Background

Shallow-water, reef-building corals are well known for associating with microalgae in the family Symbiodinaceae. All corals, including those in the deep sea, form associations with diverse groups of bacteria [1], archaea [2, 3], unicellular eukaryotes [4], and viruses [5], collectively called the coral microbiome. Many members of the microbiome are important to the health of corals, either as pathogens [6] or providing sources of nutrition [7, 8].

Common coral associates include eukaryotic apicomplexans of the order Corallicolida [9,10,11,12,13] and the bacterial genera Mycoplasma [14,15,16,17,18,19] and Endozoicomonas [1, 20,21,22]. These taxa are among the most-studied coral-associated microbes and exhibit many indications of a close relationship with their coral hosts. Endozoicomonas form dense aggregates in coral tissue [23, 24] and conduct activities relevant to the corals such as producing vitamins [25, 26], degrading dimethylsulfoniopropionate [27], and cycling phosphorus [28]. While widely assumed to be mutualists, some have proposed that Endozoicomonas may be parasitic or pathogenic since they are generally more abundant in coral species that are less capable of adapting to environmental change [29]. Similarly, corallicolids and Mycoplasma are commonly found in many coral species and reside within or on coral tissue [10, 18, 30,31,32,33]. Most apicomplexans [34] and Mycoplasma spp. [35] are parasites or pathogens. However, those found in corals have not been associated with pathology or reduced fitness [31, 36]. While these common coral associates have been the focus of extensive research, we have a limited understanding of the variability of coral-microbe associations.

The compositions of coral microbiomes can vary substantially between and within coral species and across geographic locations [20, 37, 38]. The genotype of the host coral underlies some of this variation [37, 39, 40], as do environmental factors such as water temperature [41], habitat type [42], and pollution levels [41, 43]. These factors can differ across the ranges of coral species, yet the scope of most studies is limited to a narrow selection of a species’ known geographic or depth range. Further, most studies have focused on shallow-water corals, mainly scleractinians, and to a lesser degree on black corals and the more diverse octocorals [32, 38, 44]. Overall, the extent of coral microbiome variability is likely underestimated.

Variations in coral microbiomes can have functional implications. In shallow-water scleractinian species, certain microbial assemblages are associated with increased thermal tolerance [45,46,47]. Microbiome variation could be significant for mesophotic and deep-sea corals (those living deeper than 50 m) because they often have wider depth and geographic ranges than shallow-water species [48].

Mesophotic and deep-sea corals support diverse and abundant biological communities in Earth's largest biome [49,50,51,52,53]. Most are heterotrophic because they live beyond the reach of sunlight sufficient for photosynthesis. Like shallow-water corals, they face human-induced stressors such as ocean warming, acidification, pollution, and ecosystem destruction [54,55,56,57]. In the mesophotic zone and deep sea, octocorals comprise approximately 70% of all coral species [58]. A comprehensive characterization of microbiome variation in octocorals across their geographic ranges is needed to understand the diversity and ecological roles of microbiomes in deep-water habitats.

Mesophotic and deep-sea octocorals occur worldwide along continental margins and seamounts, over a wide depth range down to at least 8000 m [48, 59]. Light intensity, oxygen concentration, temperature, pressure, pH, salinity, food availability, and habitat structure vary with depth and geography. All these factors combine to strongly influence the ecology and evolution of marine animals [60,61,62,63]. Thus, factors that vary with depth and geographic location likely drive microbiome variation in mesophotic and deep-sea octocorals as well.

Here we test the hypotheses that the microbiomes of mesophotic and deep-sea octocorals vary significantly across their ranges in correlation with environmental factors and host genotype. We characterized the microbiomes of the mesophotic octocorals Muricea pendula and Swiftia exserta, and the deep-sea octocorals Callogorgia delta and Paramuricea biscaya across their known depth and geographic ranges in the northern Gulf of Mexico (Fig. 1). All of these corals are dominant foundation species in their respective communities and are the focus of restoration efforts because at least three suffered mortality during the Deepwater Horizon oil spill [54, 55, 64,65,66]. Understanding the geographic variation of their associated microbes is essential to understanding the ecology of coral ecosystems and can also inform restoration plans.

