Molecular Cartography of a Hawaiian Coral Assemblage

Abstract

Coral reefs are declining due to multiple factors including overfishing, anthropogenic pollution, and ocean acidification. Diseases affecting corals have increased in recent decades, yet the etiology of nearly all diseases remains poorly understood. Here, we investigated three-dimensionally mapped molecules and microbes from healthy and diseased coral tissue sampled across the landscape of a coral assemblage at the Wai‘ōpae tide pools, Southeast Hawai‘i Island. A 3D molecular cartographic platform was used in combination with molecular networking tools to characterize healthy coral tissue and tissue affected by the disease growth anomaly (GA). Tissues of healthy Montipora flabellata and Montipora capitata exhibited higher microbial diversity compared to Porites lobata and GA-affected M. capitata corals. Increases in relative abundance of Ulvophyceae and sterols were observed in GA lesions, while chlorophyll decreased. Conversely, healthy coral tissues were characterized by the presence of cyanobacteria in the order of Stramenopiles, in addition to higher relative chlorophyll levels. Leveraging innovative molecular cartography provides new insight into molecular characteristics of coral colonies, and helps to better understand how diseases affect the molecular landscape of corals.

1. Introduction

As diseases continue to threaten coral reef ecosystems in conjunction with local and global stressors, there is a pressing need to develop tools that better characterize the biology and ecology of corals [1,2]. The coral holobiont is a complex assemblage of diverse organisms including the mutualistic dinoflagellate genus, Symbiodinium, and associated bacterial, archaeal, viral, and other eukaryotic microorganisms [3]. Advances in culture-independent microbial analysis has enabled investigators to characterize the microbial community of the coral holobiont, although fully understanding the functional role coral microbiomes play in the epizootiology of coral health remains challenging [3,4,5,6]. Additionally, recent advances in liquid chromatography tandem mass spectrometry (LC-MS/MS) molecular profiling have enabled scientists to characterize coral metabolomes with increased sensitivity and reproducibility [7,8,9,10]. Leveraging advanced technologies will be critical for identifying components of the coral microbiome that influence host resistance and resilience to disturbance and disease. Recent research has primarily focused on the role of microbiomes in coral health resilience, but despite the improvements in molecular research techniques, there has been minimal improvement to our mechanistic understanding of how external stressors influence the coral microbiome, and in turn the host physiology [3,11]. More research is needed to determine microbiome characteristics associated with coral host conditions to develop tools and biomarkers for assessing coral health prior to the manifestation of disease [3].

Structure-from-Motion (SfM) photogrammetry techniques are capable of producing high-resolution three-dimensional (3D) reconstructions of underwater coral reef habitats [12,13,14,15]. The resulting 3D models can be visually and structurally analyzed to enhance ecological studies investigating the spatial and structural dynamics of coral communities [12,13,14,15,16]. The models are also capable of being annotated to both visualize and analyze data onto the reconstructed reef habitat [16]. Innovative 3D cartography tools can be used to annotate and map sampling locations onto 3D reconstructions to visualize the expression of molecular data products across reef landscapes [17]. Few studies have applied these approaches on coral reefs. This study integrated molecular data derived from coral samples onto high-resolution 3D reconstructions of a coral assemblage to conduct molecular profiling and examine patterns in microbial and metabolite features associated with coral taxonomy and health conditions.

The Wai‘ōpae coastline was located at the southeastern tip of Hawai‘i Island prior to being covered by the 2018 lower Puna eruption. Since the formation of Wai‘ōpae during the 1960 Kapoho eruption, it supported a high abundance and diversity of coral in a network of submerged pools formed by permeable basalt substrate [18]. Before being lost, Wai‘ōpae received a high frequency of natural and anthropogenic disturbances, both chronic and acute [19,20]. Coral Growth Anomaly (GA) is a prominent disease affecting corals throughout the global oceans, and exhibits an abnormally higher prevalence at Wai‘ōpae than any other sites surveyed throughout the Hawaiian Archipelago [19,20]. Little is known about the etiology of GA and few studies have investigated the molecular underpinnings of this disease [21].

