Culture-dependent and culture-independent analyses reveal no prokaryotic community shifts or recovery of Serratia marcescens in Acropora palmata with white pox disease

Authors

  • Michael P. Lesser,

    Corresponding author
    1. Department of Molecular, Cellular and Biomedical Science, University of New Hampshire, Durham, NH, USA
    • Correspondence: Michael P. Lesser, Department of Molecular, Cellular and Biomedical Science, University of New Hampshire, Durham, NH 03824, USA. Tel.: 603-862-3442; fax: 603-862-2621; e-mail: mpl@unh.edu

    Search for more papers by this author
  • Jessica K. Jarett

    1. Department of Molecular, Cellular and Biomedical Science, University of New Hampshire, Durham, NH, USA
    Search for more papers by this author

Abstract

Recently, the etiological agent of white pox (WP) disease, also known as acroporid serratiosis, in the endangered coral Acropora palmata is the enteric bacterium Serratia marcescens with the source being localized sewage release onto coastal coral reef communities. Here, we show that both culture-dependent and culture-independent approaches could not recover this bacterium from samples of tissue and mucus from A. palmata colonies affected by WP disease in the Bahamas, or seawater collected adjacent to A. palmata colonies. Additionally, a metagenetic 16S rRNA pyrosequencing study shows no significant difference in the bacterial communities of coral tissues with and without WP lesions. As recent studies have shown for other coral diseases, S. marcescens cannot be identified in all cases of WP disease in several geographically separated populations of A. palmata with the same set of signs. As a result, its identification as the etiological agent of WP disease, and cause of a reverse zoonosis, cannot be broadly supported. However, the prevalence of WP disease associated with S. marcescens does appear to be associated with proximity to population centers, and research efforts should be broadened to examine this association, and to identify other causes of this syndrome.

Introduction

Infectious, transmissible diseases of humans, wildlife, and important agriculture crops are increasing and are of significant concern (Daszak et al., 2000; Harvell et al., 1999, 2002; Altizer et al., 2013; Burge et al., 2014). Most of these diseases have a well-defined set of signs, a specific etiology, and a known epidemiology or epizootiology, as well as an investigative framework for studying and understanding these infectious diseases whether they represent the emergence of a new pathogen or an increase in the prevalence of a previously described pathogen. Additionally, the epidemiological framework to study infectious disease is quickly adopting a suite of new technologies, theories and modeling to better understand the etiology, transmission, and treatment of these diseases that often occur as epidemics/epizootics. Part of this framework has been an increased understanding for the role of global climate change (GCC) in the emergence of diseases, range expansions of disease vectors, increases in pathogen virulence, and changes in host immunity (Daszak et al., 2000; Harvell et al., 1999, 2002; Sokolow, 2009; Mydlarz et al., 2010; Burge et al., 2014).

The prevalence and incidence, varying spatially and temporally, of putative infectious, transmissible diseases in marine ecosystems has also increased (Harvell et al., 1999; Lafferty et al., 2004; Ward & Lafferty, 2004; Sokolow, 2009), and outbreaks have been reported in a broad range of marine taxa including corals (Harvell et al., 2004, 2007; Sutherland et al., 2004). Corals form the structural and biological framework of some of the most diverse, productive and economically important marine ecosystems in the world. Coral reefs have been, and continue to be, degraded at an alarming rate in the last few decades as a result of the interactive impacts of over-fishing, thermal stress (i.e. coral bleaching), eutrophication, sedimentation, pollution, and ocean acidification (Hoegh-Guldberg, 1999; Lesser, 2004; Hoegh-Guldberg et al., 2007). Infectious coral diseases, however, have been implicated as a major source of coral mortality worldwide but especially in the Caribbean (Weil et al., 2002; Bourne et al., 2010; Weil & Rogers, 2011; Rogers & Miller, 2013).

