Coral Reefs

, Volume 30, Issue 1, pp 73–84

Macroecological relationships between coral species’ traits and disease potential

Authors

    • Department of Biological SciencesMacquarie University
    • Facultad de Ciencias, Departamento de Ciencias Biológicas, Laboratorio de Biología Molecular Marina (BIOMMAR)Universidad de los Andes
  • J. Madin
    • Department of Biological SciencesMacquarie University
Report

DOI: 10.1007/s00338-010-0668-4

Cite this article as:
Díaz, M. & Madin, J. Coral Reefs (2011) 30: 73. doi:10.1007/s00338-010-0668-4

Abstract

Coral disease is a growing problem for reef corals and a primary driver of reef degradation. Incidences of coral disease on the Great Barrier Reef (GBR) are increasing; however, our understanding of differences among species in their potential for contracting disease is poor. In this study, we integrate observations of coral disease on the GBR from the primary literature as well as morphological, ecological and biogeographical traits of coral species that have been hypothesised to influence “disease potential.” Most of the examined traits influence species’ disease potential when considered alone. However, when all traits are analysed together, diversity of predators, geographical range size and characteristic local abundance are the primary predictors of disease potential. Biases associated with species’ local abundance and phylogeny are tested but do not overpower relationships. This large-scale macroecological evaluation of coral disease provides insights into species-level traits that drive disease susceptibility.

Keywords

Reef coralsDisease potentialSpecies traitsMacroecologyGreat Barrier ReefPredation

Introduction

Coral reefs are a productive and economically important, yet highly threatened, global ecosystem (Lesser et al. 2007; Mumby and Steneck 2008). Studies reveal declines of 20 and 5% in coral cover per decade in the Caribbean (Gardner et al. 2003) and the Indo-Pacific (Bruno and Selig 2007), respectively. Projections suggest that up to 30% of reefs worldwide will be severely damaged by 2030 (Hughes et al. 2003). The ubiquity of reef decline over the last millennium has led to the conclusion that no pristine reefs remain (Pandolfi et al. 2003). Causes of coral decline are numerous, including climate change (predominantly increasing sea water temperature), overfishing and destructive fishing techniques, coral bleaching, coral disease, trophic level dysfunction (phase shifts), predation and pollution (Lesser 2004; Mumby and Steneck 2008; Fabricius 2005). Of these threats, coral disease is now considered a key driver of decline for corals and the reefs they build (Porter et al. 2001; Sutherland et al. 2004), and a problem that is aggravated by environmental stressors caused by human-induced change (Jones et al. 2004; Selig et al. 2006; Bruno et al. 2007; Francini-Filho et al. 2008).

A recent review found 19 described coral diseases, four of which have a worldwide distribution, 9 of which are found only in the Caribbean and 6 of which are endemic to the Indo-Pacific region (Sutherland et al. 2004). Although the Caribbean is considered the coral disease hotspot, there have been increasing reports of disease on the Great Barrier Reef (GBR) (Antonius and Lipscomb 2001; Willis et al. 2004). However, individual reports are disease-based and typically focus on few coral species and at specific geographical locations, and therefore our understanding of disease susceptibility at broader scales and across all species is limited. One exception is a study by Willis et al. (2004) in which disease observations were conducted up and down the GBR providing an assessment of reef condition through time. However, studies to date have stopped short of assessing the ecological, morphological and environmental characteristics of coral species that influence their potential for infection. Therefore, in this study, we integrate the current state of knowledge about coral disease on the GBR and paint a macroecological picture of the species-level traits that are likely to influence a species’ potential to contract disease. Given the heterogeneous data collected in coral disease studies (explained further in the methods), and subsequent lack of sampling standardised disease prevalence data, we define “disease potential” simply as whether a scleractinian coral species has been observed in the primary literature with a disease. We select a broad range of species-level traits that have been suggested to influence disease and include ecological (local abundance, number of predators, range size), life history (reproductive mode), morphological (corallite size, growth form complexity) and environmental (wave exposure, water clarity, depth range) information.

