Recovery of MAGs from A. kenti
A combination of microbiome-enrichment and metagenomics (118 Gbp post-QC, Figure S1; see Methods) was used to recover an unprecedented number of coral-associated MAGs from A. kenti of the inshore GBR (n = 22). In total, 170 A. kenti MAGs displaying quality scores ≥ 50 (completeness – 3 × contamination 30) were identified, resulting in 102 ‘medium’ to ‘high’ quality MAGs (91.2 ± 7.12% mean completeness, and 2.03 ± 2.16% mean contamination), following dereplication with additional filtering. These MAGs spanned 12 bacterial and 1 archaeal phyla including genomes from members of the Proteobacteria (n = 46), Cyanobacteria (n = 18), Bacteroidota (n = 14), Desulfobacterota (n = 7), Verrucomicrobiota (n = 5), Actinobacteriota (n = 3), Firmicutes (n = 3), Chlamydiota (n = 1), Desulfobacterota_D (n = 1), Firmicutes_A (n = 1), Nitrospirota (n = 1), Planctomycetota (n = 1), and Thermoproteota (n = 1; formerly, Thaumarchaeota), representing 60 families and 87 genera (Fig. 1A-B; Table S1). This is consistent with previous 16S rRNA gene sequencing studies which have identified these lineages in A. kenti 31 and are also in-line with the dominance of these phyla represented by cultured isolates of coral-associated bacteria more broadly 13. However, MAGs from novel putative intracellular coral symbionts including Chlamydiota (Simkania sp.), Firmicutes (Mycoplasmataceae), and non-photosynthetic Cyanobacteria of the order Gastranaerophilales, which are rarely observed within scleractinian corals 32–34, were also identified. In samples collected from the surrounding seawater (n = 6), which represented a free-living microbial community ‘control’ for phylogenetic and functional comparisons, 182 MAGs were recovered, with 99 MAGs remaining following dereplication (86.7 ± 7.25% mean completeness, and 2.89 ± 2.31% mean contamination; Fig. 1A-B; see Supplementary Note 1). The identified MAGs were estimated to represent 54% and 32% of microbial genera across the A. kenti and seawater communities, respectively (Fig. 1C). Investigation of the prevalence and relative abundances of these MAGs demonstrated their heterogeneity across the two sample types (Fig. 2A) and indicated that 20 of the A. kenti-derived MAGs were more abundant within the surrounding seawater (hereafter, seawater-specific; Fig. 2B). While the remaining 82 A. kenti MAGs were specific to the A. kenti holobiont (hereafter, A. kenti-specific; Fig. 2C; see Supplementary Note 2). Among these, Proteobacteria dominated in all but one of the A. kenti samples (Russell Island replicate 2), and collectively comprised up to ~ 90% of the total mapped reads (relative abundance) of MAGs in each sample. The relative abundances of individual A. kenti-specific MAGs ranged from a minimum of ~ 0.01% (e.g. novel Flavobacteriaceae MAG, Magnetic_MAG16), to a maximum of 70.5% of mapped reads (e.g. Chlorobium_A marina, Russell_MAG40), with all MAGs displaying site and sample specific differences in relative abundance. Collectively, the relatively high recovery of genus-level representatives (Fig. 1C) specific to A. kenti and consistency of the MAGs with previous literature, indicated that many of the key microbial members of the A. kenti colonies were recovered, allowing metabolic reconstruction of the A. kenti microbiome.
A. kenti - and seawater-specific MAGs display distinct genomic features
Statistical comparison of the A. kenti- and seawater-specific MAGs genomic features revealed distinct characteristics and functions providing insight into the coral-associated ecological niche. Overall, the A. kenti-specific MAGs displayed significantly larger genome sizes, lower GC content, greater numbers of predicted genes, and reduced coding density, despite a greater variation relative to those specific to seawater (Fig. 3A-D). This perhaps reflects the breadth of niche types experienced by microorganisms within the more nutrient-rich coral holobiont compared to seawater 32. For instance, including stable and specific symbionts, some with reduced genomes, in the skeletal matrix and tissues 35, high density coral-associated microbial aggregates (CAMAs) within tissues 34,36, epibiotic and intracellular bacteria associated with the Symbiodiniaceae 37, and a combination of stable and opportunistic symbionts within coral mucus 38 (see Supplementary Note 3). Consistent with the overall genomic differences between the A. kenti- and seawater-specific MAGs, significant differences were observed in the presence-absence of CAZy (PERMANOVA, F = 7.9, P-adj. < 0.01), KO (F = 11.8, P-adj. < 0.01), and Pfam (F = 14.3, P-adj. < 0.01) gene annotations (Fig. 3E-G). However, phylogeny was responsible for much of the variation between the recovered MAGs, regardless of their habitat type (Fig. 3E-G).
