The Earth's climate is strongly influenced by the ocean, that takes up about one third of global carbon dioxide emissions (Sabine, 2004). This uptake depends partly on the biological fixation of carbon dioxide by photosynthetic microbes that need nutrients for their activity and growth. Among these nutrients, nitrogen is considered the proximate limiting nutrient for primary production in oligotrophic waters that cover ca. 70% of the global ocean surface (Falkowski, 1997; Moore et al., 2013). In this sense, dinitrogen (N₂) fixation is recognized as an essential process for maintaining the productivity of ecosystems, as it replenishes the pool of biologically available nitrogen that is lost from the system via anaerobic ammonium (NH4+) oxidation and denitrification (Capone and Knapp, 2007). Coral reefs represent highly productive ecosystems despite often thriving in nutrient-poor tropical and subtropical waters (Odum and Odum, 1955). Their productivity also depends largely on the availability of nitrogen, that is continuously recycled (as NH4+) in the water column and sediments, either via bacterial regeneration of dead organic matter and animal catabolism (Wilkinson et al., 1984; Alongi et al., 2006), or within reef symbiotic organisms (Wang and Douglas, 1999). However, external nitrogen inputs are also required to sustain the high growth rates of reef benthos. There are several sources of external nitrogen supply (Paytan et al., 2006; Gove et al., 2016), and N₂ fixation is considered one of the major potential sources, and can thus play an important role in sustaining reef primary productivity (France et al., 1998; Bell et al., 1999). Only specialized N₂-fixing prokaryotes (so-called diazotrophs) are able to use N₂ because they possess the nitrogenase enzyme, which cleaves the triple bond of the N₂ molecule to form bioavailable NH4+. In coral reef ecosystems, diazotrophs have colonized a wide range of habitats, such as the water column (Riemann et al., 2010; Turk-Kubo et al., 2015), the sediments and dead coral surfaces (Davey et al., 2008; Bednarz et al., 2015b), the algal turfs (Rohwer et al., 2001; Haan et al., 2014; Rix et al., 2015), microbial mats and seagrass meadows (Rohwer et al., 2001, 2002; Charpy et al., 2012; Cardini et al., 2014), as well as sponge and coral tissue (Rohwer et al., 2002; Lema et al., 2014a,b). Still, their contribution to nitrogen cycling in the coral reef ecosystem as a whole, their spatial-temporal variability, level of activity and dependence on environmental factors, are poorly understood.

In the following sections, we review our current knowledge on N₂ fixation activity and diazotrophic diversity in different reef environments, assess the contribution of diazotrophically-derived nitrogen (DDN) to nitrogen nutrition and primary production of coral reefs, and give perspectives for future research with respect to the role of diazotrophs for coral reefs facing climate change.

Coral-associated bacteria were first discovered in the mucus layer (Ducklow and Mitchell, 1979; Ritchie and Smith, 1995). Because the mucus layer is highly dynamic and in constant contact with the surrounding seawater, at the dawning of coral microbiota studies it was unclear if these were opportunistic colonizations and/or interactions with water column bacteria, or if these bacteria truly lived associated to corals. However, subsequent studies found that coral-associated bacteria were more diverse and abundant than those in the surrounding seawater (Rohwer et al., 2001), and also present in other coral microhabitats such as the skeleton (Shashar et al., 1994) and the tissue (Rohwer et al., 2001, 2002), suggesting they are permanent residents despite the nature of their relationship with the host can either be beneficial or not (Rohwer et al., 2002).