Similar to human populations, coral species living at higher local abundances have been shown to be more susceptible to disease (Willis et al. 2004; Page and Willis 2008). For example, there has been an increase in the prevalence of the coral disease white syndrome in areas where coral cover is greater (Willis et al. 2004). The underlying mechanism is that disease can spread more readily within crowded populations, suggesting that locally common species have a greater disease potential than rare species (Aeby and Santavy 2006). A potential vector for disease transmission that is exacerbated in dense populations is coral predation. Predators act as vectors by oral or faecal transmission of pathogens (Aeby and Santavy 2006; Rotjan and Lewis 2008). Some diseases, such as black band disease, are thought to flourish in the presence of corallivorous fishes and the gastropod Drupella spp. that are suspected to increase the rate at which the disease is spread from infected to non-infected colonies (Antonius and Riegl 1997; Aeby and Santavy 2006). Predators also increase disease potential by making scars in corals that allow pathogens to penetrate and infect tissues (Page and Willis 2008). Certain diseases (such as skeletal eroding band) require a tissue lesion in order to infect corals (Page and Willis 2008). For example, lesions may be driving the positive association between coral polyp-eating chaetodontids and disease prevalence (Raymundo et al. 2009).

Coral polyp size is a morphological characteristic that is related to coral predation and may also be related to disease potential. For instance, most corallivorous fishes belong to the families Chaetodontidae (butterflyfishes) and Labridae (wrasses) that have small mouths often specialised for the removal of tentacles from small individual polyps (Aeby and Santavy 2006). Colony growth form is another important characteristic that is related to energy allocation among physiological processes such as growth and colony defence. For instance, branching corals invest more energy in growth and therefore allocate less energy to maintenance and disease resistance (Jackson 1979; Palmer et al. 2008). Mass transfer also tends to be reduced in more complex branching morphologies (Chamberlain 1978; Nakamura and van Woesik 2001) where lower internal water velocities slow or stop mucus and sediment shedding and increase the chance of pathogen infection. Complex colony morphology is also related with a higher degree of physiological integration between the polyps (Soong and Lang 1992) that is likely to increase the probability of disease spread within colonies.

Hydrodynamic exposure levels associated with coral species’ preferred habitats are also expected to influence the potential for corals to contract disease for a similar reason to mass transfer within colonies (i.e., the flushing of accumulated materials potentially containing pathogens). Indeed, disease prevalence in the genus Acropora is found to be greater in areas of the reef that are sheltered from wave action and lesser in exposed habitats such as the reef crest (Willis et al. 2004). Coral disease prevalence also increases with levels of nutrients and sedimentation generated by terrestrial run-off and exacerbated by human activities such as coastal development, deforestation and agriculture (Bruno et al. 2003; Francini-Filho et al. 2008). Some coral species have adaptations to cope with these conditions, including polyp retraction, lowered photosynthetic (and thus metabolic) rates and increased mucus production (Sofonia and Anthony 2008; Lirman and Manzello 2009), each of which might decrease disease infection rates in turbid water species. Therefore, as terrestrial influences spread further from land, corals without such adaptations (e.g., clear water dwelling outer reef colonies) are expected to have a greater disease potential.

Several other species’ traits are hypothesised to be important for coral disease potential, but have not yet been widely tested. First, a species with a broader geographical distribution is more likely to intercept disease and subsequently spread it to other parts of its distribution via vectors such as predators (especially if a disease and/or disease vector is strictly associated with one or few coral species). Second, several studies have shown that high temperatures increase the prevalence of disease (Jones et al. 2004; Selig et al. 2006; Bruno et al. 2007). The upper depth limit of coral species is directly associated with water temperature (Kuta and Richardson 1996), and therefore species with higher upper depth limits (closer to the ocean surface) are expected to have a greater potential for disease contraction than deeper dwelling species. Third, coral species are typically categorised as either larval brooders or broadcast spawners, representing a life history strategy that can greatly influence dispersal from parent populations and subsequently proximity to disease. Furthermore, studies have shown that broadcast spawners are more resilient to certain stressors, because recruits are capable of colonising different habitats and adults produce a higher number of propagules with higher dispersal capacity (Glynn and Colley 2008; Baird et al. 2009). For example, broadcast spawning species have been found to recover faster after bleaching events in the Indo-Pacific and in the Arabian Gulf (Glynn and Colley 2008) and therefore might be more resistant to disease infection as well.

This study tested associations between disease observations in the literature and ecological, life history, morphological, and environmental characteristics of coral species by compiling a data set of species-level traits for all known species in the GBR region. The outcome is the first ever macroecological trait-based analysis of disease potential, which provides important insights into coral traits that drive disease susceptibility globally.

