Bacterial genome size reflects evolutionary dependency of microbial species engaged in obligate symbiosis with multicellular hosts (Moran et al., 2008). Obligate symbionts exhibit the smallest genome sizes when compared to facultative endosymbionts or free-living bacteria (McCutcheon and Moran, 2012). Data on genome size reduction come from studies on insect endosymbiosis, but similar trends are likely to be observed in other tightly coupled host-microbe associations (Gerardo, 2013). While much of the evolutionary history between corals and their bacterial symbionts is yet to be elucidated, patterns of transmission are likely playing a role in host and symbiont genome evolution. If a bacterial symbiont is both vertically transmitted and an endosymbiont, its genome will be more likely to be reduced in size relative to other bacterial mutualists that have free-living stages. The coral-associated bacterial genomes that have been sequenced to date are from cultured facultative coral associates with a presumed free-living state and contain genomes that are of a similar size as their non-coral associated relatives (Reshef et al., 2008; Mavromatis et al., 2010; Santos et al., 2011; Bondarev et al., 2013; Neave et al., 2014). For example, the recently described Verrucomicrobia species Coraliomargarita akajimensis (3.7 Mb genome) was isolated from water adjacent to the coral Galaxea fascicularis (Mavromatis et al., 2010) is within the size range estimated for other recently sequenced Verrucomicrobia genomes from aquatic habitats (Martinez-Garcia et al., 2012). Other coral-associated genomes from Proteobacteria exceed 5 Mb which is typical for free-living Proteobacteria. These include an alpha-Proteobacterium Pseudovibrio strain FO-BEG1 (5.9 Mb genome) isolated from an enrichment culture of a black band diseased coral (Bondarev et al., 2013), the gamma-Proteobacteria Endozoicomonas montipora LMG24815 (5.6 Mb genome) isolated from Montipora aequituberculata (Yang et al., 2010) and Vibrio coralliilyticus a coral pathogen with a 5.5 Mb genome (Santos et al., 2011; Kimes et al., 2012).

Thanks to major developments in sequencing technology, we are no longer limited to studying the small fraction of coral-associated microbes that can be cultured. Indeed, if corals harbor co-evolved microbial symbionts these may not be readily cultured due to fastidious nutritional requirements and metabolic complementation. Culture-independent approaches such as sequencing of 16S rRNA genes has revealed high proportions of novel diversity associated with corals (Frias-Lopez et al., 2002; Rohwer et al., 2002; de Castro et al., 2010; Sunagawa et al., 2010; Fernando et al., 2014) and the dynamic genomic diversity of the microbiota has been revealed through metagenomics revealing shifts in the proportions of microbial genes for functions such as nutrient cycling and virulence following environmental perturbations (Yokouchi et al., 2006; Dinsdale et al., 2008; Vega Thurber et al., 2009). Despite these advances, identifying specific microbial activities in the holobiont and linking these to individual microbial populations that may share a co-evolutionary history with corals remains a challenge. Genome sequencing of single cells and expanded novel-cultivation-based approaches for isolation-based studies are potential tools to characterize currently uncultured microbial symbionts. Such efforts will improve the taxonomy of coral-associated microbes and will enable us to test hypothesis of coevolution and co-diversification in the coral holobiont.

In addition to the well-established mutualistic nutritional benefits between corals and Symbiodinium (Muscatine and Porter, 1977; Muscatine, 1990; Shoguchi et al., 2013), other reciprocal relationships of metabolic-complementation could extend beyond the coral-Symbiodinium spectrum for example, to include metabolic interactions between bacteria and Symbiodinium or bacteria and the coral host. Metabolic complementation, where each partner produces gene products necessary for survival of the other, may be reflected in the genome composition of the partners (Gerardo, 2013). In a recent analysis of a tri-partite insect endosymbiosis (i.e., two bacterial endosymbionts and the host), extreme genome reduction was correlated with shared metabolic pathways across three taxa (Husnik et al., 2013). The time of symbiont acquisition however played a key role in driving gene loss in the different bacterial lineages.

The mainly oligotrophic habitat of corals, makes associations driven by metabolic complementation desirable, for instance association between diazotrophic bacteria and M. cavernosa has been shown to improve nitrogen fixation (Lesser et al., 2004). Corals are open systems, in which both obligate endosymbionts and other microbes may be closely interacting with the animal host in a temporally dynamic fashion, increasing the chance for such metabolic cooperation to develop. Indeed, the recently sequenced genome of the coral Acropora digitifera lacked a key enzyme that synthesizes the essential amino acid cysteine from homocysteine or serine (Shinzato et al., 2011; Dunlap et al., 2013). Cysteine thione β synthase (CBS) is missing also from two other Acropora species but it has been reported in multiple corals (Shinzato et al., 2011), suggesting metabolic complementation by microbial symbionts within Acropora (e.g., algal or other).

