Runaway coral-algal dysbiosis may be responsible for rapid coral tissue loss

Abstract

Stony coral tissue loss disease (SCTLD) affects at least 22 Western Atlantic coral species and presents as focal or multifocal lesions, which swiftly expand across the colony, resulting in rapid tissue loss and mortality. Previous histopathological examinations have noted necrosis, body wall breakage, gastrodermal separation, exocytosis, and vacuolization of the symbiont as pathological signs; however, the same signs are present to some degree in otherwise apparently healthy coral tissues processed for histology. Here, we quantify the degree of symbiont vacuolization, symbiont size change, exocytosis, and gastrodermal separation in apparently healthy and diseased coral tissues of eight coral species in association with SCTLD transmission experiments in Florida (USA) and the United States Virgin Islands. We describe a characteristic progression of disease signs which support the hypothesis that coral-algal dysbiosis contributes to SCTLD-associated tissue loss. Progression begins with symbiont cell vacuolization, followed by symbiont cell exocytosis and gastrodermal cell lysis. The lysing of gastrodermal cells leads to separation of the gastrodermis from the mesoglea, and finally to liquifying necrosis. Disease signs varied by coral species and symbiont genus following known hierarchy in disease susceptibility across coral and algal genera. SCTLD histological dynamics were also associated with differential expression of genes considered indicative of stress and dysbiosis and further influenced by both coral species and algal symbiont genus.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-35666-4.

Introduction

Coral disease outbreaks have contributed to drastic declines in coral cover and diversity over the last several decades, particularly in the Caribbean^1–4. Histopathological signs of disruption to the symbiosome–membrane complex are strongly associated with depressed coral health states (e.g., disease, bleaching). For example, coral bleaching disrupts the relationship between coral and symbiont and can arise due to host-cell detachment⁵ or reactive oxygen species produced by the symbiont in photosystem II⁶. However, these signs can also be detected to some extent in apparently healthy corals. For example, corals are also known to “shuffle” symbionts due to subacute stress⁷ and thus the vacuole–membrane complex may appear disrupted in tissues independent of acute bleaching or disease⁷. Similarly, many of the histopathological disease signs associated with stony coral tissue loss disease (SCTLD) (e.g., vacuolization, necrosis, and exocytosis) can be observed in apparently healthy tissue, either due to subacute stress or histological processing⁸. Quantitative analysis of histological signs is necessary to parse potential infective pathology from environmental stress or even baseline cellular turnover of a healthy coral host and provide a more holistic understanding of this and other diseases.

SCTLD affects at least 22 scleractinian coral species⁹ and has remained an unprecedented epizootic event on Western Atlantic coral reefs since 2014^10,11. Bacteria have been associated with SCTLD lesions^12–14 and the disease may involve viral infection of algal symbiont cells^15–19. However, the etiology of the disease is unconfirmed at this time.

The existing model for SCTLD presentation is as follows: within the coral host, SCTLD presents with lytic necrosis first affecting the basal body gastrodermis, which then progresses to the surface body wall⁸. Symbiodiniaceae cells in the gastrodermis show necrosis, with an expansion of vacuoles (vacuolization) surrounding the symbionts^8,20,21. Vacuolization can lead to necrosis of the algal symbionts and the merging of the vacuole with other gastrodermal cells (exocytosis), which may progress to full-scale gastrodermal necrosis. This can result in the release of cellular debris and symbionts into the gastrovascular cavity, mesogleal vacuolization, and necrosis, as well as progression of vacuolization beyond the epidermal basal membrane⁸.

A recent study by²² demonstrated the utility of integrating gene expression data with quantitative histopathology to investigate cellular disease dynamics. The expression of coral rab7, whose protein product is an established marker of symbiophagy in cnidarians^23,24, was negatively correlated with symbiont size, supporting evidence that higher rab7 expression reflects increased in situ degradation of symbionts by the coral host. Interestingly, rab7 expression did not significantly correlate with symbiosome vacuolization, despite significant symbiosome vacuolization in SCTLD-exposed and infected corals relative to controls²². This suggests that symbiophagy-associated processes may preferentially target the symbiont itself, while vacuole enlargement occurs as a downstream or independent response to symbiont stress. These findings highlight that combining molecular and histological analyses can reveal nuanced relationships between gene expression and cellular morphology.

Susceptibility to SCTLD differs between coral species, with the high susceptibility observed in Colpophyllia natans, Diploria labyrinthiformis, Pseudodiploria strigosa and Pseudodiploria clivosa; moderate susceptibility in Orbicella annularis, O. faveolata, O. franksi, Montastraea cavernosa, and Siderastrea siderea; and lowest susceptibility in Porites astreoides²⁵. However, some studies have shown that within coral species, the dominant lineage of dinoflagellate algal symbiont (Family Symbiodiniaceae) can play an important role in determining susceptibility to SCTLD²⁶, though this relationship is not entirely clear. For example, associations with lineages of Breviolum occur in many high-susceptibility species²⁵. In other studies, corals dominated by lineages of Durusdinium were more susceptible to SCTLD than those hosting Cladocopium^26,27. Variation in holobiont susceptibilities to SCTLD may influence the ecology of this disease among diverse coral communities²⁸. Thus, there is a need for a cross-specific and genera study of the manifestation and progression of this disease.