In the following years, several studies have reported bacterial communities specific to different coral species (Rohwer et al., 2001, 2002; Frias-Lopez et al., 2002; Bourne and Munn, 2005; Bourne et al., 2007; Sunagawa et al., 2010), while others have reported high intraspecific variability among reef sites (Guppy and Bythell, 2006; Littman et al., 2009b). Recent high-throughput sequencing efforts have revealed that spatially separated corals harbor a core microbiome composed of the same bacterial phylotypes associated with symbiont zooxanthellae, adding to a “floating” microbiome that changes with either coral species, time and/or space (Ainsworth et al., 2015). Differences in the composition of the microbiota of spatially-separated corals of the same species have been attributed to (i) the genotypic variability of the composition of the mucus layer, which chemical composition is thought to structure the colonizing bacterial communities (Ritchie and Smith, 1995, 2004; Ritchie, 2006), or (ii) as a response to changes in the surrounding environment (Klaus et al., 2005, 2007). Corals may adapt to changing environmental conditions by changing their associated microbiota, expelling microbes that do not longer provide any metabolic benefit and replacing them for others that do (the “coral probiotic hypothesis”; Reshef et al., 2006). However, having a variable, environmentally-adapted microbiota does not exclude having also a core, obligate symbiotic microbiota present in the same coral species independently of geographical location and/or environmental conditions experienced (Kooperman et al., 2007; Chen et al., 2010). While the discussion on the species- vs. site-specific coral microbiota has so far been focused on their genetic identity, it has been also proposed that specificity may rather lay on bacterial functional diversity (Krediet et al., 2013). Molecular biology methods in the past decades have provided invaluable insights into the diversity of coral's microbiota, but only recently genomics approaches have started revealing which functional traits lay behind the genetic identity of the microbes inhabiting the coral holobiont. A high degree of species diversity in a given coral species cannot be translated into a high abundance of different metabolic pathways and microbes with different identities may perform the same metabolic functions within a symbiosis (Bourne et al., 2016). The combination of genetic identification molecular biology techniques with metabolic pathway quantification geochemical techniques is needed to elucidate the true role of microbes in the coral symbiosis (see Box 2).

Coral-associated bacteria may not only vary across large scales (e.g., one reef to another, one oceanic region to another), but also within the various microhabitats of the coral holobiont. Just as zooxanthellae adapted to different light intensities are distributed in the proximal-distal parts of the coral (Rowan et al., 1997), bacteria are likely spatially distributed throughout the microhabitats of the coral host (e.g., mucus layer, skeletal matrix, gastral cavity, etc.) according to their ecological function. Such “microzonations” are hard to assess when tissue slurry or other coral homogenates are used to obtain DNA samples, leading to a dramatic loss of information (Bourne et al., 2016).

Coral-associated bacteria can either damage the animal host by causing disease (Rosenberg and Ben Haim, 2002; Krediet et al., 2013), or provide benefits via antibiotic production and substrate acquisition facilitation (Bourne et al., 2016). Coral-associated bacteria have been found to perform various metabolic functions, including having a role in carbon, phosphorus, sulfur and nitrogen cycles (Thurber et al., 2009; Kimes et al., 2010, 2013; Mouchka et al., 2010). In addition, establishing mutualistic relationships with diazotrophs is also a way to obtain nitrogen, if the right conditions are met (see Section Coral Nitrogen Nutrition: Importance of DDN), which likely benefit both the coral host and the zooxanthellae given that both harbor NH4+ uptake genes (Yellowlees et al., 2008). Although corals have several ways to acquire nitrogen, either via heterotrophic feeding (Houlbrèque and Ferrier-Pagès, 2009) or via the uptake of inorganic nitrogen through their symbionts (Grover and Maguer, 2002; Pernice et al., 2012), nitrogen remains a recurrent limiting nutrient for the growth and health of corals (Béraud et al., 2013).