Fig. 1
figure 1

Sampling sites and their environmental and coral genotypic makeup. A In situ images of the octocoral species in this study. B Geographic map displaying sampling locations in the Northern Gulf of Mexico (BOEM bathymetry). The mean depth of all samples from a site is reported next to each site name in meters. Sites are colored based on which coral species were sampled: light gray = Muricea pendula, red = Swiftia exserta, blue = Callogorgia delta, yellow = Paramuricea biscaya. Images of C. delta and P. biscaya courtesy of the ECOGIG consortium. C Surface primary productivity of each sampling site displayed across longitude and colored by coral species sampled therein. D Bottom seawater temperature for each sampling site displayed across depth for each coral species. E UMAP projections of the genetic background of coral colonies based on RADseq data. Each point represents the genotype of a coral colony and is colored by depth

Full size image

Methods

Sample collection

Octocorals were sampled between 2009 and 2019 from multiple sites across their ranges in the northern Gulf of Mexico using Remotely Operated Vehicles (ROVs). They were placed in temperature-insulated containers at the seafloor and processed immediately after each dive. Specimens of four species were collected including the mesophotic octocoral Muricea pendula (n = 140, 10 individuals per site, 14 sites spanning 34–690 km and 53–117 m in depth), the co-occurring octocoral Swiftia exserta (n = 100, 10 per site, 10 sites spanning 13–600 km and 62–100 m in depth), the upper-bathyal octocoral Callogorgia delta (n = 60, 10 per site, 6 sites spanning 14–350 km and 439–852 m in depth), and lower-bathyal octocoral P. biscaya (n = 60, 9–11 per site, 6 sites spanning 14–655 km and 1370–2224 m in depth). Details, including research cruise, time of collection, and preservation method for each sample, can be found in Table S1.

DNA purification

Whole DNA was purified using a modified salting-out procedure [67]. To standardize DNA concentrations and reduce the concentration of PCR inhibitors, we measured purified DNA concentrations on a Qubit 4.0 fluorometer with a qubit Broad Range assay and then diluted with TE buffer to 2 ng/µl for Swiftia exserta, 20 ng/µL for Muricea pendula, 10 ng/µL for Callogorgia delta, and 10-40 ng/µL for P. biscaya. DNA extracts of C. delta and P. biscaya that were less concentrated than these thresholds were not diluted. See Supplementary Methods in Additional File 1 for details.

Microbial metabarcoding

The hypervariable V1-V2 regions of the bacterial 16S rRNA gene were amplified following Vohsen et al. [9] using the universal primers 27F and 355R. PCR products were sent to the University of Illinois Chicago for library preparation and sequencing [68] on an Illumina Miseq platform in a single sequencing lane with 250 bp paired-end reads and V2 chemistry.

Raw sequence data were imported into qiime2 [69] version 2021.2 and processed following Vohsen et al. [9] to create an amplicon sequence variant (ASV) table. Details can be found in the Supplementary Methods in Additional File 1.

Coral genotyping

To investigate the influence of the host’s genotype on the microbiome, Restriction site-Associated DNA sequence (RADseq) data were obtained from a subset of coral colonies as described by Bracco et al. [61] and Galaska et al. [62]. In brief, DNA was sent to Floragenex Inc. (Eugene, OR) for RAD library preparation using the restriction enzyme, PstI, and sequencing on an Illumina HiSeq platform. Sequence reads were demultiplexed and quality-filtered using the process_radtags program in Stacks v2.1 [70]. Read clustering and single nucleotide polymorphism (SNP) calling were performed with DeNovoGBS v4.0.1 [71] for C. delta and P. biscaya, and with gstacks v2.1 [70] for S. exserta and M. pendula after mapping to genome assemblies for each. We excluded SNP loci that: (1) had more than 10% of missing data, (2) had a minor allele frequency smaller than 0.01, or (3) were not biallelic. We also excluded individuals that were missing data for > 10% of SNP loci for S. exserta and M. pendula and > 30% for C. delta and P. biscaya or were identified as clones with R package poppr v2.8.6 [72]. To reduce the dimensionality of the genetic data, a principal component analysis (PCA) was performed on the SNP data of each species separately (S. exserta, 23,869 SNPs, n = 54; M. pendula, 9,611 SNPs, n = 100; C. delta, 14,013 SNPs, n = 51 and P. biscaya, 7,293 SNPs, n = 58) using the R package ade4 v1.7 [73]. The PCA axes were reduced to two dimensions using the uniform manifold approximation and projection method [74] using the R package umap v0.2.8. Genetic distances between individuals within species were calculated as pairwise allelic differences using the R package poppr v2.9.4.