We investigated the progression of dysbiosis, ranging from subacute signs to those more indicative of infective disease, through ex-situ disease transmission experiments of eight coral species sourced from the northeastern Caribbean and Florida. Through quantitative histopathology, we computed disease signs indicative of SCTLD and then associated them with the identities and gene expression of both coral hosts and symbionts, with the aim of better characterizing this enigmatic disease.

Methods

Sample collection

Coral samples were collected during two ex situ transmission experiments. The first was in St. Thomas, USVI in April 2019²⁰ and the second in Miami, Florida, USA in November 2020²¹. In St. Thomas, coral fragments were collected in situ from seven species: Colpophyllia natans, Diploria labrynthiformis, Montastraea cavernosa, Orbicella annularis, Porites astreoides, Pseudodiploria strigosa, and Siderastrea siderea. Fragments were utilized in laboratory transmission experiments described in²⁰, and portions were retained for histological analysis, along with samples from field collections. In the USVI study, experimental tissue samples were binned by three potential health states: an apparently healthy portion of apparently healthy coral (HH), an apparently healthy portion of a diseased coral (HD), and a diseased portion of a diseased coral (DD); those that were exposed to disease and did not present with a lesion are in HH, as they appeared grossly healthy. There was no significant difference in histological signs between field and experiment samples from USVI, so they are grouped. In Florida, samples of M. cavernosa and O. faveolata were collected in situ for a separate transmission experiment (described in²¹). Samples collected from that experiment included two tissue health states: healthy control (HH) and diseased (DD; Table 1). Samples of M. cavernosa from both locations were analyzed as if independent to determine differences between regions. A total of 182 coral samples were analyzed for histology; 102 of these were processed for symbiont identity (Symbiont subset; Table 2), and 49 were used for gene expression analyses (Table 3, Supplemental X2).

Table 1.

Histological sample sizes for eight coral species and three health states from two SCTLD transmission experiments as well as field collections of disease in USVI (182 samples total).

Species	HH	HD	DD	Total
C. natans	17	8	15	40
D. labyrinthiformis	9	0	3	12
M. cavernosa	13	0	3	16
M. cavernosa*	6	0	18	24
O. annularis	9	2	10	21
O. faveolata*	9	0	27	36
P. astreoides	13	0	5	18
S. siderea	9	0	6	15
Total	85	10	87	182

HH indicates healthy portion of a healthy colony, HD indicates healthy portion of a diseased colony, and DD indicates diseased portion of a diseased colony (along the disease margin, tissue pieces were not fully necrotic). Asterisk denotes samples from Florida experiment; all others are from USVI.

Table 2.

Sample sizes for the subset of histological samples, sourced from USVI, which were analyzed for symbiont identity and grouped by the dominant symbiont genera found in each sample.

Species	Dominant Genus	HH	HD	DD	Total
C. natans	Breviolum	2	0	1	3 (21%)
	Cladocopium	2	0	3	5 (36%)
	Durusdinium	3	0	3	6 (43%)
	Total	7	0	7	14
D. labyrinthiformis	Breviolum	3	0	2	5 (56%)
	Cladocopium	3	0	0	3 (33%)
	Durusdinium	1	0	0	1 (11%)
	Total	7	0	2	9
M. cavernosa	Cladocopium	13	0	1	14 (100%)
	Total	13	0	1	14
O. annularis	Breviolum	1	1	0	2 (12%)
	Cladocopium	6	1	8	15 (88%)
	Total	7	2	8	17
P. astreoides	Symbiodinium	12	0	4	16 (94%)
	Breviolum	0	0	1	1 (6%)
	Total	12	0	5	17
P. strigosa	Breviolum	7	2	5	14 (74%)
	Cladocopium	3	0	2	5 (26%)
	Total	10	2	7	19
S. siderea	Cladocopium	5	0	6	11 (85%)
	Durusdinium	2	0	0	2 (15%)
	Total	7	0	6	13

HH represents a sample taken from the healthy portion of a healthy colony, HD represents a healthy portion of a diseased colony in the field, and DD indicates a diseased portion of a diseased colony from the USVI experiment and field.

Table 3.

Gene expression subset of histological sample sizes for five coral species and three health states from USVI SCTLD transmission experiment (49 samples total).

Species	HH	DE	DD	Total
C. natans	4	0	5	9
M. cavernosa	5	4	2	11
O. annularis	6	0	5	11
P. astreoides	4	2	5	11
P. strigosa	4	2	1	7
Total	23	8	18	49

HH indicates a healthy portion of a healthy colony, DE means it is a sample from a coral exposed to disease but did not get a lesion, and DD means a diseased portion of a colony that developed disease lesions after being exposed to disease. All samples are from the USVI ex situ transmission experiment. DE are noted as HH in other subsets since they did not get a lesion.