Crossland and Barnes (1976) and Williams et al. (1987) were the first to report N₂ fixation activity in the stony corals Acropora cuminata and Goniastrea australensis, and Acropora variabilis, respectively. These authors found no clear evidence of epiphytic diazotrophic cyanobacteria over the coral colonies, pointing toward an active diazotrophic community living either within the coral tissue as endosymbiotic bacteria or as endolithic bacteria within the coral skeleton. Evidence for an active endolithic community was provided by Shashar et al. (1994). They measured N₂ fixation in Favia favus and found that even after removing the coral tissue, N₂ fixation activity continued in the coral skeleton when supplemented with glucose. In the late 1980s-early 1990s, while evidence of active N₂ fixation in living corals started to accumulate, the identity of the responsible diazotrophs was unknown. Lesser et al. (2004) proposed phycoerythrin and the nifH gene as indicators for the presence of diazotrophic cyanobacteria in the epithelial cells of Montastrea cavernosa. While this coral was subsequently reported to be actively fixing N₂ (Lesser et al., 2007), Oswald et al. (2007) demonstrated that phycoerythrin as analyzed in the previous study can give false-positive signals leading to an overestimation of the density of cyanobacteria in the tissue. By this time, the advent of new molecular biology techniques started to allow more detailed descriptions of the identity of coral-associated diazotrophs, but much work needed (and in still needs) to be done to establish whether these associated diazotrophs were coral species- or site-specific, whether they were permanent residents of the coral holobiont transmitted vertically through generations, or acquired horizontally when needed to confront nitrogen scarcity conditions.

Using a pyrosequencing approach Wegley et al. (2007) described the identity and metabolic function of the bacterial community associated to the coral Porites astreoides sampled in Panama, and found both filamentous and unicellular diazotrophic cyanobacteria from the orders Chroococcales and Nostocales. Olson et al. (2009) reported a wide diversity of diazotrophs in Montipora capitata and Montipora flabellata collected in Hawaii. They found a diverse community including representatives of several proteobacteria classes (Alpha-, Beta-, Gamma- and Deltaproteobacteria), being Gammaproteobacteria the dominant class (~60% of all nifH sequences recovered). The authors developed a quantitative polymerase chain reaction (qPCR) approach to quantify Vibrio-like bacteria (which were the major group within the Gammaproteobacteria nifH sequences recovered), and found that their abundance was positively correlated with the abundance of zooxanthellae. The authors interpreted this correlation as and indication that DDN sustains the productivity of zooxanthellae to some degree, but may also provide fixed DDN to the animal host, concluding that Vibrio-like diazotrophs play a key role in the nitrogen metabolism of the coral holobiont. Nevertheless, Vibrio-like bacteria are well known to induce disease in corals and hence their potential beneficial role for the coral holobiont must be interpreted with caution (Munn, 2015).

In contrast with earlier studies that considered cyanobacteria as the main diazotrophs associated to corals, the results of Oswald et al. (2007), and later Olson et al. (2009), reported a minimal contribution of cyanobacteria to the overall nifH sequences recovered (<5%). Likewise, Lema et al. (2012, 2014b) reported a clear dominance (65–71% of all sequences retrieved) of rhizobia-related bacteria (Alphaproteobacteria) in the tissues Acropora millepora, Acropora muricata, and Pocillopora damicornis, which is the diazotroph responsible for nitrogen nutrition in the roots of leguminous plants. The authors also reported differences in the diazotroph communities inhabiting the mucus layer and the coral tissue, with those in the mucus community being much more diverse and similar to that of the surrounding seawater (including Alpha-, Gamma-, Epsilon- and Deltaproteobacteria, as well as cyanobacteria representatives). This similarity of the mucus microbiota with that of the surrounding seawater is not surprising given the highly dynamic character of mucus and its constant interaction with the pelagic (Brown and Bythell, 2005). Rhizobia-related sequences were also retrieved from the tissues of Montastrea cavernosa in the Caribbean Sea (Olson and Lesser, 2013). The fact that rhizobia sequences could be only retrieved from the coral tissue could indicate a compartmental distribution of diazotrophs within the coral holobiont. For example, legume-associated rhizobia only fix N₂ in oxygen-depleted conditions, thus it is plausible that coral-associated rhizobia only succeed at fixing N₂ when the tissues experience microaerophilic conditions at night (but non-cyanobacterial diazotrophs may fix N₂ actively at a wide range of oxygen concentrations; Bombar et al., 2016). Nevertheless, high concentrations have also been found in the mucus layer of Acropora sp. corals in the New Caledonian lagoon (Camps et al., 2016), which could be either due to (i) an active expulsion of tissue-associated rhizobia toward the surrounding seawater if no longer needed (the “coral probiotic hypothesis”) or to control the growth of zooxanthellae by limiting their nitrogen resources, or conversely (ii) due to a colonization of the mucus from seawater rhizobia, which can at times be abundant in the pelagic compartment (Le Moal et al., 2011). Similarly, while Vibrio-like diazotrophs have been reported to dominate the mucus of Mussismilia hispida (Chimetto et al., 2008), they are also often found in the tissues of other corals such as Montipora capitata (Olson et al., 2009).