Environmental variables

Publicly available satellite data and model-derived values for the Gulf of Mexico were obtained to evaluate the influence of environmental variables on coral microbiomes. The satellite-derived estimates of surface primary productivity (mg/m2d), chlorophyll concentration (mg/m3), inorganic suspended particulate matter concentration (g/m3), and the particulate back-scattering coefficient at 443 nm (m−1) were obtained from the Copernicus-GlobColour Global Ocean Colour biogeochemical dataset. Model-derived values were obtained for the depth of the surface mixed layer (m), bottom seawater temperature (°C), bottom seawater oxygen concentration (mL/L), bottom seawater salinity, and bottom current magnitude (m/s) from the HYbrid Coordinate Ocean Model (HYCOM). Both datasets comprised monthly means from Jan 2011 to Dec 2018 at 4 km grid resolution. The mean and standard deviation of monthly means for each variable were extracted at the sampling location of each coral colony. Only the mean bottom seawater temperature and surface primary productivity were retained for statistical analyses. Other variables were excluded from the analyses because they were strongly correlated, not substantially variable across sites, or were not justified by a priori hypotheses. See Tables S2, S3 for further details.

Statistical analyses

Rarefied microbial ASV richness was calculated for each coral colony using the rarefy function in the R package vegan [75] v2.5.7 using the minimum library size (1220 reads) as the sampling depth following Hurlbert [76]. Pielou’s evenness was calculated for each individual using the function evenness in the R package microbiome v1.16.0 [77]. Microbial community richness and evenness were compared between coral species and across sites using a Kruskal–Wallis rank-sum test followed by pairwise two-tailed Wilcoxon rank-sum tests with a Bonferroni-adjusted α-value using the functions kruskal.test and wilcox.test in the base R package stats v4.1.2. In addition, the effects of depth and longitude on richness and evenness were evaluated for each species using a multiple linear regression model (MLR) with depth and longitude as predictors and richness or evenness as response variables.

Permutational Analysis of Variance (PERMANOVA) [78] and distance-based Redundancy Analysis (dbRDA) [79] were performed to evaluate the influence of host species, depth, geography (site and longitude), bottom temperature, and surface primary productivity on microbiome compositions. Longitude was chosen to test for regional scale patterns across geography while sampling site was used as a finer scale variable. PERMANOVA and dbRDA analyses were performed using the R functions adonis2, dbrda, and anova.cca in the package vegan with Bray–Curtis dissimilarities on ASV proportions and 9,999 permutations. The results of both analyses were congruent so only the PERMANOVA results are discussed in the text. The details of each test are provided in Tables S4 and S5. In addition, the diversity among Endozoicomonas and corallicolids associated with S. exserta and M. pendula were analyzed individually using PERMANOVA to determine the influence of host species, depth, longitude, and site.

The effect of host genotype on microbiome dissimilarity was tested using Multiple Regression on distance Matrices (MRM) [80]. Bray–Curtis dissimilarities of 4th root-transformed ASV proportions were used as the response variable while geographic distances, host genetic distances, and Euclidean differences in depth, bottom temperature, and surface primary productivity were used as predictor variables. Geographic distances were calculated from latitude and longitude coordinates of each sample using the distm command in the R package geosphere with the Haversine function. MRM was conducted using the MRM command in the R package ecodist with 100,000 permutations.

Individual ASVs whose abundance differed along depth, longitude, and across sites were identified using Analysis of Compositions of Microbiomes with Bias Correction (ANCOM-BC) [81] using the R package ancombc v1.4.0. Two separate ANCOM-BC analyses were conducted on each species individually: (1) using both depth and longitude as explanatory variables and (2) a global analysis using sampling sites as groups. ASVs were excluded from the analysis if they were present in fewer than five individual corals (half the colonies sampled at a single site). The false discovery rate was controlled using the Benjamini–Hochberg method. The global analyses were repeated using each site as the reference. Only ASVs with a maximum q-value below α = 0.05 in all permutations were considered significantly different. These analyses were repeated on the family level to inform functional differences.

In order to account for differences in preservation method and sampling year, statistical analyses were repeated including those as explanatory variables. In addition, the same analyses were applied to a separate publicly available dataset of C. delta colonies from the same sites. Please see Additional File 1 and Tables S9 and S10 for further details demonstrating the consistency of results.

