 sequencing reads with stringent quality checking and considering only OTUs represented by 15 or more reads. The microbiome of L. pauciflorum shared by at least 51% of samples was composed of 832 OTUs and represented 93% of the whole microbial community. Despite this relatively small number of OTUs, the bacterial diversity patterns of the restricted microbiome were similar to those based on all of the data; therefore, futher analysis and interpretation was based only on the former. The microbiome was dominated by the class Spirochaetes and the family Spirochaetaceae, which represented on average 43% of the retrieved sequences across all samples (Table 3). The second most abundant taxonomic class was the Gammaproteobacteria, which constituted ∼27% of the restricted microbiome. Within the Gammaproteobacteria, 85% (23% of all sequences of the restricted microbiome) were related to the family Endozoicimonaceae. Alphaproteobacteria-related sequences accounted for only 7% of recovered sequences from at least 51% of the samples, though these fell within a diverse array of different family groups, Rickettsiaceae (representing 4% of assigned OTUs shared by 51% of the samples) being the most abundant. Despite stringent sequence quality checking and comparisons to the SILVA microbial database, a high percentage of retrieved sequences (∼ 22%) could not be assigned. However, when extracted from the dataset and blasted against the NCBI database, these sequences were assigned to the kingdom Bacteria.

Direct comparisons between male and female colonies demonstrated similar alpha diversity metrics and comparable microbial profiles at the class taxonomic affiliation level (Tables 2 and 3). Although some differences were detected at class and family level (Table 3), Spirochaetes- and Rhodobacteraceae-related sequences were more abundant in samples from males than in female corals (1.4 and 4 times, respectively), these patterns were not statistically significant. Furthermore, the female microbiome was characterised by a 2-fold higher relative abundance of unassigned and Rickettsiaceae-related sequences (Table 3). A multivariate correspondence analysis of the microbiome structure shared by 51% of the samples, with constraints for both sex and sampled colony, explained 35.12% of the variation. The three sampling locations (Cattle Bay, Little Pioneer Bay and Pelorus Island) significantly (P < 0.05) separated samples along the first and second constraint correspondence axis, and explained 16.67% and 9.24% of the variation, respectively, of the association between samples and microbial community (Fig. 1). The male and female microbial communities were significantly different along the second constraint correspondance axis, P < 0.05 (Fig. 1). A similar clustering pattern but with the axis explaining less of the variation (8.51% and 7.92%, respsectively) was detected when the whole microbial community associated with male and female L. pauciflorum was analysed (Fig. S3, Supporting Information).

Figure 1.

Correspondence analysis of OTUs comprising the microbiome shared by 51% of the samples of the two sexes. Percentages in the axis titles represent the percentage described by the axis. Squares represent colonies from Cattle Bay, triangles colonies from Little Pioneer Bay and diamonds colonies from Pelorus. n = 832 OTUs, n = 8 for female samples, n = 9 for male samples.

Table 2.

Overview of alpha diversity metrics (average ± s.d.) of the restricted microbial community for both experiments.

		Observed OTUs	Unique	Alpha	Chao 1	Shannon-Wiener	Simpson
			OTUs	diversity	estimate	Index	Evenness
Sex experiment	Overall	575.29 ± 73.66	832	79.10 ± 10.65	679.64 ± 85.17	3.50 ± 0.40	0.75 ± 0.08
	Female	567.63 ± 48.55	66	79.00 ± 6.30	683.95 ± 60.00	3.56 ± 0.33	0.77 ± 0.05
	Male	582.11 ± 89.74	257	79.19 ± 13.38	675.81 ± 102.33	3.45 ± 0.44	0.73 ± 0.09
Stress experiment	Overall	161.83 ± 50.84	249	24.05 ± 8.19	173.99 ± 49.09	3.92 ± 1.51	0.79 ± 0.19
	1 day	185.79 ± 36.30a	105	27.04 ± 6.17a	197.80 ± 34.37a	4.17 ± 1.43	0.82 ± 0.14
	12 days	136.83 ± 51.79	27	20.93 ± 8.85	149.15 ± 49.83	3.66 ± 1.54	0.76 ± 0.22
	Control	182.17 ± 42.62	11	27.59 ± 7.35	192.52 ± 37.50	4.39 ± 1.49	0.83 ± 0.15
	Temperature	154.58 ± 57.21	5	23.03 ± 9.20	166.40 ± 57.10	4.12 ± 1.41	0.83 ± 0.20
	CO2	153.92 ± 43.07	4	22.51 ± 6.94	167.83 ± 43.50	3.54 ± 1.32	0.77 ± 0.14
	Temperature + CO2	156.18 ± 53.47	5	22.97 ± 8.04	168.79 ± 51.33	3.59 ± 1.65	0.74 ± 0.24