How corals acquire their microbial assemblages (zooxanthellae and prokaryotes) is unclear and likely highly variable among species and environments. Both horizontal (environmental) and vertical (genetic) transmission has been reported for zooxanthellae and prokaryotic associates (Harrison and Wallace, 1990; Chen et al., 2010; Ceh et al., 2013b). If one or the other transmission predominates may depend -at least to some degree- on the reproduction strategy of the coral species under study, i.e., brooding vs. broadcasting corals (Apprill et al., 2009; Littman et al., 2009a; Ceh et al., 2013a). Prokaryotes could offer diverse metabolic functions to the coral holobiont, and hence it is plausible that they are mostly acquired horizontally triggered by nutritional needs or environmental stress situations (Bourne et al., 2016), which is supported by the fact that horizontal transmission of bacteria often entails chemotaxis (Chen et al., 2010; Tout et al., 2015) and quorum sensing (Ransome et al., 2013).

The fact that at times cyanobacterial diazotrophs can be present or not (Lesser et al., 2007), and that spatial variability of non-cyanobacterial diazotrophs within a given coral species have been observed (Lema et al., 2014b), suggests that coral-associated diazotroph assemblages can be very dynamic. Nevertheless, the recurrent finding of nifH sequences in corals of different sites and in different seasons of the year, either in the tissues and/or the mucus layer strongly suggests that diazotrophs may important for the coral holobiont, although the actual transfer of nutrients and how it benefits the host remains to be elucidated. The studies conducted so far indicate that diazotrophs are transferred horizontally, but if such transfer only takes place upon nitrogen limiting conditions cannot be currently confirmed (Chen et al., 2010). Lema et al. (2014a) detected rhizobia-related nifH sequences in larva, 1-week old juveniles, and adult specimens (Lema et al., 2012) of Acropora millepora, and also witnessed the horizontal acquisition of diazotrophic ¹⁵N₂-labeled Vibrio bacteria in larva of the same coral species (Lema et al., 2015). These results strongly support that diazotrophs are ubiquitous members of the microbiome, hinting at their possible importance in the nitrogen metabolism of Acropora millepora. Nevertheless, the dominant associated diazotrophs may also vary depending on the health of the coral (Yang et al., 2015). Bacterial diversity decreases as corals grow, and the dominant groups also change with each life stage, likely reflecting a fine tuning of the hosted microbial community upon changes in metabolic needs in different life stages and/or when experiencing environmental stress or disease (Nissimov et al., 2009; Ceh et al., 2013b; Lema et al., 2014b; Santos et al., 2014 see Section Does Diazotrophy Enhance Coral Resilience?).