Methods

Species-level traits

Coral species-level trait information was collected for all known (406) scleractinian coral species found on the GBR. Species and family data were assigned according to Veron and Stafford-Smith (2002), Wallace (1999), and Carpenter et al. (2008). For each species, 9 different traits were recorded. Characteristic local abundance was assigned as rare, uncommon or common (following Veron and Stafford-Smith 2002; Carpenter et al. 2008), which is a general metric of local abundance developed by Veron (1986) based on ecological and taxonomic information throughout the GBR region. Given the low number of rare species, rare and uncommon categories were grouped, resulting in two local abundance classes: common and uncommon. Geographical range size was calculated as the global area occupied by a species found on the GBR measured as number of map pixels (Veron and Stafford-Smith 2002). Colony growth form was assigned following descriptions and pictures from Wallace (1999) and Veron and Stafford-Smith (2002) as either solitary, encrusting, massive, columnar, foliaceous, digitate, branching, tabulate or corymbose. Corallite size was obtained from a variety of taxonomic monographs (Veron and Pichon 1976, 1980; Veron et al. 1977) or measured directly from scaled pictures (Veron and Stafford-Smith 2002). For the genus Acropora, the corallite size information was obtained from Wallace (1999). Wave exposure was assigned as protected, exposed or broad (protected and exposed) (Veron and Stafford-Smith 2002). Preferred water clarity was assigned as either turbid, clear or both (turbid and clear) (Veron and Stafford-Smith 2002). The upper (shallowest) depth at which the coral is found was obtained from Carpenter et al. (2008). Reproductive mode was assigned as either brooder or spawner (Baird et al. 2009). The number of predatory species was determined by an extensive review of the primary literature for corallivorous animals on the GBR and the coral species they have been observed to prey upon (361 species-level observations). These predator species estimates are likely to substantially underestimate true coral predation and thereby reduce the likelihood of uncovering a strong association with disease (i.e., the number of predators is a conservative estimate).

Coral disease observations

A total of 210 species-specific disease observations (for 95 species) were found in the primary literature that included nine disease types (white syndrome, white patch, pigmentation response, tumours, black band disease, skeletal eroding band, brown band, atramentous necrosis or unidentified pathologies). Studies spanned the GBR, but were undertaken with a diverse range of objectives and tended to focus on diseases rather than details about their coral hosts (see Electronic Supplemental Material, ESM table). For example, twice as many disease observations were found in the literature (>400), but coral hosts were not described sufficiently for species-level analyses (i.e., they were recorded at higher taxonomic levels or by growth form). Furthermore, coral disease studies almost exclusively reported presence-only information, thereby preventing the extraction of more robust sampling standardised measures such as disease prevalence. The several studies that did report disease prevalence information focused on single coral diseases, which limited the generality for all disease types. There were too few studies about GBR coral disease to standardise disease observations based on the frequency at which a species is reported in the literature. Finally, observation methodologies differed dramatically among studies, ranging from sampling replicate belt transects of known area through to targeting a certain number of colonies with a given disease type.

Given the data at hand, it was not possible to look at species’ disease susceptibility using sampling standardised measures of prevalence. Therefore, to improve analytical power and use all the available information, disease data were grouped so that a species observed to have any disease anywhere on the GBR was categorised as having the “potential” to contract disease (i.e., a binomial response variable). This definition of disease potential assumes that the literature reflects species’ vulnerability to disease in the field and that species not observed with disease have a lower potential for infection (but not necessarily zero potential). This assumption results in the chance of underestimating the actual potential for species that are poorly studied, are locally rare, or have restricted geographical distributions (type II errors), which we test for by contrasting disease patterns for locally common and uncommon species separately. The data sets described earlier are available from the corresponding author upon request.

Data analysis

Relationships between coral species’ characteristics and disease potential were analysed: (1) separately to gain a better base-line understanding of patterns of disease potential within traits, (2) altogether using a generalised linear mixed-effects model (GLMM) and multiple regression with phylogenetically independent contrasts (PIC) to account for the non-independence of species and determine which traits best predict disease potential and (3) again using a GLMM after splitting the data set into common and uncommon species to test if observational biases are overpowering the results.

Separate trait analyses assessed if a given categorical grouping of species (e.g., all species with a “branching” growth form) had on average a greater potential to contract disease than all other species’ groupings. Exact binomial tests were used to determine if average disease susceptibility for a given grouping was significantly different to that of all other groups. Results are presented as the proportional difference between focal groupings relative to all other groupings (i.e., bars above zero indicate that a group is more susceptible on average, and vice versa). 95% confidence intervals illustrate any overlap with zero (“no difference”). Statistical differences are indicated for each focal group using stars (* = P-value <0.05; ** = P-value <0.01; *** = P-value <0.001; ns = not significant). Logistic regression is used for analysis of the continuous predictor variables (e.g., geographical range size). Best-fit logistic estimates are presented, because overlapping points conceal underlying trends.