a

Highlights significantly different alpha diversity metrics between period of exposure to experimental stress factors 1 day and 12 days, P < 0.05 to p < 0.05. n = 8 female and 9 male soft coral colonies. n = 24 fragments for 1 day exposure and 23 fragments for 12 days exposure. n = 12 for control, temperature and CO₂ treatment and 11 for temperature + CO₂

Table 3.

Composition of the restricted L. pauciflorum microbiome in the two independent experiments described.

		Sex experiment			Stress experiment
		Relative %			Relative %
		sequence			sequence
Class	Family	abundance	Female	Male	abundance	1 day	12 days
Bacteroidia	Marinilabiaceae	0	0	0	2.35 ± 1.10	0.28 ± 0.15a	4.51 ± 2.17
Flavobacteria	Flavobacteriaceae	0.01 ± 0.00	0.01 ± 0.00	0.01 ± 0.01	4.92 ± 1.07	7.33 ± 1.78a	2.41 ± 0.90
Alphaproteobacteria	Rhodobacteraceae	1.13 ± 0.41	0.45 ± 0.18	1.72 ± 0.71	9.96 ± 1.62	10.26 ± 2.11	9.66 ± 2.54
	Rickettsiaceae	4.19 ± 2.04	5.86 ± 3.35	2.71 ± 2.52	0	0	0
Gammaproteobacteria	Alteromonadaceae	0.14 ± 0.03	0.17 ± 0.05	0.11 ± 0.02	5.54 ± 1.02	7.20 ± 1.78	3.80 ± 0.85
	Endozoicimonaceae	22.70 ± 6.70	23.13 ± 10.01	22.31 ± 9.57	4.15 ± 1.43	1.97 ± 0.75	6.42 ± 2.76
	Enterobacteriaceae	0.02 ± 0.01	0.03 ± 0.02	0.01 ± 0.01	4.53 ± 1.71	5.82 ± 3.10	3.19 ± 1.40
	Francisellaceae	1.50 ± 0.55	1.57 ± 0.68	1.43 ± 0.88	0.24 ± 0.10	0.23 ± 0.16	0.25 ± 0.11
	Pseudoalteromonadaceae	0.68 ± 0.37	0.18 ± 0.10	1.08 ± 0.69	0	0	0
	Vibrionaceae	0.44 ± 0.20	0.25 ± 0.12	0.62 ± 0.37	1.89 ± 0.54	0.80 ± 0.36a	3.03 ± 1.01
Spirochaetes	Spirochaetaceae	42.51 ± 6.00	35.05 ± 8.81	49.15 ± 8.00	34.04 ± 5.22	31.83 ± 6.79	36.34 ± 8.10
Others		4.49 ± 1.00	3.94 ± 0.92	4.99 ± 1.74	17.23 ± 1.90	16.83 ± 2.46	17.66 ± 2.96
Unassigned		22.21 ± 4.20	29.34 ± 7.13	15.85 ± 4.05	15.15 ± 2.90	17.45 ± 4.62	12.75 ± 3.49

Mean relative abundance ± standard error of the OTUs contributing to the bacterial community structure shared by 51% of the samples at family level and distribution of relative abundances (A) between female and male colonies and (B) between sampling time points, 1 day and 12 days. All treatments were averaged for the overall relative abundance and per sampling time points. ^a highlights significantly different relative abundances between period of exposure to experimental stress factors 1 day and 12 days, P < 0.05. n = 832 OTUs for sex experiment, n = 249 OTUs for stress experiment, n = 8 female and 9 male soft coral colonies. n = 24 fragments for 1 day exposure and 23 fragments for 12 days exposure.