In summary, coral-associated diazotrophs seem to be diverse, species-specific, and distinct between different coral holobiont microhabitats. While both cyanobacterial and non-cyanobacterial diazotrophs have been identified, non-cyanobacterial groups prevail with rhizobia (Alphaproteobacteria) and Vibrio-like (Gammaproteobacteria) as the dominant types found, which are likely permanent members of the coral microbiome and hence likely play a key role in supplying fixed nitrogen to the animal host and/or the zooxanthellae. Nevertheless, the mere detection of nifH sequences in corals needs to be interpreted with some caution. Firstly, PCR-based approaches may artificially amplify the presence of a given phylotype vs. others, which cannot be directly translated into abundance. Not even using quantitative approaches (e.g., qPCR) one can assess the abundance of diazotrophs accurately, since more than one nifH copy can be found per cell. Finally, the presence of nifH genes in a given coral species does not mean the associated diazotrophs were actively fixing N₂. Function-diversity approaches are needed to elucidate which diazotrophs are actively performing N₂ fixation in corals and in which condition this newly fixed nitrogen provides profits to the corals. Indeed, while a recent study using ¹⁵N₂ (Benavides et al., 2016) showed that DDN is transferred to the zooxanthellae of corals in New Caledonia, another study (Grover et al., 2014) did not find any DDN incorporation in the symbionts or in the tissue of corals in the Red Sea, despite N₂ fixation took place in the surrounding seawater.

Despite covering a small surface on a global basis, coral reefs comprise ~25% of all marine species (Connell, 1978), and are considered the most diverse ecosystems on Earth. The health and expansion of coral reef ecosystems is currently threatened by global warming, sea level rise, ocean acidification, eutrophication and overfishing (Knowlton, 2001; Hughes et al., 2003). Adding to predicted global temperature increase scenarios, contemporary temperature increments as thus induced by El Niño events are currently provoking massive coral bleaching events (Hoegh-Guldberg and Ridgway, 2016; e.g., Shashar et al., 1994).

Coral reef loss does not only have dramatic environmental consequences, but also can severely reduce the economic benefits these ecosystems provide to their riparian countries (i.e., fisheries, coastal protection and tourism), estimated at $375 billion per year (Constanza et al., 1997). As introduced earlier, the growth and health of reef ecosystems does not only depend on the zooxanthellae-host symbiosis, but also on a variety of bacterial and archaeal microorganisms that perform important metabolic functions that benefit the animal host (Bourne et al., 2016). To date, the dynamics of coral-associated microbiota are not well understood, and it is likely that an adaptive microbiota enhances the resilience of corals facing environmental stress (Ainsworth et al., 2010; Garren and Azam, 2012; Bourne et al., 2016).

Environmental stress factors such as bleaching, thermal stress, irradiance, disease or changes in organic matter may disrupt the symbiotic relationships of the coral holobiont, often resulting in a change of the coral-associated microbiota genetic and/or functional diversity (Ritchie and Smith, 1995; Kline et al., 2006; Smith et al., 2006; Thurber et al., 2009; Garren and Azam, 2011; Krediet et al., 2013). This change can be either direct (the environment affects the microbiota) or indirect (the environment affects the coral host physiology, which in turn forces a change in its associated microbiota Ritchie, 2006; Mouchka et al., 2010). The destabilization of the coral holobiont symbioses can promote a shift of its former, healthy state microbial communities, for opportunistic and pathogenic microbes (Thurber et al., 2009; Bourne et al., 2016).