GLMMs allow for (1) both continuous and categorical predictor variables, (2) a non-linear response variable (Venables and Ripley 2002) and (3) the removal of group effects by randomising the group variable (Zuur et al. 2009). This latter feature is accomplished by removing variation due to differences among groups (e.g., taxonomic family or genus) from the error term and allowing the groups to vary randomly around the overall mean (Pinheiro and Bates 2000; Krackow and Tkadlec 2001). Other independent variables (fixed effects) can then be examined, and any significant results can then be generalised for all species independently of the randomised group (Pinheiro and Bates 2000; Jovani and Serrano 2001). PIC is an algorithm used to correct for the non-independence of species (Felsenstein 1985). For PIC analyses, we use the scleractinian coral supertree produced by Kerr (2005) that captures 286 of the 406 study species. The polytomies on the consensus tree were resolved using the function multi2di, from the R package APE (Paradis et al. 2004; R Development Core Team), which transforms all the polytomies into dichotomies. Branch lengths were treated as of unit length.

Disease potential was considered to have a binomial distribution of error (observed in the literature with disease = 1, not observed with disease = 0; these data were examined using a logit link function). GLMMs were run with the function glmmML from the R package glmmML (Broström 2009; R Development Core Team). PICs were run for each variable (including disease potential) with the function pic from the R package picante (Kembel et al. 2010; R Development Core Team). PICs map variables (continuous, binomial, and categorical) to continuous normal variables that were analysed using a simple linear model (i.e., multiple regression). To determine the best predictive model for disease potential, we compared full models with models in which one of the predictor variables are dropped (using the “drop1” function in the R base statistics distribution). If an analysis of variance found a dropped variable to have no significant effect on the model, then the variable was dropped. Interactions (up to two-way) were examined and dropped in the same fashion.

Analyses suggest that coral species that are characteristically locally common are significantly more susceptible to disease than are uncommon species (Fig. 1a). Despite possible biological explanations for this result (see “Discussion”), it raises questions about observational biases. For example, there is a higher probability of observing a common coral species on the reef, and subsequently a higher probability of observing disease in common species. Also, researchers tend to focus on locally abundant species, because they are easier to locate in the field. Therefore, to test that “species commonness” is not driving disease patterns, we repeated the GLMM analysis (with family as a random factor) separately for common and uncommon species. Any substantial divergences between full and partial analyses would suggest that observational biases are overpowering disease potential/trait analysis patterns.
https://static-content.springer.com/image/art%3A10.1007%2Fs00338-010-0668-4/MediaObjects/338_2010_668_Fig1_HTML.gif
Fig. 1

Categorical proportions of disease observation relative to expectation for species’ a local abundance, b growth form, c wave exposure, d water clarity and e reproductive mode. 95% binomial confidence intervals and significance levels show if proportions can be distinguished from expectation. See “Methods” for further details

Results

Separate trait analyses

Families

Family groupings exhibited a large spread of disease susceptibilities, ranging from ~20% less than to ~20% greater on average than the expected disease potential based on the total species mean (Fig. 2). However, the majority of differences were not significant. Acroporidae is the only family with a greater potential to contract disease. Conversely, Fungiidae and Agariciidae are the only families significantly less susceptible. Locally common families such as Pocilloporidae and Faviidae appear to have a greater disease potential on average, while the common family Poritidae appears to be less susceptible. However, these differences cannot be statistically distinguished from the total species pool.
https://static-content.springer.com/image/art%3A10.1007%2Fs00338-010-0668-4/MediaObjects/338_2010_668_Fig2_HTML.gif
Fig. 2

a The number of species observed with a coral disease in the GBR grouped by family. b Differences in family proportions of disease observation relative to the expected proportion for all families. 95% binomial confidence intervals and significance levels show if proportions can be distinguished from expectation. See “Methods” for further details

