Results
Transmission experiment
There was a significant difference in the number of days to the onset of visible disease lesions between species (ANOVA:F 1,29 = 6.22, p < 0.019), withO. faveolata significantly more likely to develop SCTLD lesions compared to M. cavernosa (log-rank: z 1,29= 3.569, p < 0.001; Figure S1). Transmission rates in the disease-exposed treatment at the conclusion of the experiment were 95% for O. faveolata with lesions occurring at a mean of 3.1 ± 0.5 days, while 55% of M. cavernosa fragments exhibited lesions at a mean of 5.4 ± 1.0 days (Table 1). No controls in either species developed lesions. Signs of SCTLD were similar between species, with tissue necrosis and lesion formation/progression being the most common (Dataset S1).
Two-way PERMANOVAs identified significant effects of coral genotype and disease status for M. cavernosa and associated Cladocopiumdatasets, while the effects of genotype and exposure group were marginally insignificant for O. faveolata (p = 0.057 andp = 0.056, respectively) and nonsignificant for associatedDurusdinium symbionts (Table 3). Pairwise differential expression tests with DESeq2 indicated the majority of differentially expressed genes (DEGs) to be attributed to diseased versus healthy samples for M. cavernosa (3,890 upregulated and 3,878 downregulated), with 1,820 DEGs between diseased and samples not exhibiting visible disease signs following exposure (NAI; 960 upregulated and 860 downregulated), and 115 DEGs between NAI and healthy samples (43 upregulated, 72 downregulated). Few DEGs were identified for any pairwise comparisons in Cladocopium symbionts, with five in diseased versus healthy samples and two in NAI versus healthy samples.Orbicella faveolata showed moderate numbers of DEGs between diseased and healthy samples (385 upregulated and 188 downregulated), with no DEGs for Durusdinium (Table 4). Ordination with PCoAs demonstrated that diseased samples were distinct from all other samples for M. cavernosa , while NAI and healthy samples were more similar (Figure 1). This relationship was reversed for Cladocopium data, where diseased and NAI samples were more closely related to each other than to healthy samples. Both O. faveolata and Durusdiniumsamples showed similarities between healthy and diseased samples, with little differentiation between exposure groups in the ordination space (Figure 1).
Functional enrichment analysis using gene ontology (GO) identified 434 GO terms across three categories (molecular function [MF]: 38; biological process [BP]: 287; cellular component [CC]: 109) for diseased versus healthy M. cavernosa samples, and 45 GO terms forO. faveolata (MF: 9; BP: 20; CC: 16; Table S1). The majority of enriched GO terms were negatively represented in diseased M. cavernosa , while the opposite was true for O. faveolata . For diseased versus NAI M. cavernosa , 371 GO terms were significantly enriched (MF: 50; BP: 244; CC: 77), while only 8 terms were significantly enriched for NAI versus healthy samples (BP: 2; CC: 6; Table S2). No GO terms were significantly enriched forCladocopium or Durusdinium symbionts, however. Functional enrichment of eukaryotic orthologous groups (KOGs) demonstrated similar patterns between diseased versus healthy and diseased versus NAI samples for M. cavernosa (Figure S2), where metabolic and transport mechanisms were generally downregulated in diseased samples, and nuclear structure and translation-related pathways were upregulated. Some nuclear structure and membrane pathways were positively enriched in NAI versus healthy M. cavernosa , with down regulation of translation, cytoskeleton, and extracellular pathways. Cladocopium diseased and NAI samples showed mild enrichment of most pathways relative to healthy controls, with nuclear structure as the most positively enriched, and cell motility as the most downregulated. Similar trends were observed for diseased versus healthy O. faveolata as compared to M. cavernosa , while Durusdinium showed an inverse pattern of KOG enrichment in diseased samples relative toCladocopium (Figure S2).
Weighted gene coexpression network analysis (WGCNA) identified ten modules with significant correlations to experimental factors (disease status and the time to onset of lesions). In general, modules that were negatively correlated with healthy samples were positively correlated with diseased samples (Figure S3). Several modules were also correlated with NAI samples that demonstrated similar expression to diseased samples (‘darkorange2’) and the time to transmission (‘thistle1’ and ‘darkorange2’), or inverse relationships with healthy samples (‘navajowhite1’). Of the modules correlated with NAI samples, only ‘darkorange2’ showed significant GO enrichment to the mRNA metabolic process (BP) and peptidase, cytosolic, and ribosomal complexes (CC; Table S4). Four modules were found to have significant trait correlations for Cladocopium , with similar expression patterns between NAI and diseased samples relative to healthy controls in the ‘pink’ module, but no significant GO enrichment (Table S4). The O. faveolata dataset had six modules with significant correlations to disease status and the time to onset of lesions; most of these were inversely related between diseased and healthy samples, apart from the ‘tan’ module that was inversely related between NAI and healthy samples (Figure S3). There were no significantly enriched GO terms for the ‘pink’ or ‘tan’ modules correlated with NAI samples. WGCNA was not successful in assigning modules to the Durusdinium dataset. Module hub genes (i.e., genes with the highest module membership) were largely unannotated for the respective reference datasets, with no apparent relationship between coral species or among associated symbionts (Table S3).
