An emerging coral disease outbreak decimated Caribbean coral populations and reshaped reef functionality

Diseases are major drivers of the deterioration of coral reefs, linked to major declines in coral 21 abundance, reef functionality, and reef-related ecosystems services 1-3 . An outbreak of a new disease 22 is currently rampaging through the populations of the remaining reef-building corals across the 23 Caribbean region. The outbreak was first reported in Florida in 2014 and reached the northern 24 Mesoamerican reef by summer 2018, where it spread across the ~ 450-km reef system only in a few 25 months 4 . Rapid infection was generalized across all sites and mortality rates ranged from 94% to < 10% 26 among the 21 afflicted coral species. This single event further modified the coral communities across 27 the region by increasing the relative dominance of weedy corals and reducing reef functionality, both in 28 terms of functional diversity and calcium carbonate production. This emergent disease is likely to 29 become the most lethal disturbance ever recorded in the Caribbean, and it will likely result in the onset 30 of a new functional regime where key reef-building and complex branching acroporids (a genus 31 apparently unaffected) will once again become conspicuous structural features in reef systems with yet 32 even lower levels of physical functionality. that SCTLD, of


86
In only a few months, SCTLD spread across hundreds of kilometers and triggered an unprecedented 87 loss of corals (Figs. 1, 2). Out of the 29,095 colonies surveyed between July 2018 and January 2020, 88 17% were already dead with signs of recent mortality (e.g., bare skeletons or covered by a thin layer of 89 filamentous algae; Fig. 1) and an additional 10% were afflicted by the disease. However, susceptibility 90 and mortality varied greatly between species. Twenty-five of the 48 recorded species were affected by 91 the disease, with disproportionate effects observed in a single morpho-functional group largely defined 92 by massive species with mid-to-large sizes, dense skeletons, low growth rates, and broadcasting sexual 93 reproduction (Fig. 1). Species from the Meandrinadae and Mussidae families were the most severely 94 affected. In particular, Dendrogyra cylindrus and Meandrina spp. experienced disease prevalence and 95 population losses greater than 80% (Fig. 1). A temporal comparison of community compositions with 96 pre-and post-outbreak data revealed that less conspicuous species, such as Dichocoenia stokesii or 97 those within the Mussinae subfamily, were noticeably less abundant after the outbreak (Fig. S1),

98
indicating that these species were more severely affected than what our post-outbreak surveys 99 suggested. This is likely because the skeletons of the small and encrusting species killed by the disease  Table S1). b) Prevalence of SCTLD for coral 106 species across 101 reef sites in the Mexican Caribbean (n = number of colonies). We included coral colonies with total mortality 107 whose deaths could be attributable to SCTLD. b) The shaded area corresponds to the prevalence of SCTLD for each morpho-108 functional group in a). The asterisk (*) indicates species with > 10% disease prevalence (these were considered highly susceptible 109 species, see Methods for more details). c) Pseudodiploria strigosa colony with the characteristic lesions produced by SCTLD in

114 115
The affectation of the SCTLD outbreak was consistent across the geography and environmental 116 gradients. The only region not afflicted by the disease outbreak (at least until the last survey in February 117 2021) was Banco Chinchorro (Fig. 2). This is an isolated offshore bank with restricted access that is 118 separated from the mainland by a deep-water channel in which the strong northward Yucatan Current 119 likely acts as a physical barrier to biological connectivity and land-based perturbations. Across all 120 surveyed sites, disease prevalence (considering both diseased and dead colonies) in highly susceptible 121 species showed no statistical differences with regard to depth, reef zone, structural complexity, wind 122 exposure, or coral density (Table S1; Fig. S1), suggesting that the main drivers of contagiousness are 123 largely controlled by the capacity of the pathogen(s) to be transported in the water column within and 124 between reef sites. However, we did find a positive relationship between disease prevalence and the 125 threat of coastal development and the age of marine protected areas (MPAs; Table S1; Fig. S2), 126 although these effects were driven by the reef sites of Banco Chinchorro and did not hold when the 127 reefs of this offshore-bank were removed from the analyses (Table S2;