Based on normalised counts, the 20 most abundant ribotypes made up ∼83% of the microbiome shared by at least 51% of the samples (77% of the overall microbial community) (Fig. 2A). OTU A1 was closely related to an Endozoicomonas sp. sequence first identified in a marine sponge, and represented 25% of the sequences corresponding to the restricted microbiome (Fig. 2A). Of the 20 most abundant OTUs, five were closely related to Endozoicomonas sp. and accounted for 29.5% of the restricted microbiome community (Fig. 2A). OTU 2 was related to a Spirochaetes sp. sequence recovered from marine sediments, and constituted 23% of the sequences in the restricted microbiome. OTUs A3, A4 and 5 were also closely related to Spirochaetes affiliates, and together accounted for ∼39% of the microbial community (Fig. 2A). OTUs A10, A14 and A17 were closely associated with a Cytophaga sp. BHI80-3 sequence and together accounted for 4% of the restricted microbiome (Fig. 2A). Although Alteromonadaceae were not represented among the most abundant families, OTU A7 (which represented 2% of the retrieved sequences) was closely related to an Alteromonadales sp. from a marine sponge (Steinert et al.2014). Ricksettiaceae-related sequences (OTU A8) were the most abundant family within the Alphaproteobacteria, but accounted for only 2% of retrieved sequences (Fig. 2A). Interestingly, OTU 2 was highly abundant in all Lobophytum samples, whereas the relative abundance of all other OTUs showed extensive variation across the samples (Fig. 2A).

Figure 2.

Taxonomic affiliation of the 20 most abundant OTUs comprising the microbiomes shared by 51% of each set of samples. Colour intensity on the heat map represents the normalised counts for OTUs in each sample. (A) n = 8 and 9 males. (B) A total of four L. pauciflorum fragments derived from six parent colonies were placed into the following tank treatments (n = 1 fragment per tank): (i)control, (ii) temperature stress, (iii) CO₂ stress; and (iv) combined temperature and CO₂ stress. One lobe of each fragment from each tank treatment was taken at sample day 1 and day 12 of the experiment. Note for one colony no sample was obtained for the combined temperature and CO₂ stress treatment on day 12. Fragments of parent colonies 1 and 2 were collected from Orpheus Island and fragments of parent colonies 4–7 were collected from Pelorus. OTUs with identical references in the SILVA database have been named identically, e.g. OTU2. OTUs highlighted in grey have identical hits in the NCBI database as OTUs in (A).

Effects of environmental stress on the microbial community and host response

Microbial community responses

The response of the Lobophytum pauciflorum microbial community to environmental stress was investigated over a period of 12 days, the treatments being (1) thermal stress (4°C above ambient), (2) pH stress (pH 7.9) and (3) both stressors simultaneously. Sequencing the 16S rRNA of samples from the three experimental treatments and controls allowed the identification of a total of 3509 OTUs derived from the 1886 381 reads, recovered after stringent quality checking and removal of OTUs represented by 15 or fewer reads. The microbiome shared by at least 51% of the samples contributed to 63% to the overall microbial community. Alpha diversity parameters were similar across experimental treatments; however, more OTUs were observed in samples collected at day 1 of the experiment compared to day 12 (P < 0.05), which resulted in a significantly higher Fisher's diversity and estimated OTUs (chao1), P < 0.05 for the first sampling time point (Table 2). Species evenness, however, was similar in samples from both sampling time points, 0.82 ± 0.14 at day 1 and 0.76 ± 0.22 at day 12. The restricted microbiome comprised 249 OTUs dominated by Spirochaetes-related sequences, assigned to the family Spirochaetaceae which represented 34% of the retrieved sequences (Table 3). The second most abundant class in the microbiome shared by at least 51% of the samples was the Gammaproteobacteria (22% of sequences shared by at least 51% of the samples) dominated by Alteromonadaceae- and Enterobacteriaceae-related sequences (6% and 5% of the sequences, respectively) though Endozoicimonaceae-related sequences were also relatively abundant (4% of sequences in the restricted microbiome, Table 3). Endozoicimonaceae-related sequences increased by roughly 4% in samples taken after 12 days of exposure to environmental stressors, while relative abundance of Alteromonadaceae-related sequences decreased by about 4% (Table 3). Vibrionaceae-related sequences contributed to a minor extent (2%) to the microbial profile of L. pauciflorum. The relative abundance of sequences assigned to this Gammaproteobacteria family increased 3-fold between day 1 and day 12 of the experiment (P < 0.05), similar to the increase of Endozoicimonaceae-related sequences (Table 3). Approximately 15% of all sequences were associated with Alphaproteobacteria, and Rhodobacteraceae-related sequences (10% of the retrieved sequences) were the most abundant at this family level. Only 5% of the sequences were assigned to Flavobacteria-related sequences, though the relative abundance was about three times lower after 12 days of exposure to environmental conditions than after 1 day (P < 0.05, Table 3). About 15% of the sequences could not be assigned to known 16S rRNA sequences in the SILVA database, however, were assigned to the kingdom Bacteria when blasted against the NCBI database.