Histological processing

Histological methods largely follow^20,21, modifications and expansions are listed below. All tissue subsamples were fixed in zinc-buffered formalin (Z-fix concentrate from Anatech Ltd. diluted with filtered seawater) for 24 hours, washed with seawater for 24 hours, then stored in 70% ethanol for transport to Louisiana State University, Baton Rouge, LA, USA. Samples were decalcified with a 1% HCl EDTA solution and stored in 70% ethanol until processed. Samples were processed (serially dehydrated and flushed with xylenes) using a Leica ASP6025, embedded in paraffin wax blocks on a Leica EG1150H embedding machine, and sectioned at 5μm thickness on a Leica RM2125RTS microtome. Five serial sections were made ~500 μm apart for each sample. Histological slides were stained with hematoxylin and eosin stain on a Leica ST5020, viewed on an Olympus BX41 microscope with an Olympus SC180 camera attachment, and analyzed using ImageJ software²⁹.

To better understand the connection between vacuolization, symbiont size, exocytosis, and gastrodermal separation in corals, we quantified the symbiont cell-to-vacuole area ratio and regressed health status over this continuous ratio in a logistic regression. Ten photomicrographs per individual were taken at 40X magnification (60,000 μm² focal area per image; 1,820 images in total) per coral individual across all five serial sections, ensuring various parts of the sample were analyzed. For every photomicrograph image, a 12-cell grid (5,000 μm² per cell) was overlaid with ImageJ software and using a random number generator, one grid-cell was selected for analysis. This process was repeated for every image used. Within the grid-cell of interest, the areas of all symbiont cells and vacuoles were measured. Twenty-five symbiont and vacuole areas (vacuolization) were measured for each coral sample. Additionally, within the grid-cells of interest, the proportion of symbiont cells exocytosed from gastrodermal cells (exocytosis) was calculated for all symbionts within the grid-cell. Finally, if present, maximum gastrodermal separation length (gastrodermal separation) was measured from the mesogleal edge to the gastrodermis. For the proportion of exocytosed symbiont cells and gastrodermal separation, all ten photomicrographs per coral were used. Symbiont and vacuole measurements required a variable number of photomicrographs for the 25 measurements.

Histological statistics

Between experiments but within disease states (HH to HH; DD to DD), there was no difference found between the two regions in the measurements (t-test, p = 0.149), so the two datasets were combined for analyses.

A Bayesian hierarchical linear model was used to quantify the effect of four histological predictors (vacuolization (proportion), symbiont size (µm), proportion exocytosis (proportion), and gastrodermal separation (µm)) on two different disease states: apparently healthy (HH) and diseased samples (HD and DD). In this “Coral Model” where the first level of the model was an unconditional binomial generalized linear model with an intercept and the four predictors. Each of the four slopes were modeled as random effects with j = 9 levels for species (taxonomically eight species, but M. cavernosa was split by source region to identify any differences in tissue pattern that were missed in the initial ANOVA). Lastly, we ran the Bayesian hierarchical model for the “Symbiont Model”, where j = 4 levels for the four genera of symbionts in place of nine coral species from the coral model.

The coral model and symbiont model evaluated potential effect of histological measurement on disease state using the following Bayesian model:

Where y_i is the binomially-distributed response variable of disease state (0 = apparently healthy and 1 = diseased), α is the intercept, β_1–4 are the slopes, and x_1–4 are the predictor values exocytosis, gastrodermal separation, symbiont size, and vacuolization (mean maximum value per individual). The subscript j indexes the random effect of species (j = 9 (coral model); 4 (symbiont model)) and the subscript i indexes the observed data (i = n). The second level of the model was unconditional; i.e., no model was applied to the random effect estimates and only grand mean estimates of the four level one parameters were estimated. Model variance-covariance was estimated using a scaled-inverse Wishart distribution³⁰. The model was run in JAGS (version 1.5.2) using the package ‘JAGSUI’³¹ in R³². All parameters were given diffuse normal priors in all models. The second model and symbiont model were initialized with a randomly selected value for all four parameters from a normal distribution with a mean of zero and a standard deviation of one. We ran three Markov chain Monte Carlo (MCMC) chains each for 40,000 iterations but removed 5,000 for burn-in and thinned by two, for a total of 105,000 iterations used for posterior analysis.

Model convergence was evaluated from the values, where < 1.1 indicated convergence³⁰. We also plotted all posteriors for visual confirmation of convergence. We interpreted predictor effects based on where 0 was in relation to their posterior distributions. All Bayesian hierarchical models converged <1.1. We used an 80% confidence interval (CI) to describe a weak effect and a 95% CI for a strong effect (e.g.,³³.