Three-dimensional reef mapping and genomic analyses have been performed independently to investigate coral communities at Wai‘ōpae, yet no studies have examined the 3D distribution of molecules on coral colonies at this site [14,21,22]. Moreover, this study employs a novel approach of combining molecular and photogrammetry tools to enable 3D molecular cartography on an assemblage of coral colonies [23]. The 16S rRNA gene amplicon sequencing and LC-MS/MS data were mapped onto a high-resolution 3D reconstruction of a Hawaiian coral assemblage composed of Porites lobata, Montipora capitata, and Montipora flabellata species (Figure 1). Samples were collected along spatial gradients where the coral colonies physically interacted, and also along gradients where tissue was affected by GA disease. Sampling along gradients where adjacent colonies were physically contacting one another provides a means to assess possible patterns in molecular profiles associated with competitive interactions between corals. Analyzing competing corals from two separate genera, different species from the same genera and from the same species can shed light on how competition interactions affect coral biology. This study identified unique molecular features that distinguish each coral species as well as molecular features associated with GA-affected coral tissue. Analyzing the data across the landscape of the 3D-reconstructed coral assemblage provides a new and useful perspective for determining spatial patterns in molecular profiles. This approach can be scaled up to determine molecular underpinnings that drive biological responses and ecological characteristics in coral communities at other locations.

2. Methods

2.1. Coral Tissue Sampling

A total of 83 samples were collected from coral colonies and the surrounding seawater at Wai‘ōpae (19°29${}^{\prime }$55${}^{″}$ N, 154°49${}^{\prime }$06${}^{″}$ W), Southeast Hawai‘i Island, in accordance with traditional Hawaiian gathering protocols and State of Hawai‘i guidelines (DAR permit number 53929). Tissue samples were collected with a 1 cm core from P. lobata (n = 11), M. flabellata (n = 12), and M. capitata (n = 56) corals (Figure A1). The colonies in this assemblage all interacted along their borders and samples were taken along spatial gradients perpendicular to the colony borders. A number of tissue samples from M. capitata were collected from tissue affected by growth anomalies (n = 10) and non-GA tissue (n = 10) within 3 cm of the GA lesions. GA lesions were characterized as having reduced polyp and tuberculae densities compared to healthy M. capitata tissue [19]. Moreover, 1.5 mL water samples (n = 4) were taken just above the coral surface. The location of each sample was recorded by photographs and video for later annotation at precise locations to the pixel coordinates on the corresponding 3D reconstruction. Each sample was subjected to molecular profiling by mass spectrometry and DNA sequencing. Samples were flash-frozen with liquid nitrogen and shipped overnight under dry ice to the University of San Diego, California. Samples were stored at −80 °C until extraction.

2.2. Three-Dimensional Reconstruction of Coral Assemblage

Images, prior to sampling, were collected using methods developed specifically for reconstructing a 3D map of the coral reef habitats that were sampled [14]. One coral assemblage was identified that contained several corals with interacting borders of different genus (Porites/Montipora), same genus different species (M. capitata/M. flabellata), same species (M. capitata/M. capitata), and diseased tissue (M. capitata GA). Ground control points (GCPs) were placed at the corners of the coral assemblage with known x, y, z coordinates to enable precise spatial rectification for the resulting 3D models. A calibration grid and a scale marker were placed at the border of the survey area to validate spatial accuracy. Overlapping photographs were taken from planar and oblique angles while swimming above the coral assemblage in a boustrophedonic pattern. All photos were taken with a Canon 5D Mark III digital SLR camera with a 24–70 mm lens set at 24 mm (Canon USA Inc., New York, NY, USA) in an Ikelite housing with an 8 in. hemispheric dome port (Ikelite Underwater Systems, Indianapolis, IN, USA). The 8 in. hemispheric dome port significantly reduces distortion due to refraction and improves the ability of the software to accurately calibrate and align the images [35,36].