Molecular mechanisms indicative of a host-associated lifestyle
Host colonisation
Functional enrichment analysis identified 24 CAZy, 1,666 KOs, and 1,950 Pfams that were significantly enriched in A. kenti-specific MAGs, compared with 7 CAZy, 454 KOs, and 467 Pfams that were enriched in the seawater-specific MAGs (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Figure S2A-C). Exploration of the significantly enriched functional genes indicated adaptations of the A. kenti microbiome to the host-associated niche and revealed potential mechanisms underpinning the establishment and maintenance of the A. kenti-microbiome symbiosis. For example, genes encoding bacterial chemotaxis and motility (e.g. che, fli, and flg operons, motA, and flaG), pili and fimbriae (e.g. pilIGLNUVZ, fimTV), and polysaccharide biosynthesis as part of self-encapsulation (alginate export, PF13372) or biofilm formation (exoP/vpsO, vpsNM, epsC, pelACEFG, gacSA), were significantly enriched in the A. kenti-specific MAGs and may help these bacteria to sense and navigate towards coral chemical cues and establish symbiotic relationships 39,40 (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). These significantly enriched motility and biofilm formation genes were present in ~ 86% (n = 71) of the A. kenti-specific MAGs, with 12 MAGs classified to the Gammaproteobacteria encoding many (70%-86%) of the identified genes associated with these phenotypes, including Endozoicomonas sp. (Russell_MAG12, Fitzroy_MAG7, Pandora_MAG22, Pelorus_MAG18), Ferrimonas sp. (Russell_MAG43), Thalassotalea sp. (Pandora_MAG24), and Alcanivorax sp. (Dunk_MAG1, Fitzroy_MAG14, Magnetic_MAG29, Pandora_MAG5, Pelorus_MAG7, Russell_MAG45). Although chemotaxis can be employed by coral pathogens 41, the ability to navigate towards a coral host may also provide a competitive advantage for commensal microorganisms 39. Indeed, these gammaproteobacterial genera have previously been identified within the A. kenti microbiome using 16S rRNA gene sequencing, and/or that of other coral species from the Indo-Pacific 27,31,42, suggesting that phenotypes and behaviours such as chemotaxis, motility, and biofilm formation likely facilitate the success of coral-associated symbionts across hosts.
Host immune system evasion
Following initial host colonisation, proliferation within a coral host may be enhanced by the ability of the microbiome to evade the host's immune responses and modulate host cell metabolism 43,44. Indeed, several eukaryote-like repeat protein domains (ERPs), such as ankyrin repeats, a leucine rich repeat, a WD40-like repeat, and tetratricopeptide repeats, involved in immune response evasion in both pathogens and symbionts were enriched in A. kenti-specific MAGs relative to those specific to the seawater (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Figure S2C). In total, 96% (n = 79) of A. kenti-specific MAGs encoded at least one ERP, with the highest number of annotations of any given ERP (PF12796, Ank_2) observed in three MAGs of the Endozoicomonas sp. (Magnetic_MAG22, Fitzroy_MAG17, Pandora_MAG11) at 85, 78, and 74 copies, respectively. Moreover, protein families implicated in cell-cell interactions in eukaryotes 45–47 were significantly enriched in A. kenti-specific MAGs, such as signalling pathways (START domain, PF01852; and phospholipase D, PF00614), and eukaryotic cell attachment, integrity, and stability domains (PF10116; PF00028; PF01875; cardiolipin formation, clsA_BC; Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). In addition, protein families implicated in bacterial virulence, such as intracellular growth, and Type III, IV, and VI secretion systems that function in the injection of effector proteins into host cells 48, were significantly enriched in A. kenti-specific MAGs (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Figure S2C). These secretion systems were encoded by 47% (n = 39) of A. kenti-specific MAGs of the Bacteroidota, Chlamydiota, Desulfobacterota, Firmicutes, Proteobacteria, and Verrucomicrobiota (Table S2). Collectively, such mechanisms may facilitate the establishment of dense areas of microbial growth within CAMAs in hospite 36, and may regulate the intracellular transport of key molecules between members of the holobiont 45. Indeed, the prevalence of these genes within the Endozoicomonas sp. MAGs which are known to predominate CAMAs36, and other intracellular microorganisms including novel Mycoplasmataceae, provide evidence for the specialisation of these lineages within the A. kenti holobiont. Taken together these findings indicate that members of the A. kenti microbiome may establish and maintain symbioses through evasion of phagocytosis, binding to host cells, and potentially engaging in direct and indirect host cell modification. Elucidation of the exact nature of these mechanisms at the molecular level is required and may provide insights into the resilience or dysbiosis of coral holobionts in the face of anthropogenic stress.
