Coral microbiome diversity reflects mass coral bleaching susceptibility during the 2016 El Niño heat wave

Abstract Repeat marine heat wave‐induced mass coral bleaching has decimated reefs in Seychelles for 35 years, but how coral‐associated microbial diversity (microalgal endosymbionts of the family Symbiodiniaceae and bacterial communities) potentially underpins broad‐scale bleaching dynamics remains unknown. We assessed microbiome composition during the 2016 heat wave peak at two contrasting reef sites (clear vs. turbid) in Seychelles, for key coral species considered bleaching sensitive (Acropora muricata, Acropora gemmifera) or tolerant (Porites lutea, Coelastrea aspera). For all species and sites, we sampled bleached versus unbleached colonies to examine how microbiomes align with heat stress susceptibility. Over 30% of all corals bleached in 2016, half of which were from Acropora sp. and Pocillopora sp. mass bleaching that largely transitioned to mortality by 2017. Symbiodiniaceae ITS2‐sequencing revealed that the two Acropora sp. and P. lutea generally associated with C3z/C3 and C15 types, respectively, whereas C. aspera exhibited a plastic association with multiple D types and two C3z types. 16S rRNA gene sequencing revealed that bacterial communities were coral host‐specific, largely through differences in the most abundant families, Hahellaceae (comprising Endozoicomonas), Rhodospirillaceae, and Rhodobacteraceae. Both Acropora sp. exhibited lower bacterial diversity, species richness, and community evenness compared to more bleaching‐resistant P. lutea and C. aspera. Different bleaching susceptibility among coral species was thus consistent with distinct microbiome community profiles. These profiles were conserved across bleached and unbleached colonies of all coral species. As this pattern could also reflect a parallel response of the microbiome to environmental changes, the detailed functional associations will need to be determined in future studies. Further understanding such microbiome‐environmental interactions is likely critical to target more effective management within oceanically isolated reefs of Seychelles.


| INTRODUC TI ON
Coral reef ecosystems are exceptionally vulnerable to anthropogenic disturbance and have been decimated by climate changedriven marine heat waves during 2015-2017, with >30% of all corals lost at many locations worldwide through bleaching (Hughes et al., 2018(Hughes et al., , 2017. Western Indian Ocean (WIO) reefs were particularly affected by Sea Surface Temperature (SST) anomalies throughout 2016 broadly exceeding 10-15 maximum degree heating weeks (DHW), driving severe bleaching and mortality throughout this region (Hughes et al., 2018). Reefs within the WIO have in fact been repeatedly impacted by heat waves throughout the last 20 years (Graham, Jennings, MacNeil, Mouillot, & Wilson, 2015;Hughes et al., 2018;McClanahan, Ateweberhan, Darling, Graham, & Muthiga, 2014), and the combination of smaller scale thermal anomalies with other stressors has increasingly limited long-term coral recovery (e.g., Graham et al., 2015;Zinke et al., 2018).
Seychelles coral reefs were among the most impacted globally during the 1998 mass bleaching, with coral cover reduced by >90% across the inner islands (Graham et al., 2006;Wilson et al., 2012).
Recovery has been limited by strong phase shifts toward algal dominance in several reefs (Graham et al., 2015;Wilson et al., 2012), and recruitment bottlenecks (Chong-Seng, Graham, & Pratchett, 2014), such that coral cover and diversity in Seychelles is now lower than for many other regions in the WIO (Harris, Wilson, Graham, & Sheppard, 2014). Reefs are characterized by patch, granitic, and carbonaceous habitats across coastal fringing and oceanic environments (Graham et al., 2006;Jennings, Grandcourt, & Polunin, 1995). However, loss of coral cover and diversity since 1998 has been greatest for the carbonaceous reefs (Graham et al., 2008;Wilson et al., 2012), resulting in reduced carbonate budgets, accretion potential, and structural maintenance (Januchowski-Hartley, Graham, Wilson, Jennings, & Perry, 2017) that underpin the critical ecosystem service value of Seychelles' reefs (Clifton et al., 2012).
Existence of functional diversity at macroecological scales of lower latitude reefs in the WIO, including Seychelles, has recently been identified as a key determinant of longer-term resilience to environmental stress (Zinke et al., 2018). However, such a role for microecological processes in the WIO remains largely unknown, in particular microbial community composition and functioning that can be critical in determining coral stress resilience in other reef regions worldwide (Putnam, Barott, Ainsworth, & Gates, 2017;Suggett, Warner, & Leggat, 2017).
Microbiome composition is important in determining coral health over space and time, yet is completely undescribed for the coral communities in Seychelles. Therefore, as part of a long-term program examining carbonaceous coral communities within Curieuse Marine National Park (Seychelles), we evaluated Symbiodiniaceae and bacterial diversity and community composition among key reef-building coral species (Acropora gemmifera, Acropora muricata, Coelastrea aspera, and Porites lutea) and for populations within a clear water versus turbid reef environment.