Correspondence analyses provided no evidence for the Lobophytum microbiome (neither the whole nor the restricted community) being significantly influenced by any of the stress treatments at either time point (Fig. 3A, Fig. S4A, Supporting Information). Moreover, at the class level, relative abundances were similar in samples taken at either the 1 day or 12 day time points, indicating that neither maintenance in the flow through system nor the experimental treatments significantly impacted the Lobophytum microbial community (Table 3, Table S2, Supporting Information). The clustering pattern of the samples was, however, explained by colony identity (Fig. 3B, Fig. S4B, Supporting Information). The explanatory power of the correspondence analysis was limited, however, with the first and second correspondence axis explaining only 4.68% and 4.24% for the whole community and 13% and 11% of the variation when the restricted community was analysed, likely due to the high similarity of the microbial community across all samples. However, small shifts in the minor groups of ribotypes associated with individual colonies appeared to be the principal driver in the separation of samples in correspondence analyses (Fig. 3B, Fig. S4B, Supporting Information). The 20 most abundant OTUs of the restricted microbiome accounted 51% of the total retrieved sequence abundance (Fig. 2B). The three most abundant OTUs (OTUs B1, 2, B3) were associated with Spirocheata sp.-related sequences and, together with OTU 5 accounted for a total of 22% of the microbial community profiles. Two ribotypes associated with a Cytophaga sp. BHI80-3 sequence contributed 4% of the community. Gammaproteobacteria- and Alphaproteobacteria-related sequences each represented ∼6% of retrieved sequences (Fig. 2B). Gammaproteobacteria was represented by Enterobacteraceae- and Alteromonadaceae-related sequences, while Alphaproteobacteria was represented only by Rhodobacteraceae-related sequences. Colony-specific OTUs were also observed in this experiment, with OTUs 5 and B14 being abundant in colony 1 and OTUs 2 and B4 being more abundant in colonies 4 and 5. Of the 20 most abundant OTUs identified in this stress experiment, 8 (OTUs B1, 2, B3, B5, 5, 19, B11 and B14) were also identified among the 20 most abundant OTUs in the sex experiment and assigned to the same sequences (Fig. 2A and B).

Figure 3.

Correspondence analysis of the microbiome shared by 51% of the samples in L. pauciflorum fragments subjected to temperature and CO₂ stress. Percentages in the axis titles represent the percentage described by the axis, n = 249 OTUs. (A) Squares denote sampling 1 day and triangles sampling 12 days after experimental stress conditions were established. Experimental stress conditions are differentiated by colour: control (C) in green, high temperature (T) in red, high CO₂ in blue and combination of high temperature and high CO₂ (TC) in purple. n = 5–6 samples for each environmental stress per time point. (B) Each colony is described by its individual colour, n = 7–8 samples for each colony.