The microbiome of corals is being studied extensively for its diversity and functional roles in corals (Rosenberg et al., 2007; Ainsworth et al., 2009; Krediet et al., 2013), as well as its potential role in preventing infectious disease (Rosenberg et al., 2007; Bourne et al., 2009). It has also been argued (Lesser et al., 2007) that the most parsimonious interpretation of the available experimental data on coral diseases is that ‘infectious diseases’ of corals are a result of opportunistic, nonspecific bacteria that exploit the compromised health state of the coral or exhibit an increase in bacterial virulence after exposure to environmental stressors such as temperature stress (e.g. Vega-Thurber et al., 2009). The dynamic balance between the host and potential pathogen can be modulated by changes in the environment in the well-known disease triad (i.e. Sokolow, 2009). When host immunity is compromised, or virulence of a normally benign bacterium increases due to changes in the environment, it can result in a state of disease (Lesser et al., 2007; Rosenberg et al., 2007; Sokolow, 2009), but under normal environmental conditions these bacteria do not cause disease and are not infectious or transmissible (i.e. contagious). In fact, numerous publications support the significant role of the environment, and specifically thermal stress, in the incidence and prevalence of coral disease (Bruno & Selig, 2007; Rosenberg et al., 2007; Mora, 2008; Ainsworth et al., 2009; Brandt & McManus, 2009; Cróquer & Weil, 2009; Miller et al., 2009; Ruiz-Moreno et al., 2012) but see Ban et al. (2013) for a dissenting view for corals from the Great Barrier Reef. Additionally, recent models have shown that most coral diseases do not fit a contagion model but are better described as disease clustering (Yee et al., 2011; Muller & Van Woesik, 2012), which is more consistent with environmental effects being the prime drivers of coral disease caused by secondary, opportunistic infections. These studies do not support the hypothesis that many coral diseases are caused by transmissible, primary, infectious agents.

One widely reported case study of infectious disease, and a reverse zoonosis, is WP disease, synonymous with acroporid serratiosis, in the endangered coral, Acropora palmata (Patterson et al., 2002; Sutherland et al., 2011). While the prevalence of WP disease is clearly correlated with thermal stress and coral bleaching (Muller et al., 2008; Rogers & Muller, 2012), a putative primary pathogen has been identified as two strains (PDL100 and PDR60) of the enteric bacterium Serratia marcescens (Patterson et al., 2002; Sutherland and Ritchie, 2004; Sutherland et al., 2010, 2011). In showing that S. marcescens is the etiological agent of WP disease, an inoculum of 109 CFU mL−1 absorbed onto calcium carbonate sediment was used as the vector in infection studies and has been criticized for lack of ecological relevance (Lesser et al., 2007), but the practice of using an unrealistically high inoculum, with an abrasive fomite that breaks the integrity of the outer epithelium, continues to be used (e.g. Sussman et al., 2008; Ushijima et al., 2012). It is well known that S. marcescens is an opportunistic pathogen and common in marine habitats impacted by sewage, as is the case in the Florida Keys where most of the research on WP disease has been carried out (Patterson et al., 2002; Sutherland et al., 2010). Serratia marcescens cannot be isolated from all A. palmata colonies with signs of WP disease in the Florida Keys (Sutherland et al., 2010), leading to the early, and inconsistent, use of the term acroporid serratiosis to identify WP disease associated with S. marcescens (Patterson et al., 2002). But S. marcescens has also not been recovered from all A. palmata affected by WP disease in the Virgin Islands (Polson et al., 2008). While an earlier contagion model provided epizootiological evidence that WP is an infectious disease (Patterson et al., 2002), another more comprehensive contagion model does not support this assessment (Yee et al., 2011). The goal of this study was to examine WP disease of A. palmata in the Bahamas where the identified source (i.e. sewage effluent) of S. marcescens, and other potential enteric pathogens, is significantly reduced. Here, we show with the same culture-dependent approaches previously used, and culture-independent approaches, that WP disease. In the Bahamas, exhibiting the signs originally described for this putative disease (Patterson et al., 2002), are caused by S. marcescens.