Three-dimensional reconstructions of the coral assemblage were rendered using Agisoft modeling software (Agisoft LLC., St. Petersburg, Russia). Images were uploaded into the PhotoScan software and camera calibration was performed using Brown’s distortion model. PhotoScan aligned all overlapping images using algorithms that detect invariant features from the overlapping image sequences. The invariant features were then used to create geometrical projective matrices and determine the exact position and orientation of the camera for each sequential image [16]. Three-dimensional geometry was constructed on the 2D image plane using the extrinsic parameters, which define the location and orientation of the camera reference frame with respect to a known world reference frame and in conjunction with the intrinsic parameters, which link the pixel coordinates of an image point with the corresponding coordinates in the camera reference frame. Iterative bundle adjustments were used to refine the 3D coordinates of the scene geometry and minimize reprojection error in order to create a sparse 3D point cloud that accurately represented the structure of the photographed coral reef plot. Markers were digitally annotated onto each of the GCPs using the PhotoScan software, and the known x, y, z values of the GCPs were used to optimize the alignment of the photos and to ensure that the resulting models were spatially rectified. After the initial point cloud was optimized, a dense point cloud was generated and used to construct a continuous mesh surface that was then triangulated and rendered with the sequential images to create textured 3D digital surface models representing the coral assemblage at Wai‘ōpae. The 84 coordinates were annotated onto the 3D model in GeoMagic 3D annotation software that represented the actual sampling sites where tissue samples were taken from each coral [37]. Coordinates were matched with each sample file from LC-MS/MS and 16S OTU feature tables.

2.3. Molecular Extraction and LC-MS/MS

Molecular extraction was performed on the frozen samples to capture small molecules that can be subjected to LC-MS/MS. Each sample was lyophilized, weighed, and subjected to 1.0 mL/mg with 70% methanol for 4 h in 10 mL glass vials [38]. The absorbed material was extracted in 500 µL of 50:50 ethanol/water (for mass spectrometric analysis). Supernatant was removed and dried until LC-MS/MS data acquisition.

Molecules were separated using chromatography and detected using a mass spectrometer. Processed coral extracts (5 µL) were subjected to UHPLC chromatographic separation using an UltiMate 3000 UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA), controlled by Chromeleon software (Thermo Scientific). Chromatographic separation was achieved using an 1.7 micron C18 (50 × 2.1 mm) Kinetex UHPLC column (Phenomenex) at 40 °C, using a flow rate of 0.5 mL/min. A linear gradient was used for the separation: 0–0.5 min 5% B, 0.5–8 min 5%B–99%B, 8–9 min 99%B, 9.01–10 min 1%B, 10–10.5 min 5–99% B, 11–11.5 min 99%B–1%B, 12–12.5 min 1% B where solvent A is water 0.1% formic acid (v/v) and solvent B is acetonitrile with 0.1% formic acid (v/v). Column eluent was introduced directly into a Bruker Daltonics maXis Impact quadrupole-time-of-flight mass spectrometer equipped with an Apollo II electrospray ionization source and controlled via otof Control v3.4 (build 16) and Hystar v3.2 software packages (Bruker Daltonics, Billerica, MA, USA). The maXis instrument was first externally calibrated using ESI-L Low Concentration Tuning Mix (Agilent Technologies, Santa Clara, CA, USA) prior to initiation of the sequence of samples, and hexakis (1H,1H, 2H-difluoroethoxy) phosphazene (Synquest Laboratories, Alachua, FL, USA), m/z 622.0295089613, was continuously introduced as an internal calibrant (lock mass) during the entirety of each LC/MS run. Data was collected in positive ion mode, scanning from 80–2000 m/z. Instrument source parameters were set as follows: nebulizer gas (Nitrogen) pressure, 2 Bar; Capillary voltage, 4500 V; ion source temperature, 200 °C; dry gas flow, 9 L/min. MS1 spectral acquisition rate was set at 3 Hz and MS/MS acquisition rate was variable (5–10 Hz) depending on precursor intensity. Data-dependent MS/MS acquisition was programmed to the top five most intense precursors per MS1 scan and any precursor was actively excluded for 1 min after being fragmented twice. Each MS/MS scan acquired was the average of 4 collision energies, paired optimally with specific collision RF (or “ion cooler RF”) voltages and transfer times in order to maximize the qualitative structural information from each precursor. The auto-acquisition of MS/MS spectra was carried according to specific settings. Precursor m/z 100, 300, 500, 1000 was selected; with an isolation width of 2, 4, 6, and 8; with a base collision energy (eV) of 10, 25, 30, 50; a sampled collision energy of (5, 10, 15, 20), (12.5, 25, 37.5, 48), (15, 30, 45, 60), (25, 40, 75, 100); a collision RF (Vpp) of 250, 500, 1000, and 1500 for all precursors; and a transfer time (µsec) of 50, 75, 100, and 150 for all samples was used, respectively.