Evidence that metabolic interaction between holobiont members shaped genome evolution may take the form of genetic systems acquired by the host genome by HGT. Predicted genes for bacterial metabolic pathways i.e., the shikimic acid pathway for biosynthesis of aromatic amino acids, and the glyoxylate pathway for biosynthesis of carbohydrates, are present in the genome of the sea anemone Nematostella due to proposed HGT (Starcevic et al., 2008; Jackson et al., 2011). Similar evidence for potential HGT of the bacterial shikimic acid pathway into the coral A. digitifera has been presented (Shinzato et al., 2011; Shoguchi et al., 2013) pointing to the possible acquisition of the pathway by a common ancestor of the Anthozoa.

The role of HGT is largely unexplored within coral holobiont but new whole genome data from multiple members associated with a single coral host will reveal how prevalent HGT is in coral reef symbioses. Indeed studies from the human gut microbiome have shown that HGT is rampant among gut microbial associates where conditions may parallel those associated with the coral polyp (e.g., high bacterial densities and selective pressures from host) (Smillie et al., 2011). We will soon see an increase in whole genome data from multiple coral hosts, algal symbionts, as well as microbial genomes from these hosts, which will open our horizons to understand the extent of metabolic complementation and rate of HGT across holobiont members.

Gamete formation in corals is not currently known to involve microbial partners, however emerging evidence of vertical and horizontal transmission supports a role for beneficial bacterial populations that associate with coral spawn. The transmission of symbionts in corals has been a subject of careful investigation for several decades and although much more is known about specific modes of transmission for the dinoflagellate algal symbiont Symbiodinium similar mechanisms of transmission may be shared with other microbial symbionts. Algal symbiont transmission is thought to occur in a mainly vertical fashion in brooding corals, with larvae acquiring Symbiodinium from the mother polyp before release (Trench, 1983; Baker, 2003). Vertical transmission has also been observed in several spawning corals that seed their eggs with Symbiodinium (Apprill et al., 2009; Padilla-Gamino et al., 2012). In contrast, externally fertilized larvae generally acquire Symbiodinium from the environment (horizontal transmission) (Trench, 1983; Baker, 2003).

Although less is known about transmission of other non-Symbiodinium microbial symbionts in the holobiont, several studies document both vertical transmission (transfer of symbionts from parent to offspring) and horizontal transmission (uptake of symbionts from the environment) of Bacterial populations. Vertical transmission in Porites astreoides was recently demonstrated. Ectoderm-associated bacteria from the parent colony, mainly Roseobacter spp., were observed in both newly released, and in 4-day old planulae, with densities increasing after settlement (Sharp et al., 2012). In contrast, many coral larvae appear bacteria-free after fertilization and only acquire bacteria as planula or polyps. Apprill et al. (2009) found that 3-day old, but not newly released, planulae of Pocillopora meandrina contained internalized bacterial cells; with Roseobacter spp., as the primary constituents. Horizontal transmission was also documented in the study by Sharp et al. (2010) which followed gametes and planulae of seven broadcast spawners and found no bacterial associations until after the first polyp stage.

Recent studies point to the water column and to the mucus layer surrounding the gametes of spawning corals as media for transfer of symbionts (Symbiodinium and bacterial species) from parent colony to larvae after gamete release (Apprill et al., 2009; Ceh et al., 2012, 2013; Sharp et al., 2012). Both brooding and broadcast spawning corals were observed to release bacteria during spawning, comprised primarily of Roseobacter spp. and Alteromonas spp. (Ceh et al., 2013). The mucus coats of gamete bundles are associated with bacterial populations similar to those found in the parent colony leading to speculation that vertical transmission of these populations may occur either prior to spawning by seeding of the mucus coat with maternally-derived bacterial symbionts, or post-spawning through uptake of microbial associates released by the parent colony into seawater (Apprill et al., 2009; Ceh et al., 2012, 2013; Sharp et al., 2012).