When inputs of zooxanthellae-derived fixed carbon are shutdown, bleached corals rely on alternative energy sources such as storage reserves or heterotrophic feeding (Grottoli et al., 2006). While the degree to which bleached or in recovery corals rely on reserves or heterotrophic feeding to recover varies among species, it has been reported that they can meet up to 100% of their carbon metabolic requirements via heterotrophic feeding (Grottoli et al., 2006; Houlbrèque and Ferrier-Pagès, 2009). The recent finding that Stylophora pistillata corals fulfill an important part of their nitrogen needs via heterotrophic consumption of planktonic diazotrophs (Benavides et al., 2016) raises the question whether this source of nitrogen acquisition would also be enhanced to face environmental stress, as observed in other coral species that increase heterotrophic feeding when the translocation of carbon fixed by zooxanthellae is disrupted due to bleaching. Indeed, the loss of zooxanthellae during bleaching events does not only imply a loss of fixed carbon availability to the coral host, but also a decrease in its nitrogen acquisition capacity (Ferrier-Pagès et al., 2010; Godinot et al., 2011; Ezzat et al., 2016). In such events, corals' resilience could be enhanced by alternative fixed nitrogen sources such as that provided by endosymbiont diazotrophs. For example, Santos et al. (2014) reported that the abundance and diversity of diazotrophs associated with the coral Mussismilia harttii increase in response to +2 to +4.5°C above in situ background seawater temperature. In the Red Sea coral-associated N₂ fixation rates have been reported to increase 3- to 10-fold when corals are artificially exposed to higher temperatures (Cardini et al., 2016b) as well as in response to natural rising temperatures in the summer, suggesting that DDN could provide alternative nitrogen sources when environmental nutrient concentrations are scarce (Cardini et al., 2015). While there is the possibility that DDN increases the coral's bleaching resilience or facilitates the recovery from bleaching events, Rädecker et al. (2015) raised a contrary hypothesis. They suggested that increased N₂ fixation in corals exposed to elevated temperatures could instead enhance their bleaching susceptibility as elevated DDN availability would cause an imbalanced nutrient supply (i.e., increased N:P ratio) to the coral host that would disrupt the nitrogen limitation of the zooxanthellae and subsequently imbalance the coral-zooxanthellae symbiosis (D'Angelo and Wiedenmann, 2014). These contrasting hypotheses highlight the urgent need to better understand the role and underlying mechanisms of DDN involved in coral bleaching. Furthermore, it has been suggested that coral bleaching followed by coral mortality can affect the nitrogen budget of the reef ecosystem due to the overgrowth of coral skeletons by epilithic communities including diazotrophs. Davey et al. (2008) demonstrated that N₂ fixation rates by epilithic communities peaked two weeks after coral mortality with up to 30-times higher rates compared to living corals. This would provide a “new” autochthonous source of nitrogen after coral mortality that may influence the biomass and nitrogen cycles on the reef. Whether this nitrogen is sufficient to reinforce a phase-shift from a coral- toward an algae-dominated reef warrants further investigation that can help managing coral reefs affected by bleaching.

Rising seawater temperatures will likely promote seawater column stratification, hindering vertical mixing and consequently impoverishing nutrient concentrations in surface waters (Passow and Carlson, 2012). In such a scenario, it is likely that nutrient-limited corals will increasingly rely on DDN to meet their nitrogen metabolic requirements. Conversely, increased pCO₂ has been reported to decrease coral-associated N₂ fixation, suggesting that nitrogen starvation of the coral host may exacerbate the long-term effects of ocean acidification (Rädecker et al., 2014). Although we are just beginning to understand the effect of single stressors on N₂ fixation, additional multiple stressor studies are needed to assess synergistic or antagonistic climate change effects on coral-associated N₂ fixation and its responsible diazotroph communities.

Coral show species-specific degrees of resilience to increased temperatures mostly due to intrinsic differences (phenotypic and genetic) among the coral hosts and their microalgal symbionts (reviewed in Hughes et al., 2003). Also, the coral-associated diazotroph community may respond in very different ways to climate change effects such as acidification or temperature (e.g., Gradoville et al., 2014). Conversely, increased anthropogenic nitrogen loading in coastal areas may act as a negative feedback on coral-associated diazotrophs, inhibiting their N₂ fixation activity (Cardini et al., 2015). All in all, it is difficult to predict how coral-associated diazotrophs' diversity and N₂ fixation potential will respond to climate change. Clearly, a more comprehensive understanding of how diazotroph-coral symbioses function in healthy conditions is needed before we can predict responses to different climate change and anthropogenic pressure scenarios (Bourne et al., 2016).