Materials and methods

Site, sample collection, and processing

Corals were sampled at 1–3 m depth from an A. palmata-dominated patch reef known as Elkhorn Reef on the east side of Lee Stocking Island (LSI, 24°15′ N, 76°30′ W), Bahamas in spring 2012. Replicate coral samples (~ 5 cm2) were collected using a hammer and chisel from diseased (N = 3; ApD1, ApD2, ApD3) and healthy (N = 3; ApN1, ApN2, ApN3) portions of coral branches. Each paired coral sample (i.e. ApD1/ApN1) was collected from the same branch separated by at least 1 m, and each branch was from spatially separated (5–10 m) colonies. Samples were transported to the laboratory where they were maintained in 10-L aquaria with flowing ambient seawater. The prevalence of WP disease signs at this site was well over 90% (Lesser, unpublished qualitative transects), but no frank mortality that could be attributed to WP disease was noted. As a result, the protocol to sample the same branch for both diseased and healthy tissue was employed rather biased the sampling of diseased tissue from 90% of the available colonies and 10% from healthy colonies in this patch reef . All sampled colonies exhibited extensive multifocal, with some coalesced, 0.5–3.0 cm lesions varying in shape from circular to oblong to irregular as described previously for WP disease or acroporid serratiosis (e.g. Patterson et al., 2002, Sutherland et al., 2011; Work & Aeby, 2006) as well as areas completely free of any lesions. Seawater temperature at the time of the collection was between 26 and 28 °C varying on a diel basis (M. Lesser, unpublished). LSI is relatively undisturbed by human habitation (~ 45 km from the nearest population center in the Bahamas) with negligible levels of sewage discharge or terrestrial run off. In addition to coral samples, mucus samples were collected in sterile 30 mL syringes with sterile 18 gauge needles attached from the margins of lesions in diseased corals (ApDMu1, ApDMu2, ApDMu3) and from nondiseased areas of diseased corals described above (ApNMu1, ApNMu2, ApNMu3). Lastly, 1 L seawater samples (N = 1 for each) were collected 1 m (SW 1 m) and 10 m (SW 10 m) away from the same patch of A. palmata described above. All samples were processed within 2 h of collection.

The coral samples were trimmed to include the margins of the lesions using sterile bone cutters under sterile conditions, and pieces (~ 1 cm2) of coral were placed in 15 mL Falcon® tubes with DNA buffer (20% DMSO, 0.25 M EDTA, saturated NaCl) (Seutin et al., 1991) and frozen at −50 °C. Mucus samples were vigorously vortexed to reduce viscosity, and 25 mL of each sample was filtered onto 47 mm 0.22 μm GFF filters. The filters were then placed in 3.0 mL cryovials with DNA buffer and frozen at −50 °C. Similarly, 500 mL of each seawater sample was also filtered onto 47 mm 0.22 μm PVDF filters, placed in 3.0 mL cryovials with DNA buffer and frozen at −50 °C. These samples (corals, mucus and seawater) were transported to the University of New Hampshire where they were maintained at −70 °C until processed for DNA extraction as described below.

Culture studies

Of the remaining material, mucus and seawater samples (0.1 mL) were plated (N = 3 plates per individual sample) onto plates of general marine media; Zobells 2216E media (2216E; 5 g peptone, 1 g yeast extract, 0.1 g ferric citrate, 15 g agar in 1 L artificial seawater) and seawater complete media with glycerol (SWC; 5 g peptone, 3 g yeast extract, 3 mL glycerol, 15 g agar in 1 L artificial seawater) as well as selective media including MacConkey Sorbitol agar with colistin (200 U mL−1) (MCSA), deoxy-ribonuclease- toluidine blue agar amended with cephalothin (0.1 mg mL−1) (DTC) and thiosulfate citrate bile salts sucrose agar (TCBS) using spread plate techniques under sterile conditions. MCSA and DTC were used to recover and identify S. marcescens, while TCBS was used to recover Vibrio sp. that are common in seawater and known to consist of both primary and opportunistic pathogens. To be ecologically relevant, all plates were incubated at 26–28 °C, the temperature of the ambient seawater, for 48 h before plates were read. Lesions from the diseased samples described above were swabbed at their leading edge, while healthy coral samples were swabbed in an area completely absent of lesions in the laboratory. Sampling was performed with sterile cotton swabs and the material transferred to MCSA, DTC and TCBS plates and streaked for isolation. These plates were also incubated at the ambient seawater temperature range of 26–28 °C of the collection site for 48 h before being read. A second set of MCSA, DTC and TCBS plates with mucus, seawater and coral swab samples was incubated at 35 °C consistent with temperatures required for the clinical use of these selective media and prior use by other investigators for the isolation of S. marcescens from A. palmata (Patterson et al., 2002; Sutherland et al., 2010, 2011). For coral, mucus and seawater samples, any positive isolates on MCSA, DTC, or TCBS were then streaked for isolation on 2216E and SWC. Both MCSA and DTC agar were verified as selective media using American Type Culture Collection (ATCC) strains of S. marcescens (PDL100), while TCBS agar was verified using Vibrio fischeri (7744). Escherichia coli (K-12) was used as a negative control for all selective media.