Beneficial activities of the bacterial holobiont members acquired during spawning are unknown and remain an important avenue for ongoing studies. Antimicrobial activity of resident microbes may serve a protective function for coral spawn or newly hatched larvae, although antimicrobial activity associated with extracts from egg bundles themselves was found in only 1 of 11 coral species (Montipora digitata) from the Great Barrier reef (Marquis et al., 2005). As pointed out by Sharp and Ritchie (2012), M. digitata uniquely among the 11 coral species tested, seeds its egg bundles with Symbiodinium, leading the authors to speculate a potentially protective role for Symbiodinium-derived metabolites. Although a low prevalence of solvent-extractable antimicrobial activity was observed in this study, it is possible that other protective compounds such as colonization inhibitors, or antimicrobial peptides (AMP), may contribute to the protection of spawn and larvae against uncontrolled bacterial colonization, as has been suggested by studies in adult corals and in the Cnidarian, Hydra (Jung, 2009; Vidal-Dupiol et al., 2011; Franzenburg et al., 2013; Krediet et al., 2013).

Despite a substantial number of studies investigating the potential role of microorganisms in the induction of larval settlement and metamorphosis, a thorough understanding of all the mechanisms and players involved is still lacking. In corals, metamorphosis and settlement are tightly coupled processes where a coral individual transitions from a motile planula larva to a sedentary polyp. Emerging evidence suggests that bacteria have a fundamental role in moderating metamorphosis and settlement of larvae in the marine environment and suggest that this event may be triggered by diffusible or potentially contact-mediated signals (Dobretsov and Qian, 2004; Hadfield, 2011; Dobretsov et al., 2013; Shikuma et al., 2014).

Enhanced larval settlement and metamorphosis has been observed in response to crustose coralline algal (CCA) surfaces (Morse et al., 1988) that represent a consortium of a red alga and a complex community of bacteria and archaea. Bacterial isolates from CCA surfaces (Pseudoalteromonas spp.) were found to form biofilms that were sufficient to increase metamorphosis rates of Acropora willisae and A. millepora by up to 50% (Negri et al., 2001). Complex bacterial biofilms formed from shallow reef waters were also sufficient to induce high levels of larval settlement and metamorphosis for A. micophthalma and were superior to biofilms formed in deep-water suggesting a connection between spatial variability of marine biofilm community composition and efficiency of larval settlement (Webster et al., 2004). Finally, antibiotic treatment of larval cultures was shown to be sufficient to inhibit settlement, suggesting that bacterial activity is necessary for settlement induction (Vermeij et al., 2009).

Efforts to uncover the molecular mechanism behind induction of coral metamorphosis by various Pseudoalteromonas strains isolated from CCA surfaces led to isolation of the inducing compound tetrabromopyrrole (TBP) (Tebben et al., 2011). TBP was sufficient for inducing metamorphosis of A. millepora at environmentally relevant concentrations. However, larvae induced by TBP, or by Pseudoalteromonas biofilms, metamorphosed without settlement (Tebben et al., 2011). In contrast, A. millepora larvae exposed to complex CCA both metamorphosed and settled, leading the authors to speculate that TBP production by Pseudoalteromonas strains has a net negative effect on coral recruitment and may have antifouling properties (Tebben et al., 2011). In contrast, a recent study of a diverse set of Caribbean corals documented that TBP extracted from Pseudoalteromonas isolates, as well as produced synthetically, stimulated both metamorphosis and settlement (Sneed et al., 2014). Thus, it is clear that TBP is a diffusible signal produced by coral reef Pseudoalteromonas strains that stimulates coral larval metamorphosis, however TBP may have divergent impacts on net larval recruitment depending on the timing of larval settlement.

Coordination of metamorphosis and settlement may involve detection of multiple types of signals produced by biofilms and CCA including both diffusible signals such as TBP and contact mediated signals located on biofilm surfaces. A contact-mediated mechanism for induction of larval settlement of Hydroides elegans, a polychaete worm, was recently established for a Pseudoalteromonas strain (P. luteoviolacea)—where settlement is triggered by larval contact with bacterial cell surface displays of phage-tail like proteins (Huang et al., 2012; Shikuma et al., 2014). Certain populations of bacteria may also deter coral settlement through diffusible signals, for example larvae were observed to avoid settling on tiles adjacent to benthic cyanobacteria hypothesized to produce toxic secondary metabolites (Kuffner et al., 2006). It is likely that the combination of CCA and biofilms comprised of specific bacterial population(s) produce multiple signals that together provide adequate settlement cues for coordinated metamorphosis and settlement in coral larvae. Such coordinated cues may help larvae from diverse invertebrate taxa identify sites optimal for recruitment, and may represent long standing relationships between larvae and bacterial partners in biofilms on marine surfaces (Hadfield, 2011).