Coral predators

Corals have a wide range of predators noted in the literature, including 15 families of corallivorous fishes and invertebrates (46 species), which prey on 11 coral families, comprising of more than 66 coral species. The dominant fish predators, Chaetodontidae and Scaridae, prey on 65 and 4 coral species, respectively. The dominant invertebrate predators, the crown-of-thorns starfish, Acanthaster planci, and the gastropod Drupella spp. targeted 8 and 4 coral genera, respectively. There was a marked relationship between the number of predator species that eat coral species and disease potential (P < 0.001; Fig. 3a). All coral species observed with more than ~3 predators have been observed with a disease regardless of their characteristic local abundance or taxonomic affinities. However, having fewer or no predators does not necessarily imply that a coral species does not have the potential to contract disease.
https://static-content.springer.com/image/art%3A10.1007%2Fs00338-010-0668-4/MediaObjects/338_2010_668_Fig3_HTML.gif
Fig. 3

Disease potential as a function of a the number of predator species, b geographical range occupied by a species (measured as number of pixels), c polyp diameter and d upper depth limit. Significance levels for logistic regressions are given in adjacent to fitted curves. See “Methods” for further details

Geographical range size and habitat preference

Logistic regression of disease potential as a function of geographical range size illustrates that coral species with larger ranges tend to be much more prone to disease than those with more restricted distributions (Fig. 3b). Species that prefer protected reef areas are found to be ~ 40% less susceptible to disease than species that prefer exposed environments or are found in a wide range of habitats (protected and exposed). Protected habitat reef species have a significantly lower disease potential, whereas the other two groups have a significantly greater potential (Fig. 1c). Clear water dwelling species are ~30% (and significantly) more likely to contract disease than turbid water dwelling species and than species found in both turbid and clear environments (Fig. 1d). Finally, species with shallower upper depth limits appear to have a greater disease potential than species preferring deeper waters (Fig. 3d). The pattern for upper depth is not significant when analysing uncommon species alone, suggesting a strong observational bias (e.g., shallower species are intercepted more by researchers).

Growth form, polyp diameter and reproductive mode

There was a general increase in disease potential as a function of morphological complexity (Fig. 1b). Robust and simple forms such as massive and encrusting have an up to ~17% reduced potential to contract disease, whereas intermediate morphologies such as columnar and foliaceous almost show no difference or are not significantly different from the proportion expected for all species. Branching and corymbose forms, which are delicate and complex, have significantly higher disease potentials than the average of all the other growth forms. Finally, solitary corals that are generally one large polyp tend to have a very low potential for disease infection. When analysing uncommon species alone, corymbose species were the only morphological group with a significantly higher potential to contract disease. A negative relationship was found between polyp diameter and disease potential (Fig. 3c), where colonies with larger polyps have a significantly lower potential for disease contraction. There was no difference in disease potential between brooding and spawning coral species (Fig. 1e).

Multiple trait analyses

Multiple trait analyses consistently illustrate that the most important predictors of coral species disease potential are geographical range size and the number of corallivorous predatory species (Table 1a). Slightly less important is characteristic local abundance. All other species’ traits (and any interactions among traits) did not significantly alter the predictive power of the model when they were removed. That is, species growth form, polyp size, upper depth limit, water clarity, hydrodynamic exposure and reproductive mode are not important predictors of disease potential when compared with the predictive power of geographical range size and number of coral predator species. The additional GLMM analyses for common and uncommon species uncovered the same general pattern as for the full model despite the loss of statistical power; although, marginal changes occurred in the significance levels for range size and predation, and growth form becomes marginally significant for uncommon species. Nonetheless, there was remarkable agreement in the model estimates for all models (Table 1b). We use the GLMM with family as a random factor to illustrate the predictions arising from the analyses (Fig. 4) because PICs map variables onto uninterpretable scales. When variance from phylogenetic relatedness between species was controlled, PIC results showed that species could be treated as independent units; therefore the phylogenetic component is not affecting the patterns found with the GLMM. The multiple regressions done with the PIC data showed that the main variables involved in disease susceptibility were the same found with the GLMM.
Table 1

Summary statistics table for multiple traits analyses

a

Variable

Family

Genus

PIC

Family (common only)

Family (rare only)

df

AIC

Pr (Chi)

df

AIC

Pr (Chi)

df

AIC

Pr (Chi)

df

AIC

Pr (Chi)

df

AIC

Pr (Chi)