Host caspase activity

Caspase activity, employed here as a proxy for the host stress response, was measured in the soft coral fragments subjected to temperature and CO₂ stress, but did not vary significantly between sampling time points (1 day vs 12 day) or between environmental stress conditions (Fig. S2A, Supporting Information). Within treatment and time point variation was higher than between treatments and time points (Fig. S2A, Supporting Information). In addition, the variability between colonies/genotypes was greater than the impact of the environmental stress; thus, the level of caspase activity is a colony property that was unaffected by stress (Fig. S2B, Supporting Information). A linear model could not explain the differences at colony/genotype level, potentially due to missing information about other physiological factors or sex.

DISCUSSION

The microbial community associated with Lobophytum pauciflorum

Despite their likely significance for colony health (Rohwer and Kelley 2004; Bourne et al.2008), the microbial communities of soft corals have received very little attention to date. In this study, the microbial profiles of L. pauciflorum showed high relative abundance of Spirochaete-related sequences across two independent experiments, highlighting the dominance and likely ubiquity of this class in the microbial communities associated with this species. Spirochaete-related sequences also dominated the microbiome of other octocorals (Holm and Heidelberg 2016; Kellogg, Ross and Brooke 2016; Lawler et al.2016; van de Water et al.2016a) and those of scleractinian corals in close proximity to a high effluent fish culture in the Philippines (Garren et al.2009). Spirochaetes have also been detected at lower abundance in some Stylophora pistillata colonies on the GBR (Kvennefors et al.2010), in the gorgonian Plumarella superba from the Aleutian Islands (Gray et al.2011), and are also minor components of the restricted microbiome of some mesophotic hard corals (Ainsworth et al.2015). Spirochaetes are common components of the hindgut microbiome of termites, establishing an obligate symbiosis that has undergone a degree of coevolution (To, Margulis and Cheung 1978; Berlanga, Paster and Guerrero 2007). Where the Spirochaetes are localised in the L. pauciflorum association is unknown, but the production of secondary metabolites with antimicrobial properties by the octocoral host (Yan et al.2010a,b, 2011) suggests that they are more likely to be internal rather than external. In the termite gut, Spirochaetes are thought to carry out nitrogen fixation (Berlanga 2015), but an additional role in structuring the microbial community by secretion of antibiotics has been suggested in the case of Corallium rubrum (van de Water et al.2016a).

Gammaproteobacteria were the second most abundant class of sequences associated with L. pauciflorum (∼26% relative abundance, Table 3) and the majority of sequences within this class were affiliated with the Endozoicimonaceae (∼23% relative abundance; Table 3). Endozoicimonaceae is a family of Gammaproteobacteria whose members are common associates of a wide range of marine invertebrates that includes hard and soft corals (Nishijima et al.2013; Bayer et al.2013b; Bourne et al.2013; Lema, Bourne and Willis 2014; La Rivière, Garrabou and Bally 2015), and recent studies suggest a degree of coevolution between Endozoicomonas strains and their gorgonian hosts, possibly extending to family level (Bayer et al.2013a; La Rivière, Garrabou and Bally 2015; van de Water et al.2016b). Aggregations of Endozoicomonas have been identified in the endoderm of the Red Sea corals Stylophora pistillata and Pocillopora verrucosa, suggesting an endosymbiotic relationship (Bayer et al.2013b, Neave et al.2017), and genome sequencing suggests a role in nitrogen cycling within the host (Neave et al.2014). Similar relationships may exist in the case of L. pauciflorum; however, in situ localisation studies combined with functional analyses are required to confirm this. Tightly packed bacterial aggregates have been identified within the gastrodermis of the Mediterranean gorgonian Paramuricea clavata, establishing the first methodology for in situ localisation in octocorals (La Rivière, Garel and Bally 2016).

Among the Alphaproteobacteria associated with L. pauciflorum, Rickettsiaceae-related sequences were identified as the most abundant (4% relative abundance, Table 2). Rickettsia-related bacteria are present as cell-associated microbial aggregates in many healthy scleractinian corals, but the significance of these is unknown (Work and Aeby 2014). In the whitefly, Rickettsia sp. are present in gut and follicle cells and are transmitted both vertically and horizontally within populations; whilst the significance of the association remains unclear, it has been suggested that infection leads to an advantageous phenotype (Gottlieb et al.2006). However, in the case of the pea aphid, Acyrthosiphon pisum, Rickettsia infection had negative effects on some aspects of host fitness (Sakurai et al.2005).