Once settlement and metamorphosis is achieved, the coral recruit (i.e., juvenile polyp) must overcome competition with other benthic organisms to grow and form a robust adult colony. Bacteria may mediate the survival of coral recruits in the proximity of adult coral colonies or other coral competitors such as macro- and turf algae (Smith et al., 2006; Marhaver et al., 2013). Enrichment of pathogenic and opportunistic bacteria in the vicinity of adult coral colonies plays a macro-scale role in reef community structure through enhanced mortality of nearby conspecific coral recruits (Marhaver et al., 2013). This effect, first described as the Janzen-Connell hypothesis for diversification of terrestrial forests and may also contribute to the high diversity of invertebrate species observed in coral reefs (Marhaver et al., 2013).

Enrichment of bacteria near primary producers such as benthic algae may also negatively impact the survival of coral recruits. Elevated mortality (up to 100%) of coral nubbins grown with algal competitors and appeared to be mediated by factors able to pass through a fine mesh (Smith et al., 2006). These negative effects were mitigated by the addition of the antibacterial compound ampicillin (Smith et al., 2006) suggesting that bacterial activities, rather than algal toxins, mediated the inhibitory effects of algal proximity on coral recruits. A similar result was obtained by Vermeij et al. (2009), who examined the post-settlement survival of Montipora capitata in the presence and absence of macroalgae and ampicillin where the addition of ampicillin increased recruit survivorship from around 30 to 60% (Vermeij et al., 2009). Exudation of organic carbon by primary producers may enrich for heterotrophic bacteria (Discussed further in Section How can Ecological Approaches Deepen our Understanding of the Multiple Levels of Coral–Microbial Interactions?), including pathogens and opportunists that may impact the survival of recruits. Early survivorship of coral recruits can be increased through gregarious settlement in a species-specific manner (Rivera and Goodbody-Gringley, 2014) and whether coral-to-coral allorecognition or competition during settlement is influenced by coral associated microbial populations is a question worthy of investigation.

Once settled, recruits grow and reach reproductive maturity. Brooding corals tend to breed year round, releasing smaller quantities of well-developed larvae, while spawning corals release gametes in large, often synchronous, spawning events a few times a year. Timing is generally thought to be mediated by the lunar cycle in most species (Babcock et al., 1986; Tanner, 1996). Over the course of spawning events the composition of bacterial assemblages associated with adult colonies remains relatively stable (Apprill and Rappe, 2011; Ceh et al., 2012) with a notable exception being an increased abundance of several microbial populations hypothesized to play a role during reproduction (Ceh et al., 2012, 2013). During spawning Ceh et al. (2012) observed increases in the proportion of α-Proteobacterial 16S rRNA gene sequences (particularly Roseobacter spp.) in the microbiota of three different coral species: Acropora tenuis, Pocillopora damicornis, and Tubastrea faulkneri, and subsequently documented increases of Roseobacter spp. and Alteromonas spp. in filtered seawater concurrent with release of gametes from A. tenuis, or larvae from P. damicornis (Ceh et al., 2013). The authors hypothesized that release of these microbial populations enriched the bacterioplankton with seed populations for colonization of gametes and larvae. Speculation on potential beneficial functions for these reproductive-associated bacteria include antimicrobial protection and roles in induction of larval settlement (Ceh et al., 2013) however the nature of mutualistic interaction, if any, remains unknown.

Coral spawn that does not proceed through development, decays and represents a source of nutrients and carbon to the water column and benthos that may stimulate growth of microbial populations. During a 4-day mass spawning event in Heron Island on the Great Barrier Reef the abundance of bacteria in the water column nearly doubled and increased 3-fold in the sediments (Patten et al., 2008). The impact of increased microbial activity and abundance stimulated by spawn decay on reef water quality is mediated by the physical characteristics of the reef region i.e., dilution capacity and flushing rate. In shallow bays of the Great Barrier Reef blooms of respiring bacteria, and subsequent hypoxia, in response to decaying coral spawn have been associated with mass mortalities of corals and other benthic invertebrates (Simpson et al., 1992). In contrast, in Kaneohe Bay, Hawaii, a bay with comparatively higher flushing rates, slight changes in bacterial production and microbial community composition in the water column were only observed when spawning coincided with a low tide where the dilution-capacity of the bay was at a minimum (Apprill and Rappe, 2011). Release of nutrients and bacterial populations during coral spawning may impact the early colonization of planulae by bacterial associates although this impact is likely to be influenced by physical process such as flow, mixing, and initial nutrient loads.