Dominant symbiont identification

Coral tissue slurry from a subset of histological samples was preserved in DNA/RNA Shield (ZymoResearch) upon collection. After incubation with Proteinase K (20 mg/ml), DNA was extracted with the ZymoBIOMICs DNA/RNA Miniprep Kit (ZymoResearch) according to the manufacturer’s instructions. The Internal Transcribed Spacer-2 (ITS-2) region of Symbiodiniaceae rDNA was then amplified using PCR with the Sym_VAR_5.8S2/Sym_VAR_REV primers ³⁴. All samples were sequenced at Oregon State University’s Center for Quantitative Life Sciences (OSU CQLS, Corvallis, OR, USA) on an Illumina MiSeq platform. The per-sample abundance of Symbiodiniaceae lineages (i.e., ITS-2 type profiles) was quantified using the program SymPortal³⁴. To identify the dominant Symbiodiniaceae genus per sample, the per-sample ITS-2 type profile absolute abundance was imported into RStudio using the package ‘phyloseq’³⁵. One sample was removed as it did not generate enough sequence reads for downstream analyses. The absolute abundances of each ITS-2 type profile were used to determine the dominant Symbiodiniaceae genus per sample (correcting for variation in ITS-2 rDNA copy number variation across Symbiodiniaceae genera³⁶ as described by¹⁵ did not influence results).

Coral and symbiodiniaceae gene expression

A subset of histological samples from the USVI transmission experiment had paired tissue samples that were used for RNA-Sequencing (RNA-Seq) analysis, which is detailed in²². To explore potential relationships between histological measurements and gene expression, all coral host and symbiont genes reported in²² were tested for correlations with histological measurements from this study. The host gene expression dataset included immune and extracellular matrix (ECM) homologs as well as orthologs showing significant differential expression between disease states. It also included highly plastic, single-copy orthologs with significant differential expression between disease states and lineage-specific, single-copy orthologs correlated with relative risk or lesion growth rate. The symbiont gene expression dataset was comprised of highly plastic, single-copy orthologs with significant differential expression between disease states, lineage-specific, single-copy orthologs correlated with lesion growth rate, and homologs correlated with host rab7 expression.

Linear regression models were applied to test the association between gene expression and histology measurements in both the host and symbiont datasets. For each gene-histology pair, a standard linear model was applied to test the relationship between gene expression and histological measurements. The general form of the model was:

where represents the histology measurement, is the intercept, is the vector of slopes (5) corresponding to the predictor variables , and is the residual error. For the host dataset, the predictors included rlog-normalized host gene expression and host species, resulting in a total of five slopes. For the symbiont dataset, two models were constructed: 1) the first model included rlog-normalized symbiont gene expression and dominant symbiont genus as predictors, resulting in a total of four slopes, and 2) the second model added host species as an additional covariate, resulting in a total of eight slopes.

A likelihood ratio test (LRT) was performed on each symbiont gene-histology pair to evaluate whether adding the host species variable significantly improved the model fit. When LRT p > 0.05, the simpler model (model 1) was retained. When LRT p ≤ 0.05, the more complex model (model 2) was retained. Given the large number of models tested (1,638 host models and 754 symbiont models), p-values were corrected using the Benjamini-Hochberg false discovery rate (FDR) procedure to reduce the risk of Type I errors. Statistical significance was set at FDR-adjusted p ≤ 0.10 to balance Type I and Type II error trade-offs and avoid discarding potentially relevant gene-histology signals in this multi-species, multi-symbiont dataset. All gene-histology regression methods were performed using R v. 4.2.2³².

Results

This study analyzed 182 coral samples from 9 species for quantitative histological traits from St. Thomas, USVI and Florida; of those 102 were compared from their dominant symbiont genus, and 49 had histological traits regressed against gene expression data. While these species have high variability in average values for each measurement (i.e size of symbionts and vacuoles, symbiont to vacuole ratio, proportion exocytosis, and gastrodermal separation; Supplementary X3) when modeled to determine disease state, disease progression patterns can be determined (Fig. 1, 2).

Fig. 1.

Histopathological stony coral tissue loss disease (SCTLD) signs in representative apparently healthy (left column) and presumed diseased (center column) Pseudodiploria strigosa tissues, and an illustration of the progression of dysbiosis in coral tissues (right). A–B represent gross images of P. strigosa colonies, C–D show representative pictographs of vacuolization, E-F are representative of degraded symbionts, G–H are exocytosis, and I–J are gastrodermal separation; g indicates gastrodermis; m indicates mesoglea; e indicates epidermis. Single-headed arrows indicate vacuolization; open squares indicate degraded symbionts; dashed circles indicate exocytosis; double-headed arrows indicate gastrodermal separation. Pictographs were taken at 40X and the scale bar is 20 μm. Righthand schematic depicts tissue from apparently healthy at the top through disease progression terminating at the bottom.

Fig. 2.