Viral defensive mechanisms in the A. kenti microbiome
In addition to interactions with coral host cells, genes significantly enriched in A. kenti-specific MAGs also indicated dynamic interactions between the microbiome and bacteriophage or foreign DNA. For instance, proteins associated with lysogenic phage, CRISPR-associated proteins (CAS1, 2, 5, and 7), toxin-antitoxin systems, and DNA binding and repair mechanisms were significantly enriched in A. kenti-specific MAGs relative to those specific to seawater (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Figure S2C). Overall, 90% (n = 74) of A. kenti-specific MAGs across all phyla encoded protein families associated with viral defence mechanisms. In particular, the phage CI repressor protein was present in 30 of the A. kenti-specific MAGs spanning 5 phyla, suggesting lysogenic phage may be prevalent within this holobiont. This is in support of the observation that lysogeny dominates over viral lysis in high microbial biomass environments such as coral mucus 49. Previous studies have revealed the extensive diversity of viruses infecting all members of the coral holobiont, and at times contributing to the maintenance and decline of reef health 50–52. Our results suggest that the A. kenti microbiome has a greater requirement for immunity against viruses and foreign DNA compared to microorganisms of the surrounding seawater, similar to previous observations within the sponge holobiont 53. A. kenti MAGs may evade viral infection through a range of molecular mechanisms, facilitating persistence within the holobiont. Interestingly, viral auxiliary metabolic genes implicated in cyanobacterial photosynthesis and encoding putative antioxidants have previously been observed in the A. kenti holobiont 54, thus our findings of bacteria-phage interactions as a discriminate feature of the A. kenti microbiome suggests this is an important avenue of future research.
Holobiont health during stress
The role of the microbiome in the maintenance of coral health has become increasingly apparent in recent years 55–57, correspondingly several genes which may support the holobiont during times of stress were significantly enriched within the A. kenti-specific MAGs. Among these were genes encoding the antioxidants catalase, glutathione synthase, superoxide reductase and superoxide dismutase, in addition to an oxidative-repair enzyme (msrA/B, msrPQ), and a heat shock protein (hslU) (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). All A. kenti-specific MAGs encoded at least one, to a maximum of 6, of these antioxidants, suggesting widespread potential to mitigate damage from reactive oxygen species within the A. kenti microbiome. This may reflect the relatively harsh environment experienced by microorganisms associated with corals, which are exposed to daily fluctuations in oxygen and UV-light levels under normal holobiont homeostasis 58. On the other hand, microbiome encoded antioxidants may help to mitigate cellular damage due to thermal stress, as has previously been shown in cultured Symbiodiniaceae-associated bacteria 58,59. Further exploration of the expression and action of the identified antioxidants in hospite is required to determine their overall impact on microbiome persistence and in the maintenance of A. kenti holobiont health.