131
Our results contrast with those of previous studies that have identified site-level differences in 132 the effects of SCTLD related to reef exposure, the distance from the coast, and community 133 composition 27,30 . This might be partly explained by the rapidness of the spread of the disease and the 134 high mortality rate once infected 4,21,32 . In addition, our surveys were conducted at different stages of 135 disease progression. In some sites, the disease was still evident in many colonies of susceptible species 136 (i.e., epidemic stage), while in others, the mortality of highly susceptible coral species was widespread 137 and the disease was also present in less susceptible species (i.e., endemic stage; 52 ). Therefore, our 138 broad-scale field surveys did not allow for definitive assessments of the roles of environmental or 139 ecological predictors on the rapid regional coral mortality that was observed. However, a critical pattern 140 depicted by our models indicated that when the disease reaches a site, coral mortality and the 141 disappearance of key reef-building species will most likely occur regardless of local-scale differences, 142 which drives regional changes in the structure and functions of reef communities.  Abrupt coral die-off radically reduced the abundance of species and the traits that support the 160 physical functionality of coral reefs. Most of the reefs shifted further away from the dominance of reef-161 building species that are key providers of three-dimensionality to depauperate assemblages dominated 162 by taxa with simpler morphological attributes and slower growth rates. The results of the Similarity 163 Percentage (SIMPER) analysis show that even before the outbreak, most coral reefs in the region were 164 already largely dominated by encrusting and sub-massive agaricids and Porites astreoides, which are 165 weedy coral species that accounted for 63.33% and 71.83% of the similarity between sites before the 166 outbreak and after the coral die-off, respectively (Fig. S4). The relative increase in the abundance of 167 these two groups accounted for 50.42% of the dissimilarity between periods (pre-and post-outbreak), 168 while decreases of highly susceptible species accounted for only 13.06% of the dissimilarity, as many 169 of these were either uncommon or rare species (Table S3).

171
Given that the species that suffered the most severe losses share key life-history traits (Fig. 1), 172 the functional space of the coral assemblages was considerably reduced at regional scales after the  Table S4; see methods). Ultimately, these losses were reflected in a 184 marked reduction of the capacity of coral communities to accumulate calcium carbonates (regional 185 mean ± SE; 4.60 ± 0.77 G = Kg CaCO3 m 2 yr -1 before the outbreak to 3.27 ± 0.53 G after the outbreak;

210 211
The ecology and physical functionality of coral assemblages in the Caribbean were undergoing 212 severe ecological changes prior to the SCTLD outbreak. Chronic and acute disturbances had 213 progressively driven a decline in the abundance of the main reef-building corals accompanied by a 214 concomitant increase in the relative or absolute abundance of opportunistic corals (Fig. 4). The pre-215 SCTLD communities were described as 'shifted' coral assemblages, and the contributions of formerly 216 dominant acroporids were often negligible given their reduced abundance, whereas large massive 217 species remained and contributed the most to ecosystem structure and functionality ( Fig. 4; 33,35-37 ).

218
However, the resulting wide-spread coral mortality described here was dictated by the vulnerability of 219 species to SCTLD, and thus caused non-random changes in community structure that further and 220 radically affected the functional integrity of the coral communities.

7
The morpho-functional groups comprised of large and massive species were the most afflicted 222 by the SCTLD outbreak (Fig. 1), whereas the species mildly affected by the disease showed relative 223 increases in abundance. The post-SCTLD coral communities are now represented by a hyper-224 domination of opportunistic corals, although this remarkably seems to be accompanied by an apparent 225 resurgence of acroporids as key functional elements (Figs. 3a-c, 4). However, the increase of 226 acroporids is primarily an artefact of the drastic reductions in the relative contributions of many other 227 species due to SCTLD (Fig. 4). Only a minor proportion of the increase in the contributions of acroporids 228 may be explained by population recovery or re-sheeting growth over relict reef structures 38-40 . However, 229 these increases are primarily artefacts of the drastic reductions in the relative contributions of many 230 other species (Fig. 4). In fact, the acroporid populations have remained low compared with their 231 historical estimations 14,15 , as these species have low biological connectivity, reduced genetic diversity, 232 impaired recovery abilities, and high vulnerability to regional and global stressors 15,41 . Although it is 233 encouraging that acroporid populations have remained relatively unchanged after the SCTLD outbreak,

234
it is unlikely that these species will significantly improve the structure and dynamics of rapidly changing

235
Caribbean coral assemblages given current conditions.

253
The grey-dashed arrows indicate that the source of stress remains, although the effects on widespread coral mortality are unclear.

8
The large-scale loss of the functionally important corals defined radical shifts in reef conditions 255 and dynamics, exacerbating further losses of ecological integrity along the entire reef track. Overall, the 256 outcomes of coral die-off from the SCTLD outbreak will compromise key geo-ecological functions, such 257 as reef framework production, sediment generation, the maintenance of reef habitat complexity, and 258 the capacity for coral reef growth that is needed to track future increases in sea level 11,46 . In this study, 259 we observed a nearly 30% reduction in the capacity of coral communities to produce calcium carbonate.

260
This is alarming because levels of community calcification were quite low when compared to those of 261 historical records or those in other regions of the world prior to the impacts of SCTLD 46 . While key 262 processes related to reef construction come to a halt with coral mortality, the structures provided by the 263 calcium carbonate skeletons remain in place for several years after the living tissues die. Thus, key 264 functional aspects associated with the tridimensionality of the system, such as habitat provision or the 265 modulation of water energy, will remain for a period after the death of the corals 11 .