Sequencing of bacterial isolates from A. palmata

Well-isolated colonies were sterilely collected from isolation streaks on 2216E or SWC media and DNA extracted using a CTAB procedure (France & Kocher, 1996). The PCR reaction of the 16S rRNA gene was carried out using primers 8F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGACTT-3′) in a mixture containing 0.25 μL Titanium Taq polymerase (Clontech, Mountain View, CA), 1 × Titanium Taq buffer, 0.2 mM dNTPs (Promega, Madison, WI), 0.4 μM of each primer and 1 μL of genomic DNA template. Reaction conditions were an initial denaturation for 5 min at 95 °C, 30 cycles of 95 °C for 30 s, 58 °C for 30 s, and 72 °C for 90 s, followed by 7 min at 72 °C. Technical replicates (N = 3) were run for all samples. PCR products were visualized on an agarose gel and identified by their predicted size (~ 1400 bp) in comparison with standard markers. PCR products were purified with a Qiaquick gel extraction kit and sequenced on an ABI3130 DNA analyzer, using the 8F primer, at the University of New Hampshire Hubbard Center for Genome Studies Sequencing Core Facility. Sequences were identified using blast (Altschul et al., 1990).

PCR-based screening for S. marcescens

Serratia marcescens PDL100 was obtained from the ATCC and grown in LB broth at 30 °C overnight with shaking. DNA was extracted from cultures using a Qiagen DNEasy tissue kit according to the manufacturer's instructions for Gram-negative bacterial cultures. DNA from cultures was used as a positive control in S. marcescens-specific PCR reactions on all coral tissue, mucus, and water samples. Additionally, for each sample, three technical replicates were run. The S. marcescens-specific primers Sm-456F (5′-GGTGAGCTTAATACGTTCATCA-3′) and UB-1492R (5′-TACGGYTACCTTGTTACGACTT-3′) and the reaction mixture described above were used with the thermal cycling protocol from Polson et al. (2008). A 10-fold dilution series of S. marcescens PDL100 DNA (60 ng μL−1) was used to determine the minimum detection limit of the reaction. PCR products were visualized on an agarose gel and identified by their predicted size (~ 1040 bp) in comparison with standard markers.

454 pyrosequencing and data analysis

To characterize prokaryotic communities associated with A. palmata mucus and water samples, the V5-V6 hypervariable regions of the16S rRNA gene were amplified with PCR using barcoded (Table S1) universal primers U789F and U1068R (Wang & Qian, 2009). Amplify (version 3.1.4) was used to test these primers in silico against 160 S. marcescens 16S rRNA sequences from the SILVA SSU database (release 115), and all but one (GenBank Accession Number DQ002385.1) of these sequences would be recovered by these primers. PCR products were purified and pooled as previously described (Fiore et al., 2013), and emulsion PCR using bidirectional multiplex 454 pyrosequencing was carried out using Titanium FLX chemistry at the W.M. Keck Center for Comparative and Functional Genomics at the University of Illinois, Urbana-Champaign. Sequencing was performed using an acyclic flow pattern (flow pattern B) initially on one half of a picotiter plate from the B adapters, and then on one half of a picotiter plate from the A adapters.