Low abundance components of the Lobophytum microbiome included Pseudoalteromonadaceae- and Vibrionacaea-related sequences (0.7% and 0.5%, respectively; Table 3). Pseudoalteromonas- and Vibrio- related sequences are frequently recovered as components of the microbiomes of scleractinian corals and, although Vibrio spp. have often been associated with compromised coral health (Bourne et al.2008; Vega Thurber et al.2009; Séré et al.2013), they are also represented in the microbiomes of ‘healthy’ corals. Analysing the microbiomes of different coral compartments, Ainsworth et al.(2015) proposed that this family might rather be associated with the surface mucus. Members of the Vibrionaceae recovered among the sequences in L. pauciflorum might therefore also be associated with the soft coral mucus. In Pocillopora meandrina, Pseudoalteromonas strain HIMB1276 was associated with the outermost ectoderm of planulae (Apprill et al.2012); as the Pseudoalteromonas metabolite tetrabromopyrrole is an inducer of settlement in Acropora millepora, Pseudomonadaceae might be important during settlement in scleractinian corals (Tebben et al.2011). Although Pseudoalteromonas might also have a beneficial role in some corals due to the high abundance in adult A. millepora (Littman et al.2009), their low abundance in Lobophytum makes this less likely, leading us to suggest that these bacteria might be epibionts associated with mucus or particles. In addition to members of Spirochaetes and Endozoicomonas, members of five families with a relative abundance below 5% were detected (Table 3) suggesting that the microbial diversity in Lobophytum is low, as in C. rubrum, where the core microbiome at 100% representation is composed of only 12 OTUs (van de Water et al.2016a), suggesting that the diversity of the octocoral microbial community might generally be low. However, the differences in number of OTUs, 832 in Lobophytum and 12 in Corallium, and the selected cut-off level for identification of the core microbiome make further comparison difficult.

The microbial communities associated with female and male colonies were very similar at the class and family levels, and the 20 most abundant OTUs that were classified taxonomically did not differ in relative abundance between the sexes. Multivariate correspondence analysis at the OTU level did, however, reveal significant differences between sampling sites and sexes of Lobophytum colonies (P < 0.05, Fig. 1). The difference between sampling sites was driven largely by variations in the relative abundance of Gammaproteobacteria and Spirochaete-related sequences; however, given the uneven replication for sampling sites (eight fragments from Cattle Bay and four from Pelorus; Table S3, Supporting Information), this result should be viewed cautiously. The apparent difference between sexes might be a consequence of differences in frequency of OTUs that were only assigned to the kingdom of Bacteria but not to any known bacterial taxon, with ∼30% of sequences in female colonies unassigned versus 16% in male colonies. Different microbial abundances between inshore and mid-shelf sampling sites have been observed in the case of A. millepora, inshore colonies being dominated by Gammaproteobacteria while Bacilli and Alphaproteobacteria were more abundant in colonies from mid-shelf reefs (Lema, Willis and Bourne 2014). The three sampling sites chosen for this experiment on L. pauciflorum can all be considered as inshore and were separated by <7 km. Considering the high overall similarity of the individuals and the low explanatory power of the correspondance analysis, it is likely that the differences detected are driven primarily by colony specific differences such as genotype (Figs 1 and 2A; Table S2, Supporting Information). Just as in the case of transcriptional responses of adult corals to various stimuli (Bay et al.2009; Seneca et al.2010; Bertucci et al.2015), individual variation may be a major factor limiting the ability to detect general patterns of host–microbiome association in the present case.