The tissue of the coral organism contains numerous secretory cells that are responsible for production of a continuous mucus layer that is typically a few hundred micrometers thick (Brown and Bythell, 2005; Jatkar et al., 2010; Garren and Azam, 2012). The formation and roles of coral mucus have been reviewed in detail (Brown and Bythell, 2005). Coral mucus serves multiple functions to maintain holobiont homeostasis that may vary in relative importance depending on the coral species. As a permeable barrier between the coral tissue and seawater that can be periodically shed by a mechanism functionally analogous to mucus clearing in other metazoans, the mucus layer is the first line of defense against biofouling, pathogen invasion, and shading by sedimentation (Brown and Bythell, 2005). In addition, the mucus layer may serve as a trap for nutrient-bearing particles (including bacteria) that are periodically ingested by the coral as a mode of suspension feeding (Coles and Strathman, 1973; Muscatine, 1973). Mucus layers are enriched in photosynthate and coral waste products and thus represent a resource-rich habitat that is readily colonized by marine bacteria (Brown and Bythell, 2005). Mucus layers accumulate dense assemblages of bacteria that share similarity with the bacterial composition of seawater (Bourne and Munn, 2005; Lema et al., 2012) and are also structured by interaction with members of the coral holobiont through both top-down process (i.e., removal/selection) (Ritchie, 2006; Franzenburg et al., 2013) and bottom-up processes (i.e., resource availability) (Nakajima et al., 2009) as discussed further in Section How can Ecological Approaches Deepen our Understanding of the Multiple Levels of Coral–Microbial Interactions? Coral mucus layers are complex habitats, defined by chemical, physical and biological gradients (Brown and Bythell, 2005), thus it is not surprising that spatial heterogeneity of mucus-associated bacterial community composition in adult corals can be observed over the scale of a single coral colony (Daniels et al., 2011). Indeed, over a diurnal cycle the mucus of some corals may cycle from being fully-oxygenated due to the photosynthetic metabolism of Symbiodinium to anaerobic at night when respiration rates exceed oxygenesis—this may be enhanced by physical compartmentalization of mucus-bearing structures on the coral colony (Carlton and Richardson, 1995; Kuhl et al., 1995; Brown and Bythell, 2005).

Although the mucus layer is densely colonized by microorganisms, microscopic examination of the coral epidermis beneath the mucus layer reveals a nearly sterile environment (Johnston and Rohwer, 2007). Surface coral cilia have been implicated in deterring the attachment of fouling organisms by generating surface currents on the order of 2 mm per second (Johnston and Rohwer, 2007; Garren et al., 2014). Although coral mucus and ciliary currents are physical deterrents to colonization of coral surfaces, it has been recently demonstrated that the coral pathogen V. coralliilyticus breaches this barrier through rapid bursts of chemotaxis (i.e., directional swimming relative to a gradient) and chemokinesis (i.e., acceleration in response to a chemical concentration) guided by gradients of infochemicals such as DMSP that are produced by the holobiont and may increase during periods of temperature stress (Garren et al., 2014). Exposure to V. coralliilyticus leads to rapid death of coral tissue (Ben-Haim et al., 2003a; Santos et al., 2011). Thus, the dynamic coral mucus layer is a critical, and traversable, barrier that is essential to maintaining a healthy balance of surface-colonizing microorganisms.

In corals, the intracellular habitat of endodermal (gastrodermal) cells is dominated by Symbiodinium, which are housed in coral generated membrane vesicles i.e., the “symbiosome” as individuals, diads, or triads (Trench, 1983, 1993). The secondary membrane around the algal endosymbiont protects the alga from coral immune responses and creates a microenvironment within the cytoplasm of low nutrients and high pCO₂/low pH conducive to exudation of photosynthate rather than algal biomass synthesis (Trench, 1983, 1993; Venn et al., 2009). To date, studies with detailed ultrastructural imaging of the symbiosome do not report the co-occurrence of bacteria (e.g., Venn et al., 2009; Pernice et al., 2012). Yet obtaining axenic (bacteria-free) cultures of Symbiodinium has proven difficult, suggesting that Symbiodinium associated bacteria may be essential for their sustained growth in culture (Ritchie, 2011). Despite efforts to remove microbial contamination from the culture, analysis of the Symbiodinium genome also revealed the presence of alphaproteobacterial sequences (Shoguchi et al., 2013). This organism matched closely to Parvibaculum lavamentivorans, and to sequences found in seawater samples (Schleheck et al., 2011). Whether Symbiodinium maintains intracellular bacteria is unknown.