Orthofinder identified 681 orthologous DEGs in diseased versus healthy samples; of these, 438 showed similar expression patterns between species (272 upregulated and 166 downregulated), and 243 showed inverse expression (Figure 2). While expression of individual genes was in some cases inversely related between species, the corresponding KOG annotations demonstrated that the number of DEGs within each KOG class that was up- or downregulated was relatively similar, suggestive of an overall similar function response to disease exposure between M. cavernosa and O. faveolata . Likewise, examination of GO categories in diseased versus healthy samples indicated 21 GO terms (MF: 2; BP: 11; CC: 8) that were commonly enriched in the same direction between species (Table 5; Figure 3).
Intervention experiment
PERMANOVAs indicated no significant effect of time or treatment on the gene expression of M. cavernosa or associated Cladocopiumsymbionts (Table 3). Visualization with PCoA revealed overlapping sample dispersion in the coral and symbiont ordinations (Figure 4). Differential expression analyses, however, identified DEGs associated with pairwise treatment/time comparisons, with a high number of DEGs between diseased and healthy corals at t0 (410 DEGs: 284 upregulated and 126 downregulated), as well as between amoxicillin-treated corals at t1 and their original diseased state at t0 (241 DEGs: 48 upregulated and 193 downregulated; Table 4; Figure 4). There were few differences between treated and healthy corals at t1 (5 DEGs: 4 upregulated and 1 downregulated) and temporal shifts in gene expression were evident (e.g., 248 DEGs in t1 versus t0 healthy corals). Few DEGs were identified across pairwise comparisons forCladocopium (Table 4; Figure 4).
Functional enrichment analysis identified 46 significantly enriched GO terms in diseased versus healthy corals at t0 (MF: 9; BP: 10; CC: 27), 39 GO terms in treated versus diseased corals (MF: 6; BP: 14; CC: 19), and 7 GO terms in treated versus healthy corals at t1 (MF: 1; CC: 6; Table S5). Similar to M. cavernosa samples from the transmission experiment, the majority of enriched GO terms were negatively enriched in diseased versus healthy corals. Conversely, treated corals showed positive enrichment of most GO terms relative to healthy controls at t1. KOG enrichment in diseased versus healthy corals at t0 was similar to the trends observed for the transmission experiment, where several metabolic and transport pathways were negatively enriched (Figure S4). Notably, the nuclear structure KOG class was highly negatively enriched in diseased corals, where it was shown to be positively enriched for both diseased M. cavernosa and O. faveolata samples from the transmission experiment. Comparison of treated versus diseased corals demonstrated an opposite response in the same pathways as compared to diseased versus healthy corals, showing an increase in metabolic, transport, and nuclear structure classes. No KOG classes were strongly enriched when comparing treated and healthy corals at t1. Cladocopium KOG enrichment was similar in diseased versus healthy corals to the transmission experiment samples, with strong positive enrichment of the nuclear structure class and negatively enrichment of cell motility. Following treatment with amoxicillin, those patterns were reversed, even when comparing treated to healthy corals at t1 (Figure S4).
WGCNA returned 8 modules significantly correlated with treatments/time points, with the strongest patterns of correlation reversal between healthy corals at t0 to t1 (modules ‘darkred’ and ‘salmon;’ Figure S5). Modules ‘greenyellow’ and ‘purple’ showed inverse correlations between diseased and healthy corals at t0, and only the ‘pink’ module showed an inverse relationship between treated and diseased corals. Only the ‘darkred’ module contained significantly enriched GO terms, all of which were related to endoplasmic reticulum complexes (CC; Table S4). One module had a significant correlation to experimental traits inCladocopium , but it was driven by one sample and therefore disregarded. Visualization of expression patterns for the 30 most significant DEGs in modules ‘greenyellow’ and ‘grey60’ demonstrated a shift in expression patterns from the original disease state being distinct to healthy controls, to overall similarities between treated and healthy corals (Figure S5). Of the module hub genes with annotations, none matched among reference transcriptomes for the transmission experiment, although the ‘green’ module hub gene belongs to the heat shock protein 70 family (Table S3).
Cross-experiment comparisons
Comparison of diseased versus healthy samples from the transmission and intervention experiments revealed 1,458 genes that were shared between experiments. The majority of common genes showed the same directionality between experiments (452 upregulated and 489 downregulated), while 517 genes showed inverse relationships between experiments (Figure 5). When comparing diseased versus healthy samples in the transmission experiment to treated versus diseased corals in the intervention experiment, however, the majority of genes were found in inverse relationships between experiments (1,041 genes), whereas only 285 genes were found to share the same direction of expression (121 upregulated and 164 downregulated). Gene representation within KOG classes was largely similar across experiments, with the exception of unannotated genes that contributed to variability between experiments and among disease status/treatment comparisons (Figure 5).