Effects of environmental stress on the microbial community and host response

Subjecting L. pauciflorum fragments to different environmental conditions including increased seawater temperature and lower pH did not shift the structure of the Lobophytum microbial community over the 12-day experimental period (Table 3), and caspase activity in host tissue, used here as a proxy for the stress response, likewise did not change significantly over the course of the experiment. These results indicate that, at both the host and microbiome levels, L. pauciflorum is resilient to short-term environmental stress. In the case of Porites compressa, the microbiome shifted dramatically from a Cyanobacteria- and Proteobacteria-dominated community to become dominated by Bacteriodetes and Fusobacteria when adult colonies were subjected to pH and temperature stress (1.4 units below and 5°C above ambient, respectively) (Vega Thurber et al.2009). On the other hand, juvenile A. millepora subjected to elevated CO₂ (∼680 μatm) for a ‘prolonged’ (9 day) period showed the capacity to acclimate to stress conditions via suppression of apoptosis and elevated expression of specific heat shock proteins (Moya et al.2015). In gorgonians under heat stress, rapid induction of the capacity to protect proteins from thermal denaturation (Morimoto, Kroeger and Cotto 1996; Nathan, Vos and Lindquist 1997) has been observed (Wiens et al.2000; Kingsley et al.2003). A similar capacity in Alcyoniidae (to which L. pauciflorum belongs), combined with the ‘moderate’ bleaching susceptibility of this family (Goulet, LaJeunesse and Fabricius 2008), could explain the apparent resilience of L. pauciflorum in this study.

The lower relative abundance of Endozoicimonaceae-related sequences in the stress experiment compared to the sex experiment may be due to the fact that, in the former case, sampled fragments were held in a recirculating system for 2 months prior to conducting the experiment. However, OTUs affiliated with Endozoicomonaceae were common and still represented 4% in the microbiome of the aquarium manipulated corals. Transferring sponges from the wild into an aquaculture system induced a significant shift in microbial communities between wild and cultured sponges (Mohamed et al.2008). Notably the abundance of Alpharoteobacteria and Gammaproteobacteria was reduced in sponges kept in culture for up to 2 years though the shift occurred after 6 months of culturing. Returning the coral fragments to an open system of ocean water appeared to have reversed this effect as Endozoicimonaceae-related sequences increased over the course of the experiment. However, for a full ‘return’ to the initial microbiome the experiment might not have been conducted long enough. Shifts in microbial communities as response to changing environmental conditions were also observed in scleractinian corals (e.g. Bourne et al.2008; Vega Thurber et al.2009) and are commonly observed when transferring coral fragments from the wild into experimental tanks (Morrow et al.2017).

The clustering patterns of the microbial profiles in the stress experiment were explained by the fragment origin, i.e. the colony the fragment derived from, supporting the idea that different genotypes are the main drivers of the differences detected. The idea that differences in the minor ribotypes drive the differentiation between colonies is supported by the fact that colony 1 is largely different from all other colonies when the whole microbial community is considered (Fig. S4B, Supporting Information); however, when looking only at the restricted microbiome, colony 1 groups with colonies 2, 6 and 7 (Fig. 3B). The colony-specific microbial profiles were in line with individual environmental parameters, a common observation when working with marine invertebrates. To overcome the problem of limited detection and statistical power, future studies should aim to increase the variety of genotypes and numbers of replicates (3–5) sampled per genotype. To date, limited microbiome data are available for octocorals, and these hosts potentially harbour many novel microbes. This study provides a baseline understanding of the microbial community associated with the gonochoric soft coral L. pauciflorum and its susceptibility to environmental stress but did not attempt to fractionate coral tissues. It is possible that some of the apparent differences detected between sampling sites are due to particles attached to the outside of the fragments, and future studies should investigate distribution across different tissue microhabitats. OTUs that are rare overall may have much higher relative abundance within tissues, as has been demonstrated for some hard corals (Ainsworth et al.2015), and the location of specific ribotypes carries functional implications, so it is important to localise the microbiome in Lobophytum through the application of fluorescent in situ hybridisation approaches.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSEC online.

Acknowledgments

The authors thank Anthony Bertucci and Natalia Andrade for field work assistance and the staff of the Orpheus Island Research Station, particularly Marta Espinheira, for their support. The authors also thank the two anonymous reviewers for their very constructive comments and suggestions which greatly improved the manuscript.

FUNDING

This work was supported by the Australian Research Council via the Centre of Excellence for Coral Reef Studies. WW gratefully acknowledges receipt of James Cook University Postgraduate Research Scholarship and James Cook University Graduate Research Scheme Grant funding.

Conflict of interest. None declared.