267
In the absence of recovery, the ultimate consequences of coral mortality will thus be modulated

278
The widespread coral die-off associated with SCTLD has affected the populations of many 279 important reef-building species. In the Caribbean, the wide-spread consequences of this outbreak are 280 yet to be known. However, the rapid movement of the disease across the region 22 and the overlapping 281 distribution ranges of most species within the Greater Caribbean region 34,53 , suggest that the outbreak 282 will affect the entire region as has occurred with previous disease outbreaks 1,10 . Therefore, some 283 species will rapidly be at a clear risk of extinction across their distribution ranges (e.g., D. cylindrus; 284 Fig.1a; 29 ), while other susceptible species that underwent comparatively lower declines (30-70%; e.g.,

285
brain corals) will also have compromised abilities to overcome future sources of stress. For example, 286 the evident declines in the populations of species belonging to the Meandrinidae and Mussidae families 287 could reduce their levels of genetic diversity (e.g., 54 ), putting these species at risk of bottleneck events 288 that would limit their ability to cope with environmental change 55 . In addition, the levels of isolation of 289 the remaining colonies will reduce or hinder the capacity for sexual reproduction (e.g., allele effect) and 290 genetic recombination, further diminishing the abilities of populations to adapt to rapidly changing 291 conditions 56 . Moreover, many afflicted species are slow growing (Fig. 1, Table S5), and the replacement 292 of dead corals will undoubtedly take decades while many acute and chronic stressors operate on 293 smaller temporal scales 57 . This is particularly relevant given that corals weakened by SCTLD are likely 294 to be more susceptible to subsequent disease outbreaks and to other threats like bleaching.

296
One key question for the coming years is whether populations of highly afflicted species will be 297 able to recover and sustain key geo-ecological functions. To date, we have little evidence in this regard.

298
Empirical observations during our surveys have shown that small (< 5 cm) Meandrina, Diploria, or 299 Pseudodiploria corals have apparently remained unaffected by SCTLD, even in sites that underwent 300 severe losses in adult populations (Fig. S6). Furthermore, recent evidence shows that coral colonies 301 with evident lesions associated with SCTLD can spawn and produce viable gametes 58 . However, the 302 replacement of dead corals through larval recruitment or the growth of small juveniles will take several 303 years given low larval survival and the slow growth rates of most species 59,60 . In addition, coral 304 recruitment (i.e., the successful settlement of coral larvae) and subsequent survival will largely depend 305 on suitable ecological conditions, such as low densities of harmful fleshy macroalgae 61 . Unfortunately, 9 in our study region and elsewhere in the Caribbean, there is extensive evidence indicating that 307 macroalgae cover is progressively becoming a dominant component of benthic reef communities 62,63 . It 308 is likely that macroalgae will rapidly overtake the free space left by recently deceased corals (e.g., 28 ), 309 hindering coral recovery by impeding the settlement of new recruits and reducing the likelihood of 310 colonies that were able to slow or halt disease progression to recolonize their own structures or 311 neighboring substrates (e.g., 31,64 ). Natural processes might therefore be insufficient to restore the 312 severe population losses of many coral species due to the SCTLD outbreak. Rather, it is likely that 313 human interventions in the form of rescuing colonies of vulnerable species, preserving their genetic 314 material, and implementing restoration efforts will be needed to facilitate recovery and prevent the 315 region-wide extinction of some species (e.g., 29 ). We believe, however, that these actions will only 316 succeed if they are accompanied by stringent controls that take into consideration climate change, 317 coastal development, and wastewater treatment to improve local conditions and ecosystem resilience.

322
Field surveys and SCTLD prevalence

324
To assess the spread and impacts of SCTLD, extensive surveys were conducted across the Mexican 325 Caribbean between July 2018 and January 2020 (post-outbreak period). In total, 101 sites were 326 surveyed (82 fore-reefs, 19 back-reefs, and four reef-crests) in depths ranging from 1-24 m (Fig. 2).

327
Thirty-five of these sites were also surveyed in 2016 and 2017 (pre-outbreak period) as part of a 328 separate effort 45 and were used to investigate the ecological and functional consequences of the 329 SCTLD outbreak. These 35 resampled sites are also distributed across the Mexican Caribbean (Fig. 2) 330 and cover similar habitats and depth gradients (thirty reef sites are fore-reefs and five are back-reefs).

331
For the pre-outbreak period, white plague-type disease prevalence is reported as there were no reports 332 of SCTLD for that period.