Symbiont cell to vacuole ratio between A) healthy (HH) and diseased (DD) samples across all species; B) healthy (HH), healthy portion of diseased colony (HD), and diseased portion of diseased colony (DD) with all species combined; C) and including all samples in a logistic regression of disease state (healthy or diseased) on symbiont-to-vacuole ratio (bars indicate one standard error and the black dashed line represents the point where the model predicts a disease probability of 0.5; symbiont-vacuole ratio = 0.55; grey vertical bar indicates Bayesian 95% confidence interval of inflection point). M. cavernosa from USVI and FL are separated for visualization purposes. Box plots represent distribution of symbiont-to-vacuole ratio interquartile range and line depicting median, with dots representing outliers. Codes in A denote species: CNAT: C. natans; DLAB: D. labyrinthiformis; MCAV: M. cavernosa; OANN: O. annularis; PAST: P. astreoides; PSTR: P. strigosa; SSID: S. siderea; OFAV*: O. faveolata (FL); MCAV*: M. cavernosa (FL). Asterisks above panel A indicate where vacuolization was significantly different between disease states.

Coral model

Of the four quantitative histological measurements, symbiont-to-vacuole ratio had the strongest relationship with disease state across most coral species. According to logistic regression, a symbiont cell-to-vacuole ratio between 0.52 and 0.56 predicted an equal likelihood that the colony was healthy vs. diseased (Fig. 2). A ratio <0.52 and approaching 0 (i.e., large vacuoles and/or shrinking symbionts), had an increasing probability the coral was diseased. Whereas a ratio >0.56 to approaching 1.0 (i.e., close contact between Symbiodiniaceae and gastrodermal cells) had an increasing probability that the colony was healthy. The coral model revealed a weak effect of the symbiont-to-vacuole ratios in health condition prediction for D. labyrinthiformis, M. cavernosa (USVI and FL), O. annularis, O. faveolata, P. astreoides, and S. siderea when modeled separately (80% CI did not overlap zero), but not for C. natans or P. strigosa (Fig. 3).

Fig. 3.

Logistic regression analysis of disease state versus quantitative histology metrics for each species. Colors denote trend strength determined by model standards at 80% CI. Species codes: CNAT: C. natans; DLAB: D. labyrinthiformis; MCAV: M. cavernosa; OANN: O. annularis; PAST: P. astreoides; PSTR: P. strigosa; SSID: S. siderea; OFAV*: O. faveolata (FL); MCAV*: M. cavernosa (FL). In disease state, a one indicates diseased and zero indicates apparently healthy.

The coral model detected a weak relationship between size of symbiont and disease state for three species, O. annularis, M. cavernosa (FL), and P. strigosa (80% CI did not overlap zero; Fig. 3), where diseased samples had larger symbionts in O. annularis and P. strigosa but smaller symbionts in diseased samples from M. cavernosa from FL (Fig. 3). There is no relationship between disease state and symbiont size for C. natans, D. labyrinthiformis, M. cavernosa, O. faveolata (FL), P. astreoides, S. siderea.

The coral model identified that exocytosis was weakly associated with disease state for O. faveolata (80% CI did not overlap zero). For some of the boulder corals (C. natans, O. annularis, S. siderea, and M. cavernosa [FL]), diseased samples had higher proportions of exocytosis than healthy samples. In contrast, healthy D. labyrinthiformis, M. cavernosa (USVI), P. astreoides, and P. strigosa samples exhibited higher proportions of exocytosis (Fig. 3).

The coral model detected weak evidence for gastrodermal separation associated with disease state for D. labyrinthiformis, O. annularis, and P. strigosa (80% CI did not overlap zero), where healthy samples had more gastrodermal separation than diseased.

Symbiont model

Most coral samples were dominated by a single genus; either Symbiodinium, Breviolum, Cladocopium or Durusdinium. Of all samples included in the symbiont analysis, 23.5% were dominated by Breviolum, 52.0% were dominated by Cladocopium, 8.8% were dominated by Durusdinium, and 15.7% were dominated by Symbiodinium. Breviolum- and Cladocopium-dominated samples were from C. natans, O. annularis, P. strigosa, D. labyrinthiformis, and P. astreoides, with Cladocopium-dominated samples also including S. siderea. Durusdinium-dominated samples include C. natans, D. labyrinthiformis, and S. siderea, whereas P. astreoides was the only Symbiodinium-dominated species. The symbiont Bayesian model detected a strong correlation between disease state and vacuolization for corals dominated by symbionts in (respectively) Symbiodinium, Cladocopium, or Durusdinium (95% CI did not overlap zero; Fig. 4), and a weak correlation for corals dominated by Breviolum (80% CI did not overlap zero).

Fig. 4.

Logistic regression analysis of disease state versus quantitative histology metrics for corals dominated by Symbiodiniaceae from either Symbiodinium (A), Breviolum (B), Cladocopium (C), or Durusdinium (D) from subset analysis. Colors denote trend strength and direction determined by model standards at 80% CI. In disease state, one indicates diseased sample and zero indicates apparently healthy samples.