Nutrient cycling contributing to A. kenti holobiont function
Coral mucus metabolism
Several genes significantly enriched in the A. kenti-specific MAGs belonged to key nutrient cycling pathways within the holobiont, providing another example of specialisation between the A. kenti- and seawater-specific microorganisms. This included the metabolism of polysaccharides and photosynthates common to coral mucus 59,60, such as transcriptional regulators for the uptake and catabolism of arabinose (PF12625, PF14525, PF06719), CAZy families targeting alginate (PL6), mannan (GH130), fucose (GH29, GH141), starch and glycans (GH15, GH13_38), and genes encoding the metabolism and transport of the photosynthates aspartate (nadB, asnA, asd, aspA, panD, pcm) and maltose (maa, malEFG; Fisher’s exact test, P-adj. ≤ 0.05; Table S2; Figure S2A-C). These genes were collectively encoded by 99% (n = 81) of the A. kenti-specific MAGs, with the Bacteroidota, Desulfobacterota, Proteobacteria, and Verrucomicrobiota encoding the greatest total numbers. One Draconibacterium sp. MAG (Russell_MAG41; Prolixibacteraceae, Bacteroidota) appeared to be a specifically adapted to this niche, encoding 73% of the identified genes associated with coral-mucus metabolism. This organism is a facultative anaerobe, suggesting potential adaptation to the oxygen-variable niche of the surface boundary layer to exploit localised polysaccharides and photosynthates 58,61. Collectively, these results reflect the notion that coral mucus represents a rich source of organic compounds within the oligotrophic reef seascape that is an important driver of holobiont community structure at the coral-seawater interface 62.
Nitrogen metabolism
Nitrogen availability is considered a primary determinant of Symbiodiniaceae population density and controls the release and translocation of photosynthates to the coral host 63, as such nitrogen fixation is believed to be an important function mediated by the coral microbiome 64. Key genes encoding nitrogen fixation (nifHDK and nifENB) were significantly enriched within 11% (n = 9) of the A. kenti specific-MAGs, of the Desulfobacterota (n = 5), Bacteroidota (n = 2) and Proteobacteria (n = 2; Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Fig. 4). Ammonia derived from nitrogen fixation, or remineralised nitrogen, could provide an energy source for the microbiome through the sequential steps of nitrification 65. However none of the MAGs encoded complete nitrification, although the Thermoproteota JACEMX01 sp. (Pandora_MAG19) encoded ammonia monooxygenase subunit A (amoA) and the novel Nitrospirota Bin75 sp. (Magnetic_MAG5) encoded nitrite oxidoreductase (nxrAB), reflecting previous functional gene studies of coral nitrification 66. Nitrite can be utilised by other holobiont members for assimilation or energy conservation, with three different nitrite reductases (nrfAH, nirBD, nirK) of the dissimilatory nitrate reduction to ammonia (DNRA) and denitrification pathways significantly enriched in the A kenti-specific MAGs (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Figure S2B). Consistently, nitrate reductases of DNRA, denitrification, and the assimilatory nitrate reduction to ammonia (ANRA) pathways (napAD, narGHI, and nasA, respectively) were also significantly enriched (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2; Figure S2B). Of the MAGs encoding assimilatory pathways, Nitratireductor aquibiodomus (Pandora_MAG31; Rhizobiaceae, Proteobacteria) displayed prevalence across 60% of A. kenti samples at relative abundances between 0.1–7.4%, suggesting a key role in holobiont nitrogen cycling. The greater number of lineages with complete assimilatory as opposed to reduction pathways (Fig. 4), suggests low nitrogen levels and limited nitrogen loss from the A. kenti holobiont. Although the distribution of several of these lineages across the A. kenti colonies was patchy (Fig. 2; see Supplementary Note 4) and functional validation of these largely anaerobic processes is required, this would be in-line with previous observations of low denitrification rates in corals relative to assimilatory pathways or nitrogen fixation 65. Thus the activity of these MAGs may contribute to A. kenti holobiont homeostasis, by retaining nitrogen within the system 10,67.
Autotrophic carbon fixation
Acroporid corals are autotrophic, and therefore primarily dependent upon carbon derived from Symbiodiniaceae photosynthesis 65. Here, four autotrophic carbon fixation pathways were also identified within the A. kenti-specific MAGs spanning 16 families from 5 different phyla (see Supplementary Note 5). These included the reverse tricarboxylic acid cycle, the modified hydroxypropionate-hydroxybutyrate cycle, the Calvin-Benson-Bassham cycle, and the Wood-Ljungdahl (WL) pathway, suggesting prokaryotic carbon fixation may supplement the holobiont carbon pool (Fig. 4), as has recently been suggested for the P. lutea holobiont 67. Of these, genes encoding key steps within the WL pathway (cooSF, cdhED, and acsE) were significantly enriched within the A. kenti-specific MAGs compared to those specific to seawater (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). The complete WL pathway was restricted to MAGs of the Desulfobacter sp. (Fitzroy_MAG3, Fitzroy_MAG4, Magnetic_MAG13) and Desulforhopalus sp. (Pandora_MAG12), all of which encoded the key WL enzyme carbon monoxide dehydrogenase. It is notable that 6 of the A. kenti-specific MAGs, including Desulfobacter sp. and Desulforhopalus sp., in addition to Chlorobium_A marina and Ferrimonas sp., encoded two autotrophic carbon fixation pathways. This is unusual in both symbiotic and free-living bacteria and has been suggested as a mechanism to increase the efficiency of carbon fixation, for example within sulfur-oxidising gammaproteobacterial symbionts of marine tubeworms 68. Redundancy in autotrophic carbon fixation pathways may provide a functional advantage if fixed carbon becomes limiting during times of stress 67. In the future, the expression of these pathways and quantification of their rates in situ are required to validate the importance of microbiome contributions to the A. kenti carbon pool.