2.4. Microbial DNA Extraction and Sequencing

DNA extraction was performed on the frozen samples using the Qiagen® MagAttract® PowerSoil® DNA KF kit following the Earth Microbiome Project (EMP) high-throughput DNA extraction protocol to capture microbial DNA from each sample that can be subjected to sequencing [26,27]. Extractions followed manufacturers protocols, with an added 10 min water bath incubation at 65 °C after Lysis buffer addition and retention of the 10 min incubation at 4 °C after lysate addition to IR solution [26]. Extracted DNA from samples was stored at −80 °C until amplification and sequencing.

Extracted DNA was amplified following the EMP protocols adapted for the Illumina MiSeq platform [39]. The V4 region of the 16S rRNA gene was amplified with region-specific primers, including Illumina flowcell adapter sequences [39]. Amplified DNA was confirmed on agarose gel and quantified using the Quant-iT PicoGreen dsDNA Assay Kit and cleaned using MoBio UltraClean PCR Clean-Up Kits following the manufacturers protocols.

Amplified DNA was diluted to 2 nM in a serial dilution and then balanced with 15–30% PhiX to reduce bias in samples [39]. Samples were then subject to V4 paired-end sequencing following the Earth Microbiome Project Protocols for Illumina MiSeq (University of San Diego, La Jolla, CA, USA) to determine which bacteria were present in each sample [27,39]. Amplicon sequences were demultiplexed and quality-controlled using the defaults provided by QIIME 1.9.1.

2.5. Data Processing

Raw MS data were converted to .mzXML formatting using Bruker Data Analysis 4.1 software and uploaded into MZmine 2.31 [30,33]. Peak detection thresholds for MS1 and MS2 were set at 5.0E3 and 1.0E3, respectively. Chromatogram builder thresholds were set at a minimum height of 1.0 E4 with an isolation tolerance of 0.01 min and m/z tolerance of 20 ppm. Chromatogram deconvolution was performed with a minimum peak height of 3.0 E3 and a peak duration tolerance of 0.1–1 min. MS2 scan pairing was performed at 0.01 da and RT scan pairing was performed at 0.4 min. Chromatograms were deisotoped with a RT tolerance of 0.2 min, an m/z tolerance of 20 ppm, and selected to have a maximum charged state of [M+2H]2+. Peaks were aligned with up to a 20 ppm error, with 75% of the alignment priority given to m/z accuracy and 25% to 0.5 min RT overlap. Gap-filling was performed, allowing for a 10% intensity tolerance, a 20 ppm m/z tolerance, and a 0.4 min RT tolerance. Lastly, duplicate peak filtering was performed with a m/z tolerance of 0.01 da and a RT tolerance of 10 min. Aligned peaks were exported as .csv files and .mgf files for further analysis.