The activity of Symbiodinium may shape several intracellular and interstitial physicochemical gradients in the coral holobiont. Reactive oxygen production is associated with the activity of coral phagocytotic immune cells (Mullen et al., 2003) and with oxygenic photosynthesis by Symbiodinium that generates free oxygen radicals in and around coral gastrodermal cells (Kuhl et al., 1995; Banin et al., 2003). Resulting gradients of reactive oxygen within the holobiont may structure microbial populations according to their ability to mitigate oxidative stress through enzymatic activity [e.g., superoxide dismutase (SOD) and catalase]. Virulence of Vibrio shiloi, a pathogen of Symbiodinium in the coral Oculina patagonica required SOD to initiate infection (Banin et al., 2003). While a mutant V. shiloi strain with inactivated SOD could adhere to and penetrate cells of O. patagonica, intracellular survival was impaired, preventing establishment of infection (Banin et al., 2003). These results suggest that high reactive oxygen concentrations proximal to Symbiodinium are a barrier to microbial colonization of the intracellular niche, and may play a role in mediating colonization of microorganisms elsewhere in the holobiont where reactive oxygen is produced.

A second class of physicochemical gradients may be shaped by exudation of organic molecules by intracellular Symbiodinium. In addition to photosynthate, which may represent a source of nutrition for associated bacteria, an increasing diversity Symbiodinium-derived of chemical compounds has been recognized. For instance, Symbiodinium produce microsporine amino acids, which localize to the coral mucus and are hypothesized to protect the holobiont against UV radiation (Banaszak et al., 2000) and to serve as antioxidants (Yakovleva et al., 2004). Symbiodinium also produce DMSP, which is metabolized by a wide variety of microbial taxa (Raina et al., 2009) and as previously discussed, serves as an infochemical for chemotactic pathogens (Garren et al., 2014). How the profile of compounds released by Symbiodinium varies based on clade or sub-clade is not well-known but may influence the composition of the associated microbiota. Coral microbial communities have been shown to differ based on the clade of Symbiodinium present in juveniles of two Acroporid coral species under both stressed and non-stressed conditions (Littman et al., 2009, 2010) supporting the hypothesis that unknown factors associated with different Symbiodinium genotypes influence the coral microbiota composition.

Residence of bacterial populations in intracellular habitats affords close proximity to host- or Symbiodinium-derived ATP and nutrients, and protection from wandering immune cells and predators. In most known cases intracellular occurrence of bacteria in corals has been linked to pathogenesis (e.g., V. shiloi, V. coralliilyticus) (Ben-Haim et al., 1999, 2003a,b; Banin et al., 2003). However, intracellular association of non-pathogenic bacteria has been observed in two Cnidarian classes—the hydrozoans Hydra magnipapillata and Hydra oligactic and the anthozoan coral Montastrea cavernosa (Lesser et al., 2004; Fraune and Bosch, 2007; Chapman et al., 2010) and has been implicated in cultivation-independent surveys of holobiont microbiota composition. The genome sequence of H. magnipapillata revealed a near complete genome for a betaproteobacterial strain (family Comamonadaceae) with some evidence of genome-reduction associated with symbiosis (Chapman et al., 2010). In H. oligactic, stable associations with Rickettsiales were observed in both laboratory and field polyps that were harbored in secondary membranes within epithelial cells suggesting these were endosymbionts (Fraune and Bosch, 2007). In M. cavernosa, a nitrogen-fixing cyanobacterium was found within coral cells (Lesser et al., 2004) and subsequently shown to supplement the nitrogen demand of the host (Lesser et al., 2007). The abundance of Symbiodinium has been positively correlated with the abundance of nitrogen-fixing bacteria (Olson et al., 2009) which may provide a substantial nutritional benefit to all members of the coral holobiont. Association of nitrogen-fixing bacteria with corals have been reported in several other studies (e.g., Chimetto et al., 2008; Lema et al., 2012; Olson and Lesser, 2013). Nitrogen-fixing taxa include various rhizobia closely related to intracellular nitrogen-fixing symbionts of legumes, suggesting rhizobial coral associates may play a similar role (Lema et al., 2012).