An analysis of homogeneity of multivariate dispersions between sites was performed for each species using the command betadisper with group centroids in the R package vegan (Table S12). Principal coordinates of each species were plotted showing centroid differences between sites (Fig. S8).

Results

Microbiomes are distinct among octocoral species.

The microbiomes of all four coral species were dominated by ASVs that represented more than 50% of the microbial community in at least one colony (Fig. 2B, Table S6). M. pendula had the greatest diversity of dominant ASVs, including two Mycoplasma, two Marinoplasma, an Endozoicomonas, an unclassified alphaproteobacterium, and a corallicolid apicomplexan. S. exserta hosted only two dominant ASVs: one classified as BD1-7 clade (Spongiibacteraceae) and another classified as Mycoplasma. Both C. delta and P. biscaya hosted three dominant ASVs. All three dominant ASVs in C. delta were recently described [30] as novel members of the class Mollicutes: two Ca. Oceanoplasma and one Ca. Thalassoplasma. The dominant ASVs in P. biscaya were classified as Mycoplasma, Endozoicomonas, and Pseudoalteromonas. Dominant ASVs belonging to the same taxon were given short identifiers (ex: M1, M2, S2) to distinguish which coral species they were associated with (first letter of coral genus) and the rank of abundance (1 being most abundant).

Fig. 2
figure 2

Characterization of coral microbiomes. A Non-metric multidimensional scaling plot of microbiome compositions using Bray–Curtis dissimilarities of ASV proportions. B Barplots of dominant ASVs in each coral species. Each horizontal bar represents the relative abundance of ASVs in an individual coral colony which sum to 100%. Colonies are grouped by site which are sorted by depth. ASVs that comprised 50% or more in any sample are designated with different colors. All other ASVs are collapsed and shown in light gray. Ca. Oceanoplasma and Ca. Thalassoplasma were recently described by Vohsen et al. (2022). C Rarefied ASV richness and D Pielou’s evenness between different coral species. MP = Muricea pendula, SE = Swiftia exserta, CD = Callogorgia delta, PB = Paramuricea biscaya. E Rarefied ASV richness and F Pielou’s evenness in Muricea pendula with depth. Numbers over boxplots designate significantly different groups

Full size image

Each coral species had a distinct microbial community (Fig. 2A, B; Table S4A; PERMANOVA R2 45%, p ≤ 10–4). All dominant ASVs were abundant in only one coral species and were rare (< 4%) in all others. The only exception was the corallicolid ASV M1 in M. pendula which was also abundant, but less so, in S. exserta along with other corallicolid ASVs. Coral species also differed in microbiome richness and evenness (Kruskal–Wallis rank-sum test, p < 2.2 × 10–16, Table S7A). The microbiomes of M. pendula and P. biscaya had higher rarefied ASV richness and Pielou’s evenness than S. exserta and C. delta (Fig. 2C, D; Table S7A; post-hoc pairwise Wilcoxon rank-sum tests, p ≤ 5.4 × 10–6).

Factors shaping microbiomes

The microbial communities of all coral species differed significantly across sites which accounted for 33–59% of the variation within species (Table S4B, PERMANOVA p ≤ 10–4). This variation was driven by significant trends across depth (p ≤ 10–4) and longitude (p ≤ 0.0251), which simultaneously accounted for 11–17% and 3–12% of the variation, respectively, depending on coral species (Table S4C). The only exception was C. delta, whose microbial community composition was not significantly correlated with longitude. After accounting for depth and longitude, sampling site remained associated with a significant (p ≤ 0.016) and substantial amount of additional variation (13–44%) for each coral species (Table S4D). However, sampling site no longer explained a significant amount of variation in C. delta after accounting for preservation method and sampling year (Table S9D).

Temperature decreased with depth and was associated with a significant amount of microbiome variation in all species (Fig. 1D; Table S4E; PERMANOVA p ≤ 0.0156, 5–14%) except P. biscaya in which it was not tested due to low variation between sites (< 0.6 °C). However, this did not account for the entire structure with depth which retained a significant correlation with microbiome dissimilarity after accounting for temperature (Table S4E; PERMANOVA p ≤ 0.0006, 4–9%, Table S4E). Similarly, surface primary productivity was higher in the east and could explain a significant amount of microbiome variation in all species except C. delta (Fig. 1C; Table S4F; PERMANOVA p ≤ 0.0154, 1–7%). After accounting for primary productivity, longitude was still associated with a significant amount of variation (Table S4F; PERMANOVA p ≤ 0.0063, 4–11%,).