Sequences were analyzed using the Quantitative Insights Into Microbial Ecology (qiime) pipeline version 1.7.0 (Caporaso et al., 2010) on the Amazon Elastic Compute Cloud (EC2) except where noted. Sequences were filtered and clustered, and taxonomy was assigned to operational taxonomic units (OTUs). OTUs were checked for contaminants and chimeras as in Fiore et al. (2013). Additionally, OTUs assigned only to a domain (i.e. ‘Bacteria’, ‘Archaea’) or ‘unassigned’ OTUs were searched with blast against the NCBI nr database and removed from analysis if there were no significant hits (e-value < 0.001) or if the top hit was not against a16S rRNA sequence.

OTUs assigned to the phylum Cyanobacteria were subject to additional checking to ensure that chloroplast sequences were discarded and cyanobacterial sequences were retained. Cyanobacterial OTUs were searched against a custom database of cyanobacterial and chloroplast sequences (Janouskovec et al., 2012) using blast. Sequences with ambiguous results were aligned to a reference alignment (Janouskovec et al., 2012) using the profile–profile alignment option in muscle, and a tree was created using raxml-hpc v7.6.3 (Stamatakis, 2006). The placement of the OTU sequence in the tree was compared with the results from blast and ribosomal database project (RDP) to determine a final classification of the OTU. If a conclusive identity as cyanobacteria or chloroplast could not be reached, the sequence was retained in the dataset.

Rarefaction curves of observed OTUs were created for each sample type (coral tissue, coral mucus, and water), and then, samples were rarefied to an even sampling depth of 3500 sequences per sample prior to β-diversity analysis. A phylogenetic tree was created with fasttree 2 (Price et al., 2010) in qiime, and weighted unifrac distance values (Lozupone et al., 2007) were calculated. Unifrac distances were imported into PRIMER and used in permanova analyses, and to generate and plot principal coordinates (Clarke & Gorley, 2006).

Results

Culture-dependent results

Mucus and seawater samples had rare growth (< 5 colonies per plate) on MCSA or TCBS agar and no growth on DTC agar when grown at 26–28 °C or 35 °C. These same samples did exhibit significant growth on 2216E and SWC media with multiple colony types including both swarming and pigmented colonies. For both normal and diseased coral samples, MCSA agar recovered numerous colorless (negative fermentation of sorbitol), mucoid colonies with occasional swarming noted while no growth on DTC agar occurred. This is not indicative of the presence of S. marcescens. Similarly, TCBS agar had numerous colonies, but no colonies were TCBS positive (i.e. yellow-edged colonies indicating the fermentation of sucrose). This suggests the presence of sucrose-negative species of Vibrio. Similar colony morphologies were observed on 2216E and SWC for diseased and healthy corals as was observed for mucus and seawater. Where possible, individual bacterial colonies, or portions of swarming bacterial colonies, were selected from the plates incubated at 26–28 °C, streaked for isolation on 2216E and SWC media, and individual colonies picked for sequencing using 16S rRNA general eubacterial primers as described above. Visually, the colonies observed on plates incubated at 35 °C were not different from those grown at 26–28 °C and were not processed further.

Sequences of 16S rRNA (~ 1400 bp) for cultured isolates (N = 10) from diseased areas belonged to three genera (99–100% identity); Pseudomonas (N = 7), Xanthomonas (N = 1), and Stenotrophomonas (N = 2) all known to contain representatives of primary or secondary pathogens of humans and plants. Cultured isolates (N = 10) from healthy areas of A. palmata belonged to the genera Pseudomonas (N = 7), Sediminibacterium (N = 1), Sphingobacterium (N = 1), and a member of the Family Saprospraceae (N = 1). Despite the fact that Vibrio sp. were identified on TCBS agar, no 16S rRNA sequences were identified as Vibrio sp.

Culture-independent results

The PCR screening for S. marcescens was positive for the WP disease control isolate PLD100, and the detection limit was in the fg range (Fig. S1; Lane 6 [6 fg]). All samples of mucus, seawater, and coral (healthy and diseased) were screened for the presence of S. marcescens, and all were negative. It is important to note that this PCR-based approach would have detected both, but not distinguished between, the PDL100 and PDR 60 strains of S. marcescens.