Bacterial aggregates associated with coral gastrodermal tissue have been documented in early studies of both reef-building corals and anemones (reviewed in Work and Aeby, 2014). Among 21 coral genera examined, prevalence of cell-associated microbial aggregates (CAMA) was linked to coral genera; notably CAMA were absent in all samples examined from the genus Montipora (Work and Aeby, 2014). Based on histological staining profiles the authors speculated that some of the bacteria resident in CAMA were Rickettsia or Chlamydia—two taxonomic groups containing obligate intracellular species. CAMA are currently not known to be intracellular or extracellular based on current ultrastructural characterization. Widespread association of bacterial aggregates with cells of adult corals, whether intra- or extra-cellular, may point to functional mutualism, and further work to elucidate the species composition and activity associated with these aggregates will shed light on this interesting hypothesis.

The coral gastric cavity and associated lumen form an enclosed environment with properties distinct from surrounding reef waters and other coral tissues. The gastrovascular cavity is exposed to periodic influxes of both nutrients and a rich assemblage of microbial populations borne in seawater, detritus consumed via suspension feeding, and on coral prey items. Microorganisms in ingested detritus or coral prey may represent transiently associated populations and are hypothesized to contribute to the large diversity of microorganisms observed associated with corals (Fernando et al., 2014). Similar to guts of other organisms, the microorganisms in the coral gastrovascular cavity may provide their host with essential nutrients such as vitamins and amino acids, while participating in food digestion. Using fiber optic microsensors in the only study of its kind to date, Agostini et al. (2012), measured the physical properties at different depths in the gastric cavity of G. fascicularis and observed gradients in dissolved O₂, pH, alkalinity, and nutrient concentrations that varied with diel oxygenic photosynthesis of associated Symbiodinium. Elevated levels of vitamin B₁₂, phosphate, and nitrogen species were observed in the gastric cavity, where vitamin B₁₂ was hypothesized to be a nutrient produced by bacteria for coral or Symbiodinium uptake. Dissolved oxygen decreased from the coral mouth to the bottom of the cavity, where consistently low or anoxic conditions prevailed, perhaps creating a permanently anoxic microenvironment (Agostini et al., 2012) (Figure 3) and overall, pH was lower in the cavity compared to overlying seawater. The microbial community in gastric fluids was 100-fold more concentrated than in the surrounding seawater (i.e., >10⁷ cells/ml) and the community composition included ribotypes observed in gut microbiomes of other marine and terrestrial taxa (shrimp, bees, and humans) (Agostini et al., 2012). This data suggests that the coral gastrovascular cavity may be more similar to other animal guts in terms of microbiota structure, and possibly function, than previously expected.

The coral skeleton provides a sheltered environment, offering protection from stressors such as UV light and predation (Shashar et al., 1997). Several studies have examined microbial composition of coral skeletons, finding diverse, and distinct communities of microbial eukaryotes (Shashar et al., 1997), and bacterial communities that contain many of the same bacterial types observed in coral fragments with live tissue (Frias-Lopez et al., 2002; Sweet et al., 2011; Fernando et al., 2014). In a study of bacterial diversity of Mussismilia spp. corals, Fernando et al. (2014) observed that 8.9% of the bacterial 16S rRNA operational taxonomic units (OTUs) from 12 coral individuals were also observed in a bare skeleton sample. Interestingly, skeleton-associated OTUs made up a greater proportion of the bacterial richness for sequence types found in multiple coral samples, up to 77% of the richness of OTUs recovered from at least 11 of the 12 corals (Fernando et al., 2014), suggesting that skeleton-associated microbial populations are a significant source of diversity in the coral holobiont and may contain stable associates. Coral skeletons from three coral species maintained in aquaria revealed bacterial community profiles that did not cluster in a host species-specific manner, in contrast to mucus and tissue samples (Tremblay et al., 2011) raising interesting questions about the extent to which conditions in the coral skeleton are influenced by interactions with the other habitats in the coral holobiont. In 11-day old juvenile polyps of P. astreoides bacterial cells observed by fluorescent in situ hybridization are seen adjacent to tissues surrounding developing coral skeleton (Sharp et al., 2012). The reason for this localization is unclear, though the authors suggest future work should explore whether bacteria may influence deposition of the carbonate skeleton in developing corals. Indeed, the mineral nucleating properties of bacterial cell wall material is well-established and can be both induced in the laboratory, and observed in natural mineral formations (Aloisi et al., 2006).