Microbiome dissimilarity was significantly correlated with host genetic distances while controlling for covariates in P. biscaya (Table S8A, MRM p = 0.006) and weakly in M. pendula (Table S8A; MRM p = 0.041). In C. delta, genetic distance was strongly correlated with depth (Table S8B; Partial Mantel p = 10–4) and was significantly correlated with microbiome dissimilarity when ignoring depth (Table S8C; MRM p = 0.01).

Microbiome patterns with depth

Depth was correlated with microbiome diversity in various ways. In Muricea pendula, both richness (Fig. 2E; MLR p = 6 × 10–6) and evenness (Fig. 2F; MLR p = 0.012) decreased significantly with depth. The decrease in diversity in M. pendula was great enough that the average richness among colonies from the deepest sites was comparable to the richness observed in S. exserta and C. delta (Fig. S1 in Additional File 1). Conversely, both richness and evenness in S. exserta increased with depth (MLR p ≤ 6 × 10–6), as did richness in C. delta (MLR p = 0.038).

The relative abundances of multiple microbial taxa strongly correlated with depth including many dominant taxa. In M. pendula, Endozoicomonas ASV M1 increased in abundance with depth and so did the total abundance of all Endozoicomonas ASVs (Fig. 3A, ANCOM-BC, q = 4 × 10–9, 10–12). Similarly, the abundance of the dominant corallicolid ASV M1 increased with depth in M. pendula (ANCOM-BC, q = 3 × 10–57). This was also true for the total abundance of all corallicolid ASVs in Muricea pendula (Fig. 3B; ANCOM-BC q = 6 × 10–48) and Swiftia exserta (Fig. S2A in Additional File 1; ANCOM-BC q = 7 × 10–8). In contrast, other dominant ASVs decreased in abundance with depth: BD1-7 ASV S1 in S. exserta (Fig. 3E; ANCOM-BC, q = 5 × 10–5), Ca. Marinoplasma ASV M1 in M. pendula (Fig. 3C; ANCOM-BC, q = 1 × 10–7), and Ca. Thalassoplasma C1 in C. delta (Fig. 2B; q = 2 × 10–52).

Fig. 3
figure 3

Examples of depth patterns in coral microbiome composition. The relative abundances along depth in Muricea pendula of A the sum of all Endozoicomonas ASVs, B the sum of all corallicolid ASVs, C Ca. Marinoplasma ASV M1, D the sum of top 20 microbial families that decreased significantly with depth. E Relative abundance of BD1-7 ASV S1 with depth in Swiftia exserta. F NMDS ordination of Endozoicomonas communities in M. pendula showing patterns with depth. The community composition of G Endozoicomonas and H Shewanella communities in Callogorgia delta. Each horizontal bar is a single coral colony whose composition reflects the relative abundances of each ASV when restricted to only Endozoicomonas or Shewanella ASVs. Colors denote different ASVs. Colonies are arranged by site in order of increasing mean depth

Full size image

Many other taxa that comprised a smaller proportion of the microbial community also differed with depth (Fig. 3D). M. pendula hosted many of these taxa and the majority decreased in abundance with depth (54/80 non-dominant ASVs and 36/42 families, ANCOM-BC q ≤ 0.05). The most abundant of these non-dominant, depth-decreasing taxa included photosynthetic groups such as Cyanobiaceae and chloroplasts but also ASVs in the families Spirochaetaceae, Helicobacteraceae, Flavobacteriaceae, Francisellaceae, Burkholderiaceae, and unclassified bacteria that clustered with Spirochaetia and Epsilonproteobacteria. Together with the dominant Ca. Marinoplasma M1, these taxa averaged 70% of the microbial community among M. pendula colonies from the shallowest site (Sonnier Bank, 55 m) and decreased to a mean of 3% at the deepest site (Bright Bank, 116 m, Fig. 3D). Although many of these taxa were also detected in S. exserta, their abundances were either not significantly correlated with depth or increased with depth.