No symbiont genera had a correlation between cell size and health state. Proportion of exocytosis had a strong correlation with disease state for corals dominated by Cladocopium (95% CI did not overlap zero) and a weak correlation in Breviolum (80% CI did not overlap zero), where there was a higher proportion of exocytosed symbionts in diseased samples than in healthy. A similar trend was seen in corals dominated by Durusdinium or Symbiodinium, but not to the same degree (Fig. 4). Gastrodermal separation had a weak correlation with disease state and Breviolum or Cladocopium (respectively, 80% CI did not overlap zero; Fig. 4). Only corals dominated by Durusdinium exhibited a positive relationship between disease state and gastrodermal separation (80% CI did not overlap zero).

Gene expression and histology

A total of 8 host gene-histology associations and 2 symbiont gene-histology associations were significant after applying the p-value threshold (Supplemental Data X1). Among the host associations, the histology measurements most frequently associated with differences in gene expression were average and maximum symbiont-to-vacuole ratios. For example, host Toll-like receptor 6 (TLR6) expression was significantly positively associated with average symbiont-to-vacuole ratio (Fig 5). In addition, basement membrane-specific heparan sulfate proteoglycan core protein (HSPG2) expression showed a positive association with average symbiont-to-vacuole ratio (Fig. 5). Significant symbiont gene-histology associations included genes such as DnaJ homolog subfamily C member 7 (DNAJC7) and exonuclease mut-7 homolog (MUT-7). DNAJC7 expression was positively associated with average symbiont size, while MUT-7 expression was associated with a higher symbiont-to-vacuole ratio (Fig 5).

Fig. 5.

Associations between gene expression and quantitative histological metrics across coral species and dominant Symbiodiniaceae genus. Each panel displays a significant gene-histology pair identified through linear regression, with panels A and B illustrating coral expression-histology associations and panels C and D showing dominant Symbiodiniaceae expression-histology associations. Colored lines represent coral species- or dominant Symbiodiniaceae genus-specific regression lines, while the dotted black line shows the overall trend across all data points, with the shaded area representing the 95% confidence interval on the overall trend. Each panel contains the FDR-adjusted p-value for the gene’s coefficient, testing the effect of gene expression on the histology measure, and the effect size, which reflects the change in histology measure per one rlog increase in gene expression. Shapes represent health condition at the end of the SCTLD transmission experiment.

Discussion

Describing and quantifying disease states in corals by utilizing histological techniques is important for understanding their pathology and relating tissue-level disease conditions with gross lesion presentation. In this study, we characterized the histopathological signs of SCTLD in different health states, highlighted differences in disease progression and severity that may be due to coral species or dominant symbiont identity and correlated histological signs with holobiont gene expression.

We confirmed the progression of SCTLD signs in coral tissues, beginning with apparent coral-algal dysbiosis characterized by the vacuole expanding to over twice the area of the corresponding algal symbionts (Fig. 1). This is correlated to algal cell degradation and eventually exocytosis (i.e., free-floating partially degraded symbionts in the gastrodermis), which may contribute to reduced communication between the coral host and symbiont. The lysis of gastrodermal cells leads to the separation of the gastrodermis from the mesoglea, and structural integrity of the host is lost. This results in the gross sign of tissue sloughing²⁰. We demonstrated that the magnitudes of these histopathological signs are correlated with the reduced expression of genes responsible for managing tissue degradation, which could lead to higher levels of degradation within the host and symbiont. These conclusions point to SCTLD pathology as a positive feedback of dysbiosis and the failure of the coral host to remove the harmful byproducts of dysbiosis from surrounding tissues. Observed differences in coral species and dominant symbiont genus susceptibility to this disease may indicate variation in strength of this positive feedback among corals across life history characteristics.

A recent study³⁸ noted that microscopic disease signs (vacuolization and exocytosis) were different between Florida and the Virgin Islands. Through quantification of specific disease signs, we were able to identify similar patterns in microscopic progression of SCTLD between regions. Clearly, more of this type of work linking quantitative histopathology to molecular datasets) is needed to better characterize this unprecedented disease. Thus, with continued sampling and metric development, our holistic understanding of coral disease will improve.

While the overall observations were consistent with previous histological observations, our ability to quantify these disease signs has greatly increased our ability to assess SCTLD. Specifically, we identified some important processes in the progression of SCTLD that may be variable among coral hosts and symbiont genera. The symbiont-to-vacuole ratio (vacuolization) measure has a specific inflection point between apparently healthy and diseased colonies that is predictive of disease condition. While vacuolization is a general stress response⁵ that can be exacerbated by histological processing, we show that vacuolization due to SCTLD exposure is more severe than from general stress or processing.

This relationship was first observed qualitatively as larger vacuoles in diseased corals^8,37, but our quantification removes potential bias when viewing histological sections. As these histological metrics strengthen, our ability to identify SCTLD becomes more streamlined. For example, vacuolization appears to be the dominant histological sign of SCTLD exposure and can be used to quickly differentiate this disease from white plague and other rapid tissue loss diseases which present with similar gross lesions.