Four hundred and fifty-four pyrosequencing resulted in 949 758 total reads, 247 841 of which passed quality filters and 145 174 of which were nonsingletons. Rarefaction curves of OTUs (Fig. S2) were curvilinear and generally reached an asymptote for all samples at the common sampling depth used for all downstream analyses. The sequence length of the nonsingleton reads was 247 ± 22 bp (SD). Clustering at a similarity threshold of 97% produced 1081 OTUs; five of which were flagged as chimeras by UCHIME, 291 could not be aligned or were not 16S rRNA sequences, one was an apparent contaminant, and 27 were chloroplast OTUs for a total of 324 discarded OTUs. The remaining 757 OTUs were used in all downstream analyses. The observed OTUs in each sample ranged from ~ 100 to 330 (Table S2). Subsampling of OTUs after normalization to a common sampling depth resulted in a similar range of OTUs for all samples (~ 80–230, Table S2). Both the Shannon and Simpson indices of β-diversity reveal that the mucus and seawater samples are more diverse than the coral tissues, whether from diseased or healthy tissues (Table S2).

A total of 25 phyla and 76 families of prokaryotes were identified from all samples. With even sampling, four phyla and 12 families were lost, resulting in the analysis of 21 phyla and 62 families. The resulting communities in A. palmata tissues consisted of ~ 41 families (as opposed to 47 families without incorporating even sampling) of Eubacteria and Archaea. All coral tissue samples were dominated by Gammaproteobacteria (80–89% of reads); the family Endozoicimonaceae, most likely representing the Genus Endozoicomonas, was the most abundant, but Vibrionaceae, including Vibrio and Photobacterium, and Alteromonadales were also prominent representatives of the Gammaproteobacteria (Fig. 1). Among the other Proteobacteria, the Camplyobacteraceae (Epsilonproteobacteria) were the most numerous, but a small complement of Alphaproteobacteria was also observed. Gracilibacteria (formerly candidate phylum GN02 (Rinke et al., 2013) were a minor but consistently recovered component of the community. The microbial communities of mucus and seawater samples consisted of ~ 53 families of Eubacteria and Archaea (as opposed to 70 families without incorporating even sampling) and were dominated by Alphaproteobacteria with typical members of the seawater community such as Pelagibacteraceae (SAR11 clade), Order Rickettsiales, and Cyanobacteria (Prochlorococcus) (Fig. 1).

Figure 1.

Taxonomic classification of 16S rRNA OTUs from Acropora palmata tissue, mucus and the adjacent water column using 454 pyrosequencing. OTUs were classified at a confidence threshold of 80%. For sample abbreviations, see 'Materials and methods'.

The statistical analysis of the weighted unifrac distance values revealed no significant difference (permanova: F1,13 = 0.58, = 0.647) in the microbial communities between diseased and healthy corals. The mucus and seawater samples were significantly different than the coral tissues (permanova: F1,13 = 324.8, = 0.0002), with pairwise comparisons revealing a significant difference between the mucus and tissue samples (= 0.003). Additional pairwise testing of unweighted unifrac distances using two sample, two-tailed t-tests with Monte Carlo permutations (n = 10 000) resulted in nonparametric, Bonferroni-corrected P values, which showed that the communities from the seawater and mucus were not significantly different from each other (> 0.05), but were significantly different from all tissue samples (= 0.002 for all comparisons), which were also not significantly different from each other (> 0.05) (Fig. 2). These differences were clearly seen by an ordination of principal coordinates (Fig. 3) and were not due to a difference in dispersion between sample types (PERMDISP: F2,11 = 1.2731, = 0.6132). Pyrosequencing reads and the resulting OTU sequences were submitted to the CAMERA (Cyberinfrastructure for Microbial Ecology Research and Analysis, http://camera.calit2.net/) website under project Accession Number CAM_P_0001135.

Figure 2.

Boxplot of weighted unifrac distance values showing the pairwise analysis using two sample, two-tailed t-tests with Monte Carlo permutations (n = 10 000). Common superscripts indicate groups not significantly different from each other.

Figure 3.

Principal coordinates plot from pyrosequencing of prokaryotic communities from Acropora palmata tissue and mucus and water samples. Analysis is based on weighted unifrac distance values derived from 16S rRNA OTUs.