At the ecosystem level (Box 2), there is a significant lack of knowledge on the nitrogen budget in reefs, and on the extent of the importance of its different pathways such as N₂ fixation, nitrification, denitrification and anammox. This is mostly due to the unknown contribution of the different reef symbioses to these processes. For example, it has only recently been demonstrated that sponges can be major players of nitrogen cycling in reefs, together with other planktonic symbioses such as those of diatoms and protists (Fiore et al., 2010; de Goeij et al., 2013). In addition, the impact of environmental changes, such as water acidification, warming and eutrophication is still unknown, despite the fact that all factors directly influence N₂ fixation and can have antagonistic effects. For example, increased acidification can enhance the nitrogenase activity and thus N₂ fixation of some organisms such as Trichodesmium (Barcelos e Ramos et al., 2007), but also poses threats to calcifying organisms (Orr et al., 2005), some of which are hosts to N₂-fixing symbionts (Thompson et al., 2012). Rising sea surface temperatures also increase N₂ fixation by diazotrophs in sediments or rocks, but leads to coral bleaching, which in turn induces a change in the associated microbial communities (Thurber et al., 2009).

At the organism level, the main pending questions taking the coral holobiont as a model are also highlighted in Box 2. We should first focus on focus on better characterizing of the existing symbiotic diazotrophs, using advanced molecular techniques such as metagenomics and metatranscriptomics. This last tool can provide information on the microbial metabolic functions present in addition to their phylogenetic identity, providing a first link between biodiversity with functional investigations. However, transcriptomics results do not provide information on the molecular composition of the substrate being synthesized or metabolized, or at which rate these processes happen. To gain this information genomics and transcriptomics should be combined with metabolomics. Metabolomics is the scientific discipline that characterizes the metabolites (or low molecular weight dissolved organic matter produced by a given microbe. The metabolome is the result of a metabolic function, encoded in the microbe's DNA, expressed through messenger RNA to synthesize proteins, which are then used to metabolize substrates and synthesize metabolites which are eventually released out of the cell, and used for a variety of ecological functions (e.g., nutrient acquisition, cell signaling, species competition) (Kujawinski, 2011). A remaining challenge however is to accurately separate the DNA/RNA/protein fractions of the different constituents of the holobiont (prokaryotes, zooxanthellae and host tissue), which is not easy since mitochondrial DNA often dominates the DNA dataset (Vega Thurber et al., 2008; Littman et al., 2011). There is thus a critical need for the development of coral-specific ’omics protocols (Bourne et al., 2016).

Another challenge will be to locate the diazotrophs within the different compartments of the symbiosis and assess which of these diazotrophs contribute most importantly to the DDN production in the symbiotic association. For example, the ubiquitous presence of diazotrophs in both the mucus layer and the skeleton of most coral species surveyed, and within the tissue of some other species suggest a high functional relevance of these groups for the coral health. However, the extent to which these diazotrophs contribute to the nitrogen needs of the coral host and its associated zooxanthellae is still unknown. To achieve this localization, however, a substantial attention will have to be paid in separating the different compartments of the symbiosis, i.e., mucus, tissue, zooxanthellae, skeleton. A direct localization of the diazotrophs can be achieved using specific primers and the FISH technique. Finally, the advent of single-cell isotope mass spectrometer technology such as nanoSIMS (nanoscale secondary ion mass spectrometry), now allows to detect ¹⁵N enrichment (from ¹⁵N₂ fixation) in different compartments of the symbiosis without physically separating them prior to analysis (Pernice et al., 2014), an approach that promises elucidating how nitrogen is transferred between different coral microhabitats, as well as how nitrogen is used by the different partners (Pernice and Levy, 2014).

Finally, all the classical and newer techniques should be applied in combination to address the effect of climate change stressors and eutrophication on coral reef organisms but also to assess what is the role of diazotrophs for the functioning and resilience of the coral symbiosis facing climate change. As reviewed in the above sections, stressors may indeed have contrary effects on the diazotroph activity/abundance and the animal host and a better knowledge on the effect of each stressor on the symbionts, host and the whole association is needed.

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

Funding was provided by the Centre Scientifique de Monaco.