Exocytosis of symbionts was a consistent feature in diseased tissues in seven of the nine species tested. Vacuolization of symbionts followed by the surrounding of the algal cells by autophagic membranes has been observed in bleaching experiments⁵. This indicates that during stress, the symbiosis-maintaining mechanism is destabilized, and autophagy is activated. The potential for symbiophagy as a strategy for disease defense in corals deserves further investigation. Gastrodermal separation was not consistent across coral species, possibly as a result of histological processing or timing of sample collections. If SCTLD progresses through polyps of different coral species at different rates²⁰, the rates that tissue-level disease signs appear is likely also unique to species. Samples were pulled from disease experiments at the first appearance of gross lesions, where microscopic signs such as gastrodermal separation may not be as apparent.

Our study spans many coral species that have known differences in susceptibility to SCTLD. For example, C. natans is one of the most susceptible coral species and experiences high mortality²⁰. Colpophyllia natans diseased tissue samples showed higher vacuolization, smaller symbionts, and more exocytosis than healthy samples. When C. natans is affected by SCTLD, the vacuole in the host cell expands and communication between the symbiont and host cell may be limited, adding further stress to the symbionts, leading them to shrink, degrade, and exhibit a reduction in photosynthetic effectiveness. This is then followed by exocytosis of the symbionts from the gastrodermis. The susceptibility and mortality of C. natans to SCTLD may be explained in part by symbiont degradation and expulsion. Conversely, P. astreoides has a low susceptibility to SCTLD, and the opposite responses in symbiont size and gastrodermal separation. It is possible that P. astreoides can recognize the disease-affected tissue and expel symbionts upon detection more efficiently than the higher susceptibility species. Lower-susceptibility species may have a better coordinated immune response or are otherwise able to avoid runaway dysbiosis and inflammation that may be responsible for tissue loss in other species. While no disease agent has been identified for SCTLD, consensus is growing in recognizing the pathogen(s) is primarily infecting the algal symbionts within the coral host^19,22,26. Given that the strongest relationship between histological signs and disease state was symbiont-to-vacuole ratio, we provide additional evidence that SCTLD is primarily targeting the algal symbionts rather than the coral host. Continued investigation into relationships between disease signs and algal symbiont identity is critical.

This study evaluated coral disease at the tissue-level through light microscopy. While we could address disease signs and general differences in symbiont and vacuole size, differentiating between different symbiont genera is not possible with this technique. Differences in size between Symbiodiniaceae taxa are on the scale of fewer than five microns³⁹, which is within a reasonable error for the magnification used here. As our observations were correlated to dominant symbiont genus in the Symbiont model, this assumes every cell measured in histological sections was a member of the dominant symbiont genus in that sample; single-cell RNAseq following histological measurement would be more accurate. Future studies should collect samples for transmission electron microscopy or fluorescent in situ hybridization or immunofluorescence (e.g., Coy et al.,⁴⁰) with genus-based symbiont resolution to more closely address these questions.

For the symbiont model, the difference in disease signs across corals harboring different dominant symbiont genera is apparent (Fig. 4). Contrary to evidence that it enhances SCTLD susceptibility²⁵, Breviolum (the genus most often found to dominate P. strigosa, D. labyrinthiformis, and C. natans) did not have recognizable tissue responses in disease samples. On the other hand, samples predominantly hosting Cladocopium, Durusdinium, and Symbiodinium behaved similarly to one another, and exhibited strong relationships between measured tissue signs and disease state (Fig. 4).

Symbiont size exhibited a strong positive correlation to disease state in corals dominated by Symbiodinium. Symbiodinium cells were larger in diseased than in healthy samples, indicative of swelling; this trend was also found in corals dominated by Breviolum or Cladocopium, but with a weaker correlation (no correlation was observed in corals dominated by Durusdinium). Symbiont cells can swell during pyknosis prior to lysing, and this swelling could be how the coral host modulates a response to SCTLD within its tissues. Disease samples had smaller gastrodermal separation distances, but this could be reflective that the gastrodermal separation marker is indicative of coral host, rather than dominant symbiont genus.

By connecting quantitative histopathological traits with transcriptomic patterns, we can better understand the mechanisms driving cellular changes, such as cell expansion or shrinking and elucidate disease progression through the tissue to understand pathogenesis.

Among the host associations, Toll-like receptor 6 (TLR6) expression was positively associated with average symbiont-to-vacuole ratio. Toll-like receptors (TLRs) are highly conserved transmembrane pattern recognition receptors that have ancient roles in pathogen recognition and immune signaling^41,42. The relationship between higher TLR6 expression and reduced vacuolization around the symbiont suggests that TLR signaling helps maintain a more stable and functional symbiotic relationship. Similarly, the positive association between basement membrane-specific heparan sulfate proteoglycan core protein (HSPG2) expression and average symbiont-to-vacuole ratio indicates an important role of extracellular matrix (ECM) integrity in symbiosis maintenance. HSPG2 encodes a protein essential to the structure and function of basement membranes⁴², which are dense, selectively permeable ECM layers that separate cell types and provide structural support to epithelial tissues⁴³. Taken together, the relationship between elevated TLR6 and HSPG2 expression and higher symbiont-to-vacuole ratios suggests that these genes play beneficial roles in coral-symbiont functioning by reinforcing immune defenses and maintaining cellular structural integrity in the coral host.