2.6. Data Analysis

Principal coordinate analyses (PCoA) were conducted using Bray–Curtis measures of dissimilarity for the metabolomic data. HCA grouping methods were applied to identify clusters among the various tissue samples. Probabilistic quotient normalization of 30% of the samples was used. Plots were visualized in Emperor [32]. PCoA plots for microbiome datasets were constructed in Qiita by using the processed untrimmed 100 rarefied sample set. Ordinates were directly observed on the Qiita interface. Alpha diversity experiments were also performed and observed in Qiita using the untrimmed Wai‘ōpae BIOM table rarefied to 100. An adonis PERMANOVA analysis was performed using the vegan package in R studio [40]. Random forest analysis of both the MZmine output data and Qiita OTU feature table was performed in Metaboanalyst to identify top features that describe the metabolome and microbiome species variation [29,41,42]. Peak intensity tables were uploaded and normalized by sum and by mean intensity values. Sum normalization was performed on each sample and mean-centered data scaling was also performed.

Global Natural Products Social Molecular Networking was used to identify MS/MS spectral matches of molecular features in each sample set [28]. Parameters for creating molecular networks include a parent mass tolerance = 0.1 Da, Min matched Peak = 6, Ion Tolerance = 0.5, and Score Threshold = 0.70 with default advance and filter search options. Using these parameters, 64, 804 MS/MS spectra were merged into 5006 nodes.

Molecular feature visualization was performed to visualize microbiome and metabolome features on a 3D coral model. The open-source web application ‘ili was used by incorporating feature-finding tables and STereoLithography (.stl) file, OBJ 3D file (.obj), and Material Library File (.mtl) created in Photoscan [31].

3. Results

3.1. Metabolomic Profiling

LC-MS/MS data were subjected to molecular networking into GNPS and visualized in Cytoscape [34]. Using the networking parameters described in the methods, 64,804 spectra were considered to create nodes or unique molecular ions based on their MS/MS fragmentation patterns (Figure 2). Of the 5006 nodes, 205 ($\tilde{4}$%) had matching fragmentation patterns (min match peak = 4, cos score = 0.70) to the GNPS reference library and were issued level two or three annotations in accordance with the 2007 metabolomics standard initiative [43]. GNPS provided an estimated FDR value of 1% for annotations above 0.63 cos score value [44,45]. Unannotated nodes are colored in gray. Node borders were colored based on their origin. A majority of the detected nodes were unannotated and shared among the different corals, as indicated by the gray nodes with a dark orange border (Figure 2).

The putatively annotated metabolome of molecules shared among the different sampled coral species include long and short chain fatty acids, steroids, aromatics, sugars, phospholipids, nucleosides, amino acids, chlorophyll, a pesticide, and three unlabeled but structurally defined cyclic compounds (Figure 2). The annotated metabolome is considered putative, at level two or three according to the metabolomics consortium, as they have not been confirmed using synthetic standards or isolated and verified using other means. Using the recent false discovery rate (FDR) estimation method, the FDR is estimated to be 1% [44,45]. Fatty acids include hexadienoic acid, linoleic acid, pinolenic acid, docosenamide, docosatetraenoic acid, docosatetraenoic acid, eicosanoids, monolinolein, pentanoic acid, and propanoic acid. Steroids include androsterone and campesterol. Sugar constituents include glucuronic acid and glucocerebroside. Phospholipids include phosphatidylethanolamine, phosphatidylcholine, lyso-PAF, 1-octadecyl-sn-glycero-3-phosphocholine, palmitoyl-sphingosine, and phosphatidylserine. Thymidine nucleoside and decalactone were also annotated, as well as a pesticide known as fuberidazole. A degradation product of chlorophyll was annotated as pheophytin A. Amino acids acetyl-L-lysine and tryptophan were annotated. Four clusters with m/z values of 628.295, 415.212, and 247.167 were annotated only by structure.

3.2. Microbiome Characterization

Closed-reference OTUs were used to characterize the microbiome of the sampled coral assemblage to family-level classification. Relative abundances of OTUs were visualized against all samples (Figure A2 shows a bacterial phyla-level summary). The top 10 most abundant bacterial families present across the coral community in descending order are as follows: Rhodobacteraceae, Piscirickettsiaceae, Verrucomicrobiaceae, Hyphomicrobiaceae, Endozoicimonaceae, Pirellulaceae, bacteria in the order Stramnopiles, koll13, Erythrobacteraceae, and Clostridiaceae (Figure 3). Bacteria from these families were found in all samples, including the water column. Porites lobata tissues appear to be dominated by the family Endozoicimonaceae, making up nearly 90% of the microbiome composition. Montipora tissues of both species consisted of a more diverse microbiome, with the family Rhodobacteraceae having the largest relative abundance.