The composition of ASVs within microbial taxa also shifted with depth. For example, Endozoicomonas ASVs in M. pendula (Fig. 3F, Table S4G; PERMANOVA p ≤ 10–4) and corallicolids in S. exserta (Fig. S2K in Additional File 1; Table S4G, PERMANOVA p = 0.0029). Notably, M. pendula and S. exserta each hosted a unique corallicollid ASV at the deep site Elvers Bank (96 m). Similar depth transitions were also prevalent among the associates of C. delta. Ca. Oceanoplasma ASV C1 dominated the microbial community of most C. delta colonies from most sites except the deepest two sites (Fig. 2B, GC249 and GC290). At these sites, the closely related Ca. Oceanoplasma ASV C2 dominated some colonies. Shifts in ASVs were also seen among Endozoicomonas and Shewanella associates of C. delta (Fig. 3G, H, ANCOM-BC by ASV, 2 × 10–38 ≤ q ≤ 0.002). A single Endozoicomonas ASV was present in most colonies from all sites (Fig. 3G, gray bars), while other ASVs were mostly restricted between three depth zones. More strikingly, Shewanella communities between colonies from two depth zones (< 637 m and > 794 m) were completely non-overlapping, failing to share a single ASV (Fig. 3H).

Microbiome patterns with geography

Microbiome richness was higher in the east for both M. pendula and S. exserta (Table S7B; MLR p < 2 × 10–8). Multiple microbes exhibited differences in abundance across the geographic range of their host coral (Fig. 4A-F; ANCOM-BC, 9 × 10–11 ≤ q ≤ 0.05). M. pendula colonies from eastern sites had higher abundances of Mycoplasma ASV M2 and ASVs classified as Endozoicomonas, corallicolids, BD1-7, Spirochaetaceae, Alteromonas, Pseudoalteromonas, Pseudomonas, Marinobacter, Ruegeria, Caedibacter, Sphingomonadaceae, and Cyanobiaceae. The mean total relative abundance of all ASVs that were more abundant in the east ranged between 0.8% and 26%, excluding the dominant Mycoplasma ASV M2 and dominant corallicolid ASV M1. Some of these same ASVs and related taxa were also more abundant in S. exserta in the east including Pseudomonas, Pseudoalteromonas, Alteromonas, Ruegeria, Endozoicomonas, Cyanobiaceae, BD1-7, and Spirochaetaceae (Additional File 3, ANCOM-BC 0.0018 ≤ q ≤ 0.05).

Fig. 4
figure 4

Examples of geographic patterns in coral microbiome composition. Barplots of the relative abundance of microbial ASVs from Muricea pendula that A increased and B decreased from West to East across its range in the Northern Gulf of Mexico. ASVs are colored by family or other taxonomic level detailed in the legend. Each coral colony is represented by a vertical bar, and colonies are grouped by site sorted from westernmost to easternmost starting from the left. The total relative abundances of these ASVs are shown in C and D plotted against longitude to better reflect relative distances between sites. The relative abundance of E Mycoplasma ASV M1 in M. pendula and F Mycoplasma ASV S1 in S. exserta along longitude. Samples where zero reads were detected are represented with open circles. G Barplots showing the five most abundant ASVs that differed between sites in P. biscaya

Full size image

Other Cyanobiaceae, Mycoplasma, and Endozoicomonas ASVs were more abundant in M. pendula from western sites (Fig. 4B, D, ANCOM-BC, 9 × 10–11 ≤ q ≤ 0.03). Similarly, Ca. Marinoplasma and other photosynthetic taxa including several chloroplast ASVs were more abundant in the west. Notably, two ASVs classified as Prochloron comprised over 50% of the microbial community in some corals from adjacent sites in the west (Fig. 4B). Only two ASVs associated with S. exserta were significantly more abundant in the west including an unclassified gamma-proteobacterium (Fig. S3G in Additional File 1; ANCOM-BC q = 0.0018) and the dominant Mycoplasma ASV S1 which was completely absent from all samples from the three easternmost sites (Fig. 4F; q = 4 × 10–19).

Remaining site scale microbiome variation

The abundances of many microbial taxa were not correlated with depth nor longitude but still differed across sites. These comprised up to 54% of the community in M. pendula and 36% in S. exserta. Many ASVs were only detected in colonies from a few sites or even a single site. Examples include three close relatives of Endozoicomonas averaging a total relative abundance of 26% in M. pendula at McNeil bank (Fig. S3E in Additional File 1) and potential parasites within Rickettsiales in all coral species. Other examples include a corallicolid ASV only detected in C. delta colonies from VK826, GC234, and MC885 and another corallicolid ASV only detected in colonies from GC234 (Fig. S3H, I in Additional File 1).