Number of predators

1

316.1

0.000***

1

308.4

0.000***

1

2.86

0.000***

1

212.1

0.000***

1

121.6

0.029*

Geographical range size

1

316.8

0.000***

1

312.0

0.000***

1

6.09

0.000***

1

203.5

0.004**

1

133.2

0.000***

Local abundance

1

300.8

0.007**

1

291.0

0.021*

1

−6.94

0.002**

NA

NA

NA

NA

NA

NA

Growth complexity

8

293.1

0.098

8

273.9

0.973

1

−16.6

0.925

8

191.2

0.281

8

119.5

0.034 *

Water clarity

2

296.1

0.111

2

285.9

0.332

1

−16.5

0.710

2

195.1

0.446

2

120.3

0.065

Upper depth

1

296.2

0.111

1

289.4

0.053

1

−16.4

0.664

1

196.5

0.293

1

120.0

0.075

Polyp size

1

294.1

0.511

1

287.1

0.236

1

−16.1

0.474

1

195.6

0.693

1

119.1

0.135

Wave exposure

2

292.1

0.816

2

284.1

0.813

1

−16.4

0.638

2

194.2

0.700

2

115.7

0.665

b

Variable

Family

Genus

PIC

Common only (family random)

Rare only (family random)

Est.

SE

P-val

Est.

SE

P-val

Est.

SE

P-val

Est.

SE

P-val

Est.

SE

P-val

Intercept

−6.96

1.09

0.000***

−8.64

1.42

0.000***

NA

NA

NA

−6.82

1.51

0.000***

−7.84

1.49

0.000***

Number of predators

1.63

0.32

0.000***

1.87

0.41

0.000***

0.17

0.03

0.000***

1.55

0.36

0.000***

1.93

0.65

0.002**

Geographical range size

0.0006

0.0001

0.000***

0.0008

0.0001

0.000***

0.00007

0.00001

0.000***

0.0006

0.0002

0.003**

0.0006

0.0002

0.003**

Local abundance

−1.00

0.34

0.003**

−1.05

0.41

0.011*

−0.16

0.05

0.001**

NA

NA

NA

NA

NA

NA

The first two columns summarise GLMMs with either taxonomic family or genus as the random variable. The middle column summarises a multiple linear regression with variables following phylogenetic corrections using PICs. The final two columns summarise GLMMs with family as the random variable for common and uncommon species separately. (a) The effect of dropping each predictor variable separately from full models (showing both AIC and chi-square test statistics). (b) The final predictive model (showing estimates, standard error, and P-values)

https://static-content.springer.com/image/art%3A10.1007%2Fs00338-010-0668-4/MediaObjects/338_2010_668_Fig4_HTML.gif
Fig. 4

A contour plot of predicted disease potential as a function of both species’ geographical range sizes and number of predatory species, illustrated separately for characteristically a common and b uncommon species

Discussion

With the exception of reproductive mode, all coral species traits collected and analysed in this study displayed clear and significant trends with respect to disease potential when analysed separately (Figs. 1, 3). In general, species observed in the literature as having fewer predators, restricted geographical ranges, lower abundances, simple growth forms, larger corallites, and a preference for protected, turbid and deeper reef habitats tend to be observed with disease less frequently than species that have more predators, broad geographical ranges, high abundances, complex growth forms, smaller corallites, and a preference for exposed, clear and shallower reef habitats. However, when all traits were analysed together, geographical range size, the number of predatory species and characteristic local abundance were consistently the most significant predictors of a coral species’ potential to contract disease (Table 1). Figure 4 illustrates the predictions for disease potential for a coral species stemming from the study’s analyses.

Locally common species have a higher potential to contract disease than uncommon species (Table 1a; Fig. 1a). This conspicuous result prompted the investigation into possible observational biases associated with sampling scientific literature (i.e., a higher probability of observing common species within a sampling area). Distinguishing between observation biases and a genuine biological signal is not possible without a standardised sampling approach from the literature (which is not possible given limitations of the disease literature; see “Methods”). Nonetheless, the GLMM and separate trait analyses show no substantial changes in either pattern or statistical significance when characteristically common or uncommon were analysed separately (Table 1). This result suggests that species’ abundance does not interact substantially with most traits and subsequent trait/disease potential patterns are presumably genuine macroecological signals. In support of this result, several studies have found that population density is an important driver of disease susceptibility (Willis et al. 2004; Bruno et al. 2007), which is likely facilitated by increased rates of disease transmission as a result of diminished distances between individual colonies of the same species. Host density can also be related to vector abundance (e.g., coral predators; Bruno et al. 2007; Raymundo et al. 2009). The allocation of resources to competition in more densely populated reefs can also render species more vulnerable to pathogen transmission and infection (Bruno et al. 2007). For example, there is a distinct relationship between coral cover and disease incidence for white syndrome (Bruno et al. 2007), whereas coral cover has no effect on disease prevalence for black band disease (Page and Willis 2006). Despite these purported differences in disease transmission within coral assemblages, the results presented here suggest that, as a whole, local abundance is a significant predictor of a species’ potential to contract disease.