Sulfur cycling
Sulfate is unlikely to be limiting nutrient within the coral holobiont, as it is readily available within marine waters and can be assimilated by the host and Symbiodiniaceae 69, and is often utilised by marine microorganisms as a source of sulfur through its assimilatory reduction to sulfide 70. Nevertheless, genes encoding extracellular sulfate transport (cysPUWA), and key steps within assimilatory sulfate reduction (ASR; e.g. cysD, cysH, cysJI) and dissimilatory sulfate reduction pathways (DSR; e.g. aprAB, dsrAB) were significantly enriched in the A. kenti-specific MAGs relative to those specific to seawater (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). This suggests that differences in sulfur utilisation or availability may occur between the A. kenti and seawater niches. As an example, methylated sulfur compounds may represent preferential sulfur sources in hospite 71 as the metabolism of reduced sulfur compounds is less energetically expensive 72. Indeed, Acropora sp. are known to contain high concentrations of methylated sulfur compounds, such as dimethylsulfoniopropionate (DMSP) which is produced by both the animal and Symbiodiniaceae 73,74. Within the genes significantly enriched in A. kenti-specific MAGs were those encoding the anaerobic dimethyl sulfoxide (DMSO) reductase enzyme (dmsABC; Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2), which catalyses the reduction of DMSO to the climate active gas dimethyl sulfide (DMS). Moreover, phylogenetic placement of genes involved in DMSP lysis via the cleavage (ddd genes) and demethylation (dmd genes) pathways, identified putatively functional dddD (n = 14), dddL (n = 3), dddP (n = 6), and dddW (n = 8) in 28 A. kenti-specific MAGs of the Proteobacteria (n = 27) and Desulfobacterota (n = 1; Table S3). Notably, the dddD gene, which converts DMSP to DMS, acetate, and 3-hydroxypropionate 75, was present in MAGs recovered independently from each sampling location. All 9 Endozoicomonas sp. MAGs encoded dddD, extending recent reports of the potential for DMSP metabolism by cultured strains of Endozoicomonas acroporea 9. DMSP demethylation to methanethiol was encoded by 10 MAGs of the Gamma- (n = 6) and Alphaproteobacteria (n = 4), including representative MAGs from each sampling site (Table S3). Three MAGs encoded genes for both DMSP cleavage and demethylation, including Gammaproteobacteria belonging to the QNFE01 family (Magnetic_MAG15) and Luminiphilus sp012270045 (Pandora_MAG34), and CABZJG01 sp. within the Rhodobacteraceae (Pelorus_MAG20). Interestingly, this suggests these taxa may be able to switch between the two pathways in a concentration dependent manner 76,77, or in response to changing environmental conditions 78. Functional validation of these genes is required to truly define the role of these microorganisms in modulating the production of DMS within the A. kenti holobiont.