In P. biscaya, the most abundant ASVs differed substantially across sites but did not correlate strongly with depth or longitude (Fig. 4G). For instance, SUP05 had a high abundance in colonies from GC852, Endozoicomonas ASV P1 was abundant at DC673 and KC405, Pseudoalteromonas ASV P1 was abundant at MC297, and Mycoplasma ASV P1 was abundant at most sites but its mean relative abundance ranged from 0.1 to 74% across sites.

Discussion

Differences between coral species

Coral species was the strongest factor differentiating microbial communities. Species differed markedly in which microbes they associated with and also the diversity of their microbial communities. Variants of shared taxa such as Endozoicomonas and Mycoplasma were species-specific suggesting possible adaptation to different coral hosts. Other dominant microbes belonged to different phyla, such as BD1-7 in S. exserta and Ca. Oceanoplasma in C. delta, suggesting that the metabolic function of microbial associates may differ between coral species.

Microbiome diversity differed strongly across coral species which may reflect different ecological strategies [82]. Some corals like M. pendula and P. biscaya appear to be microbial generalists that foster more diverse and flexible microbial communities. In contrast, other corals like S. exserta and C. delta appear to have specialized communities that are less diverse and less variable [82, 83]. Alternatively, S. exserta and C. delta may have higher absolute abundances of their dominant associates reducing the richness and evenness of their communities. Measuring absolute microbial abundances would aid the interpretation of differences in richness and evenness between species. For example, Endozoicomonas dropped in relative abundance in a Mediterranean coral in response to a heatwave, but this was shown to be due to the growth of opportunistic bacteria while the absolute abundance of Endozoicomonas remained the same [84].

These comparisons between species rely on the wide sampling range of this study. M. pendula and P. biscaya had the highest microbial diversities and greater variation across sites (Fig. S1 in Additional File 1). However, at some sites their diversities were comparable to both S. exserta and C. delta. This demonstrates the importance of sampling widely across a host species’ range in order to fully assess microbial diversity.

The influence of depth

Depth strongly structured the microbial communities of all coral species investigated. We observed differences in the composition, diversity, and abundance of specific microbial ASVs and families. The number and variety of taxon-specific patterns with depth reflect the likelihood that depth structures microbial communities through multiple mechanisms.

For instance, some variation in microbial communities may be due to temperature which decreases from > 30 °C at the surface to 4 °C below ~ 1500 m in the Gulf of Mexico. Depth and temperature were correlated with community composition in coral species living shallower than 1500 m. Previous studies have shown that shallow-water, scleractinian corals from locations with different temperature regimes host different microbial communities [45, 85]. Specifically, the stress of high temperatures shapes corals’ microbiomes [86], especially if they bleach [87,88,89,90]. The corals in our study may experience different frequencies or intensities of thermally induced stress which may influence their microbiomes. While temperature may be the underlying cause of some of the variation with depth, it does not appear to be the only factor.

The genetic divergence of coral populations may further contribute to differentiation in microbial communities across depths. In the scleractinian Pocillopora damicornis, genetically distinct coral populations harbor distinct microbial communities [40]. The turnover of the Shewanella community in C. delta is consistent with its population genetic structure [60, 61]. Coral colonies from the deepest two sites were genetically distinct and hosted unique Shewanella communities. Thus, some of the differentiation in microbial communities across depth may be caused by genetic differentiation of the coral host.

Other patterns with depth appear to be driven by different causes. For instance, the diversity of microbial associates of M. pendula decreased continuously with depth and may be due to a decrease in bacterioplankton (Fig. 2E, F). Indeed, some bacteria from mostly free-living taxa decreased in abundance with depth in M. pendula including Colwelliaceae and Marinobacteraceae. Further, light intensity may play a role since photosynthetic taxa associated with M. pendula predictably decreased with depth.

Many symbiotic taxa differed in abundance with depth such as Endozoicomonas, Mycoplasma, and corallicolids. However, the reasons for this are unclear because many of these microbe’s roles and interactions with their coral hosts are unknown or not well-characterized. Interestingly, corallicolids increased in abundance with depth in both M. pendula and S. exserta, but no depth pattern was found in corallicolids that infect C. delta. This is in contrast to previous studies that found a decrease in abundance with depth in the shallow-water coral, Montastraea annularis, and across dozens of species spanning a depth of 2 km [9, 91]. These inconsistencies may be explained if corallicolids, like their close relatives, infect intermediate hosts with differing depth distributions. Some of the closest relatives of corallicolids are found in fish [92,93,94], opening the possibility that corallicolids may use fish as intermediate hosts and could be limited by their depth ranges.