3.3. Metabolomic and Microbial Diversity

Metabolomic diversity (H) was significantly lower (PERMANOVA, p < 0.001) in P. lobata compared to both species of Montipora. Metabolomic diversity (H) was also lower in GA-affected M. capitata tissue compared to healthy tissues of and M. flabellata. Significant differences in the microbial diversity (H) were found when comparing diversity (H) among all sampled corals (Figure 4). GA-affected M. capitata tissue had significantly less microbial diversity when compared to healthy M. flabellata and M. capitata tissue (PERMANOVA, p < 0.05). Similarly, P. lobata tissue had significantly less microbial diversity than M. flabellata and M. capitata tissue (PERMANOVA, p < 0.001).

A principal coordinate analysis (PCoA) using Bray–Curtis values of dissimilarity was conducted to identify patterns among the groups of coral tissue (M. capitata, M. flabellata, and P. lobata). The ordination plots provide evidence of clusters, which are also supported by the PERMANOVA analysis described above (Figure 5).

Random forest classification was used to identify features that describe the metabolome and microbiome species variation with an out-of-bag (OOB) error of 0.0056 and 0.362, respectively. Although random forest provides an inventory of up to “species-level” (defined as 97% OTUs) microbial features that describe microbiome variation, the only annotated discriminatory metabolite feature was a 1-octadecyl-sn-glycero-3-phosphocholine within the top 15 features among the sample coral species. Additionally, random forest made it possible to identify features that describe variation between GA-affected M. capitata tissue and healthy M. capitata tissue with an OOB error of 0.149 for microbiome datasets and 0.350 for metabolomic datasets. The top microbial taxonomy feature that appears to be absent in diseased tissue and present in healthy tissue is classified as a cyanobacteria belonging to the order of Stramenopiles. The top three molecular features that describe the variance between healthy and disease tissue are 226.191 m/z, 455.340 m/z, and 412.263 m/z. None of the top 15 features were identifiable via GNPS.

The top features identified from random forest analysis that describe the difference in molecular profiles between GA-affected tissue and healthy M. capitata tissue were also visualized in the ‘ili 3D cartography platform (Figure A3). Pheophorbide abundances were visualized and identified. Cyanobacteria of the order of Stramenopiles were mapped onto the 3D coral assemblage at each sample location to visualize the higher abundance at all sample locations compared to GA-affected tissue. Similarly, pheophorbide abundances in M. capitata were undetectable on GA-affected sample site locations when visualized with the ‘ili platform but present in other sites in tissue distal to M. capitata GA. A fully interactive visualization can be viewed on the ‘ili platform interface for the metabolome (see Data Availability).

4. Discussion

This study used an innovative 3D cartography approach to conduct a quantitative and visual characterization of the molecular profiles associated with an assemblage of Hawaiian corals. Differences in the composition and diversity of microbial and metabolomic profiles among M. capitata, M. flabellata, P. lobata, and GA-affected M. capitata coral tissue were identified, including significant differences in the microbial and metabolomic signatures among the three coral species and between healthy and diseased tissues. Integrating genomic and metabolomic data into a 3D cartographic platform provides a powerful tool for visualizing the molecular profiles of corals on the physical reefscape created by coral colonies.