Discussion

For Elkhorn Reef on Lee Stocking Island, the signs of WP disease at the time of these collections were significant, but only represent a single snapshot in time. The bacterial communities of the healthy and diseased tissues from these A. palmata colonies were not significantly different from each other as observed for other coral diseases (Sunagawa et al., 2009), but were distinct from the mucus and water column samples using 454 pyrosequencing of the 16S rRNA gene. The microbiome observed in A. palmata tissues from Lee Stocking Island was composed of many of the same phyla found by previous investigators (Sunagawa et al., 2010), but differences in sample processing, primers, and analysis in our study most likely led to the recovery of a more diverse prokaryotic community with more Proteobacteria, especially Endozoicomonaceae and Vibrionacaeae, and less Bacteroidetes and Firmicutes than observed in A. palmata from Panama. Endozoicomonas has been shown to be tissue associated in corals (Bayer et al., 2013), so this difference may be attributable to the different DNA extraction protocol used by Sunagawa et al. (2010).

The absence of any change in the bacterial communities between diseased and healthy tissues is similar to that reported for Acropora sp. from the Great Barrier Reef (Ainsworth et al., 2007). This observation is consistent with programmed cell death (i.e. apoptosis) of the host cells reported by Ainsworth et al., 2007 and warrants additional investigation. Additionally, the population of A. palmata studied here has the same Symbiodinium phylotype (A3) identified in populations throughout the Caribbean and Bahamas including the Florida Keys (Thornhill et al., 2006) such that its effect on the host should not be a potential confounding feature of this disease in the Bahamas. Furthermore, the host population genetics shows a clustering of populations from the Bahamas and Florida Keys, indicating some level of connectivity across this region (Baums et al., 2005). This suggests that the two populations are not significantly differentiated genetically and that host-related differences in susceptibility to WP disease are unlikely.

Not only was the putative pathogen of WP disease, S. marcescens, not recovered, but the mucus and surrounding seawater did not contain any taxa that might indicate the presence of human sewage (e.g. Lipp et al., 2002). As a result, the observed WP disease on LSI should not be referred to as acroporid serratiosis sunsu stricto as suggested by Patterson et al. (2002) despite the fact that the terms are often used interchangeably (e.g. Sutherland et al., 2011) and only add confusion rather than clarity in trying to identify the cause(s) of WP disease. Despite the reported impact of infectious coral diseases on the mortality of corals, the etiologies of most coral diseases remain unknown (Richardson, 1998; Sutherland et al., 2004; Lesser et al., 2007; Bourne et al., 2010). The etiological agent (s) of only a few coral diseases are known such as black band disease (Voss & Richardson, 2006). But the diagnosis of black band disease requires both robust microbiological practices and relevant physiological and environmental measurements (Richardson et al., 2001) for its diagnosis. All investigations of coral disease should consider a similar approach and measurements regardless of whether some or all of these practices originate from the veterinary or biomedical community, or not (sensu Work et al., 2008).

As previously reported by Lesser et al. (2007), the prevalence of coral disease throughout the Caribbean is low and the loss of acroporid corals from other causes, especially in the Florida Keys, is significant (e.g. Porter et al., 1982, 1999, 2012) and consistently overlooked. Despite the fact that coral reefs in Florida experience seasonal increases in seawater temperatures above their historical maximum mean monthly average, and increases in WP disease are associated with these elevated seawater temperatures (Patterson et al., 2002), this coral disease continues to be identified as a primary infectious disease associated with sewage that is transmissible (Sutherland and Ritchie, 2004; Sutherland et al., 2004, 2010, 2011). Additionally, it is important to recognize that recent evidence strongly supports a significant role for anthropogenically related mortality in corals that predates both coral bleaching and disease (Cramer et al., 2012) and probably involved multiple tipping points and subsequently unrecognized regime shifts in coral communities (Hughes et al., 2013).