Our analyses of coral families corroborate reports that members of the family Acroporidae are more likely to contract disease on the GBR (Willis et al. 2004; Ulstrup et al. 2007; Page and Willis 2008). Despite this higher susceptibility and the shear number of species in Acroporidae, the study’s trait/disease potential patterns did not change appreciably when phylogenetic relationships were taken into account, indicating that this family is not driving the study’s broader results. Other families were no different or less susceptible to disease than expected (Fig. 2). The susceptibility of Acroporidae species may explain the observed declines in coral communities since acroporids dominate GBR reefs and are found in almost every reef habitat, making them one of the most important reef builders (Wallace 1999; Willis et al. 2004; Page and Willis 2008). Furthermore, the two species of Acropora found in the Caribbean (A. palmata and A. cervicornis) were almost driven to extinction by disease in the 1990s (Richardson 1998). Page and Willis (2008), however, found pocilloporids to be more susceptible to disease than acroporids, concluding that pathogenic microorganisms disproportionately target faster growing corals and not necessarily the most spatially dominant species. While our regional-scale analysis finds that pocilloporids have a higher disease potential on average, this proportion is not significantly different to expectation due to this family’s small number of species. Previous work has also found that species from the families Poritidae and Faviidae have a reduced disease potential on average (Willis et al. 2004; Page and Willis 2008); however, our analyses suggest that their potential to contract disease is not significantly different from expectation. Interestingly, the family Fungiidae has a significantly reduced potential to contract disease compared with other families. Fungiids are typically solitary, single-polyp forms and therefore are likely to invest in disease resistance strategies to defend this single polyp. Additionally, this family does not have any reports of predators, which we found to be the primary trait driver of disease potential.

Geographical range size emerges as a highly significant factor in determining disease potential, where more widespread species are expected to be more susceptible to disease in general (Table 1; Fig. 4). Most species range more widely than the study’s coverage (the GBR, approximately 200 map pixels) and study locations in the literature (typically the same research stations and surrounding reefs), indicating that this result is not a geographical sampling bias. One explanation is that greater ranging species have a higher probability of intercepting pathogens and passing them to con-specifics. However, because most coral diseases are not host-specific, they can be passed readily from broader to narrower ranging species. A more plausible explanation is that predators tend to target broader ranging coral species in order to increase the probability of having local resources following dispersal. Indeed, there is a positive triangular relationship between the number of coral predator species and geographical range size (Fig. 5), suggesting that coral predation is likely to be the primary causal driver because it tends to operate increasingly on broader ranging coral species.
https://static-content.springer.com/image/art%3A10.1007%2Fs00338-010-0668-4/MediaObjects/338_2010_668_Fig5_HTML.gif
Fig. 5

The triangular relationship between geographical range size of coral species and number of predator species. A generalised linear model (with a Poisson error distribution) illustrates the underlying trend

Although species’ growth form was not a significant predictor of disease potential when all traits were analysed together, the separate analysis of growth form here confirms the current hypothesis that massive and encrusting corals tend to have a reduced potential to contract disease. Simple growth forms have more effective mass transfer, by which toxic metabolites produced by the pathogenic microorganism can be easily expelled from the polyp and water flux subsequently removes pathogens and metabolites lingering on the surface (Nakamura and van Woesik 2001). It has been reported that the pathogens responsible for black band disease are usually found in sediments within the crevices and folds of coral colonies (Richardson 1998), making more convoluted morphologies more prone to sediment accumulation and therefore more prone to infection (e.g., digitate and undulating colony patterns). Complex morphologies have been suggested to be associated with faster growing species that spend more energy on growth and less on defence mechanisms (e.g., Acropora millepora; Palmer et al. 2008). Meanwhile, slow growing, simpler forms tend to invest more energy in the implementation of stronger defence (e.g., production of melanin, phenoloxidase and special proteins; Palmer et al. 2008) and regeneration systems since they cannot recover as quickly from partial colony mortality (Jackson 1979). Finally, corals with these complex growth forms (especially acroporids) tend to have higher colony integration than massive corals, where polyps are physiologically independent (Soong and Lang 1992). This is hypothesised to affect bleaching susceptibility (Baird and Marshall 2002) and could also act as a mechanism for pathogen transmission within the colony. Despite the plausible mechanisms for greater disease potential in more complex growth forms, the analyses presented here suggest that growth form is not a dominant predictor when compared with coral predation and geographic range size.