| Site description and benthic sampling
Coral communities have been examined from two fringing carbonaceous reef sites within Curieuse Marine National Park (CMNP) since 2009 (Supporting Information Figure S1)

| Microbiome sampling and DNA extraction
All colonies were sampled 6-7 days after strong bleaching was first observed. Triplicate fragments were taken from independent colonies for each of the four key coral species found throughout the upper reef slope at both turbid (Praslin) and clear water ( (Pogoreutz et al., 2018;Salter et al. 2014) were performed in triplicate reactions with Qiagen Multiplex PCR Kit (Qiagen; see below).
To amplify the bacterial 16S rRNA gene, we used the primers
Briefly, the multicopy nature of the rRNA gene means that every Symbiodiniaceae genome contains hundreds to thousands of copies of it. Each of these gene copies is able to accrue mutations somewhat independently. As such, considerable intragenomic sequence diversity that may be leveraged for purposes of taxonomic delineation is found within every Symbiodiniaceae cell (Hume, D'Angelo, Burt, & Wiedenmann, 2018). SymPortal aims to make use of this diversity, using next-generation amplicon sequencing data, to resolve between genetically differentiated taxa. SymPortal works by identifying specific sets of defining intragenomic ITS2 sequence variants (DIVs) that are used to define the taxonomic unit of SymPortal, the ITS2 type profile, indicative of genetically differentiated Symbiodiniaceae taxa. Demultiplexed and paired forward and reverse fastq.gz files outputted from the Illumina sequencing were submitted directly to SymPortal. Sequence quality control was conducted as part of the SymPortal pipeline using Mothur 1.39.5 (Schloss et al., 2009), the BLAST + suite of executables (Camacho et al., 2009), and minimum entropy decomposition (MED; Eren et al., 2015).
Over the years, development and variation in the range of techniques employed to genetically differentiate within the Symbiodiniaceae have led to a range of different terms being used to describe the genotypic units of resolution. For example, "type," "ITS2 type," "ITS2 profile," "ITS2 fingerprint," "clade," "subclade," "subtype," and most recently "ITS2 type profile" are among the most common.
Some of these are used interchangeably in one setting, while representing different entities in others. To clarify and limit ambiguity, for the purposes of this study, we will restrict our use to "type" and "ITS2 type profile." A type refers to Symbiodiniacea taxa that have a specific sequence or set of sequences as their most abundant sequence. An ITS2 type profile is a set of sequences that are used to define either a putative or defined taxa. For example, Durusdinium trenchii is a D1 type and has an ITS2 type profile of D1-D4.

| Data analysis
To initially examine trends of total live coral cover, we pooled data according to species of Acropora and Pocillopora versus "other species" since, as in previous bleaching episodes in Seychelles (e.g., see Wilson et al., 2012), these two genera are typically most susceptible to bleaching. Unpaired t-tests were used to analyze the percent of coral cover data, using Welch's correction for variable standard deviations (GraphPad Prism v.6  Total live coral cover has been steadily increasing within CMNP at both sites between 2009 and 2016 (from 35% ± 5% to 49% ± 9% and from 30% ± 2% to 42% ± 7%, in East Bay and Praslin, respectively; Figure 1a). However, following the 2016 heat wave, total live coral cover at East Bay significantly declined from 49% ± 9% to 22% ± 2% (t = 2.91, df = 4, p < 0.05) and at Praslin from 42% ± 7% to 13% ± 3% (East Bay) and 29% ± 4% (Praslin; Figure 2b).
Of the 49% total live cover at East Bay before the heat wave in 2016, almost half (21% ± 4%) was Acropora sp. and Pocillopora sp.