Altogether, depth may structure microbial communities through multiple mechanisms, including changes in free-living microbial communities, decreasing light levels, and genetic differentiation of host corals. When comparing different coral species, a recent review found that octocoral and scleractinian species from shallow water had more diverse microbial communities than their relatives from deeper depths [95]. This suggests that some of these mechanisms may influence microbial community structure on greater scales.

Variation across geography

In addition to variation across depth, we found that the microbial communities of all four species differed geographically. Corals from different sites differed strongly in microbial communities in all species. Yet nearby sites showed similarities reflecting regional-scale geographic variation.

Regional scale variation was identified in both mesophotic coral species. Both M. pendula and S. exserta had higher diversity in eastern sites where many likely free-living microbial taxa were detected at higher abundances. This may be due to the influence of the Mississippi River on bacterioplankton. There are marked longitudinal differences in surface primary productivity in the northern Gulf of Mexico, being higher in the area of influence of the Mississippi River. Microbial communities of shallow-water corals appear to be significantly influenced by surface primary productivity [96]. This is consistent with our observation that octocoral microbiomes were significantly correlated with surface primary productivity.

Symbiotic taxa also differed in abundance regionally including Endozoicomonas, Mollicutes, and corallicolids. This is consistent with other studies that found geographic variation in Endozoicomonas and Mycoplasma associates in shallow-water corals [15, 20]. Some of the regional patterns may result from limited dispersal since some ASVs were restricted to nearby sites like Mycoplasma ASV S1, which was completely absent in all S. exserta colonies from the three easternmost sites.

For every coral species, microbial communities differed strongly between sites after accounting for depth and geographic region. This is likely due to additional differences in local environmental conditions between sites that influence microbial community structure. For instance, the high abundance of the SUP05 cluster in P. biscaya at GC852 may reflect the presence of seeps nearby [97] (< 450 m). Similarly, the abundance of Pseudoalteromonas in P. biscaya at the Mississippi Canyon sites may reflect the lasting impact of the Deepwater Horizon oil spill [55, 65, 98]. Consistent deviations from depth trends at certain sites suggest that other factors play a role. For example, the total corallicolid abundance in M. pendula at Roughtongue Reef was higher than expected given the trend with depth (Fig. 3B). Conversely, corals from some subsets of sites displayed similar microbial communities despite being far apart and inhabiting different depths. For instance, P. biscaya from KC405 and DC673 were dominated by Endozoicomonas ASV P1, but it comprised no more than 0.12% in any colony from any other site. These sites are the farthest apart of all sites and do not have overlapping depths. Interestingly, coral populations from these two sites are closely related [62] suggesting an influence of the coral host’s genotype on the presence of Endozoicomonas.

Implications for restoration efforts

Characterizations of coral microbiomes based on samples from a single site or even a few sites may miss substantial variation across a species’ range. This is important to consider in restoration plans that utilize collections from different locations. For example, a restoration plan could use this information to avoid inadvertently transmitting parasites to new sites through transplants or to avoid rearing infected and non-infected corals together in aquaria. Examples of geographic variation in potential parasites discovered in this work include corallicolids in C. delta at VK826, MC885, and GC234 and the unique corallicolids in M. pendula and S. exserta at Elvers Bank.

Considerations for microbial taxa at low abundance

Some methodological factors may have influenced the patterns among microbial taxa at low relative abundances therefore these should be interpreted carefully. Low levels of contamination between samples cannot be ruled out. In some instance, multiple coral samples were collected in the same container, sometimes involving different species. However, all corals were transferred to fresh seawater individually before preservation. Also, all samples were sequenced on the same lane allowing for potential sample bleeding [99]. Because of these considerations, this manuscript focused on patterns involving microbes with high relative abundance.

Conclusions

This work characterizes the microbial communities of four dominant coral species in the Gulf of Mexico and demonstrates substantial variability across their ranges. These results provide promising targets to study the interactions between corals and their associated microbes. They also reveal substantial variability associated with depth and geography that could only be uncovered through extensive sampling across the hosts’ range. The numerous patterns associated with depth demonstrate that in addition to its ecological and evolutionary effects on corals, it also strongly structures their microbial communities.