Essential amino acid and vitamin metabolism
Scleractinian corals display distinct patterns of essential and non-essential amino acid metabolism 79,80. Functional annotation of A. kenti and C. goreaui proteins demonstrated incomplete pathways for a range of amino acids, including histidine, tryptophan, and lysine, in addition to vitamins, such as thiamine, riboflavin, biotin, and cobalamin (Fig. 4). Instead, these pathways were widespread across the A. kenti-specific MAGs, spanning a diversity of different phyla (Fig. 4). Moreover, significantly enriched functional genes within the A. kenti-specific MAGs included those encoding the biosynthesis of the polyamine spermidine (nspC and speE), and the amino acids tyrosine/phenylalanine (pheA, tyrA), tryptophan (trpAB), and serine (sdaAB; Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). In addition, genes encoding five B-vitamins, namely thiamine (B1; thiCDEGHLM, THI45), nicotinate (B3; nadABCD, pncB, ushA, yrfG), pantothenate (B5; e.g. panCD), biotin (B7; bioABMN), and folate ((B9; e.g. folAK) were also enriched in the A. kenti-specifc MAGs (Fisher’s Exact Test, P-adj. ≤ 0.05; Table S2). These findings suggest that amino acid and B-vitamin metabolism are an important functional trait of the A. kenti microbiome that may be essential to support the nutrient requirements of the A. kenti holobiont. This is highlighted by the absence and incomplete nature of these pathways in A. kenti and C. goreaui (Fig. 4), and across several other coral host and Symbiodiniaceae genomes 18,81,82.
Latitudinal heterogeneity in the A. kenti microbiome underpinned by water quality
Significant differences in microbial community composition, and the associations between environmental variables underpinning community-level and functional gene dissimilarity, were explored using constrained ordination and functional gene enrichment. Prior to statistical analyses, additional dereplication was performed to group closely related MAGs recovered independently from the different sites (see Methods). This resulted in 63 A. kenti-specific and 49 seawater-specific MAGs, which displayed significant differences in community composition between them, and across the inshore reefs (P = 0.001, F = 6.06, n = 28; Goodness of fit, r2 = 0.78, P = 0.001; Table S4). Up to ~ 80% of the variation in MAG community structure was driven by the separation between the seawater and A. kenti samples, followed by the latitudinal influence of the northern (Dunk Island, Russell Island, and Fitzroy Island) and southern (Magnetic Island, Pandora Reef, and Pelorus Island) sampling gradients (spanning poor to improved water quality, respectively; Fig. 5A). Significant differences in A. kenti community composition were driven by the water quality categories of these sites (P = 0.001, F = 2.3, n = 22; Goodness of fit, r2 = 0.68, P = 0.001; Table S5), primarily reflecting the differences between the North Marine (Fitzroy and Russell Islands) and degraded Coastal (Magnetic Island) reefs (Fig. 5B; see Methods). Post-hoc testing identified the North Marine and Coastal A. kenti microbiomes as being significantly distinct to each other (Table S5). While the North Marine communities were also significantly distinct to the degraded North Plume (Dunk Island) and less-impacted South Marine (Pelorus Island) samples (P-adj. < 0.05). However, the differences between the North Plume, South Marine, South Plume (Pandora Reef), and Coastal communities were not significant (Table S5), displaying similarities in their composition that suggests an interplay between latitude and water quality driving A. kenti microbial community dissimilarity.
The water quality parameters identified as driving the heterogeneity in the A. kenti microbiome community composition encompassed physical, chemical, and biological variables (Fig. 5C-D). Subsequent ordination with the four variables displaying the strongest correlation (R2 > 0.65, P-adj. = 0.001) revealed that these variables explained ~ 72% of the variation in the first two axes, with Si driving the separation of the North Marine samples, and NH4, temperature, and salinity driving the partitioning of the Coastal samples from the other categories (Fig. 5D). Corresponding with the structuring nature of temperature and dissolved inorganic nitrogen (NH4 and NO3), MAG-based gene clusters significantly enriched between the water quality categories included those identified as heat-shock proteins, molecular chaperones, and nitrate reductases (Mann Whitney U, P < 0.05; Table S6). The prevalence of clusters associated with reducing or repairing protein mutations and misfolding (e.g. cluster_84722, cluster_55609, cluster_45348, cluster_11927) and periplasmic nitrate reductase subunit napA (cluster_54900), specifically in the Coastal category, suggests adaptation of the microbiome to increased temperature stress and nutrients at this site. This is consistent with the typically warmer, relatively more nitrogen-rich waters of the inshore GBR, which may be more directly impacted by anthropogenic activity 83. The ability to overcome protein damage to maintain important phenotypes is also indicative of resilience in the face of environmental perturbation 84, which may suggest one mechanism of A. kenti holobiont adaptation within the typically more stressful coastal inshore GBR. These findings build upon recent observations of acquired resilience in A. kenti recruits through parental environmental history 24,81, suggesting a role for the microbiome in environmental adaptation.