Among the symbiont gene-histology associations, DnaJ homolog subfamily C member 7 (DNAJC7) expression was positively associated with symbiont size across the four genera. DNAJ proteins, such as the one encoded by DNAJC7, are known to enhance the activity of molecular chaperones like HSP70, ensuring proper protein folding and homeostasis during cellular stress⁴². This relationship suggests that symbiont cellular health and size may be partially maintained through protein quality control and repair mechanisms. Additionally, the expression of exonuclease mut-7 (MUT-7), which is involved in the RNA interference (RNAi) pathway that targets unwanted or viral RNA for degradation⁴⁴, was associated with a higher symbiont-to-vacuole ratio. If SCTLD is indeed caused by a viral pathogen or exacerbated by opportunists^15–19, higher expression of Symbiodiniaceae MUT-7 could benefit the coral holobiont by degrading viral RNA and slowing viral spread. Taken together, these results indicate that disruptions in symbiont molecular chaperone activity–leading to the accumulation of misfolded proteins and impairments in targeted RNA degradation–are likely to negatively impact the health of the coral holobiont.

Relationships between histological traits and gene expression were also coral species- and algal symbiont-specific. For example, in C. natans, the relationship between expression of the basement membrane-specific heparan sulfate proteoglycan core protein and symbiont-to-vacuole ratio was highest when the coral is healthy, and lowest when diseased. Alternatively, P. astreoides has a low susceptibility to SCTLD and shows higher expression of basement membrane-specific heparan sulfate proteoglycan core protein in diseased samples. This potentially means that while the immune defense is lowered, P. astreoides’ immune defense is not lowered to the same point as C. natans, so the coral and symbiont may still be able to communicate and defend themselves, resulting in lower gross susceptibility.

Conclusions

These findings support the model that SCTLD pathogenesis begins with coral-symbiont dysbiosis and symbiont vacuolization in gastrodermal cells. While the etiological agent or agents of SCTLD have yet to be identified, this study supports the notion that SCTLD could be a disease of the symbionts rather than of the coral animal itself. Variations in coral species-specific SCTLD susceptibility could result from subtle differences in this pathology, dominant symbiont identity, or both. Quantification of histopathological disease signs between and across coral species and symbiont genera supported the application of a Bayesian model that has uncovered species-specific trends in SCTLD pathogenesis that would otherwise have been difficult to ascertain. Further, analyses revealed important associations with gene expression, helping to contextualize both tissue characteristics and gene expression data. We demonstrate the different ways that SCTLD can affect corals depending on how those corals associate with their symbionts, how they manage their immune responses and inflammation, and which genera of symbionts they host. As this disease continues to devastate reefs in the Western Atlantic and becomes endemic, studies that probe the multifocal response to SCTLD will be crucial to mitigation.

Author contributions

Experimentation: SM, EVB, AV, AA, MB, AMSC, ICE, EMM, LM, MSS. Data collection/analysis: AMR, KMB, CEK, JB, SM, GEC, BG, EVB, AV, SRM, MSS, TBS, DMH . Writing: AMR, KMB, CEK, SM, GEC, BHF, EVB, AA, MB, AMSC, ICE, SRM, EMM, LM, TBS, MSS, DMH.

Funding

USVI experiments were supported through NSF (Biological Oceanography) award numbers 1927277 to D.M.H, 1928753 to M.E.B. and T.B.S., 1928609 to A.M.S.C., 1928817 to E.M.M., 1928771 to L.D.M., as well as 1928761 and 1938112 to A.A., NSF EEID award number 2109622 to M.E.B., A.A., L.D.M., D.M.H., and A.M.S.C. , and through the Louisiana Board of Regents Research Support Fund Research Competitiveness Subprogram (#LEQSF(2020-23)-RD-A-06) to D.M.H. Florida experiments were supported by the NOAA Coral Reef Conservation Program (#31252) and NOAA ‘Omics OAR (NO_0020) to I.E. and M.S.

Data availability

The dataset and models generated during these analyses are available by corresponding author upon request. The raw RNAseq data used in this study is deposited in the NCBI SRA database under accession code PRJNA860922. The raw ITS-2 data used in this study can be found in the NCBI SRA database under accession code PRJNA1230223. Reviewers can access with (https://dataview.ncbi.nlm.nih.gov/object/PRJNA1230223?reviewer=qus6fu35vkqr47deoi4s9bpj30) .

Declarations

Competing interests

The authors declare no competing interests.