If you cannot isolate the putative pathogen in all cases of WP-disease-affected colonies of A. palmata, what does that mean? What is the utility of using the terms WP disease and acroporid serratiosis that are described by the same signs and can only be differentiated by the presence or absence of S. marcescens? First, and foremost, if you are adhering to the tenets of Koch's postulates, then not being able to isolate the agent using either culture or nonculture-dependent approaches from every case with similar signs suggests that other agent(s) and causes should be investigated. The problem is, and has always been, that many coral diseases are described by a limited number of very similar signs and open to wide diagnostic interpretation (Work & Aeby, 2006). It is therefore possible that many of the diagnosed cases of WP disease are the result of multiple primary pathogens, opportunistic pathogens secondary to stress, or some stress response not involving any pathogen (e.g. apoptosis). Is there other evidence that may be more parsimonious in explaining the prevalence of WP disease represented by a specific set of associated signs with and without significant mortality? It is clear from the available data that the majority of WP disease is associated with coastal reefs and sewage input, but is that factor alone ‘necessary’ and ‘sufficient’ (sensu Sokolow, 2009) to cause disease? Fecal contamination from sewage on coral reefs has been shown for the Florida Keys (Lipp et al., 2002; Lapointe et al., 2004), the US Virgin Islands (Kaczmarsky et al., 2005), and Puerto Rico (Hernández-Delgado et al., 2008; Bonkosky et al., 2009). This suggests that human activities are a common theme with WP disease, and those coral reefs in nearshore waters close to significant population centers are likely to have more microorganisms, including potential pathogens, analogous to what has been observed along a gradient of human population densities in the Northern Line Islands (Dinsdale et al., 2008). The WP disease reported here is from the Bahamas (Lee Stocking Island) and is 45 km from the nearest population center with very low concentrations of dissolved inorganic nutrients (Voss & Richardson, 2006; Gochfeld et al., 2012). Also, the reefs around Lee Stocking Island have significantly lower concentrations of dissolved organic matter (DOM) compared with the Florida Keys (Lesser & Mobley, 2007; Zepp et al., 2008), a potential source of physiological stress for corals that is believed to potentially lead to opportunistic coral diseases (Barott & Rohwer, 2012). The water quality around the Bahamas compared with the Florida Keys, and the absence of S. marcescens from the tissues and mucus of A. palmata and the surrounding waters, in colonies with clear signs of WP disease, suggests that exposure to sewage is not ‘necessary’ to cause disease as previously described and should stimulate continued research into alternate hypotheses about the etiology of WP disease. The investigation of human activities (e.g. sewage) should still figure prominently in these studies, but other components of sewage (e.g. DOM loading) should also be investigated.

Understanding whether a disease is caused by a primary infectious agent or an opportunistic one secondary to stress should be part of any microbiological investigation of coral disease (Lesser et al., 2007), and doing so would not be controversial (Weil & Rogers, 2011), generate confusion (Work et al., 2008) nor is it a diversion (Pollock et al., 2011). It is essential for not only understanding the etiology and pathobiology of a disease state but for what, if anything, can be done about it. The data presented here show that for a population of A. palmata exhibiting signs of WP disease from a relatively pristine site in the Bahamas, the putative causative agent could not be recovered using the same techniques used to initially identify S. marcescens as the causative agent (Patterson et al., 2002; Sutherland and Ritchie, 2004; Sutherland et al., 2010, 2011). This result is similar to other coral diseases putatively caused by an identified etiological agent of an infectious nature where that agent could not be isolated from all corals with the same signs of the disease (Ainsworth et al., 2008; Sweet & Bythell, 2012; Apprill et al., 2013; Cook et al., 2013). Describing WP disease as being of known etiology and a case of reverse zoonosis cannot be supported as S. marcescens cannot be isolated or identified from all locations or from all colonies with the same set of signs for WP disease. Therefore, and until more data become available, corals identified with signs of WP disease should be referred to as having WP syndrome (sensu Lesser et al., 2007), and alternate hypotheses on its etiology investigated.

Acknowledgements

This research was supported by NSF (IOS-1231468) and complies with all applicable laws of the United States and the Bahamas. We thank Seana Fraser for help with sequencing of bacterial isolates. This article was improved by comments from Marc Slattery, Deborah Gochfeld, and three anonymous reviewers.

Ancillary