| Bacterial community structure
Overall, the data set comprised 43 16S rRNA gene libraries (three replicates × four coral species × two sites × two coral conditions) totaling 1,412,100 sequences with a mean length of 294 bp. After quality filtering and exclusion of chimeras, 1,065,414 sequences were annotated to bacteria. Clustering of these sequences at the 97% similarity level resulted in 2,362 OTUs (Supporting Information Table   S7), presented as a taxonomy stacked column plot to the phylogenetic level of family (Supporting Information Figure S3). Significant interactions in bacterial diversity were found between species, site, and colony condition for the number of OTUs per sample, phylogenetic diversity and Chao1 (Univariate GLM; see Table 1). There was a significant difference between coral species for the number

| D ISCUSS I ON
Microbiomes play a key role in contributing to coral fitness over space and time (Putnam et al., 2017;Suggett et al., 2017) and are known to exhibit broad changes across reefs persisting under different environmental conditions (Roder et al., 2015) and when subjected to atypical stress (Grottoli et al., 2018;Röthig et al., 2016;Ziegler, Seneca et al., 2017b). Here, we provide the first characterization of the microbial community composition (i.e., Symbiodiniaceae and bacteria) for key reef-building coral taxa of Seychelles across two different environments. In addition, we characterize the microbial communities associated with states of coral health (i.e., bleached and unbleached), collected during the 2016 marine heat wave that induced mass coral bleaching and mortality. In examining coral species that have previously been shown to be broadly heat stress sensitive (A. muricata, A. gemmifera) and tolerant (P. lutea, C. aspera) in Seychelles (Harris et al., 2014) and WIO (McClanahan et al., 2014, we have shown that bleaching susceptibility among coral species is indeed broadly consistent with differences in microbiomes. However, conserved microbiome signatures observed for bleached and unbleached colonies of all coral species suggest complex regulation of bleaching severity by the coral holobiont and genotype.

| Decline in coral cover during 2016 mass bleaching
Heat stress is recognized as the most common cause of coral bleaching, and high record temperatures between 2015 and 2017 triggered the third global mass bleaching event, the most damaging to date (Hughes et al., 2018(Hughes et al., , 2017. At CMNP, about 40% of all corals  in April 2017. Specifically, decline of Acropora sp. and Pocillopora sp. from ca. 15%-20% (2016) to 1%-2% of total benthic cover was recorded (2017; Figure 2c). Such dramatic loss of these same taxa was similarly observed during previous recent marine heat wave events in Seychelles (Graham et al., 2008;Wilson et al., 2012) and other sites in the WIO (Baker, McClanahan, Starger, & Boonstra, 2013;McClanahan et al., 2014McClanahan et al., , 2007, reflecting the typically stress-sensitive "boom and bust" nature of commonly fast-growing branching taxa (Darling, Alvarez-Filip, Oliver, McClanahan, & Côté, 2012;Zinke et al., 2018). The significant decline in abundance of Acropora sp. and