The activity of the endolithic microbial community is elusive; far fewer studies have focused on skeletal associates than on tissue- and mucus-associated bacteria. Skeleton-associated microbes have been implicated in nitrogen fixation and as a potential source of food and nutrients during bleaching events (Shashar et al., 1994; Le Campion-Alsumard et al., 1995a; Fine and Loya, 2002). Endolithic algae can be observed macroscopically, and blooms are triggered when coral paling or bleaching reduces competition with Symbiodinium for light. Endolithic algae can be recognized as green bands a few millimeters beneath the coral tissue layer (Fine et al., 2006; Carilli et al., 2010). The endolith Ostreobium is distributed among a wide diversity of corals and as a micro-borer can negatively affect the robustness of the coral skeleton (Lukas, 1974; Gutner-Hoch and Fine, 2011). Fungi are also observed in microscopic examination of coral skeletons (Le Campion-Alsumard et al., 1995b). Coral associated fungal assemblages are diverse (Wegley et al., 2007; Amend et al., 2012), can change composition based on environmental conditions (Le Campion-Alsumard et al., 1995b; Bentis et al., 2000; Amend et al., 2012; Yarden, 2014), increase relative to other holobiont members in stressed corals (Vega Thurber et al., 2009; Littman et al., 2011), and in some cases cause coral disease (Bourne et al., 2009). While it is likely that microbial communities in the coral skeleton are distinct from those in tissue-associated habitats due to different physicochemical conditions and localized primary production by endolithic algae, the nature of interactions between endolithic microbiota and the rest of the holobiont is unclear as there are not sufficient studies with both the spatial and taxonomic resolution to distinguish microbial populations associated with live tissue from those associated with healthy coral skeletons.

The microbial communities of several species of corals are hypothesized to harbor stable associates of the coral host, including strains of the nitrogen-fixing cyanobacteria Cyanothece (Lesser et al., 2004, 2007; Olson and Lesser, 2013), alphaproteobacterial strains, such as Roseobacter spp., which are transferred from parent to offspring during reproduction, (Sharp et al., 2010, 2012; Ceh et al., 2013), antimicrobial members persisting in the coral mucus such as Exiguobacterium spp., Photobacterium spp., and Halomonas sp. (Ritchie, 2006; Krediet et al., 2013) and populations such as Endozoicomonas spp. that are found at high proportion across multiple types of corals (Yang et al., 2010; Morrow et al., 2012, 2014; Bayer et al., 2013a,b; Pike et al., 2013; Neave et al., 2014). Resources for maintenance metabolism coupled with favorable physical and chemical conditions are necessary for any organism to persist in a habitat. If the organism is subject to removal (e.g., by predation, or physical dilution) then biomass growth must balance removal rates for persistence. In the coral holobiont, like other ecosystems, competition for resources including light, inorganic nutrients (N, P, and Fe), electron acceptors (O₂, SO₄), and organic carbon mediates the structure of the populations that persist and coexist. Perturbations that relieve resource limitation have both holobiont and ecosystem scale consequences.

Top-down forcing of holobiont structure through immune interactions, predation, and interference-based competition allows certain microbes to grow/persist in the holobiont while others are eliminated. Like all animal tissues, corals represent rich stores of nutrients kept out of bounds from potential microbial consumers by tissue barriers and immune functions. The epithelium and immune system of healthy corals are critical lines of defense to prevent indiscriminant microbial digestion of viable tissues and to allow stable coexistence of coral tissue and a microbiota. In addition, emerging research indicates that other components of the holobiont (i.e., the microbiota and perhaps Symbiodinium) contribute to this stable coexistence through a complex chemical ecology of antimicrobial compounds and signals that modulate microbial activity. Environmental stresses that disrupt coordinated activities of the coral holobiont pose a great risk to coral homeostasis by weakening top-down control of microbiota structure and enabling growth of opportunistic microbes that proliferate at the expense of the coral host.

Variation in physicochemical conditions can have distinct impacts on the coral holobiont community. Among healthy corals the bacterial community composition has been found to vary seasonally (Koren and Rosenberg, 2006) suggesting that shifting ecological drivers (bottom up and top down control) as well as shifting composition of the source assemblages from which immigration occurs leads to a dynamic microbiota that is compatible with a state of homeostasis in the holobiont. In contrast, numerous studies document shifts in coral microbiota following perturbation of physicochemical conditions that may lead to the destabilization and ultimately death of the holobiont (Table 1). Little is known about factors that determine whether the holobiont can maintain homeostasis during, or recover from, physicochemical perturbations. A general trend in coral perturbation studies is that opportunistic pathogens including vibrios and alteromonads increase following exposure to a stressor (Vega Thurber et al., 2009; Littman et al., 2011; Meron et al., 2011) suggesting that activities of virulent microbes may accelerate the impact of a stressor, and contribute to the breakdown of holobiont functioning due to changes in host response or due to pathogenicity (Rosenberg et al., 2007; Bourne et al., 2009).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.