334
All sites were surveyed using the Atlantic and Gulf Rapid Reef Assessment protocol 65 . At each 335 site, coral assemblages were surveyed in 10 x 1 m transects. For the pre-outbreak period, 1-7 transects 336 (mean = 2.8; SD = 1.4) were evaluated in each site. For the post-outbreak period, we considerably 337 increased the effort to ensure the representation of uncommon and rare species that we knew were 338 affected by SCTLD, and between 3-23 transects (mean = 8; SD =3.71) were conducted in each site.

339
The following information was recorded for each coral colony within each transect: species name, 340 colony size (maximum diameter, diameter perpendicular to the maximum diameter, and height), 341 bleaching percentage, mortality percentage (new, transition, and old), and the presence of SCTLD or 342 other diseases 65 . For this study, we also recorded colonies with 100% mortality that could be attributed 343 to SCTLD (i.e., recent or transient mortality was still evident; e.g., Fig. 1e).

345
We focused on exploring the prevalence and geographical and temporal trends of the most 346 'highly susceptible species,' which were species with more than 10% disease prevalence (considering   We model the percentage of afflicted colonies as a function of coral colony density (prior to the impacts 361 of the disease), reef structural complexity, reef zonation, depth, and the degree of exposure to dominant 362 winds. In addition, we evaluated the influence from land-based human activities using the Coastal 363 Development level (World Resources Institute database, 66 ) and protection status using MPA age.

364
These variables were selected based on their importance to coral reef health and the known 365 susceptibility of coral assemblages to disturbance ( 45,67 , see Table S4 for details), depending on the 366 availability of information for the 101 sites. Water temperature or thermal stress were not included, as 367 remote sensing data do not capture local variation at the necessary resolution (4 km; 68 ). Furthermore,

368
previous studies have shown that temperature is a poor predictor of the spatial variation in coral reef 369 conditions in this ecoregion 69 and that high water temperatures do not affect SCTLD prevalence or 370 virulence 4,23,27,28 . We used generalized linear models (GLMs) with the glm function in R 70 , setting the 371 percentage of afflicted colonies as the response variable and the aforementioned factors as the 372 predictive variables. A gaussian distribution with an identity link function was adopted as the goodness 373 of fit test and indicated a normal distribution for the response variable. The data were not transformed, 374 but all numerical predictive variables were scaled (i.e., z-scores with the scale function in R). Statistical 375 analyses were carried out using a 95% confidence interval (α = 0.05), and model assumptions were 376 validated with residual plots. Regression models were constructed in R v. 3.6.1.

378
Coral morpho-functional groups and community shifts

380
The functional diversity of coral communities was estimated using six different traits: skeletal density, 381 growth rate, rugosity index, colony size, reproduction strategy, and corallite width (Tables S5, S6).

382
Some or all these traits have been previously used in other studies to represent the functioning of reef-383 building corals in a multidimensional space 34,71 . The information on species-level traits was obtained 384 from different sources that provide comprehensive details for each selected trait ( 33,72,73 ; Table S1). To 385 better compare traits considering the contribution disparities between species (see 74,75 ), the traits were 386 categorized into numerical groups (1-5). Hierarchical clustering was performed to identify groups in the 387 data set and estimate trait similarity. We then grouped the reef corals into morpho-functional groups 388 using a Gower dissimilarity matrix ('vegan' package in R; 76 ) and average-linkage hierarchical clustering, 389 which calculates the average distance between clusters before merging ('cluster' package in R; 77 ).

390
Groups were defined using 65% dissimilarity because it was the most evident grouping and provided a 391 concise number of groups (Fig. 1).

393
We measured the contribution of each species to compositional similarity within periods and 394 dissimilarity between periods with a Similarity Percentage analysis (SIMPER; 78 ). SIMPER identifies the 395 species that are most responsible for the observed patterns (e.g., the species that typify each factor 396 level and those that contribute the most to dissimilarity between levels) by disaggregating the Bray-

397
Curtis similarities between samples. The more abundant a species is within a group, the more it 398 contributes to intra-group similarity; species with consistently high contributions to the dissimilarity 399 between groups are good discriminating species 79 .

401
To explore temporal changes in coral composition and the traits of those assemblages, a

414
Coral community calcification

416
To estimate the effects of SCTLD on the physical functionality of coral reefs, we calculated the potential 417 calcification (i.e., kg CaCO3 m -2 yr -1 ) of the coral assemblages for each period using the sum of the 418 calcification rate of each colony proportional to the sampled reef area (m 2 ). We used a reef-level 419 estimation to calculate mean coral community calcification for each period. For this, we estimated the 420 calcification rate of each colony within each study site considering the size, mean annual growth rate 421 (cm yr -1 ), mean skeletal density (g cm -3 ), and morphological growth of each species following the