| Species-specific Symbiodiniaceae composition
As expected, the major symbiont communities for corals sampled in CMNP broadly reflected those sampled from other Indo-Pacific regions. For example, ITS2 type C3 (and additional types characterized by sequences from the C3 radiation; Thornhill, Lewis, Wham, & LaJeunesse, 2014) for Acropora sp. in Eastern Africa (Chauka, 2012), the Chagos Archipelago (Yang et al., 2012), Red Sea (Ziegler, Eguíluz et al., 2017a) and Persian-Arabian Gulf (Hume et al., 2013(Hume et al., , 2016Smith, Vaughan et al., 2017). The C3 sequence represents one of several major radiations within the genus Cladocopium (Thornhill et al., 2014) and our C3z type profile among Acropora sp. sampled here is 2 bp different from the C3 basal sequence. As such, it is currently unclear how genetically and/ or phenotypically comparable our C3 type(s) are to these reports from elsewhere in the WIO previously (but see Ziegler, Eguíluz et al., 2017a for the Red Sea). Heat stress tolerance is clearly highly variable among the C3 radiation (Hume et al., 2013(Hume et al., , 2015. However, the mass bleaching response for Acropora sp. within the CMNP commonly hosting C3z-C3 would suggest that this Cladocopium type is inherently heat stress sensitive, and a highly conserved Acropora sp. host-symbiont association. While we did not assess symbiont types associated with Pocillopora sp. that also experienced mass bleaching in CMNP, previous observations from the Chagos Archipelago (Yang et al., 2012) and Tanzania (Chauka, 2012) have identified almost synonymous associations with the C1 group (Pocillopora damicornis, Pocillopora verrucosa, Pocillopora eydouxi), another major radiation within Cladocopium (Thornhill et al., 2014).
Alternate host-symbiont associations were observed among the coral taxa that exhibited comparatively little mass bleaching within CMNP, P. lutea and C. aspera. As with elsewhere in Eastern Africa (Chauka, 2012), the Red Sea and Persian-Arabian Gulf Smith, Vaughan et al., 2017;Ziegler, Eguíluz et al., 2017a) and However, the mass bleaching observed at both our relatively turbid and clear sites (of similar extent, Figure 2) would thus suggest small-scale environmental variability afforded through complex reef habitats (e.g., shading from overhangs, Cacciapaglia & Woesik, 2016) are more important in providing refuge from heat stress. Bleaching of P. lutea and loss of C15 type cells at both sites similarly suggests localized small-scale amplification of heat stress, for example, by high light (Hoogenboom et al., 2017 (Bourne et al., 2008;Cárdenas et al., 2012;Roder et al., 2013;Roder, Arif, Daniels, Weil, & Voolstra, 2014;Vega Thurber et al., 2009). Community shifts (such as a decrease in Endozoicomonas sp.) can occur in visibly healthy corals in degraded ecosystems (Ziegler et al., 2016) and changes to the bacterial community prior to visual signs bleaching (e.g., increase in Vibrio related sequences) have been shown to occur (Bourne et al., 2008). In contrast, we did not observe such obvious bacterial community shifts between bleached and unbleached conspecifics for heat stress susceptible or resistant coral species in Seychelles. This outcome is surprising since bacterial communities are expected to exhibit more rapid responses to stressors than Symbiodinium due to faster metabolism and generation times (Pogoreutz et al., 2018). As such, the observed bleaching susceptibility in our study may be more likely driven by differences in Symbiodiniaceae (via stress susceptibility) rather than by the hosts' bacterial assemblages.  Neave, Apprill, Ferrier-Pagès, & Voolstra, 2016;Neave, Rachmawati, et al., 2017a;Pogoreutz et al., 2018;Pootakham et al., 2017). While the function of Endozoicomonas has not yet been defined, their genomes are significantly enriched in genes for carbohydrate transport and recycling as well as for protein and amino acid provision (Neave, Michell, Apprill, & Voolstra, 2017b) and phenotypic assays confirm a high metabolic versatility in vitro (Yang et al., 2010). Despite the high abundance of Endozoicomonas in apparently healthy corals (Apprill, Hughen, & Mincer, 2013;Roder et al., 2015), strongly reduced abundances have been reported for stressed, diseased or bleached corals, suggesting they may be an indicator of coral health or habitat suitability (Bourne et al., 2008;Cárdenas et al., 2012;Meyer, Paul, & Teplitski, 2014;Röthig et al., 2016;Ziegler et al., 2016). However, similar to previous findings in bleached Red Sea P. verrucosa (Pogoreutz et al., 2017 Table   S5). Given their presence in a range of hosts and environments (including seawater), it suggests this taxon is metabolically flexible and may provide important functions to the coral holobiont (Röthig et al., 2016).
Vibrionaceae are an opportunistic and potentially pathogenic bacterial family commonly associated with coral disease and have previously been linked to bleaching (Bourne et al., 2008;Garren et al., 2015;Tout et al., 2015). We found Vibrionaceae in all corals, albeit in lower abundance, ranging between a contribution of 1.12% (P. lutea) to 2.61% (A. muricata). The reported association of Vibrionaceae associated with unimpaired, apparently healthy corals has been reported (Bourne & Munn, 2005). Nevertheless, the consistent association of these bacteria suggests that corals in Seychelles harbor stable, and presumably locally adjusted microbiomes, allowing the corals to cope well in the ambient environment (sensu Hernandez-Agreda, Leggat, Bongaerts, & Ainsworth, 2016).