When all traits were analysed together, polyp diameter was found not to be an important driver relative to other species-level traits. Nonetheless, when analysed separately, a species’ polyp diameter also shows a clear association with disease susceptibility, where smaller corallite species tend to be more susceptible to disease (Fig. 3c), and might explain why acroporids and pocilloporids are particularly prone to infection. Colonies with larger polyps and lower densities per unit surface area might be more resistant, because each polyp contributes in greater proportion to the colony’s energy budget. Larger corallites also appear to be less targeted by coral predators, which are hypothesised to be a primary transmission vector of coral diseases (Aeby and Santavy 2006; Raymundo et al. 2009), and there is evidence that the scars left by A. planci on the coral tissue might promote the transmission of some coral disease (Nugues and Bak 2009). The most common predators (from reef fish families Chaetodontidae, Labridae, and Monacanthidae) have small mouths specifically designed for removing tissue from individual polyps and therefore specifically target colonies with small corallites (Rotjan and Lewis 2008). A recent study found that chaetodontids are the only fish family to be significantly and positively related with coral disease prevalence, supporting the hypothesis that they act as vector of disease (Raymundo et al. 2009). Moreover, chaetodontids have been found to specifically target diseased tissue, decreasing the subsequent disease spread within colonies, but illustrating the potential for the fishes to interact with disease and act as vectors (Cole et al. 2009).

Although our multi-trait analyses found that habitat preferences are not significant relative to other traits, coral species preferring turbid, protected reef habitats and/or greater depths show a tendency to be less likely to be observed with a coral disease when analysed individually (Fig. 3a, b). This result might be because these species are already adapted for living in environments where pathogen levels are higher and sediment-shedding water motion is slower. It is hypothesised that corals living in turbid water are less sensitive to sedimentation and have mechanisms for dealing with excess sediment loads that might contain disease pathogens, decreasing infection rates (Lirman and Manzello 2009). On the other hand, the increased disease potential in species that prefer clear water and/or high wave exposure might be explained by the fact that they do not have the necessary adaptations to cope with increasing levels of pathogens in their environment. It is suggested that the majority of disease-causing pathogens are of terrestrial origin, either from terrestrial sediments or are coliforms coming from faecal contaminated waters that are dumped into the oceans without proper treatment (Frias-Lopez et al. 2003; Lesser et al. 2007; Francini-Filho et al. 2008). Depth showed no significant effects on disease potential in the analyses, but when analysed individually it had a significant effect. Studies have suggested that depth could potentially be an important variable because shallower waters tend to be warmer, which increases the spread and degree of virulence of the disease/pathogens (Jones et al. 2004; Selig et al. 2006; Bruno et al. 2007). Some diseases such as black band disease are more prevalent in shallower environments (<6 m; Page and Willis 2006), and so the grouping of disease types in our analyses would have obscured such patterns. Overall, habitat preference was not found to be an important driver of general disease susceptibility in corals on the GBR.

A lack of understanding about the causes of coral disease has prevented the detection of mechanistic links between disease occurrence and environmental perturbation (Work et al. 2008). Our species-level trait-based analysis is an important step towards better understanding disease potential and could be used to support conservation initiatives by highlighting species-level characteristics that render corals more prone to disease infection, as well as providing insights into modelling community vulnerability. Given the general lack of coral taxonomic resolution and disease prevalence metrics reported in the disease literature, and our subsequent grouping of diseases to increase statistical power, our analyses were not able to tease apart disease-specific patterns. Moreover, despite our best efforts, some uncertainty remains about how observational and literature biases might influence the reported patterns. Improved taxonomic resolution and the integration of disease aetiologies and ecologies and coral defence mechanisms (immune response) into our macroecological framework would further improve our understanding of coral disease ecology.

Acknowledgments

We thank the Juan Armando Sánchez for helpful input about conducting the phylogenetic analysis. We thank the Andrew Allen and Melanie Bishop for analytical advice, and Andrew Baird, Simon Davy, and three anonymous reviewers for insightful manuscript reviews. The study was supported by an Australian Research Council Discovery Project grant to JSM (DP0987892).

Supplementary material

338_2010_668_MOESM1_ESM.doc (69 kb)
Supplementary material 1 (DOC 69 kb)

Copyright information

© Springer-Verlag 2010