Identifying and visualizing the distribution of specific compounds within the coral tissue provides useful insight for understanding the biological implications of interference competition between coral colonies and the impacts of disease on the molecular composition of coral tissue. Sterols, the family to which the level-3 annotation campesterol belongs, have been previously found in tissue tumors and appear to be in high abundance in one of the GA-affected tissues [46]. The 3D visualization of the Wai‘ōpae coral assemblage illustrates the distribution of this molecule between healthy and GA-affected tissue within the same M. capitata colony (Figure A3). Considering that the biology and ecology of coral diseases are still poorly understood, it is valuable to be able to search for molecules known to be associated with other animal diseases to improve the characterization of conditions affecting corals and visually assess their distributions within a coral colony [47].

Ubiquitous fatty acid and phospholipid molecules detected within the metabolome (Figure 2) of the Wai‘ōpae coral assemblage are consistent with previously identified primary metabolites in scleractinian corals, soft coral, sponges, and dinoflagellates. Although lipid-based immune response pathways have been identified in marine metazoans, it remains unclear how these signaling molecules are selectively expressed in the current model [48,49,50].

No significant molecular differences were detected among the healthy coral tissue of separate M. capitata colonies, yet tissue affected by the GA disease exhibited significantly decreased molecular diversity (Figure 2). The observation suggests that a decrease in molecule abundance within GA-affected tissue could potentially decrease metabolic activity in the organism. Microbial diversity was found to be significantly lower in P. lobata than the other sampled coral species (Figure 2 and Figure 3) and mirrors previous studies that found lower bacterial diversity across Porites coral species in other locations [5,51]. The lower microbial diversity found in the GA-affected tissue is also consistent with many microbial studies investigating the health of organisms in an array of environmental and ecological settings [52,53,54]. The significant decrease in diversity of both microbial and molecular features in GA-affected M. capitata and P. lobata tissue highlights the need for further research using these techniques to determine whether molecular profiles can serve as biomarkers for discriminating among coral species and health states.

A derivative of chlorophyll, pheophorbide, was distributed throughout the sampled coral assemblage, but was absent from GA-affected M. capitata tissue. This finding highlights the potential reduction in photosynthetic activity occurring in GA-affected tissue in relation to healthy tissue [20]. This finding supports the previous studies showing that GA has detrimental impacts on the organism’s overall metabolism as the light-harvesting symbiotic microorganisms within coral tissue provide much needed nutrients for the host [55,56]. It is unclear whether the lack of photosynthesis contributes to, or is a result of, the diseased tissue, but our findings support prior research indicating a distinct photophysiological reduction in coral tissue affected by this disease [20].

Algal DNA was detected on several coral samples, which may result from chloroplast matches to the 16S rRNA gene sequence. Although some have viewed chloroplast contamination as a problem, the 3D visualizations in this study allows for visual confirmation that increased amounts of microboring algae are present within GA-affected tissue (Figure 6) [57,58]. Previous studies have investigated the relationship of turf algae growth on GA tissue at Wai‘ōpae and found growth to occur on both diseased and healthy tissue [58,59]. The detection of Ulvophyceae suggests that this algae has found an opportunity to infiltrate colony defense and begin algal turf succession on live coral tissue. Other bacterial families have also been identified as key descriptors of the variance between healthy and diseased tissue. Although accumulations of Rhodobacteraceae have been associated with coral disease in other parts of the world, levels of Rhodobacteraceae and overall bacterial diversity were lower in GA-affected corals [60,61].

5. Conclusions

As reef communities around the world exhibit declines due to anthropogenic disturbances and disease, it is critical to integrate innovative technologies to improve our understanding of the biological underpinnings that drive the ecology of these valuable coral reef ecosystems. Combining molecular and microbial profiling with 3D cartography enabled this study to create transformative visualizations to couple complex molecular datasets with high-resolution 3D reconstructions of the sampled coral colonies that can be used to identify patterns and associations that are less apparent in traditional 2D representations. This technique allows researchers to “see” patterns in molecular profiles on the physical reefscape used for sampling, which provides critical insight into the biological mechanisms associated with competition and disease. This approach can be applied to other sites to work toward detecting and characterizing molecular signatures associated with healthy, stressed, and diseased coral tissue. Identifying molecular biomarkers of stress and disease can enhance conservation strategies aimed at protecting coral reef ecosystems.