| Bacterial diversity aligns with bleaching susceptibility
Healthy corals generally comprise specific, stable, and uneven microbial assemblages indicating host-selected microbiomes Bourne et al., 2008). We observed the highest species richness (Chao1), evenness (Simpson's), and bacterial diversity (Shannon's) in the more heat-tolerant massive corals (P. lutea, followed by C. aspera) compared with stress-sensitive branching corals (A. muricata higher than A. gemmifera), consistent with findings from F I G U R E 4 Average relative abundance (%) of bacterial community composition for (a) Acropora gemmifera, (b) Acropora muricata, (c) Coelastrea aspera and (d) Porites lutea classified as bleached or unbleached from East Bay and Praslin in Seychelles as a taxonomy stacked column plot to family level. Remaining taxa are grouped as "other". Values displayed are mean relative abundances (n = 2-3). Each color represents one of the 10 most abundant bacterial families. (e) Principle Coordinate Analysis (PCoA) for the dominant bacterial taxa found in four coral species at East Bay and Praslin in Seychelles. Data was fourth-root transformed, and a Bray-Curtis similarity matrix was used with a correlation of 0.2. Ellipses denote similarity clusters of 20% (green dashed line). Percentages on axes indicate variation explained by the two coordinates Liang et al. (2017). Acropora muricata also exhibited a highly une-  Recent work has shown coral host intraspecific differences in Symbiodiniaceae composition correlated with disease susceptibility (Rouzé, Lecellier, Saulnier, & Berteaux-Lecellier, 2016).

Specifically, the predisposition to disease and infection by
Vibrio spp. was positively correlated with Symbiodiniaceae genus Symbiodinium (formerly clade A), but negatively correlated with Durusdinium in Acropora cytherea (Rouzé et al., 2016). In our study, we found Durusdinium types D4 and D9 unique to P. lutea (and bleached samples of C. aspera) and previous work has shown increased occurrence of Durusdinium types is consistent with increased turbidity. Furthermore, Durusdinium-infected corals have shown shifts in the associated bacterial community under heat stress, while no shifts were reported for Cladocopium within the same coral host (Littman et al., 2010). It might therefore be informative to examine for bacterial associates that co-occur with Symbiodiniaceae types in future studies.
In characterizing the microbiome composition of four species of Seychelles' corals during the most severe mass bleaching event on record, we have shown that susceptibility to stressors is reflected by underlying microbiome community structures. Specifically, bleaching susceptibility among coral taxa corresponds largely to differences in specific host-Symbiodiniaceae associations, while the bacterial microbiome community remains largely stable. As such, unbleached colonies of bleaching-susceptible corals likely persist through availability of small-scale ("micro") environmental refuges that can dampen the effect of heat stress, (e.g., shading from surrounding substrate). While bacterial communities were highly similar between bleached and unbleached corals, differences were observed between species, adding to previous evidence for species-specificity. Thus, microbiome profiling of both Symbiodiniaceae and bacterial communities may provide new capacity to more broadly identify stress susceptible versus tolerant coral populations, which is needed to aid targeted management in reef systems such as Seychelles.

ACK N OWLED G M ENTS
We wish to extend our greatest thanks to the Seychelles National

CO N FLI C T O F I NTE R E S T
None declared.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section at the end of the article.